Computer methods

in applied
mechanics and
engineering
ELSEVIER Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155

Three:-dimensional solution for simply-supported

piezoelectric cylindrical shell for axisymmetric load
Santosh Kapuria a, S. Sengupta a, PC. Dumir b,*
a Engineers India Ltd., Bhikaijee Cama Place, New Delhi-110066, India
b Applied Mechanics Deptt., Indian Institute of Technology, Hauz Khas, New Delhi-110016, India

Received 23 January 1996; revised 12 April 1996

Abstract
This work presents a three-dimensional solution for simply-supported piezoelectric cylindrical shell subjected to axisymmetric
electromechanical load. The variables are expanded in Fourier series to satisfy the boundary conditions at the ends. The solution
of the governing differential equations with variable coefficients is constructed as a product of an exponential function and a
power series in the thickness coordinate. The coefficients of terms of all degrees in the governing equations are set to zero. This
yields a linear eigenvahle problem for the exponent and recursive relations for the coefficients of the power series. The arbitrary
constants in the general solution are determined from the boundary conditions. The solution of the inverse problem of inferring
the applied pressure field from the measured electrical potential difference between the surfaces of the shell is also presented.
Numerical results are presented showing the effect of nature of loading and the geometrical parameters on the response. These
results would enable development and assessment of two-dimensional piezoelastic shell theories.

1. Introduction

The coupling existing between the elastic and electric fields in piezoelectric materials is used in various
engineering applications. The direct piezoeffect is used in sensors in electromechanical transducers to
infer the deformation from the induced electrical potential difference. The converse piezoeffect is used
in electromechanical actuators for controlling an entity by the application of appropriate electrical po-
tential difference. Few analytical solutions of three-dimensional field equations are available for coupled
response of piezoelectric elements to electromechanical load. These analytical solutions are needed to
assess the accuracy of two-dimensional plate and shell theories. Ray et al. [1,2] presented exact solutions
for static analysis of a simply-supported piezoelectric plate and a layered intelligent flat panel under
cylindrical bending. Similar solution was presented by Brooks and Heyliger [3] for cylindrical bending
of simply-supported piezoelectric laminates with distributed and patched actuators. Three-dimensional
exact static analysis of simply-supported rectangular plate with sensing and actuating layers has been
presented by Ray et al. [4] and Heyliger [5]. Mitchell and Reddy [6] have presented a power series solu-
tion for axisymmetric composite cylinder with surfacebonded or embedded piezoelectric laminae under
axial load.
We present a three-dimensional solution for finite, simply-supported circular cylindrical piezoelectric
shell subjected to axisymmetric electromechanical load. The displacements, stresses, electric potential
and electrical displacement are expanded in Fourier series in the axial coordinate to satisfy the bound-
ary conditions at the simply-supported ends. The prescribed electromechanical functions on the lateral
boundary are similarly expanded in Fourier series. The governing equations reduce to ordinary dif-

*Corresponding author.

00457825/96/$15.00 (Q 1996 Elsevier Science S.A. All rights reserved

PII sn045-183S~9h~~10lS-4
140 S. Kapuria et al. /Cornput. Methods Appl. Mech. Engrg 140 (1997) 139-155

ferential equations in the radial coordinate with variable coefficients. These are solved by the method
developed in [7] for thermoelastic problems. The general solution is constructed in the form of a product
of an exponential function and a power series in the thickness coordinate. This expansion is substituted
in the governing equations and in the resulting equation, the coefficients of terms of different degree
are set equal to zero. This yields a linear eigenvalue problem for the exponent with six eigenvalues and
eigenvectors and recursive relations for the coefficients of the power series. The six arbitrary constants
in the general solution are determined from the boundary conditions at the inner and outer surfaces.
The solution of the inverse problem of inferring the applied pressure field from the given measured
distribution of electrical potential difference between the lateral surfaces of the shell is also presented.
The present method of solution involving products of exponential functions and power series yields
converged results for each Fourier component with the number of terms in the power series depending
on the type of loading and the accuracy desired. Since the three-dimensional elasticity and electrostatic
field equations are used, the series solution presented is valid for thin as well as thick shells. Numerical
results are presented to illustrate the effect of the nature of loading and the geometrical parameters,
radius-to-thickness ratio and span-to-thickness ratio, on the response.

2. Governing equations

The linear constitutive equations of a piezoelectric medium are given by

c=Su+dTE, D=dcr+EE (1)
where superscript T denotes matrix transpose and the components of stress u, strain 6, electric field E
and electric displacement D are given in cylindrical coordinate system (r, 8, z) by
U = [a, Ue uz QZ 7zr WIT, E = [Er Eti EzlT,

E = [&T Eg &z Yez Yzr YrelT, D = [Or Do DzlT.

S, d and E denote, respectively, the matrices of elastic compliance, piezoelectric strain constants and
dielectric constants of the piezoelectric material. Materials like PVDF, PZT4 possess class mm2 symmetry.
For orthotropic materials of class mm2 symmetry, with poling in the radial direction, the matrices of
material constants with respect to principal axes along r, 8, z are given by Tiersten [8]:

I 1
311 s12 s13 0 0 0 rd, 0 0
s12 s22 s23 0 0 0 d2 0 0
El 0 0
s13 s23 s33 0 0 0 d3 0 0
s= , dT= 0 0 0 > E =: 0 q 0 )
(2)
0 0 0 s44 0 0
0 0 E3
0 0 0 0 s55 0 0 0 d5
0 0 0 0 0 s@j 0 d6 0

where
E = v+edT, e=dC, c = s-l.
C is the matrix of elastic constants, e is the piezoelectric stress matrix and 77is the permittivity matrix
for constant strain field. The equations of force equilibrium and charge equilibrium without body force
and internal charge source are
ur,r + w,o/~ + rZr,z + (6 - fl~e>/r = 0, 7,~)~+ u0,elr + Tez,z +27,9/r = 0,
(3)
7Zr,r+ r02,8/r + a,,, + rzr/r = 0, D,,, + D,/r + De,0 /r + D,,z = 0,
where a subscript comma denotes differentiation.
Consider a finite simply-supported circular cylindrical piezoelectric shell of mean radius R, thickness
h and length a. The inner and outer radii are RI = R - h/2 and Rz = R + h/2. Let it be subjected to
axisymmetric electromechanical load. Let the radial and axial displacement components of the axisym-
metric solution be u and W.The strains and the electric field are, respectively, related to the displacements
and the electric potential 4 by
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 141

-5 = u,r, &(j = U/Y, &z = w,z Xr = u,z+ w,r, Yrs = Yez = 0, (4)
Er = -4,r, Ez = -4,z> E8 = 0. (5)
Using expressions (2), (4) and (5), the constitutive equation (1) can be written as
u/r = S12ar + jS222co+ S23cz - d24,r, (6)
W,z = S13ar + :63go + S33uz - d34,n (7)
u ,r = Sllar + S12ae + S13cz - dt4,rr (8)
W,r = -u,, + Ssrzr - d54,z > (9)
rre = r(jz = 0, Do =O, Dz = d57zr - ~34,zr (10)
D, = d,a, + dZ2cro+ d3az - q4,r. (11)
We introduce dimensionless coordinates 5 and 5 as follows:
5 = z/a, i: = (r - Rl)/h. (12)
The ends of the cylindrical shell are electrically grounded. The shell is simply-supported in such a way that
the supports allow displacement normal to the boundary but prevent inplane tangential displacement. At
the lateral surfaces of the shell, the electrical potential or the electrical displacement D, is prescribed.
Let the prescribed axisymmetric pressure and electrical potential or electric displacement D, at the
inner and outer surfaces be p1 (t), 41(t) or DI([) and p2(5), &2(~) or 02(t), respectively. The boundary
conditions are

at 5 = 0, 1:
u = 0, a, =o, 4 =o; (13)
at Y= Ri:
gr(Ri, 5) = -I?i(t), Tzr(Ri, 6) = 0, 4(Ri, 6) = 4i((5)
or D,(I;!i, 5) = Di(t); i = 1,2. (14)
For numerical work, the governing equations are nondimensionalised in terms of dimensionless enti-
ties:
S = R/h, ui* = Ui/h, U; = aiiS/ET, 4* = 4ldrl/h, (CT, VT) = (Ei, ~~)/ETId~12~
a* = a/R, E,; = EijS, ET = EildTIS, 0: = DiS/ETldTI, l= (r - Rl)/h,
h* = h/R, SG = SijET, e,> = eij/ETJdTl, G,*i = Gij/ET, r* = rfR, z* = z/R,
RF = Ri/R, df = di/JdT\, Cl; = Cij/ET, YT = Yi/ET, 8’ = 6, 5’ = z*/a*. (15)
where Yi are Young’s moduli, Gij are shear moduli and ET, dT are Young’s modulus and piezoelec-
tric coefficient in a. specified direction. For simplicity, in the sequel, we drop the subscript * from the
dimensionless entities. The dimensionless form of Eqs. (l)-(14) is exactly the same as given earlier.

3. General solutiou of governing equations

The solution of the boundary value problem, satisfying the boundary conditions (13) is taken in the
separable form of the following Fourier series:

(u, err, ~0, cz, 4, Dr) = g&n, a,,, “0, I uzz., 4,,, Drn) SinnmF,
n=l

(w, rzr, Dz> q = ~(wn>

rzrn,Dzn>
II=1
~0s nd. 06)

In order to satisfy lthe conditions (14) termwise, the functions pi(t), 4i (0, Di([) are similarly expanded
as
I42 S. Kapuria et ai./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155

(pi,&iDi) =~(pi~,~i~,Dil)sinn?r~. (17)

i?=l
In order to solve three nontrivial equilibrium equations (3) and six nontrivial constitutive equations
(6)-(ll), we transform them into a combined system of six first-order differential equations for the six
independent variables u, w, crrr,rzr, 4, D, and three algebraic equations for ue , 0; and D,. Substituting
expansions (16) into Eq. (10) yields an algebraic equation for Dzn:
Dzn = d5rzro - qic#+, with ri = nrla. (18)
Substituting +.r from Eq. (11) into Eqs. (6)-(8) an d using expansions (16), we obtain the following
algebraic expression for Q,, u,~ and differential equation for u,,:
ae,, = &,u,lr + bi2fiwn + blsarn + bl@r,, (19)
o-zn = &iUn/r + &fiwn + &urn + b&r,, (20)
with
bii = $3/‘% h3 = (s;,s;, - s;3s;,)/s, b,6 = (S;,d; - S;3d;)/S,

b12 = -b2, = S;,/S, b23 = <$3$2 - $2$3)/S, b26 = (S;,d; - S;,d;)/s,

b22 = -SG2/S, S’ij = Sii - didj/El, s = si2si3 - s;2,, (21)

b31 = -bn, b33 = S;zbn + S;jbz + S;, , d,! = di/El,

b32 = bzr byj = S;,bl6 + Si3b26 + d;;

and
Lln,r = b31U,/r + b32sWn+ b33or, + b3&r,, . (22)
Substituting expansions (16) in (9), (11) and (3) and using expressions (18) to (20), we obtain
W n,r = --nuti + S557zr, - d5f%,, (23)
4n,r = b52fiWn + b53ur,, + b5@rv + b5lunlr, (24)

0;” >T = nrzrn + @lSW, + b63Gn + h&,)/r + hlun/r2, (25)

rzrn ,r = -b2yi2w,, - b2giurn - bz,giD, - (bzlfiu,, + rzrn)/r, (26)

D rnr = d@rz, - •5~4, - D, f r, (27)

with
b5r = -b16, bgj = d;b16 + d;b26 - l/El, b52 = b26, b53 = bx, be3 = b13 - 1. (28)
The six first-order differential equations (22)-(27) are expressed in matrix form as
X,,r = (Ao + Al/r + &/r2)& (29)
where
X, = [G wn urn rzr, & Drn] T ,

-0 bati b33 0 0 b36 b31 0 0 00 0

-ii 0 0 s55 -d$i 0 0 0000 0
0 0 0 ii 0 0 b12ti b63 0 0 h
Ao= o Al= ’
-b22ii2 -b23ti 0 0 -b&i ’ -b2# 0 0 -1 0 0
0 bszfi bss 0 0 bs6 bsl 0 0 00 0
0 0 0 d5fi -e3fi2 0 0 0000 -1

(Az)ij=O for i#3 and jfl, @2)31 = hl. (30)

The general solution of homogeneous linear differential equation (29) with variable coefficients is
taken as a product of an exponential function and a power series in the thickness coordinate 5 as
follows:
 S. Kapuria et aL/Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 143

Xn,r = eAi1 [AYi + (i + l)Yi+j] J’/h (31)

j=:O i=O
We substitute these into Eq. (29). In the resulting equation, the coefficient of each power of { is equated
to zero. The coefficient of the zeroth-order term, p”, involves Y. and Y1. Choosing Y, = 0 and equating
the coefficient of the zeroth-order term to zero yields the following linear algebraic equation for Yo:
AYo = hYo, with A = [AoR +A, +AJR,]/s, s = RI/h (32)
Hence, the exponent A and Y, are eigenvalue and eigenvector pair of 6 x 6 real matrix A. The solution
of the characteristic equation det(A - AZ) = 0, where Z is a 6 x 6 identity matrix, and the procedure for
determination of the eigenvectors are presented in Appendix A. The complementary solution is the sum
of the six solutions for the six eigenpairs of A. Equating the coefficient of 5’ to zero for i > 0, we obtain
the following recursive relationship for Y! corresponding to each eigenpair (Aj, Yi), j = 1,. . . ,6:

Y!+l = [do(hj, i)Y,’ + dl(hj, i)Y/_, + dz(Aj)Y/_2]/(i + l), i 3 1 (33)

with
&(A, i) = A -- (A + 2i/s)Z, dl(A, i) = [2R1Ao +A1 - (2sA + i - 1)11/s*, &(A) = (hAo - AZ)/s2
(34)

and Y{ = 0, YI, = 0. The eigenvalues of real matrix A are either real or occur in complex conjugate pairs.
The complementary solution for pair of complex conjugate eigenvalues Al, A2 = (Yf ip with complex
eigenvector Y,’ corresponding to A,, can be expressed in terms of two real arbitrary constants C; and
C; as

X,(L) = Z5V)CI” + Z%V)C; (35a)

1,
with

F;(J) = ea5 cos~[~!R(Y~)[’ - sinPie3(Y,‘)5’ (35b)

1. WC)
i=O i=O

F;(c) = em5 sinpc5 LR(Y/){’ +cos&Y#

i=O i=o

ZR and S indicate .the real and imaginary parts of a complex number. The complementary solution for
distinct real eigenvalue, say A3 = p with eigenvector Yi, in terms of arbitrary constant C;, is

Xi([) = F$(&‘)C; with F;(l) = epe (36)

Complementary solutions for multiple eigenvalues are not listed since these do not arise for any II for
the material and the problems considered in this study. Thus, the complete general solution X, of Eq.
(29) can be expressed in terms of six arbitrary constants CJ’(j = 1 to 6) as
6

(37)
j=l

where the functional form of F,?(l) is given by Eqs. (35) or (36) according to the nature of hi. The
infinite power serlles in F]!‘(l) is truncated to finite number of terms such that the contribution of the
first neglected term as determined by the recursive relation is less than a stipulated small number 77.

4. Solution of direct and inverse problems

The direct problem is solved by determining the constants (C;)‘s. Expressions of u,, w,, urn, r,,n, &
144 S. Kapuria et al. /Comput. Methods Appl. Mech. Engrg. 140 (1997) 1.?9-155

and D, from Eq. (37) and the expansions (16) and (17) are substituted into six electromechanical
boundary conditions (14). This yields, for the nth Fourier component, six linear algebraic equations
for six arbitrary constants (Cy)‘s. Thus, the arbitrary constants CT’s can be determined from the given
boundary conditions for each Fourier component by taking a finite number of terms in the power
series of 5 as per the accuracy specified by the value of v. X, = [u, W, err,, rZr,, $n D,,lT is then
determined from Eq. (37). Substitution of these entities in Eqs. (lS)-(20) yields D,,,, CT@,,, u,,~. Finally,
the displacements, electrical potential, stress and electric displacement components can be computed at
any point in the shell using expressions (16) by taking finite number of terms, say N, in the Fourier
series. A convergence study will determine the value of N to be used for given electromechanical load.
Consider an inverse piezoelastic problem in which the cylindrical shell is subjected to a known pressure
pi(t) at the inner surface and is subjected to an unknown pressure at the outer surface. The electrical
displacement component D, is prescribed at the inner and outer surfaces. The pressure profile ~~(6)
is to be determined from the known measured potential difference V(t) between the outer and inner
surfaces:

$(&,5) - $(Rl,5) = V(S) = e V,sinnr5. (38)

n=l
Substituting expression of & from Eq. (37) and expansion of $J from Eq. (16) in Eq. (38) yields
r6 1 r6 1

(39)
Lj=l J Li=l J
The six electromechanical boundary conditions (14) and the potential difference condition (39) yield
seven linear algebraic equations for six constants Cy’s and pz,. Their solution yields CT’s and the nth
Fourier component ~2, of ~~(6). The unknown pressure distribution on the outer surface can then be
determined from expansion (17) as: ~~(5) = C,“=I p2, sinnr.$.

5. Numerical results and discussion

Numerical computations have been carried out for circular cylindrical shells of polyvinyledene fluoride
(PVDF) polarised along the radial direction and drawn along the circumferential direction. The material
properties of this piezoelectric polymer are taken as.[9,10]:
Y, = YZ = Ys = 2.0 GPa, v12 = v23 = v331 = l/3; n1 = q = n3 = 0.1062 x lop9 F/m,
dI = -30 x lo-l2 CN-’ 7 d2 = 23 x lo-l2 CN-’ , d3=3x10-‘2CN-1, d,=dg=O;
l1 = 0.3141 x lo-” F/m, l2 = 0.3078 x 1O-1o F/m, c3 = 0.3078 x lo-” F/m.
where Yi’s are the Young’s moduli, Yij’Sare the Poisson’s ratios, vi’s are the permittivity constants and
ei’s are the computed values of the dielectric constants. Six cases of electromechanical loads considered
are

1. h(5) = 0, h(5) = 0, Pl(G = 0, P2W = PO,

2. d+(5) = 0, +22(0 = 0, Pl(&) = 0, ~2(5) = PO sin ~5,
3. &(5) = 0, D2(5) = 0, Pl(5) = 0, P2W = PO?
4. h(5) = 0, +22(0 = ho, Pl(5) = 0, P2M = 0,
5. h(5) = 0, 4~45) = 41 sin ~5, Pl (8 = 07 P2G) = 0,
6. DIW = 0, D2(0 = Do, PlM = 0, P2(0 = 0.

Superposition of results for various cases is admissible since the governing equations are linear. Results
are computed for cylindrical shells with radius-to-thickness ratio S = R/h = 2, 4, 6, 10, 20, 50, 100 and
length-to-radius ratio a/R = 1, 2, 4, 10. The results for cases of po # 0, C#JO
# 0, DO # 0 are, respectively,
expressed in dimensionless form for uniform and sinusoidal mechanical or electrostatic loads as follows,
with ET = Yl and dT = dl:
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 145

l.po#O:
u= u*/sp;, w = w*/sp;;, or = qp;, *o = a;/sp;, 0, = D;/Sp;,,
for uniform pressure: GZ = a;/SpT,, ;izr = ~;~/p$, CjJ= 1ooo$J*/p;, 0, = D;/p;;
for sinusoidal pressure: @, = a,‘/~;, 7Zr = r$S/p& tj = looo~*s/p~, 0, = lOD;S/p;;
2. $0 # 0:
2 = u*/s+,*, G = w*/s+;, G0 = ~;I+;,
+Z = u,‘/40’, 3 = 4*/G> D,, = D;/S+,*,
for uniform potential: cYr= cr;S2/100&, TZr= T;~/~O*, D,, = D;/lOOO~;,
for sinusoidal potential: & = cr,?S/+,*, +Z,r= r$S/$;, D,, = D;/lOO~;,
3. Do # 0:
ii = looou*/D;,, W = lOOOw*/D;, cYr= lOOo;S/D;, cYe = lOOOa;/D;;,
cYZ= lOOOa;/D;, TZr= lOOO+S/D;, 4 = lOOO+*/SD;,
D, = D;/D;, & = D;/SD;;,

The convergence ratio, 7, for terminating the power series, has been taken as lo-” for each Fourier
term. A study of the number of terms, K, required in the power series to maintain the specified accuracy
q has been conducted for case 1 of uniform pressure. Numerical computations have revealed that for
given values of index n of the Fourier term and the ratio a/R, K increases with decrease in the value
of radius to thickness ratio S. For example, for y2= 1 and a/R = 4, the values of K are 6, 11, 21, 66 for
S = 100, 10, 4, 2, respectively. For given values of S and a/R, the value of K increases with increase in
the value of the index n of the Fourier term. For example with a/R = 4, as y1increases from 1 to 31, K
increases from 6 to 9 for the case of S = 100 and from 66 to 143 for the case of S = 2. The length-to-
radius ratio was found to have relatively little effect on the value of K. For n = 1, as a/R changes from
1 to 10, K remains constant at 9 for the case of S = 20.
Only one term solution is required for sinusoidal loading. For nonsinusoidal loadings, convergence
studies have been conducted for various values of radius to thickness ratio S and length-to radius ratio
a/R to arrive at the number of Fourier terms, N, required for given accuracy. The results of a typical
convergence study are presented in Table 1 for a cylinder with S = 50 and a/R = 4 subjected to uniform
pressure loading of case 1. It is observed that very good accuracy is achieved at N = 189. The value of
N needed to achieve a desired accuracy, decreases as the radius-to-thickness ratio S decreases as well as
the length-to-radius ratio a/R decreases. The values of N used in this study for cylinders with a/R = 4
are 289, 189, 99, 69, 35 for S = 100, 50-10, 6, 4, 2, respectively. For cylinders with S = 20, the values of
N used are 99, 119, 189 for a/R = 1, 2, 4-10, respectively.
The present three-dimensional exact piezoelastic solution reduces to exact elasticity solution if di’s
are zero. The piezoelastic response of a simply supported cylindrical shell with a/R = 1 and S = 200
subjected to uniform electromechanical loading of case 1, is compared in Table 2 with the exact elastic
response (di = 0) and the elastic response obtained using the two-dimensional classical thin shell theory
(CST) [ll]. For this case of thin shell of PVDF, it is observed that in the absence of any potential
difference between the two surfaces, the exact piezoelastic solution differs little from the exact elasticity
and the CST solutrons.

Table 1
Convergence study
N U(O.5, 0.5) 00 (0, 0.5) az (0, 0.025) D,(O.5, 0.5)
40 1.0061 1.009 -0.6260 0.7642
90 1.0066 1.010 -0.5851 0.7642
110 1.0066 1.010 -0.5911 0.7642
130 1.0066 1.010 -0.5913 0.7642
150 1.0066 1.010 -0.5892 0.7642
170 1.0066 1.010 -0.5891 0.7642
190 1.0066 1.010 -0.5901 0.7642
210 1.0066 1.010 -0.5901 0.7642
146 S. Kapuria et al./Cmnput. Methods Appl. Mech. Engrg. 140 (1997) 139-155

Table 2
Comparison of exact piezoelastic, exact elastic and classical shell theory results
qo.5, 0.5) az (0, 0.025) iFz(1, 0.025)
Exact piezoelastic - 1.0019 0.5109 -0.5121
Exact elastic (di = 0) -1.0019 0.5105 -0.5116
Classical elastic thin shell theorv -1.0002 0.5105 -0.5105

The effect of radius-to-thickness ratio S on the response of a cylindrical shell with a/R = 4 has been
investigated for the six cases l-6. The results for the simply supported piezoelectric cylindrical shell
with a/R = 4, subjected to load of case 1, i.e. a uniform pressure on its outer surface with the inner
and outer surfaces held at zero potential, are presented in Table 3. The predominant stress component
induced is the circumferential stress a*. The longitudinal stress aZ in the cylinder has a very small value
except near the two ends. The through-the-thickness variation of various entities, at locations where
they are high, is shown in Figs. l-5 for seven values of S. It is observed from Fig. 1 that at midspan,
the radial displacements U of the outer and inner surfaces differ by more than 5% for S < 4. The axial
displacement W at the end varies nonlinearly across the thickness for S < 6. The distributions of the
radial stress a,. at the midspan and the transverse shear stress ;iZI at the end are given in Fig. 2. 5, is
maximum on the loaded surface for all values of S and the location of the maximum value of Ttr shifts
from the mid-surface towards the outer surface for S < 10. The distributions of the circumferential stress
e0 at the midspan and the longitudinal stress dZ at 5 = 0.025 are shown in Fig. 3. The circumferential
stress is uniform across the thickness for very thin shells, resulting in a membrane state. However, the
stress distribution becomes nonuniform as the shell becomes thicker indicating the presence of bending.
The difference of the values of c?~ at the outer and inner surfaces relative to the mean value increases
from 10% to 45% as the value of S decreases from 10 to 2. The distribution of the longitudinal stress tiZ
is almost linear with almost zero value at the mid-surface for S 3 4. The values of the electrical potential
4 and the electric displacements 0, and fi, induced in the shell for this case of loading without any
applied voltage are small. It can be observed from their distributions given in Figs. 4 and 5 that the signs
of 4 and 0, induced in very thick shells are opposite to those in thin shells.
The results for case 2 of sinusoidal pressure applied at the outer surface of the cylindrical shell with
grounded surfaces are presented in Table 4. The distributions of U, err, tie, ;izr and 0, across the thickness
are similar in nature to those in case 1 and are not depicted for brevity. However, unlike case 1 of uniform
pressure loading, through-the-thickness distributions of W and 5, have been found to be linear and that
of I$ to be parabolic for all values of S.
The results for case 3 of uniform pressure applied on the outer surface of the cylindrical shell with
uncharged surfaces are presented in Table 5. A comparison of the values in Tables 3 and 5 reveals that
the displacements and the stresses for the case of loading with uncharged surfaces are almost the same
as for the case of loading with grounded surfaces.

Table 3
Effect of radius-to-thickness ratio S for case 1
S 2 4 6 10 20 50 100
-U(O, 0.5) 1.2146 1.1043 1.0723 1.0474 1.0243 1 DO99 1.0050
-U(l, 0.5) 1.1560 1.0460 1.0274 1.0177 1.0085 1.0034 1.0017
-W(O, 0) 1.4302 1.1148 1.0004 0.9011 0.8161 0.7525 0.7243
-W(l, 0) 0.4798 0.4552 0.4668 0.4913 0.5284 0.5714 0.5965
-@,(0.5, 0.5) 0.6904 0.5952 0.5626 0.5376 0.5192 0.5090 0.5048
-;i,,(O.5, 0) 0.6805 1.026 1.287 1.701 2.449 3.821 5.444
-bs(O, 0.5) 1.633 1.271 1.172 1.103 1.051 1.020 1.010
-@o(l, 0.5) 1.076 1.007 1.002 1.002 1.000 1.000 1.000
-sz (0, 0.025) -0.2900 -0.3928 -0.4496 -0.5104 -0.5753 -0.5901 -0.4927
-&(0.5, 0.025) -0.0623 -0.0339 -0.0182 -0.0077 -0.0040 -0.0017 -0.GQO7
-4 (0, 0.025) 0.6381 0.4207 0.4417 0.4929 0.5641 0.5852 0.4906
-&(0.5, 0.025) 0.1942 -0.0596 -0.2437 -0.4733 -0.7149 -0.8271 -0.7105
-B,,(O, 0.5) 0.8859 0.8177 0.8028 0.7915 0.7795 0.7719 0.7693
-d,(l, 0.5) 0.5292 0.6352 0.6799 0.7160 0.7414 0.7566 0.7617
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 147

1.01

1 .o

c
0.5

0.0
0.4 0.6 1.0 1.2 1.4 0 2
-5 -?zr

Fig. 1. Distributions of 17 at 5 = 0.5 and W at 5 = 0 across the thickness for case 1.

Fig. 2. Distributions of 8, at e = 0.5 and ;izr at ( = 0 across the thickness for case 1.

_0.9 - 0.5 - 0.1 0.3

-a

1.0

3
0.5

0.0
-0.6 - 0.2 0.2 0.6
- 6== -6,

Fig. 3. Distributions of de and 8z at 5 = 0.025 across the thickness for case 1.

Fig. 4. Distributions of 4 at 5 = 0.025 and 0, at 5 = 0.5 across the thickness for case 1.
 148 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155

Fig. 5. Distribution of oz at 5 = 0 across the thickness for case 1.

Table 4
Effect of radius-to-thickness ratio S for case 2
S 2 4 6 10 20 50 100

-Ci(O, 0.5) 1.1619 1.1049 1.0747 1.0470 1.0243 1.010 1.005

-U(l, 0.5) 1.1113 1.0466 1.0297 1.0173 1.0085 1.003 1.002
-W(O, 0) 0.8893 0.6455 0.5685 0.5091 0.4660 0.4409 0.4326
-W(l, 0) 0.4664 0.4373 0.4316 0.4282 0.4261 0.4251 0.4247
-@,(0.5, 0.5) 0.6680 0.5859 0.5575 0.5346 0.5173 0.5069 0.5035
-?&OS, 0) 0.0709 0.0712 0.0705 0.0696 0.0687 0.0681 0.0679
-as (0, 0.5) 1.485 1.232 1.153 1.091 1.045 1.018 1.009
-&(l, 0.5) 1.111 1.042 1.025 1.013 1.006 1.002 1.001
-ez (0, 0.5) -0.4059 -0.3842 -0.3728 -0.3623 -0.3537 -0.3482 -0.3463
-“(l, 0.5) 0.3408 0.3478 0.3481 0.3473 0.3462 0.3452 0.3448
-C#J(O.5,0.5) 1.540 1.581 1.586 1.587 1.586 1.584 1.584
-O,(O, 0.5) 0.8414 0.8200 0.8053 0.7913 0.7795 0.7719 0.7693
-&(l, 0.5) 0.5011 0.6372 0.6813 0.7159 0.7415 0.7566 0.7617

Table 5
Effect of radius-to-thickness ratio S for case 3
S 2 4 6 10 20 50 100
-
-U(O, 0.5) 1.2029 1.0936 1.0620 1.0372 1.0144 1.000 0.9952
-U(l, 0.5) 1.1499 1.0383 1.0191 1.0087 0.9992 0.9938 0.9920
-W(O, 0) 1.4291 1.1148 1.0008 0.9020 0.8175 0.7543 0.7264
-W(l, 0) 0.4835 0.4593 0.4708 0.4951 0.5319 0.5745 0.5995
-&(0.5, 0.5) 0.6897 0.5948 0.5623 0.5374 0.5191 0.5090 0.5048
-& (0, 0.5) 1.628 1.269 1.170 1.102 1.051 1.020 1.010
-&(l, 0.5) 1.078 1.009 1.003 1.003 1.001 1.000 1.000
-I?~ (0, 0.025) -0.2898 -0.3921 -0.4485 -0.5088 -0.5730 -0.5869 -0.4892
-ci;-(1, 0.025) 0.6371 0.4199 0.4407 0.4915 0.5620 0.5821 0.4872

The response of the piezoelectric cylindrical shell subjected to electromechanical loading of case 4
with uniform electrical potential applied on its outer surface is presented in Table 6 and illustrated in
Figs. 6-8. It is noted from Fig. 6 that the radial displacement Li is almost uniform over the thickness
for thin shells but follows a linear to nonlinear nonuniform distribution as the shell gets thicker. The
difference in the values of B at the outer and inner surfaces increases from 13% to 63% as S decreases
from 10 to 2. The distributions of G and SZ become nonlinear, particularly near the outer surface, for
S 6 10. The circumferential stress So at the midspan, depicted in Fig. 7, has nonlinear variation across
the thickness for S < 20. It is revealed from Fig. 8 that though 3 follows a linear distribution for S > 4,
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 149

Table 6
Effect of radius-to-thickness ratio S for case 4
S 2 4 6 10 20 50 100

t-2(0,0.5) -1.0179 -0.8847 -0.8456 -0.8160 -0.7915 -0.7767 -0.7717

G(l, 0.5) -0.5292 -0.6352 -0.6799 -0.7160 -0.7414 -0.7566 -0.7617
w, 0) -0.2445 -0.0799 -0.0155 0.0431 0.0962 0.1390 0.1582
WO) 0.5402 0.4687 0.4268 0.3812 0.3287 0.2815 0.2579
Cqo.5, 0.5) -0.0027 -0.0063 -0.0064 -0.0162 -0.1318 -0.6321 -1.455
?U(O.5, 0) -0.6611 -1.207 -1.656 -2.538 -3.181 -3.133 -4.308
$0 (0, 0.5) -0:7912 -0.6329 -0.5846 -0.5768 -0.5614 -0.4930 -0.4481
BB(1, 0.5) -0.0225 0.9414 0.0882 0.9534 0.9697 0.9709 0.9728
&(O, 0.025) o.r39s5 2.362 3.322 4.297 8.558 22.88 38.10
&(l, 0.025) 3.OOO -1.089 -0.9489 -2.013 -6.980 -20.86 -36.33
&(O. 0.5) -78.57 -68.45 -65.53 -63.33 -61.74 -60.88 -60.58
d,(l, 0.5) -32..54 -60.58 -50.54 -60.25 -60.21 -60.21 -60.21

Fig. 6. Distributions of li at .$ = 0.5 and GJ at 5 = 0 across the thickness for case 4.

Fig. 7. Distributions of c?@at 5 = 0.5 and c?~ at 5 = 0.025 across the thickness for case 4.

the radial electric displacement br does not remain uniform over the thickness for shells with S < 10.
It can be noted from Figs. 7 and 8 that the distributions of ~20and fir show a reversal in curvature near
the outer surface as the shell gets thicker.
The results for case 5 of sinusoidal potential applied to the outer surface of the cylindrical shell with
no pressure loading are presented in Table 7. The distributions of 2, &, 3 and 6, at midspan are given
in Figs. 9 and 10. It is observed that, unlike case 4 of uniform potential, no reversal of the curvature
of the distributions of e0 and fir occurs near the outer surface. The distributions of other entities are
similar to those for case 4.
150

Fig. 8. Distributions of 4 and 8, at f = 0.5 across the thickness for case 4.

The results for case 6 of uniform charge density applied to the outer surface of the shell with no
pressure loading are presented in Table 8. The distributions of ii and G across the thickness, not shown
for brevity, are similar to those for case 4 of uniform applied potential. Unlike case 4, the distributions
of 50 and c+~at the midspan have been found to be almost linear across the thickness for the entire
range of S from 2 to 100. The distributions of 4 and fi, at the midspan have revealed that 4 is uniform
over the thickness for S > 6 and that D, varies linearly for S > 4.
The effect of length-to-radius ratio a/R has been investigated for a cylindrical shell with S = 20 for
the loading cases 1 and 4. The response of a shell for case 1 of uniform pressure loading without applied
potential is presented in Fig. 11 and Table 9. It is observed that the distributions of U, W, &,, Tzr and
tie are not affected significantly by the value of a/R. However, the distribution of 6, deviates from the
linear one for very short cylinders with a/R 6 1. The effect of a/R on the response of cylindrical shell
for case 4 of uniform potential with no pressure loading is illustrated in Figs. 12, 13 and Table 10. It is

Table 7
Effect of radius-to-thickness ratio S for case 5
S 2 4 6 10 20 50 100
li(0, 0.5) -0.9889 -0.8858 -0.8475 -0.8158 -0.7915 -0.7766 -0.7717
q1, 0.5) -0.5011 -0.6372 -0.6813 -0.7159 -0.7415 -0.7566 -0.7617
w, 0) -0.0413 0.0460 0.0742 0.0961 0.1120 0.1213 0.1243
w, 0) 0.2388 0.1938 0.1738 0.1561 0.1421 0.1333 0.1303
qo.5, 0.5) -0.1324 -0.1238 -0.1208 -0.1185 -0.1168 -0.1157 -0.1154
C+‘e(O,0.5) -0.6597 -0.5490 0.5172 -0.4934 -0.4764 -0.4666 -0.4634
Ba(1, 0.5) 0.4619 0.4562 0.4561 0.4570 0.4583 0.4593 0.4597
ez;o, 0.5) 0.0986 0.1235 0.1282 0.1308 0.1319 0.1324 0.1324
*z 0, 0.5) -0.0540 -0.0974 -0.1113 -0.1198 -0.1264 -0.1301 -0.1313
Ij,(O, 0.5) -76.60 -68.02 -65.35 -63.28 -61.73 -60.82 -60.51
6,(1,0.5) -50.15 -53.49 -55.79 -57.42 -58.74 -59.62 -59.91
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 151

t t

-1.0 -0.9 -0.6 - 0 .? -0.6 - 0.5

;

-0.8 - 0.4 as 0.0 0.4 -80 -70 -60 * -50 -40 -3

r

Fig. 9. Distributions of Li and I?@at & = 0.5 across the thickness for case 5.
Fig. 10. Distributions of 4 and fi,, at 5 = 0.5 across the thickness for case 5.

Table 8
Effect of radius-to-thickness ratio S for case 6
S 2 4 6 10 20 50 100

qo, 0.5) 7.9979 7.1732 6.9339 6.7338 6.5536 6.4424 6.4046

C(l, 0.5) 3.9744 5.1079 5.5513 5.9018 6.1377 6.2760 6.3214
q0, 0) 2.2954 0.9454 0.4035 -0.1050 -0.5863 -0.9954 -1.1939
q1, 0) -4.5941 -4.0379 -3.6963 -3.3068 -2.8667 -2.4435 -2.2217
qo.5, 0.5) 0.2949 0.2424 0.2231 0.2103 0.2055 0.2049 0.1993
fz,(0.25, 0) 2.085 4.496 6.561 10.29 17.98 22.92 30.21
&* (0, 0.5) 12 56 9.720 8.948 8.395 7.975 7.726 7.645
&(l, 0.5) -10.99 -9.782 -8.866 -8.451 -8.040 -7.785 -7.682
&(O, 0.5) 5 287 4.264 3.874 3.662 3.513 3.416 3.388
&(l, 0.5) -4 630 -4.120 -3.741 -3.627 -3.508 -3.428 -3.395
i(0, 0.5) -4152 -37.94 -36.64 -35.56 -34.74 -34.24 -34.07
i(L 0.5) -43 73 -38.48 -36.88 -35.65 -34.76 -34.24 -34.07
D,(OS, 0.5) 0.5468 0.5273 0.5191 0.5119 0.5062 0.5032 0.5019

Table 9
Effect of length-to-radius ratio a/R for case 1

alR 1 2 4 10

U(O.5, 0.5) 1.1255 1.0107 1.0160 1.0160

qo, 0) 0.2908 0.4659 0.8161 1.866
w, 0) 0.0035 0.1782 0.5284 1.5792
Cr(0.5, 0.5) 0.5147 0.5191 0.5192 0.5203
%(0.5, 0) 2.441 2.459 2.449 2.300
a0 (0, 0.5) 1.143 1.048 1.051 1.051
$I (1, 0.5) 1.126 0.9927 1.000 1.000
ez(O, 0.025) -0.2323 -0.4032 -0.5753 -0.4437
az (1, 0.025) 0.2382 0.3985 0.5641 0.4348
152 S. Kapuria et al. /Comput. Methods Appl. Mech. Engrg. 140 (1997) I_%I.2

Fig. 11. Effect of a/R on the distributions of W at 6 = 0 and 8, at 5 = 0.025 for case 1.
Fig. 12. Effect of a/R on the distributions of k at 5 = 0 and c?~ at 5 = 0.5 for case 4.

1.0

c
05

0.0
- 3.5 -2.5 -,.5 - 0.5 0.5
h

0.5-

-10 -5 R 5 IC

Fig. 13. Effect of a/R on the distributions of k* and c’+~at f = 0.5 for case 4.
 S. Kapuria et al./Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 153

Table 10
Effect of length-to-radius ratio a/R for case 4
alR 1 2 4 10

Li(O.5,0.5) -0.8477 -0.7621 -0.7661 -0.7661

qo.5, 0) 0.0697 0.1201 0.2202 0.5190
&(OS, 0.5) 0.0147 0.0970 -0.1318 -0.0652
t&0.5, 0) -4.427 -3.733 -3.181 -1.805
C?o(0, 0.5) -1.937 -0.5181 -0.5614 -0.5390
Ba(l, 0.5) -3.076 -0.2190 0.9697 0.7066
ez (0, 0.025) 8.959 7.262 8.558 6.733
ez(l, 0.025) 0.8871 -1.740 -6.980 -5.867
fir(O.5, 0.5) -60.26 -60.17 -60.23 -60.34

Fig. 14. Inferred pressure distribution at the outer surface for inverse problem.

revealed that the ratio a/R has no significant effect on the distributions of t and ci, across the thickness.
However, the distribution pattern of the radial normal stress & undergoes drastic change, including
reversal in its sign, as the shell becomes shorter. It is observed from Fig. 13 that the distributions of c?~
and kZ are highly nonlinear for very short shells (a/R < 2) and tend to linear one for the longer shells
(a/R > 10).
The method of solution for the inverse problem has been illustrated by taking the distribution of
potential differencle between the outer and inner surfaces of the shell, obtained in the direct problem of
case 3, as the known distribution of potential difference and inferring the pressure distribution applied
on the outer surface from this. The inferred pressure profile is compared in Fig. 14 with the applied
pressure distribution for a simply-supported shell with a/R = 4 and S = 20. There is very good agreement
between the inferred and applied pressure distributions.

6. Conclusions

Three-dimensional exact analytical series solution has been presented for simply-supported piezoelec-
tric cylindrical shell subjected to axisymmetric electromechanical loading. The solution is constructed as
a product of an exponential function and a power series in the thickness coordinate. The solution can
be obtained to any desired degree of accuracy by truncating the power series to appropriate number of
terms. Numerical results are presented for simply-supported piezoelectric shells of PVDF subjected to
uniform and sinusoidal pressure, applied potential or electical displacement. The effect of geometrical
parameters on the response has been investigated for shells with radius-to-thickness ratio ranging from 2
to 100 and the span-to-radius ratio varying from 1 to 10. It has been observed that the radius-to-thickness
ratio has a predorninant effect on the response. In the absence of any applied potential, the results for
piezoelastic shells subjected to pressure loading differ little from those of elastic shells. Solution of an
1.54 S. Kapuria et al./Cornput. Methods Appl. Mech. Engrg. 140 (1947) 139m1.5.~

inverse problem has been illustrated. The solution is being extended to hybrid laminated shells with
surface bonded or embedded piezoelectric layers. The detailed numerical results and the exact distri-
butions of several entities across the thickness, presented in this study, would also enable development
and assessment of two-dimensional piezoelastic shell theories.
The method of solution used in this study is not applicable to other boundary conditions and geometries
of shells with varying curvature. A three-dimensional finite element method or a two-dimensional one
based on piezoelastic shell theory would be needed to analyse the general case. The authors intend to
pursue such work in future. The present solution would serve as benchmark against which the accuracy
of two and three-dimensional finite element solutions would be assessed.

Appendix A: Solution of eigenvalue problem

The characteristic equation of the eigenvalue problem AY” = AY, is det(A - AZ) = 0. Expanding the
determinant explicitly using relations (30) and setting p = sh(sh + l), the characteristic equation can be
reduced to the following cubic equation:
p3 + c2p2 + c1p + co = 0 (A.11
where

co = IsAl, c2 = -al,(l + %> - a13a31 + @al2 + a24442 - 2425U46 - &6a65)b=d2,

Cl = [(I +a11>@ 12 - alla24a42 + 2alla25a46 + @la56a65) + a12(a12a24”31 - 2a16a25u31)

+al3a32(1 - a24a32) - al2a24a32(1 + 2%) - al6a25a32(l - 2all)

+a2$36(3@2 + &1a12 + %3a32) + a,3a31(2a2Sa46 - a24a42) - a25446 + al6a65@36 - a31”16)

+a36a65@llal6 + al3a36) +~13”31u56u651@Rl)2 + b:, - 413u42 + 2a25(al6a42 - ul2446)

+a& (ai6 - a42a56) + a46a65 @al6 + a244461 - u56a6S @l2 + u24u42)l cnRl I4 G4.2)

Eq. (A.l) can be solved, by Cardan’s method, by transforming it into

A3+DA+F=0 64.3)
where
A = p + C2/3, D = (3q - c;),‘3, F = (2~; - !klC2 + 27co)/27 (A-4)
The nature of the roots of Eq. (A.3) depends on the sign of G = F2/4 + D3/27 ([12]):
1. If G > 0, then one root is real and the other two are complex conjugate as follows:

A, = w - p, AZ, A3 = -A,/2 & i&(w + p)/2, with

(A.9
w = (v% - F/2)1’3, p = (& + F/2)“3
2. If G 6 0, then the roots are real. If G < 0, then the roots are distinct; if G = 0 and D # 0, then one
root is a double root. The roots are given by

Aj = 2(-D/3)1/2~~~[{y + 2n(j - 1)}/3], j = 1,2,3, where cos y = --v!ZF/~(-D)~/~

(fw
If G = 0 and D = 0, then all the three real roots are equal:
A, = 0 64.7)
The roots pi’s are then computed from Eq. (A.4) as pj = Aj - c2/3, (j = 1,2,3). The roots A’ (= sh)
corresponding to a root pj can be determined by solving the quadratic equation
A’2+ A’ - pi = 0 WV
For a real pi, if (1 + 4pi) > 0 then the two roots A’ = [-1 i (1 + 4pj)“2]/2 are real, and if (1 + 4pj) < 0
then the roots A’ = [-1 fill + 4pj11j2)/2 are complex conjugate. For a pair of complex conjugate roots
pj = rj f iSj (with 6 > 0), the four roots of Eq. (A.8) are
 S. Kapuria et al. /Comput. Methods Appl. Mech. Engrg. 140 (1997) 139-155 155

A’ = [-1 f (II f iv)]/2 where p = [{l + 4y + [(l + 4~)~ + 1662]‘/2}/2]‘/2, u = 28/p. (A.9)
Finally, the eigenvalues hj (j = 1,. . . ,6) are obtained as hj/s.
Let Y0 = [YO, Yo2 . . . YoJT be the eigenvector corresponding to a distinct eigenvalue A. Without loss
of generality we set Ya, = 1 and rewrite AYo = AYo after deleting the kth equation as
(A’ - AZ)Y’ = -R’ (A.lO)
where Y’ and R’ are obtained from the eigenvector Y0 and the kth column of matrix A after deleting
the elements in the kth row. A’ is the reduced matrix obtained from A by deleting the kth row and
kth column. Choosing k such that IA’ - All # 0, we can solve Eq. (A.lO) for Y’. If eigenvalue A is real
then vector Y’ is also real since the matrix A is real. However, for complex eigenvalue A = (Y+ ifi, the
eigenvector is complex and we let Y’ = Yfi + Yi. Yk and Y; can be determined using only real arithmetic.
The real and imaginary parts of Eq. (A.lO) yield
(A’ - aZ)Y; t- PY; = -R’, -/3Yf, + (A’ - aZ)Y, = 0 (A.ll)
_ -
=+ Y; = --P[A.A + p2Z]-lR’, Y; = Ayj/p, with A = A’ - al (A.12)

The elements of eigenvector Y0 are now obtained as

Ya, = Y; for i < k, Yo, = 1, YOi = Y,:I for i > k. (A.13)

References

[l] M.C. Ray, K.M. Rao and B. Samanta, Exact analysis of coupled electroelastic behavior of a piezoelectric plate under cylindrical
bending, Comput. Struct. 45 (1992) 667677.
[2] M.C. Ray, K.M. Rao and B. Samanta, Exact solution for static analysis of an intelligent structure under cylindrical bending,
Comput. Struct. 47 (1993) 1031-1042.
[3] P. Heyliger and B. Brooks, Static behavior of piezoelectric laminates with distributed and patched actuators, J. Intelligent
Mat. Systems Struct. 5 (1994) 635-646.
[4] M.C. Ray, K.M. Rao and B. Samanta, Exact solutions for static analysis of intelligent structures, AIAA J. 31 (1993) 16841691.
[5] P Heyliger, Static behavior of laminated elastic/piezoelectric plates, AIAA J. 32 (1994) 2481-2484.
[6] J.A. Mitchell and J N. Reddy, A study of embedded piezoelectric layers in composite cylinders, J. Appl. Mech. Trans. ASME
71 (1995) 166-173.
[7] W.S. Burton and A.K. Noor, Three-dimensional solutions for thermomechanical stresses in sandwich panels and shells, J.
Engrg. Mech. ASCE 120 (1994) 20442071.
[8] H.F. Tiersten, Line,sr Piezoelectric Plate Vibrations (Plenum Press, New York, 1969).
[9] S.S. Rao and M. Sunar, Analysis of distributed thermopiezoelectric sensors and actuators in advanced intelligent structures,
AIAA J. 31 (1993) 1280-1286.
[lo] C.K. Lee, Piezoelectric laminates: theory and experiments for distributed sensors and actuators, in: H.S. Tzou and G.L.
Anderson, eds., Intelligent Structural Systems (Kluwer Academic Publishers, Dordrecht, 1992) 75-167.
[ll] S. Timoshenko and S. Woinowsky-Kreiger, Theory of Plates and Shells, 2nd edition (McGraw-Hill, New York, 1959) 477.
[12] VI. Smirnov, A Cc’urse of Higher Mathematics, Vol. 1 (Pergamon Press, London, 1964) 491-497.