Nonlinear vibration of piezo-laminated plates

Nonlinear free vibration analysis of simply supported
piezo-laminated plates with random actuation electric

potential dierence and material properties
K. Jayakumar
a
, D. Yadav
a,
*
, B. Nageswara Rao
b
a
Department of Aerospace Engineering, Indian Institute of Technology Kanpur, Kanpur 208 016, India
b
Structural Analysis and Testing Group, Vikram Sarabhai Space Centre, Trivandrum 695 022, India
Received 12 January 2007; received in revised form 4 June 2007; accepted 10 February 2008
Available online 21 February 2008
Abstract
Studies are made on nonlinear free vibrations of simply supported piezo-laminated rectangular plates with immovable
edges utilizing Kirchos hypothesis and von Karman straindisplacement relations. The eect of random material prop-
erties of the base structure and actuation electric potential dierence on the nonlinear free vibration of the plate is exam-
ined. The study is conned to linear-induced strain in the piezoelectric layer applicable to low electric elds. The von
Karmans large deection equations for generally laminated elastic plates are derived in terms of stress function and trans-
verse deection function. A deection function satisfying the simply supported boundary conditions is assumed and a
stress function is then obtained after solving the compatibility equation. Applying the modied Galerkins method to
the governing nonlinear partial dierential equations, a modal equation of Dungs type is obtained. It is solved by exact
integration. Monte Carlo simulation has been carried out to examine the response statistics considering the material prop-
erties and actuation electric potential dierence of the piezoelectric layer as random variables. The extremal values of
response are also evaluated utilizing the Convex model as well as the Multivariate method. Results obtained through
the dierent statistical approaches are found to be in good agreement with each other.
2008 Elsevier B.V. All rights reserved.
PACS: 46.40.f; 05.40.a; 85.50.n; 05.10.Ln
Keywords: Convex model; Dungs equation; Free vibration; Induced strain actuation; Monte Carlo simulation; Multivariate concept;
Piezo-laminated plate; Random variables
1. Introduction
Composite materials are widely used in various types of engineering structures. These materials are in
general subject to certain amount of scatter in their elastic moduli due to many factors. Some of these are: mis-
alignments of bers, ber pull out, imperfect bonding between the bers and the matrix, voids, run out, etc
1007-5704/$ - see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.cnsns.2008.02.003
*
Corresponding author. Tel.: +91 512 2597951; fax: +91 512 2597626.
E-mail address: dy@iitk.ac.in (D. Yadav).
Available online at www.sciencedirect.com
Communications in Nonlinear Science and Numerical Simulation 14 (2009) 16461663
www.elsevier.com/locate/cnsns
along with variation in the material properties of the matrix and the ber. Properties of composite materials
manufactured by the same process inherently demonstrate dierences in their elastic properties. It is very dif-
cult to completely control all the parameters involved in the manufacturing process. For design purposes, it is
essential to knowthe potential variations in the structural response due to the uncertainties in the elastic moduli
[1]. This assumes special importance in designs with low margin for error like aerospace applications.
The design margin is small in sensitive applications. Lack of adequate knowledge of the system behavior
may result in failure of the design. This has motivated many researchers to study problems with randomness
in structural properties. The vibration and buckling response due to variation in the ber spacing were studied
by Leissa and Martin [2]. Shinozuka and Astill [3] obtained statistical properties of eigenvalues of spring sup-
ported columns with the support and axial loading along with material and geometric properties as random.
Chen and Soroka [4] studied multi-degrees of freedom system with random system properties subjected to
deterministic external excitations. Yadav and Verma [5] obtained the buckling response of thin cylindrical
shell with random material properties, using classical laminate theory. Second-order response statistics was
evaluated by adopting rst-order perturbation technique. All the studies of composite structure in random
environment are conned to small displacements in the linear domain. Large amplitude dynamic response
of composite plates with random material properties has not been addressed fully.
Application of smart structure technologies to aerospace and other systems are expanding rapidly. One of
the basic elements of adaptive structures is a thin composite plate with surface bonded or embedded sheet
actuators. With tailored laminated plate, induced strain actuation can control its extension, bending and
twisting behavior. Plates with distributed-induced strain actuators can be used: to control pointing of preci-
Nomenclature
a and b length and width of the plate
A, B and D matrices of membrane, coupling and exural stiness of the plate
f(f) restoring force function
h thickness of plate in z direction
h
i
thickness of ith lamina
h
p
thickness of piezoelectric layer
I(f) potential energy function
M moment resultants
M
K
actuator moments
N stress resultants
N
K
actuator forces
Q matrix of reduced stiness coecients
Q
T
transformed reduced stiness matrix
t time
u, v and w displacements in x, y and z directions
W
mn
(t) time dependent amplitude
e total strain
e
0
reference plane strain (i.e., strain at z = 0)
f amplitude to thickness ratio
f
S
specied amplitude to thickness ratio
j curvature changes
K actuation strains
q
i
density of ith lamina
r stress
u Airys stress function
d/
a
applied electrical potential dierence
x nonlinear frequency
x
L
linear frequency
K. Jayakumar et al. / Communications in Nonlinear Science and Numerical Simulation 14 (2009) 16461663 1647
sion instruments in space; to control structural noise; and to change aerodynamic shape for vibration reduc-
tion [6], utter suppression and gust alleviation. Classical laminated plate theories have been developed to pre-
dict exural response of laminated plates with surface bonded or embedded-induced strain actuators utilizing
the advanced models based on assumed strain eld [7,8].
In the present study, nonlinear free vibration analysis is carried out for a generally laminated composite
plate (Fig. 1), with any arbitrary layer as actuator/sensor that is composed of piezoelectric material. The
von Karmans straindisplacement relationship is used to obtain the governing dierential equation in terms
of transverse displacement and stress function. A linear model is considered for induced strain in the piezo-
electric layer applicable to low elds. The boundary conditions of the plate are simply supported along all
the four immovable edges. The formulation of the theory is based on the assumption that strains and square
of the rotations are small compared to unity. von Karman plate theory predicts deections and stresses in a
thin plate with reasonable accuracy for deection having the order of the plate thickness. This can be used for
small strains but moderately large rotations, and can be termed as the moderately large deection theory. A
deection function satisfying boundary conditions is assumed and a stress function is then obtained after solv-
ing the compatibility equation. Modied Galerkins method is applied to the governing nonlinear partial dif-
ferential equations to study the nonlinear vibration characteristics of piezo-laminated thin rectangular plates.
The solution is validated by comparing published results in the literature on isotropic and conventional com-
posite plates with the corresponding degenerate case of the present formulation for piezo-laminated plate (i.e.,
in the absence of smart piezoelectric layer). Case studies are performed on piezo-laminated plates. Monte Car-
lo simulation has been carried out considering the material properties and actuation electric potential dier-
ence as random variables to examine the nonlinear free vibrations of a simply supported plate. The mean and
variance of the system frequency response are calculated for a specied material properties and actuator elec-
tric potential dierence statistics. The extremal statistical properties in the response of the piezo-laminated
plate are also examined following the Multivariate method and Convex modeling.
2. Formulation
2.1. System equations
Adaptive structures or smart structures are usually constructed of a thin composite plate with surface
bonded or embedded sheet actuators. Fig. 1 shows a thin rectangular plate of length a in the x direction, width
Fig. 1. Conguration of a simply supported piezo-laminated plate with immovable edges.
1648 K. Jayakumar et al. / Communications in Nonlinear Science and Numerical Simulation 14 (2009) 16461663
b in the y direction and thickness h, in the z direction. The z-axis is directed normal to the reference plane
(z = 0). The thickness of the plate is assumed small in comparison with its smallest lateral dimension (i.e.,
h (a or h (b). The plate is composed of an arbitrary number of anisotropic layers, arrangement and thick-
ness. With respect to the reference plane, u, v and w are displacements in x, y and z directions, respectively. The
governing equations are based on the assumptions of von Karmans elastic thin plate theory. Three additional
assumptions are added. First, there is no slip between the adjacent layers of the laminated plate. Second,
rotary inertia and transverse shear eects are neglected. Third, kinematics relations (ou/ox)
2
and (ov/oy)
2
are neglected as compared with (ow/ox)
2
and (ow/oy)
2
terms.
For a generic coupled laminated plate with surface bonded or embedded-induced strain actuators (piezoce-
ramic sheets) placed at arbitrary locations in z direction, the total strain is expressed using Kirchos hypoth-
esis of classical thin plates as
e e
0
zj 1
The matrices of strains (e
0
) and curvature changes (j) are written by considering von Karman type of geomet-
ric nonlinearity, as [9,10]
e
0
e
0
x
e
0
y
e
0
xy
_
_
_
ou
ox
1
2
ow
ox
_ _
2
ov
oy
1
2
ow
oy
_ _
2
ou
oy
ov
ox
ow
ox
ow
oy
_
_
_
_
2
j
j
x
j
y
j
xy
_
_
_
_

o
2
w
ox
2
o
2
w
oy
2
2
o
2
w
oxoy
_
_
_
_
3
The constitutive relation for any ply of a laminated plate is
r Q
T
e K 4
with components in the stress matrix, r = [r
x
, r
y
, r
xy
]
T
and the actuation strain matrix, K = [K
x
, K
y
, K
xy
]
T
. A
linear model for induced strain in the piezoelectric layer is utilized in Eq. (4), which is a reasonable approx-
imation applicable to low elds. Proper care is taken when constitutive equations for piezoelectric materials
are used to model-induced strain actuation, as they are poling direction dependent. Referring to Fig. 1, the
planar isotropy of poled ceramics is expressed by their piezoelectric strain constants, such that d
31
= d
32
.
The applied static electric eld within the piezoelectric actuator is assumed to be constant as the thickness
of this layer is relatively small. When an electric potential dierence is introduced across the thickness of
the piezoelectric layer (actuator layer), it is strained in its plane. For PZT-5A the in-plane free strain compo-
nents K
x
and K
y
developed in the kth actuator layer, having a charge constant d
31
and thickness h
p
, when an
electric potential dierence d/
a
is applied are: K
x
K
y

d
31
d/
a
h
p
and K
xy
= 0.
Matrix Q
T
is the transformed reduced stiness of the plate and Q
T
K represents an equivalent stress due to
actuation. Integrating the stressstrain relations through the thickness of the plate for net forces and moments,
one obtains
N
M
_ _
A B
B D
_ _
e
0
j
_ _
N
K
M
K
_ _
5
where
N
k
; M
k

_
h=2
h=2
1; zr
k
dz k x; y; xy 6
N
x
, N
y
, N
xy
are membrane forces per unit length and M
x
, M
y
, M
xy
are the bending and twisting moments per
unit length. The element A
ij
, B
ij
and D
ij
(i, j = 1, 2, 6) in the 3 3 symmetric matrices A, B and D in Eq. (5) are
dened as
A
ij
; B
ij
; D
ij

_
h=2
h=2
1; z; z
2
Q
ij
dz i; j 1; 2; 6 7
For symmetric anisotropic laminates, the material coupling does not occur between transverse bending and in-
plane stretching, viz., B
ij
= 0, whereas in the case of non-symmetric laminated plates B
ij
0 [9,10].
The components in the matrices of the actuator forces and moments are dened as
N
Kk
; M
Kk

_
h=2
h=2
1; zQ
T
Kdz k x; y; xy 8
Nonlinear equations of motion of generally laminated plates are
oN
x
ox

oN
xy
oy
0 9
oN
xy
ox

oN
y
oy
0 10
o
2
M
x
ox
2
2
o
2
M
xy
oxoy

o
2
M
y
oy
2
N
x
o
2
w
ox
2
2N
xy
o
2
w
oxoy
N
y
o
2
w
oy
2

n
i1
q
i
h
i
o
2
w
ot
2
11
where q
i
and h
i
are the density and thickness of the ith layer, and n is the number of layers in the plate.
The partially inverted form of constitutive equation (Eq. (5)) is written as
e
0
MM
K
_ _
T
D
_ _
NN
K
j
_ _
12
where A
*
= A
1
, B
*
= A
1
B, and D
*
= D + BB
*
.
The Airy stress function u, which satises Eqs. (9) and (10) is dened by
N
x

o
2
u
oy
2
13
N
y

o
2
u
ox
2
14
N
xy

o
2
u
oxoy
15
The compatibility equation is derived from relation (2) as
o
2
e
0
x
oy
2

o
2
e
0
y
ox
2

o
2
e
0
xy
oxoy

o
2
w
oxoy
_ _2
o
2
w
ox
2
o
2
w
oy
2
16
Using Eqs. (12)(15) in Eqs. (11) and (16), one obtains
n
i1
q
i
h
i
o
2
w
ot
2
L
1
w L
3
u Lu; w 0 17
L
2
u L
3
w
o
2
w
oxoy
_ _2
o
2
w
ox
2
o
2
w
oy
2
0 18
where the dierential operators are:
L
1
D
11
o
4
ox
4
4D
16
o
4
ox
3
oy
2D
12
2D
66
o
4
ox
2
oy
2
4D
26
o
4
oxoy
3
D
22
o
4
oy
4
L
2
A
22
o
4
ox
4
2A
26
o
4
ox
3
oy
2A
12
A
66
o
4
ox
2
oy
2
2A
16
o
4
oxoy
3
A
11
o
4
oy
4
L
3
B
21
o
4
ox
4
2B
26
B
61
o
4
ox
3
oy
B
11
B
22
2B
66
o
4
ox
2
oy
2
2B
16
B
62
o
4
oxoy
3
B
12
o
4
oy
4
L/; w
o
2
/
oy
2
o
2
w
ox
2

o
2
/
ox
2
o
2
w
oy
2
2
o
2
/
oxoy
o
2
w
oxoy
Eqs. (17) and (18) are the two coupled governing equations of arbitrarily piezo-laminated thin plates.
2.2. Solution approach
The solution for large amplitude vibrations of simply supported generic coupled laminated rectangular
plate with surface bonded or embedded-induced strain actuators are examined here by applying the Galerkins
method.
Boundary conditions for simply supported rectangular plates are
w 0; M
x
0 at x 0; a 19
w 0; M
y
0 at y 0; b 20
The in-plane boundary conditions for immovable edges are
_
a
0
e
x
1
2
ow
ox
_ _
2
_ _
y0;b
dx 0 21
_
a
0
fN
xy
g
y0;b
dx 0 22
_
b
0
e
y
1
2
ow
oy
_ _
2
_ _
x0;a
dy 0 23
_
b
0
fN
xy
g
x0;a
dy 0 24
The system of Eqs. (17) and (18) in terms of the transverse deection w and stress function (u), are to be solved
in conjunction with the boundary conditions (19)(24). A deection shape function for w satisfying the geo-
metric boundary conditions of the plate is assumed. A stress function (u) is then obtained from the compat-
ibility equation (18). Galerkins method is applied to the governing nonlinear partial dierential equation to
yield a second-order ordinary nonlinear dierential equation of motion in time variable. The details of the
solution steps of the problem are outlined below.
The transverse supporting conditions given in (19) and (20) are satised by assuming deection functions of
the laminate in the separable form corresponding to the axial and transverse wave numbers m, n as
w W
mn
t sina
m
x sinb
n
y 25
Here, a
m

mp
a
and b
n

np
b
.
Substituting the transverse deection, w into Eq. (18), and utilizing the in-plane boundary conditions, the
stress function (u) is obtained as
u u
0
x; y W
mn
u
1
x; y W
2
mn
u
2
x; y 26
The expressions for the functions u
0
, u
1
and u
2
are given in Appendix A. It should be noted that for symmet-
ric anisotropic laminates, the bendingstretching coupling matrix B = 0 and D
*
= D in Eq. (12). For this case
u
1
(x, y) as dened in Appendix A, vanishes and Eq. (26) for Airys potential is independent of linear term in
W
mn
.
When w and u are substituted into Eqs. (19) and (20), the force boundary conditions are not satised when
B
*
does not equal zero. In such a situation usage of the modied Galerkins method [11] is more appropriate.
In the case of simply supported rectangular plates assuming w sin a
m
x sin b
n
y, the residual force and moment
have the following relation:
_
a
0
_
b
0
L
R
/; w wdxdy
_
b
0
M
x
o w
ox
_ _
x0
dy
_
b
0
M
x
o w
ox
_ _
xa
dy
_
a
0
M
y
o w
oy
_ _
y0
dx
_
a
0
M
y
o w
oy
_ _
yb
dx 0 27
where L
R
(u, w) is the residual force.
Letting L
N
(u, w) = L
R
(u, w) + L(u, w), Eq. (17) becomes
n
i1
q
i
h
i
o
2
w
ot
2
L
1
w L
3
u L
N
u; w 0 28
Applying Galerkins method to Eq. (28), the modal equation is then obtained as
n
i1
q
i
h
i
d
2
W
mn
dt
2
a
K
aW
mn
bW
2
mn
cW
3
mn
0 29
The constants a
K
, a, b and c in the equation of motion (Eq. (29)) are dened in Appendix B. For symmetric
anisotropic laminates, B
*
= 0, which implies that the constant coecient b = 0 in the modal equation (29).
Hence, the quadratic term in the restoring force function of equation of motion vanishes for symmetric lam-
inates. Then, Eq. (29) has a closed-form solution in terms of Jacobian elliptic function; the values of which can
be obtained from the mathematical tables of the Handbook [12] or MATLAB
.
Dening f = W
mn
/h and s = xt, the modal equation (29) can be written in the form
x
2
f x
2
L
f f 0 30
where the restoring force function is
f f d
0
f d
1
f
2
d
2
f
3
31
and x
L

n
i1
q
i
h
i
_
is the linear frequency of the plate, d
0

a
K
ah
, d
1

bh
a
, d
2

ch
2
a
, x is the nonlinear frequency
and over-dot denotes dierentiation with respect to s.
The initial conditions for Eq. (30) are
f f
S
;
_
f 0 at s 0 32
Here f
S
is the specied maximum amplitude to thickness ratio for the plate.
Solution for the nonlinear frequency of the plate, which is a function of material properties, electric poten-
tial dierence, dimensions of the plate and the amplitude of vibration is obtained through exact integration.
If d
0
= 0 and d
1
= 0, then f(f) becomes an odd function, and the magnitudes of maximum positive and neg-
ative amplitudes in the periodic motion will be equal. In the case of mixed-parity (i.e., d
0
0 and d
1
0), f(f)
is a non-odd function and the magnitudes of maximum positive and negative amplitudes in the periodic
motion will be dierent. Hence, for non-odd function f(f) the behavior of oscillations is dierent for the posi-
tive and negative amplitudes, and consequently, the frequency values for the specied maximum positive and
negative amplitudes having the same magnitude, will be dierent.
The relationship between the maximum positive and negative amplitudes, viz. f
+
and f
, can be found by
equating the potential energies in either position,
If
If
33
where
In
_
n
0
f gdg d
0
n
1
2
n
2
d
1
3
n
3
d
2
4
n
4
34
Multiplying Eq. (30) by df/ds and integrating,
1
2
x
2
df
ds
_ _
2
x
2
L
fIf If
g 0 35
The initial conditions used while integrating Eq. (30) to obtain (35) are at s = 0:
f f
and
_
f 0 36
The solution curve on the f
_
f plane is the integral curve or the phase trajectory. In the periodic motion of the
system, the solution curve on the f
_
f plane is a closed trajectory [13]. If d
0
= 0 and d
1
= 0, f(f) is an odd
function and I(f
+
) is an even function. The solution curve of Eq. (35) will be symmetric on f and
_
f axis. If
d
0
0 and d
1
0, the solution curve of Eq. (35) will be symmetric only on the f axis.
Integrating Eq. (35) from s = 0 to s = p, one gets
x
L
x

1
2
p
p
_
f
df
If
If
_ 37
For a specied maximum positive amplitude-to-thickness ratio f
+
the corresponding maximum negative
amplitude-to-thickness ratio f
is obtained from Eq. (33) and vice versa. By substituting for f
and f
+
in
Eq. (37) the nonlinear frequency x is obtained. Since f
and f
+
are the roots of I(f
) I(f) = 0, one can

express
If
If
1
2
f f
fc
0
c
1
f c
2
f
2
38
The constants c
0
, c
1
and c
2
in Eq. (38) are obtained in terms of d
1
, d
2
, f
+
and f
by comparing the coecients

of f
2
, f
3
and f
4
. The integrand in Eq. (37) has poles at the end of integration (i.e., at f = f
and f
+
), which may
adversely aect the accuracy of an integration rule. Hence, the integrand in Eq. (37), is modied by using
f f
1
f
2
cos
1
2
p1 n
_ _
39
where f
1

1
2
f
; f
2

1
2
f
.
That eliminates the singularities and yields a form
x
x
L
1
2
_
n1
n1
dn
c
0
c
1
f c
2
f
2
_
_ _
1
40
where c
0
1
4
3
d
1
f
1
1
2
d
2
3f
2
1
f
2
2
, c
1

2
3
d
1
d
2
f
1
, c
2

1
2
d
2
.
The negative amplitude-to-thickness ratio f
corresponding to the positive amplitude-to-thickness ratio f

+
obtained from Eq. (33) is as detailed below.
Eq. (33) can be written in the form
1
2
f
a
3
f
3
3a
2
f
2
3a
1
f
a
0
0 41
Since, f
+
f
, the unknown f
has to be obtained from the cubic equation:

a
3
f
3
3a
2
f
2
3a
1
f
a
0
0 42
One real root of the cubic equation (42) is [12]:
f

s
1
s
2
a
2
a
3
43
where
s
1;2

q
3
r
2
_
3
_
; q a
3
a
1
a
2
2
; r
1
2
3a
3
a
2
a
1
a
2
3
a
0
a
3
2
a
3
c
2
; a
2

1
3
2
3
d
1
a
3
f
_ _
; a
1

1
3
a
2
f
and a
0
3a
1
f
2d
0
It should be noted that for the present problem, q
3
+ r
2
is always greater than zero and hence, the cubic equa-
tion (42) gives only one real root, which is given in Eq. (43). The other two roots are complex-conjugate and
are not considered.
Monte Carlo simulation (MCS) has been carried out for the nonlinear free vibration of simply supported
piezo-laminated plates, assuming the material properties and the actuation electric potential dierence as ran-
dom variables. Multivariate method [14] and Convex method [1] are utilized to nd the extremal values of the
piezo-laminated plate frequency response. Gaussian distribution is assumed for material properties and electric
potential dierence. The details of simulations carried out in the present study are briey highlighted below.
2.3. Mean and variance of nonlinear frequency x Monte Carlo simulation
Monte Carlo simulations are conducted to obtain second-order frequency statistics with random material
properties and actuation electric potential dierence d/
a
. Material properties E
11
, E
22
, m
12
, G
12
along with d/
a
are taken as the basic variables and are assumed as independent random variables (RV). This assumption
helps to make the solution approach simpler as the covariance matrix is diagonal and only these need to
be accounted for in developing the sample set for the random variables. However, most materials have some
correlation between the properties and the correlation matrix is expected to be fully populated. It is possible to
dene new variables as combinations of the physical variables that are independent of each other [15]. With
these new basic variables, the solution approach would work unchanged.
Each input RV is considered and a sample set, with size N, of the random variables are generated with the
specied statistical characteristics. The numerical experiments are carried out with this sample set using Eq.
(40). The coecient of variation of the solution sample set with this size is obtained. For convergence check,
the sample size of the RV is increased to M (i.e. M > N) and the corresponding solution set statistics obtained.
Table 1
The nonlinear period (T) of the fundamental mode of a simply supported isotropic square plate with immovable edges (wave numbers:
m = 1, n = 1) E
1
/E
2
= 1; G
12
/E
2
= 0.385; m
12
= 0.30
Amplitude
ratio (f
S
)
T/T
L
(x
L
/x) Present
study
Chu and
Herrmann [16]
Wah
[17]
Mei [18] and Mei and
Decha-Umphai [19]
Chia and
Prabhakaran [20]
0.2 0.9809 0.9783 0.9821 0.9818 0.9809
0.4 0.9297 0.9210 0.9338 0.9331 0.9291
0.6 0.8602 0.8451 0.8673 0.8670 0.8602
0.8 0.7853 0.7653 0.7943 0.7958 0.7853
1.0 0.7131 0.6901 0.7233 0.7271 0.7130
Table 2
Frequency ratio
x
x
L
_ _
of various vibrating modes of a simply supported rectangular plate (a/b = 2; b/h = 100) and lay-up sequence [0/90/
90/0] with amplitude ratio f
S

W mn
h
_ _
f
S
Frequency ratio (x/x
L
)
Fundamental mode (m = 1, n = 1) Various vibration modes of present analysis
Singh et al. [21] Onkar and Yadav [22] m = 1, n = 1 m = 1, n = 3 m = 1, n = 2 m = 2, n = 2 m = 2, n = 3
0.3 1.22 1.18 1.1737 1.2307 1.2248 1.1737 1.2136
0.6 1.63 1.61 1.5776 1.7402 1.7238 1.5776 1.6925
0.9 2.18 2.16 2.0795 2.3523 2.3250 2.0795 2.2728
Table 4
Frequency ratios of various vibrating modes of the simply supported piezo-laminated square thin plate having lamina sequence [PZT-5A/
0/90/90/0/PZT-5A] with applied electric potential dierence, d/
a
for the specied amplitude ratio f
S

W mn
h
_ _
f
S
Frequency ratio (x/x
L
) for various vibrating modes
m = 1, n = 1 m = 1, n = 3 m = 1, n = 2 m = 2, n = 2 m = 2, n = 3
Applied electric potential dierence, d/
a
= 0
0.3 1.0448 1.0696 1.0594 1.0448 1.0232
0.6 1.1678 1.2531 1.2186 1.1678 1.1908
0.9 1.3461 1.5074 1.4430 1.3461 1.3904
a
= +1067 V
0.3 1.0537 1.0720 1.0708 1.0468 1.0530
0.6 1.1989 1.2613 1.2411 1.1746 1.1960
0.9 1.4056 1.5225 1.4770 1.3593 1.3998
a
= 1067 V
0.3 1.0385 1.0673 1.0512 1.0430 1.0499
0.6 1.1452 1.2454 1.1996 1.1615 1.1860
0.9 1.3021 1.4931 1.4132 1.3339 1.3814
Table 5
Nonlinear frequencies (Hz) of fundamental vibration of simply supported piezo-laminated square plate with the amplitude ratio f
S
(a = b = 100 mm; h = 2 mm)
f
S
[PZT-5A/0/45/
60/0/PZT-5A];
asymmetric
[PZT-5A/0/30/
30/0/PZT-5A];
symmetric
[PZT-5A/30/30/
30/30/PZT-5A];
balanced
[PZT-5A/0/
90/PZT-5A];
cross-ply
[PZT-5A/0/30/
30/0/PZT-5A];
angle-ply
[PZT-5A/0/90/90/
0/PZT-5A]; symmetric
cross-ply
a
= 0 V
x
L
? 669.2 669.0 678.6 658.4 669.7 667.1
0.3 696.2 696.2 705.7 688.7 697.1 697.0
0.6 771.0 771.4 780.7 771.6 772.9 779.1
0.9 880.4 881.5 890.7 891.5 883.8 898.0
1.2 1013.0 1014.8 1023.9 1035.3 1017.9 1040.9
1.5 1160.7 1163.3 1172.5 1194.3 1167.3 1199.2
a
= +1067 V
x
L
? 610.2 609.9 620.5 598.4 610.8 607.9
0.3 639.7 639.7 650.1 631.5 640.7 640.6
0.6 720.2 720.7 730.7 720.8 722.3 728.9
0.9 836.1 837.3 846.9 847.7 839.7 854.6
1.2 974.5 976.4 985.8 997.5 979.6 1003.4
1.5 1127.0 1129.7 1139.2 1161.6 1133.8 1166.6
a
= 1067 V
x
L
? 723.3 723.1 732.1 713.4 723.8 721.4
0.3 748.4 748.4 757.3 741.4 749.3 749.2
0.6 818.5 819.0 827.7 819.1 820.4 826.2
0.9 922.5 923.6 932.3 933.2 925.8 939.4
1.2 1050.0 1051.8 1060.6 1071.6 1054.8 1077.1
1.5 1193.3 1195.9 1204.9 1226.1 1199.8 1230.9
Table 3
Mean values of the basic variables for glass-epoxy and PZT-5A
Property Glass-epoxy PZT-5A
Moduli
E
11
(GPa) 143 63
E
22
(GPa) 9.7 63
G
12
(GPa) 6 24.2
Poissons ratio m
12
0.3 0.3
Density q (kg m
3
) 2000 7800
Potential dierence d/
a
(V) 534
If the statistics corresponding to sample sizes N and M are in agreement with each other within a desired limit,
then the statistics obtained are the converged quantities. Otherwise, the exercise is continued by setting M as N
and selecting a new M. The inuence of each RV or their combinations on the nonlinear frequency can be
assessed with MCS. For the present study convergence was obtained for a sample set size N = 25,000 for each
of the basic RVs.
2.4. Variability in the nonlinear frequency x due to uncertain material properties and d/
a
by Multivariate method
The nonlinear frequency x is functionally related to the ve basic random variables E
11
, E
22
, m
12
, G
12
and
d/
a
. For mathematical simplicity these variables are denoted by X
1
, X
2
, . . ., X
5
, respectively. It is assumed that
these basic variables have small dispersions about their mean valuesX
1
; X
2
; . . . ; X
5
, respectively. This implies a
good quality control during the manufacturing process, which is true for most aerospace applications. Under
this assumption, x lends itself to Taylor series expansion about the mean values. Keeping only up to second
power of small quantities, x can be expressed as:
xX
1
; X
2
; . . . ; X
5
xX
1
; X
2
; . . . ; X
5
i1
ox
oX
i
X
i
X
1
2
5
i1
5
j1
o
2
x
oX
i
oX
j
X
i
X
i
X
j
X
j
44
All the derivatives of x are evaluated at the mean values of the basic RVs. The mean and the variance of the
nonlinear frequency x are obtained from Eq. (44) as:
l
x
Ex xX
1
; X
2
; . . . ; X
5

1
2
5
i1
VarX
i
o
2
x
oX
2
i
45
Table 6
Nonlinear frequencies (Hz) of fundamental vibration of simply supported piezo-laminated square plate with the amplitude ratio f
S
(a = b = 200 mm; h = 2 mm)
f
S
[PZT-5A/0/45/
60/0/PZT-5A];
asymmetric
[PZT-5A/0/30/
30/0/PZT-5A];
symmetric
[PZT-5A/30/30/
30/30/PZT-5A];
balanced
[PZT-5A/0/
90/ PZT-5A];
cross-ply
[PZT-5A/0/30/
30/0/PZT-5A];
angle-ply
[PZT-5A/0/90/90/
0/PZT-5A]; symmetric
cross-ply
a
= 0 V
x
L
? 408.2 406.7 414.8 414.6 411.3 409.4
0.3 426.0 422.3 431.3 438.6 427.9 433.8
0.6 475.0 465.7 476.8 503.0 473.8 498.9
0.9 546.2 529.4 543.6 594.1 540.9 590.8
1.2 631.9 606.9 624.6 701.3 622.3 698.7
1.5 727.0 693.5 715.0 818.5 712.9 816.4
a
= +1067 V
x
L
? 345. 7 343.9 353.5 353.2 349.4 347.1
0.3 366.5 362.2 372.6 381.0 368.7 375.5
0.6 422.2 411.8 424.3 453.4 420.9 448.9
0.9 500.7 482.4 497.9 552.4 495.0 548.9
1.2 592.7 566.1 585.0 666.1 582.5 663.4
1.5 693.0 657.9 680.5 788.3 678.3 786.2
a
= 1067 V
x
L
? 462.4 461.1 468.2 468.0 465.1 463.4
0.3 478.2 474.9 482.9 489.4 479.9 485.1
0.6 522.4 513.9 524.0 548.1 521.3 544.3
0.9 588.0 572.5 585.6 632.9 583.1 629.8
1.2 668.6 645.0 661.7 734.8 659.5 732.2
1.5 759.3 727.3 747.8 847.4 745.8 845.4
Varx Ex l
x
5
i1
VarX
i
ox
oX
i
_ _
2
1
4
5
i1
5
j1
o
2
x
oX
2
i
o
2
x
oX
2
j
VarX
i
VarX
j
46
The positive square root of variance is the standard deviation.
2.5. Variability in the system response x by Convex modeling
If a sucient amount of experimental data on material properties is available, the average values of these
data and their variance can be computed. These data can be utilized in Eqs. (45) and (46) to obtain variability
in the system response. For a case where variation ranges of material properties are known it is possible to
obtain the extreme values of the system response by Convex modeling [1].
The maximum and minimum values of the nonlinear free vibration frequency (viz., x
max
and x
min
) are
given by [1]:
x
max
x
1 a
S
5
i1
e
i
x
ox
oX
i
_ _
2
_
_
_
_
_
47
x
min
x
1 a
S
5
i1
e
i
x
ox
oX
i
_ _
2
_
_
_
_
_
48
Table 8
Extremal values of frequency ratio by Monte Carlo simulation method, Convex method and Multivariate method
Amplitude ratio f Monte Carlo method Convex method Multivariate method
Maximum Minimum Maximum Minimum Maximum Minimum
10% coecient of variation in d/
a
0.3 1.0891 1.0576 1.0989 1.0416 1.0806 1.0602
0.6 1.3175 1.2122 1.3513 1.1592 1.2900 1.2216
0.9 1.6251 1.4309 1.6892 1.3335 1.5756 1.4489
10% coecient of variation in E
11
0.3 1.0787 1.0593 1.0702 1.0702 1.0772 1.0632
0.6 1.2834 1.2183 1.2553 1.2553 1.2785 1.2316
0.9 1.5632 1.4423 1.5114 1.5114 1.5543 1.4675
10% coecient of variation in E
22
0.3 1.0709 1.0691 1.0702 1.0702 1.0708 1.0696
0.6 1.2577 1.2514 1.2553 1.2553 1.2573 1.2532
0.9 1.5158 1.5042 1.5114 1.5114 1.5151 1.5076
10% coecient of variation in m
12
0.3 1.0703 1.0699 1.0702 1.0702 1.0704 1.0701
0.6 1.2556 1.2540 1.2553 1.2553 1.2559 1.2547
0.9 1.5120 1.5091 1.5114 1.5113 1.5124 1.5103
10% coecient of variation in G
12
0.3 1.0713 1.0688 1.0702 1.0702 1.0711 1.0694
0.6 1.2588 1.2506 1.2553 1.2553 1.2582 1.2524
0.9 1.5178 1.5027 1.5114 1.5114 1.5168 1.5060
Table 7
Mean frequency ratio (x/x
L
) with amplitude ratio f
f (x/x
L
)
0.3 1.0702
0.6 1.2553
0.9 1.5114
where x
*
and
ox
oX
i
_ _
are evaluated at the nominal values of the independent RVs given as X
i
X
U
i
X
L
i
=2,
X
U
i
and X
L
i
are upper and lower bounds of X
i
and the semi-axes of the ellipsoid of the RVs e
i
X
U
i
X
L
i
. The
scaling parameter a
S
is assumed as unity as suggested in Ref. [1]. These values can be obtained if the charac-
teristics of the random variable are known.
Fig. 2. Coecient of variation of the frequency ratio for various independent RVs considered when f = 0.3, 0.6 and 0.9 using Monte
Carlo simulation. Electric eld direction and the PZT-5A crystal poling direction are same.
3. Results and discussion
3.1. Validation
To the best of the authors knowledge no published literature is available for comparison of response sta-
tistics of piezo-laminated plates with random material properties and actuation electric potential dierence.
Hence, the deterministic solution is validated by comparing the results of conventional isotropic and compos-
ite plates that are available in published literature when the induced actuation is not considered in the piezo-
laminated model.
The present analysis results in Table 1 are found to be in good agreement with those of Chu and Herrmann
[16], Wah [17], Mei [18], Mei and Decha-Umphai [19] and Chia and Prabhakara [20], obtained by dierent
numerical schemes for isotopic simply supported square plates with immovable edges.
Frequency ratios of various vibrating modes of the simply supported rectangular plate with the amplitude
ratio (f
S
) have been computed. These results are obtained by taking:
m = 1, n = 1 for fundamental mode;
m = 1, n = 3 for a symmetricsymmetric mode;
m = 1, n = 2 for a symmetricanti-symmetric mode;
m = 2, n = 2 for an anti-symmetricanti-symmetric mode; and
m = 2, n = 3 for an anti-symmetricsymmetric mode.
The variation of frequency ratio (x/x
L
) with the amplitude ratio (f
S
) for the above ve modes are shown in
Table 2 for a rectangular plate (a/b = 2; b/h = 100) and lay-up sequence [0/90/90/0]. The properties used in
the analysis are [21]: E
11
= 40E
22
, G
12
= 0.5E
22
and m
12
= 0.25. The results obtained by Singh et al. [21] and
Onkar and Yadav [22] for the fundamental vibration of a symmetrically orthotropic rectangular plate are
found to be in good agreement with the present analysis results. The frequency ratio for all the modes
increases with increase in the amplitude ratio.
Frequency ratios are computed for various vibrating modes of the piezo-laminated square plate having
lamina sequence [PZT-5A/0/90/90/0/PZT-5A] with applied electric potential dierence, d/
a
for the spec-
ied amplitude ratio, f
S
. The properties of glass-epoxy and PZT-5A in Table 3 are considered in the analysis.
Table 4 gives the analysis results for all the ve vibration modes. For all the cases, the frequency ratio increases
with increasing amplitude ratio.
Tables 5 gives nonlinear frequencies of fundamental modes of vibration with dierent applied electric
potential dierence for piezo-laminated rectangular plates of six lamination schemes asymmetric, symmetric,
balanced, cross-ply, angle-ply and symmetric cross-ply. Table 6 presents similar results for a square plate. The
thicknesses of the PZT-5A and glass-epoxy layers are taken as 0.5 and 0.25 mm, respectively. It is possible to
control (enhance/reduce) the frequency of vibration with the applied electric potential dierence.
3.2. Second-order statistics of nonlinear frequency
The nonlinear free vibration of simply supported piezo-laminated rectangular thin plate is examined with
moderately large deection considering uncertain material properties and d/
a
. Numerical results for simply
supported square plates with immovable edges are presented for wave numbers m = n = 1. The piezoelectric
Table 9
Eect on the coecient of variation (%) of the frequency ratio (x/x
L
) due to scatter in the basic RVs
Amplitude ratio f Coecient of variation (%) in all RVs
5 10 15 20
0.3 0.1929 0.3859 0.5789 0.7719
0.6 0.5524 1.1048 1.6573 2.2097
0.9 0.8493 1.6987 2.5480 3.3974
layer can be surface bonded or embedded in one of the layers. Here, top and bottom layers are taken as com-
posed of PZT-5A. Glass-epoxy laminate is sandwiched between them. Table 3 gives the mean values of the
material properties taken as the basic random variables for glass-epoxy and PZT-5A utilized in the study.
Fig. 3. The bounds for the frequency ratio x/x
L
of fundamental vibration in terms of coecient of variation, when the material properties
of the base structure and d/
a
are random using Convex method. Electric eld direction and the PZT-5A crystal poling direction are same.
The piezoelectric charge constant d
31
is assumed to be 154 10
12
m/V. The strength of PZT-5A is
75.9 MPa; the rated stress is 20.7 MPa. Lamina sequence [PZT-5A/0/90/90/0/PZT-5A] and 100% rated
d/
a
is considered for the present study. The applied electric eld within the piezoelectric actuator is assumed
to be constant as the layer is considered thin. The nonlinear vibrations are studied for specied amplitude to
thickness ratio. The thickness of the PZT-5A and glass-epoxy layers are taken as 0.5 and 0.25 mm,
respectively.
The mean value of the frequency ratio (x/x
L
) is given in Table 7. Comparison of the upper and lower
bounds obtained using the three statistical methods followed in this paper is given in Table 8. For the purpose
of illustration 10% coecient of variation is considered in the material properties and d/
a
. The Convex
method is useful when the distributions of the input parameters are not known but its extremal values are
known. MCS involves large computational time. The values obtained by MC method are comparable with
those obtained by Multivariate method. Convex method and Multivariate method are very simple techniques
that can be applied to study stochastic response of the structure.
The eect of scatter in the material properties and the actuation electrical potential dierence on the non-
linear frequency ratio is obtained using MCS. For amplitude ratios 0.3, 0.6 and 0.9, the coecient of variation
of the nonlinear frequency is plotted against the coecient of variation of the ve individual RVs. These are
shown in Fig. 2. When variation in one basic variable is being considered the values of the other variables are
kept constant. From Fig. 2, one can infer that the nonlinear free vibration is more sensitive to scatter in E
11
and d/
a
compared to other basic variables. The Poissons ratio scatter inuences the nonlinear frequency least.
For sensitive applications with piezoelectric actuation the dispersion in E
11
and d/
a
should be kept to the min-
imum for better controllability as shown in Fig. 2a and e.
Table 9 gives the coecient of variation in the frequency when scatter in all basic RVs are present simul-
taneously using the Multivariate method. The results show that at higher amplitude ratios the eect of scatter
in the basic RVs are more pronounced thus reiterating the plots shown in Fig. 3 that were obtained using Con-
vex method for each RV. Fig. 3 depicts the results for the Convex modeling for the upper and lower bounds of
the nonlinear frequency response for a given amplitude ratio with 5% coecient of variation of the specied
basic random variable. This also shows the dominance of scatter in E
11
and d/
a
on the nonlinear frequency of
free vibration.
4. Concluding remarks
Studies are made on nonlinear free vibrations of simply supported piezo-laminated rectangular plates with
immovable edges. Frequency ratios of various vibrating modes are computed. In all the modes of vibration
considered, the frequency ratio increases with increasing amplitude ratio. Nonlinear frequencies of fundamen-
tal vibration of piezo-laminated plates of six lamination schemes (viz., asymmetric, symmetric, balanced,
cross-ply, angle-ply and symmetric cross-ply) are also computed. Appreciable variation is noticed in frequen-
cies with the applied electric potential dierence. It is possible to control (enhance/reduce) the frequency with
the applied electric potential dierence.
Three statistical model Monte Carlo simulation, Multivariate method and Convex modeling are utilized
to evaluate the variability of system response of a simply supported piezo-laminated rectangular plate. For the
specied variations in material properties and applied electric potential dierence the lower and upper bounds
of the frequencies are evaluated. These bounds are essential and very useful in practice, which can be incor-
porated directly into the design.
Appendix A
The expressions for the functions, u
0
, u
1
and u
2
in Eq. (26) are given below:
/
0
x; y w
1k
x
2
w
2k
y
2
u
1
x; y w
1
sin a
m
x sin b
n
y w
2
cos a
m
x cos b
n
y
u
2
x; y w
3
x
2
w
4
y
2
w
5
cos 2a
m
x w
6
cos 2b
n
y
w
1
w
11
w
21
w
12
w
22
=w
2
11
w
2
12
w
2
w
11
w
22
w
12
w
21
=w
2
11
w
2
12
w
3

1
16
b
2
n
A
11
a
2
m
A
21
A
11
A
22
A
12
A
21
_ _
w
4

1
16
a
2
m
A
22
b
2
n
A
21
A
11
A
22
A
12
A
21
_ _
w
5
b
2
n
=32A
22
a
2
m
w
6
a
2
m
=32A
11
b
2
n
w
11
a
4
m
A
22
a
2
m
b
2
n
2A
12
A
66
b
4
n
A
11
w
12
2a
3
m
b
n
A
26
2a
m
b
3
n
A
16
w
21
a
4
m
B
21
a
2
m
b
2
n
B
11
B
22
2B
66
b
4
n
B
12
w
22
a
3
m
b
n
B
61
2B
26
a
m
b
3
n
B
62
2B
16
w
1k

1
2
N
Kxy
A
16
A
21
A
26
A
11
A
11
A
22
A
12
A
21
_ _
N
Ky
_ _
w
2k

1
2
N
Kxy
A
12
A
26
A
22
A
16
A
11
A
22
A
12
A
21
_ _
N
Kx
_ _
Appendix B
The constants a, a
K
, b, c
K
and d in the equation of motion (29) are given below:
a w
0
w
01
2a
2
m
w
2k
2b
2
n
w
1k
c
1
16
a
4
m
A
11
b
4
n
A
22
_ _
w
0
a
4
m
D
11
2a
2
m
b
2
n
D
12
2D
66
b
4
n
D
22
w
01

w
11
w
2
21
w
2
22
2w
12
w
21
w
22
w
2
11
w
2
12
For m and n both odd:
b
8a
m
b
n
3ab
_ _
4w
1
21
A
22
12
A
11
_ _
a
K
0
For m even and n odd:
b
2a
m
b
n
ab
_ _
B
21
A
22
a
2
m
B
11
3b
2
n
A
11
_ _
32
ab
a
m
b
n
B
11
w
4
B
21
w
3
a
K

16
ab
a
m
b
n
B
11
2w
2k
N
Kx
B
21
2w
1k
N
Ky
B
61
N
Kxy
M
Kx
For m odd and n even:

b
2a
m
b
n
ab
_ _
B
12
A
11
b
2
n
B
22
3a
2
m
A
22
_ _
32
ab
b
n
a
n
B
12
w
4
B
22
w
3
a
K

16
ab
b
n
a
m
B
12
2w
2k
N
Kx
B
22
2w
1k
N
Ky
B
62
N
Kxy
M
Ky
For m and n both even:
b 0; a
K
0
References
[1] Li YW, Elishako I, Starnes Jr JH, Shinozuka M. Prediction of natural frequency and buckling load due to uncertainty in material
properties by Convex modeling. Field Inst Commun 1996;9:13954.
[2] Leissa AW, Martin AF. Vibration, buckling of rectangular composite plates with variable ber spacing. Compos Struct
1990;14:33957.
[3] Shinozuka M, Astill CJ. Random eigen value problems in structural analysis. AIAA 1972;10(4):45662.
[4] Chen PC, Soroka WW. Multi-degree dynamic response of a system with statistical properties. J Sound Vibr 1974;37(4):54756.
[5] Yadav D, Verma N. Buckling of composite circular cylindrical shells with random material properties. Compos Struct
1997;37:38591.
[6] Bailey T, Hubbard J. Distributed piezoelectric polymer active vibration control of a cantilever beam. J Guid Control Dynam
1985;8:60511.
[7] Dimitriadis EK, Fuller CR, Rogers CA. Piezoelectric actuators for distributed excitation of thin plates. J Vib Acoust 1991;113:1007.
[8] Wang T, Rogers CA. Laminate plate theory for spatially distributed induce strain actuators. J Compos Mater 1991;25:43352.
[9] Chia CY. Nonlinear analysis of plates. New York: McGraw-Hill Book Company; 1980.
[10] Reddy JN. Mechanics of laminated composite plates: theory and analysis. New York: CRC Press; 1996.
[11] Chandra R, Raju BB. Large amplitude exural vibration of cross-ply laminated composite plates. Fiber Sci Technol 1975;8:24365.
[12] Abramowitz M, Stegun IA. Handbook of mathematical functions with formulas, graphs and mathematical tables. US Department of
Commerce; 1972.
[13] Srinivasan P. Nonlinear mechanical vibrations. New Delhi: New Age International (P) Ltd. (Formerly Wiley Eastern); 1995.
[14] Bowker AH, Lieberman GI. Engineering statistics. Englewood Clis, NJ: Prentice Hall Inc.; 1972.
[15] Nigam NC. Introduction to random vibration. Cambridge: The MIT Press; 1983.
[16] Chu HN, Herrmann G. Inuence of large amplitude free exural vibrations of rectangular elastic plates. J Appl Mech
1956;23:53240.
[17] Wah T. Large amplitude exural vibrations of rectangular plates. Int J Mech Sci 1963;5:42538.
[18] Mei C. Finite element displacement method for large amplitude free exural vibrations of beams and plates. Comput Struct
1973;3:16374.
[19] Mei C, Decha-Umphai K. A nite element method for nonlinear forced vibration of rectangular plates. AIAA J 1985;23:110410.
[20] Chia CY, Prabhakaran MK. A general mode approach to nonlinear exural vibrations of laminated rectangular plates. J Appl Mech
1978;45:6238.
[21] Singh G, Venkateswara Rao G, Iyengar NGR. Large amplitude free vibration of simply supported anti-symmetric cross-ply. AIAA J
1991;29(5):78490.
[22] Onkar AK, Yadav D. Nonlinear free vibration of laminated composite plate with random material properties. J Sound Vib
2004;272:62741.

Nonlinear vibration of piezo-laminated plates

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nonlinear vibration of piezo-laminated plates

Uploaded by

Copyright:

Available Formats

Nonlinear free vibration analysis of simply supported

piezo-laminated plates with random actuation electric

is obtained from Eq. (33) and vice versa. By substituting for f

) I(f) = 0, one can

by comparing the coecients

corresponding to the positive amplitude-to-thickness ratio f

has to be obtained from the cubic equation:

For m odd and n even:

You might also like