Sie sind auf Seite 1von 24

Int. J. Solids Structures Vol. 23, No.4, pp. 441-464, 1987 Printed in Great Britain.

0020-7683/87 $3.00 +.00 Pergamon Journals Ltd.

ANALYTICAL, THREE-DIMENSIONAL ELASTICITY SOLUTIONS TO SOME PLATE PROBLEMS, AND SOME OBSERVATIONS ON MINDLIN'S PLATE THEORY

W. H. WITTRICKt The Old Forge, Ebrington, Chipping Campden, Gloucestershire GL556NL, U.K.

(Received 15 July 1985; in revised form 8 February 1986)

Abstract-Solutions are obtained, using the exact three-dimensional theory of elasticity, to (i) the eigenvalue problem of buckling under biaxial compression or the free lateral vibration of a simply supported rectangular plate with orthotropic stress-strain properties, and (ii) the static response of the same plate to a lateral load that varies sinusoidally in two directions. The eigenvalue in problem (i) and the lateral deflections of both the surface and middle plane of the plate, as well as the bending strains, in problem (ii) are obtained in the form of series expansions in even powers of the plate thickness. Exact algebraic expressions are presented for the first two coefficients in the case of orthotropic plates; additional coefficients can be obtained if required, and are given for the much simpler case of isotropic plates. In all cases the first term agrees with classical plate theory. The solutions are compared with those obtained from Mindlin's plate theory. In neither problem is it found to be possible, in general, to choose values for Mindlin's effective shear moduli to make the Mindlin solution agree with the first two terms of the exact solution. There are, however, two exceptions to this, namely a restricted class of orthotropic materials, embracing all isotropic ones, in which the elastic constants satisfy a certain condition, and the case of cylindrical bending when the Mindlin plate reduces to a Timoshenko beam of wide rectangular cross-section. In both these exceptional cases appropriate values for the effective shear moduli are obtained.

1. INTRODUCTION

Classical plate theory (CPT) assumes that (i) the lateral displacement w is constant through the thickness, (ii) normals to the middle surface remain both straight and normal after deformation, and usually (iii) in the dynamic case the rotational inertia, arising from u, v displacements parallel to the middle surface due to bending, is negligible. With the possible exception of edge effects, the error in CPT is O(h2 I A, 2) where h is the thickness and A, a typical "half-wavelength", or characteristic length, of the bent surface. For isotropic plates the error is usually very small in practice, though it can be significant at the shorter wavelengths associated with high frequency vibration modes. Nevertheless there has been much attention paid to the development of higher order plate theories with the implicit aim of reducing the error to O(h4/A,4) or less. Notable among these are the theories of Reissner[l] and Mindlin[2]. There have also been other, more recent, theories proposed but they will not be mentioned further here. In recent years, however, the development of fibre-reinforced composites has resulted in greater interest being shown in higher order theories. The reason is that in such plates the shear modulus associated with transverse (i.e. "through-the-thickness") shearing stresses is often very small compared with the elastic moduli associated with the bending stresses; this results in CPT becoming inadequate at much smaller hi A, ratios than in the isotropic case.

In order to assess the accuracy of approximate higher order theories it is obviously advantageous if some "exact" solutions of plate problems, based upon three-dimensional elasticity theory, are available for comparison. Such solutions have been derived for certain problems associated with the buckling, vibration or static loading of simply supported, orthotropic rectangular plates, of both homogeneous and laminated types, notably by Srinivas and co-workers[3-6] and Pagano[7]. They obtained many numerical solutions in this way and compared them with various approximate theories. In this paper we shall obtain algebraic solutions for the same set of problems considered by Srinivas and co-workers except that, in order to keep the algebra within manageable

t Professor W. H. Wittrick, FRS, died in June 1986. Please address any correspondence concerning this paper to the Department of Civil Engineering, University of Birmingham, P.O. Box 363, Birmingham B15 2TT, U.K.

441

442

W. H. WITTRICK

bounds, we shall confine ourselves to homogeneous plates only. The problems considered are (i) the eigenvalue problem arising from either the buckling under biaxial compression or the free vibration of the plate, and (ii) the response to a static, lateral loading with intensity varying sinusoidally in both the x- and y-directions. Both problems result in a doubly sinusoidal mode of deformation, enabling a solution to be obtained by separation of the variables. We shall start with the basic equations of three-dimensional, orthotropic elasticity, but whereas Srinivas and co-workers[3-6] resorted to computation at an early stage and obtained numerical solutions for specific cases our solutions will be entirely algebraic. The eigenvalue in problem (i) and the deflections and bending strains in problem (ii) are in the form of series expansions in powers of h 2 Specific expressions are given for the first two terms in the case of a general orthotropic material; more terms are obtainable but they become increasingly complicated algebraically. It is, however, rather easy to obtain additional terms of the series for an isotropic plate, and this has been done. These solutions are then used to discuss the accuracy of Mindlin's plate theory. What appear to be new results are obtained for the effective shear modulus in the case of isotropic plates, enabling the error to be reduced from O(h2j22) to O(h4j24), and smaller still for all practical purposes in the eigenvalue problem. But for orthotropic plates it is shown that in general it is impossible to select values for Mindlin's two effective shear moduli which reduce the error to O(h4j24) for all values of the ratio between the wavelengths in the x- and y-directions. The only exception to this is a class of orthotropic materials, embracing all isotropic ones, in which the elastic constants satisfy a very restrictive condition. It is possible also in the case of cylindrical bending of an orthotropic plate, in which case the Mindlin plate behaves effectively as a Timoshenko beam of wide rectangular cross-section.

2. THREE-DIMENSIONAL THEORY OF ELASTICITY SOLUTIONS

2.1. The basic equations of three-dimensional elasticity

Consider a homogeneous elastic plate, of uniform thickness h, subjected to uniform static compressive stresses CT~ and CT~ parallel to axes x and y lying in the middle plane of the plate. Suppose that small displacements u(x, y, z, t), v(x, y, z, t) and w(x, y, z, t) occur from this datum state of uniform biaxial compression. Then, if there are no body forces, the equations of motion of an element in the x-, y- and z-directions are

OCTx + Ot XY +

ox

Ot xz =

.!l'(u)

oy

OZ

Ot xy + OCTy + Ot yz = .!l'(v)

ox

oy

OZ

Ot xz + ot yZ + OCT z = .!l'(w) ox

oy

oz

where the differential operator .!l' is defined by

0 2

0 2

0 2

.!l' = CT~ox2 + CT~oy2 + p ot2

(1)

(2)

and p is the density. The stresses CT x , CT Y ' CT z ' t xy , txz and tyz are additional to those in the datum state. The operator .!l' which appears on the right-hand sides of eqns (1) takes account not only of the inertia forces but also of the destabilizing effect of CT~ and CT~ in all three directions. The simplest way of deriving these terms is via a variational principle, using Green's strain tensor to relate the linear strains in the x- and y-directions to the

Analytical, three-dimensional elasticity solutions to some plate problems

443

displacements u, v, w by the quadratic expressions (iJu/iJx) + e~ and (iJv/iJy) + e;, respectively, where

e~ = ~[(:~J + (::J + (~:JJ. e; = ~[e;J + (:;J +

e;J}

(3)

The constitutive equations for an orthotropic material with principal axes parallel to x, y, z are

(1x

= cllex

+

cl2 e y +

C13 e z'

'yz = C44Yyz

(1y

= c 12 e x

+ cney + C23ez'

'xz = CssYn

(1z = C13 e x + c 23 ey + C33ez,

'xy = C66Yxy

where the strains are given by

(4)

 

iJu

iJv

iJw

 

ex =

iJx'

e y = iJy'

e z =

iJz'

 

(5)

iJv

iJw

iJu

iJw

iJu

iJv

Yyz -

--+-,

iJz

iJy

Y

--+-,

iJz

iJx

xz -

Yxy =

iJj + iJx'

Substitution of eqns (4) and (5) into eqns (1) gives three equations of motion expressed in terms of the displacements, as follows:

(

iJ2

iJ2

iJ 2

Cll iJx 2 + C66 iJy2 + css iJ z 2 -

)

ft'

iJ 2 v

iJ 2 w

u + (C12 + C66 ) iJxiJy + (C13 + c ss ) iJxiJz = 0

iJ 2 u (C12 + C66) iJxiJy +

(

.iJ2

iJ2

iJ

2

C66 iJx 2 + Cn iJy2 + C44 iJ z 2 -

ft'

) iJ 2 w v + (C23 + C44) iJyiJz = 0

iJ 2 u

iJ 2 v

(C13 + css) iJxiJz + (C23 + C44 ) iJyiJz +

2

(iJ c ss iJx2

iJ2

iJ2

+ C44 iJy2 + C33 iJ z 2 -

ft'

)

w =

O.

(6)

In the special case of an isotropic material we have

and

 

(1 -

v)E

_

2(1 -

v)G

c ll

=

C 22

=

C 3 3

= (1 + v)(1 -

2v) -

(1 -

2v)

vE

2vG

c 12 = c 13 = C 2 3 = (1 + v)(1 -

2v) = (1 -

2v)

 

C 44

E

= css = C 66 = 2(1 + v)

=G

(7)

where E, G and v are the Young's modulus, shear modulus and Poisson's ratio, respectively. Equations (6) then take the concise form

{

iJ

iJx; iJy; iJz

iJ

iJ}

e + (1 -

2v)Jf'{u;v;w} = 0

(8)

444

W. H. WITTRICK

where e is the dilatation, defined by

e = (ou/ox) + (ov/oy) + (ow/oz)

and the differential operator Jf is defined by

0 2

Jf = ox2 + oy2 + OZ2 - G O'~ox2 + O'~oy2 + p ot2

0

2

0 2

0 2

2

0 2

1(0

(9)

) (to)

.

2.2. The general solution for a doubly sinusoidal mode of displacement

The problems for which we shall obtain solutions are all concerned with a mode of displacement of the form

u(x, y, z, t) = U(z) cos IXX sin f3y cos rot

v(x, y, z, t) = V(z) sin IXX cos f3y cos rot

w(x, y, z, t) = W(z) sin IXX sin f3y cos rot.

(11)

Using eqns (4) and (5) it can then be seen that w = 0, v = 0 and O'x = 0 on the edges x = 0 and a, whilst w = 0, u = 0 and O'y = 0 on the edges y = 0 and b of a rectangular plate, provided that IX and f3 are such that lXa/rr and f3b/rr are integers. For convenience we shall refer to edges with these boundary conditions as simply supported. Clearly the conventional definition of a simply supported edge in classical thin plate theory conforms to this more general definition. If eqns (11) are substituted into eqns (6) we obtain the following simultaneous differential equations for U, V and W

(EC2 -

as)

- (CSS/C44)1/2b3

-(CSS/C33)1/2b4EC

r

where EC is the operator d/dz

- (C44/CSS)1/2b3

(EC 2

-

a 4 )

-(C44/C33)1/2bsEC

(C33/ C SS)1/2 b 4 EC ] r u ]

(C33/C44)1/2bsEC

V

(EC 2

-

a 3 )

W

=

a3 = (CsslX2 + C44f32 - ~)/C33'

a4 = (C661X2 + C22f3 2 - ~)/C44'

as = (cll1X 2 + C66f3 2 - ~)/CSS,

b 3 = (C12 + C66)1Xf3/(C44CSS)1/2

b 4 = (C 1 3 + Css)IX/(C33CSS)1/2

b

s = (C23 + C44)f3/(C33C44)1/2

and ~ is defined by

~ = 1X20'~+ f320'~+ pro 2

.

0

(12)

(13)

(14)

If U and V are eliminated from eqn (12) we obtain a sixth-order differential equation for W(z), namely

where

Ql =

(EC 6 -

Q 1 EC 4 + Q 2 EC 2 -

a 3 + a 4 + as -

b~

-

b;

Q3)W = 0

Q2 = a 3 a 4 + a 4 a S + a S a3 + 2b 3 b 4 b s - b~ - a4b~ - asb;

Q3 = a 3 (a 4 a S -

b~).

(15)

(16)

Analytical, three-dimensional elasticity solutions to some plate problems

445

Let the roots of the auxiliary equation of eqn (15) be ± q I' ± q2 and ± q3' Then

Q I =

qi + q~ + q~

Q2

=

qiq~ + q~q~ + q~qi

Q3 = qiq~q~·

(17)

In the analysis that follows it is fortunately never necessary to obtain expressions for the individual roots, ql' q2 and q3' Only eqns (17), and the values of certain other combinations of the three roots that can be deduced from them, are required, and it is immaterial whether the individual roots are real, imaginary or complex. The other combinations that appear in the analysis are always reducible to the form I""ft or J""ft' where

I""ft = I* [q~"'(qfft- qJ")]

J""ft = I*[qf"'qJ"'(qfft

-

qJ")]

(18)

(19)

and m and n are positive integers. The symbol I* denotes a summation extending over

three terms, formed by allocating the values (1,2,3), (2,3, I) and (3, 1,2) to the suffixes (i,j, k) in the general term shown. Certain relations between the various I""ft and J""ft quantities are given in Appendix A, and in particular it is shown that they can all be expressed in the form J I, I times a polynomial function of QI' Q2 and Q3' The general solutions of the differential equations, eqns (12), are

 

U(z)

=

3

- I qje~Kjsinh qjZ

+ Ki cosh qjz)

 
 

j=

I

 

V(z)

=

-

I

3

qj!.{K j sinh

qjZ + Ki cosh qjz)

(20)

 

j= t

 
 

W(z) =

I

3

(Kjcoshqjz + Kisinqjz)

 
 

j=

I

where the K j and Ki are constants of integration, and

 
 

ej

= (C33/CSS)I/2[b3bs + b.(qi -

a.)]/Zj

 

(21)

 

h

=

(C 33 /C•• )1/2[b 3 b.

+ bs(qf -

as)]/Zj

in which

 

Zj =

 

(qf -

a.Xqf -

as) -

 

(22)

[Note that U = -q •.ej, V = -qjh is the solution of the first two of eqns (12) with ~

replaced by qj and W =

The stresses acting on planes perpendicular to the z-axis, which are required later, are obtainable from eqns (4), (5), (11) and (20). They are given by

1.]

(1z =

8 z sin ax sin {3ycos wt

r xz = f xz cos ax sin {3y cos wt

ryz = fyz sin ax cos {3ycoswt

(23)

446

W. H. WITTRICK

where the quantities bearing a circumflex are functions of z given by

8% =

fx% =

fy% =

3 L q,{c dxei + C 23 PJ; + C 33 )(Kj sinh qiz + K; cosh qiZ)

i= 1

3

C ss L

i= 1

(tX -

3

C44 L (P -

i= 1

qfei)(Ki cosh qiz + K; sinh qiZ)

qf J;)(K i cosh qiz + K; sinh qiZ).

2.3. The eigenvalue problem and its solution

(24)

We shall now solve, simultaneously, the problems of:

(a) initial buckling of a simply supported rectangular plate under the uniform biaxial,

static, compressive stresses CT~ and

CT~, and

(b) free lateral vibration at a frequency w of a simply supported rectangular plate

about the datum state of uniform biaxial, static, compression CT~and CT~.

In both cases the mode is of the form of eqns (11). In the buckling problem we have w = 0, and e, defined in eqn (14), is then a load parameter. In the vibration problem we postulate that CT~and CT~are specified but are not large enough to cause the plate to buckle, and eis then a frequency parameter. Both are eigenvalue problems, and since CT~, CT~ and w appear only in combination in the form of the parameter e, its eigenvalues simultaneously give the buckling loads in problem (a), and the natural frequencies in problem (b). Moreover, the buckling and vibration modes are identical. It is clear that the lowest buckling load and the lowest natural frequency, for given values of the wavelength parameters tX and p, will be associated with modes in which W(z) is an even-valued function whilst U(z) and V(z) are odd-valued functions. Hence, referring to eqns (20), we may assume that K; = 0 (i = 1,2,3). We now impose the condition that the plate surfaces, Z = ± th, are free from traction, so that

Equations (24) then give

[

q~FlTl

(tX - e 1 qi)

(P -

flqi)

8%=fx%=fy%=0

atz=±th.

q~F2T2

(tX -

e2q~)

(P - f2q~)

q~F3T3J[Kl COShthq

(tX -

e3q~)

(P - f3q~)

1 j

K2 cosh thq2

cosh thq3

K3

=

0

where

 

Fi =

c 13 tXe i + c 23 PJ; + C 33

and

1i = tanh Hhqi)/Hhqi)·

(25)

(26)

Alternatively 1i can be expanded into the form of a series

1i = 1 -

(

h2) 6(h2)2

12

qf + 5

12

qt -

51(h 2 )3

35

12

62(h 2 )4

12

q~ + 35

q~

-

4146(h 2 )5

1925

12

qtO +

(27)

Analytical, three-dimensional elasticity solutions to some plate problems

447

For a non-trivial solution of eqn (25) to exist, the determinant of the matrix must vanish and this gives the following characteristic equation

L* [qf FjT;{rx(qfh - qUk) + P(qfek -

qfej) + qfqf(ej!k -

ekh)}J = 0

(28)

where the summation is as defined in the sentence following eqn (19). If eqns (26) and (27) are substituted into eqn (28) the resulting equation can be expressed as

where

1/1(0) _ (h2)1/I(1) + ~(h2)21/1(2) _ ~(h2)31/1(3) +

12

5

12

35

12

= 0

I/I(r) = L*[qfr+2(c 13 rxej + c 23 P/; +

+ qfqf(eJk -

ekh)}].

C 33 ){rx(qfh -

qUk) + P(qfek -

qfe)

(29)

(30)

Note that the coefficient of I/I(r) in eqn (29) is the same as that of qfr in the series expansion of T; in eqn (27). After a cyclic modification of the suffixes i, j, k on some of the terms in eqns (30) it can be written as

where

,/,(r) _

'I'

-

c

13

rx{rxH(r) + pH(r) + H(r)} + c

11

12

13

+ C33{rxH~l + PH~1 + H~1}

23

p{rxH(r) + pH(r) + H(r)}

21

22

23

H (r)

11

=

"'* 2

L

qj qj

2(

qj

2r

f

e j j -

2r r)

qj ejJj

H<t1 = L*qfqf(qf' -

qfr)eiej

H<t1 = Q3L*qfrek(ejh -

ej/;)

H~l = L*qfqf(qfr - qfr)/;h

H (r)

22

-

-

"'*

L

qj qj

2

°2(

qj

2r

e

i

f

j

-

2r r)

qj ejJj

H~1 = Q3L*qf'h(e i h - ej/;)

H~l = L*qfqf(qf1j - qf'J;)

H~1 = L*qfqf(qf'e i - qfre)

H~1 = Q3L*qfr(e i h

-

ej/;).

(31)

(32)

By using eqns (16), (17), (21) and (22) all of the Ht~ summations defined in eqns (32) can be expressed in terms of the Im,n or Jm,n types, as defined in eqns (18) and (19). These in turn can all be expressed in terms of J 1 ,1 as explained in Appendix A, and this enables the Ht~ summations to be evaluated. It turns out that they all contain a common factor, ", defined by

" = (C33/Z1Z2Z3)(C44C55)-1/2(b; - a4a5)[b4b5(a4 - a 5 ) + b 3 (bi - bmJl,l

(33)

448

W. H. WITIRICK

Table I. Values of the H~~ summations, defined by eqns (32), for r = 0, 1 and 2

r

0

2

 

-1

(b~ - a~)

Q,(b~ -

a~) +

Q2 -

a 3 a 4

K-'Hn

-1

(b; - a 4 )

Q,(b; -

a 4 ) +

Q 2 - a3a~

(C~~/C44)'/2K-'Hn

o

o

(C~~/C33)'/2K-'Hn

o

(C44/C33)'/2K - 'H~~

o

b 4

(C33/C44)'/2K-' Hn

K-'Hn

-a 3

(b4b~ -

(b4b~ - b 3 )

b 3 )

-a 3 b 4

-a3b~

[b 4 (Q, - a 4 ) + b3b~]

[b~(Q, - a~) + b 3 b 4 ]

-a 3 (a 3 -

b~ - b;)

It is now convenient to define 'I'(r) by

'I'(r)

= K- 1 1jJ(r).

Q,(b4b~ -

Q,(b4b~ -

-a 3 (Q,b 4

-a3(Q,b~

b 3 ) + b 3 ) +

+

+

a 3 b 3 a 3 b 3 a 4 b 4 )

- b 3 b 4 - a~b~)

b3b~

Q,[b 4 (Q, - a 4 ) + b3b~] - b 4 Q2

Q,[b~(Q, - a~) + b 3 b 4 ] - b~Q2

-Q,a 3 (a 3 -

b~

-

b;) + a 3 Q2 -

Q3

(34)

Expressions for '1'(0), '1'(1) and '1'(2) are readily obtainable using Table 1 and eqns (13), (16) and (31); they are as follows

\IJ(l)

T

=

'1'(0)

= ~

[-

CliO(

4

+

2(-

C 12

+

2)

C660(

2f32

+ C 22

-

f34]

_ 0(2

~[

+ f32 + 2(C130(: + C23f32~ _

33

J

~

C33

(35)

(36)

'1'(2) = [(Cll _2C 13 )0(2 + (C 22 _ 2C23)f32 _ ~(_1+ _1 + _1 - ~)J'I'(l)

where

+

C 55

(

0(2

C 44

-

C 33

f32

+ -

C ss

C 44

C 33

C 33

~)[ 2f32{ (C12 - + 2C 66 )

0(

- --

C 44 C 5S

2

-

C 44

C 11 C

22

-

-}

C ss

C 44 C S5

t{(-

+.,

C ll -

2C

13

C

66

C 3 3

)

C1.

2

+ (C22 - 2C~::66)p2} - ~2(1 + ~::)] - ~Q2

C

ll

= C ll

-

(ci3/ C 33)'

C 22

=

C 22

-

(d3/ C 33)'

C 12 = C 12 -

(c13C23/C33)'

(37)

(38)

By putting (1z = 0 in eqns (4) and eliminating the corresponding value of C z from the equations for (1x and (1y it can be seen that C 11 ' C22 and c 12 are modified elastic constants associated with a state of plane stress (see eqns (68». It can be seen from eqn (15) that

q? =

Ql

-

Q2qi- 2 + Q3Qi- 4 .

Equations (32) and (31) respectively then show that

and

H(r+l) =

IlV

Q

1

H(r) _

IlV

'I'(r+l) = Ql'1'(r) _

Q

2

H(r-l) + Q

IlV

H(r-2)

3

IlV

Q2'1'(r-l) + Q3'1'(r-2).

This recurrence equation, in conjunction with eqns (35)-(37), enables '1'(3), '1'(4),

(39)

(40)

to

Analytical, three-dimensional elasticity solutions to some plate problems

449

be obtained sequentially if required. Note the remarkably simple result for 'P(O) in eqn (35), as a result of which eqn (29) becomes

e= h 2 'P(1) _

12

~

120

'P(2)

17h 6

'P(3) _

+ 20,160

.

(41)

It is apparent from this equation that eis of O(h2) and a first approximation for e can be obtained from the first term of eqn (41) alone, with 'P(1) calculated from eqn (36) with e = O. Thus

h

2

e= 12 [C ll 0(4 + 2(C 1 2 + 2C 66 )0(2p2 + C22P 4 ] + O(h4).

(42)

It will be seen later than this agrees with the result of classical thin plate theory for an orthotropic material. A second approximation for ecan now be obtained from the first two terms of eqn (41), using the first approximation (42) to adjust the expression for 'P(1) in eqn (36), but with 'P(2) calculated from eqn (37) with e = O. After some algebraic manipulation this second approximation can be written as

2

2

e= 12 [ 1 - 12(0( +

h

h

2

p2 ]

)

(0(

2

h 2 [( 1

--

--

120

Css

C13)

_

3CllC33

+

IX

2P2(

C23 1X

2

P2)

-- -_- + -_-

3C 33

C 22

C ll

C13

'111 +

P2)

'122

0(

2

2

'111 +

(1

--

C 44

C23

3C22C33

)p2

2

'122

{(C12 + 2C 66 ) - C ll C22} ] + O(h )

-

2

-

-

6

(43)

where, to save space, we have introduced the quantities

'111 = C ll IX 2 + (C 12 + 2C 66 )P 2

and

'122 = C 22 P 2 + (C12 + 2C 66 )1X2.

(44)

In principle this process for obtaining approximations of progressively higher order can be continued indefinitely, but we shall not take it any further in the case of an orthotropic plate, partly because of the algebraic complexity and partly because eqn (43)

is sufficient for our purposes. However, the equations for an isotropic plate are algebraically

much simpler and in that case we have taken the process one stage further and obtained

a third approximation for e. The result can be expressed very concisely if we introduce

two dimensionless parameters <p and A, given by

and

<p =

(0(20'~+ p20'~+ pro 2 )

G(1X2 + p2)

.1= (1X2 + P2)h 2 /12.

(45)

(46)

450

W. H. WITT RICK

In terms of these parameters, the third approximation is

¢

2d

= 1_ )1 -

J.i.1d + J.i. 2 d 2 ]

+ O(d 4 )

where

 

17 -

7v

 

J.i.l

= 5(1 -

v)

and

J.i.2

62

= 35

+

(62 -

5(1

_

42v)

V)2

.

(47)

(48)

(49)

Although eqn (47) has been derived by reduction of the equations for an orthotropic plate, it can be derived in an alternative and more direct way if isotropy is assumed ab initio, thereby providing a welcome check on the correctness of the algebra in the orthotropic case. An outline of this alternative derivation is given in Appendix B. Equations (43) and (47) provide a yardstick against which the accuracy of higher order plate theories can be measured. In particular we shall use them in Section 3.2 to investigate the accuracy of Mindlin's plate theory.

2.4. The static loading problem and its solution

We now turn to the problem of a simply supported rectangular plate, with sides a

and b and with zero in-plane loads (i.e.

the z-direction of magnitude p per unit area given by

O"~ = O"~ = 0), subjected to a static lateral load in

p =

P sin IXX sin f3y

where IX and f3 are such that lXa/n and f3b/n are integers. We shall obtain a solution in the form of eqns (11) with w = o. In classical plate theory, and indeed in many higher order plate theories, it is immaterial whether the lateral loading is applied to just one surface, or shared between the two surfaces z = ± tho The reason for this is, of course, that w is assumed to be independent of z. In the three-dimensional theory, however, it is necessary to specify precisely how the load is applied. We shall assume that it is shared equally between the two surfaces, as a pressure tp on one surface and a tension tp on the other, as shown in Fig. 1(a). This results in an enormous simplification of the analysis, because the displacements are now antisymmetrical with respect to the middle plane z = O. The boundary conditions to be satisfied are

O"z =

±tp,

'xz =

'yz

0

atz = ±th.

(50)

It should be noted that if the load is applied as a pressure p on the surface z = - th only, the exact solution could be obtained by superposition of the solutions for the antisymmetric and symmetric loadings of Figs 1(a) and (b). Clearly the middle plane deflection in Fig. 1(b) is zero, so the middle plane deflection is due entirely to the antisymmetric load component of Fig. 1(a). On the other hand, if the lateral deflection of the loaded surface is required there is a small error incurred in ignoring the change in the half-thickness in Fig. 1(b). If we assume that o"z = -tp throughout the thickness and that the strains ex and e y are very small compared with e z in Fig. 1(b), it is easily seen that the amplitude of the change in the half-thickness, tbh, is approximately equal to Ph/4c 3 3. Now it will be found later that the amplitude of the lateral deflection of the plate due to the

Analytical, three-dimensional elasticity solutions to some plate problems

lp

2

Ip

2

(a)

tp

(

Ip

2

b)

451

Fig. 1. (a) An antisymmetrical and (b) a symmetrical lateral loading.

lateral load p according to classical plate theory is w.,( where

Hence

w.,( = (12P/h 3 )[c 11 a 4 + 2(c 12 + 2c 66 )a 2 f32 + C 2 2rr 1

.

!-bh/w.,( = (h4/48c33)[c11a4 + 2(c 12 + 2c66)a2f32 + C 22 f3 4 J

(51)

(52)

and we see that the change in half-thickness due to the symmetrical loading of Fig. 1(b) is

is a length typifying a half-wavelength of the mode. But

we shall also find that the correction to the classical deflection due to transverse shear

deformation is of order (h 2 /

is therefore legitimate to replace the loading by that of Fig. 1(a). [An exception to this

might, however, occur

The general solution for the displacements due to the antisymmetric loading is given by eqns (20) with K; = O. The boundary conditions (50) lead to three simultaneous equations for the Ki cop.stants which are identical with (25) except that the zero on the right-hand side of the first equation is replaced by P/h. The solution of these equations can conveniently be written as

2 )w.,(, and to the accuracy that we shall be concerned with it

of the order (h4/

1.4)w.,(,

where

1.

1.

if C 3 3 is small compared

with c 11 , C 22 or (c 12 + 2C 66 ).J

(}K k cosh !-hqk = (P/h)[a(/;qf - h-qJ) + f3(ellJ -

eiqf) + (eih- -

ej/;)qfqJJ

(53)

1, 2), and () is the determinant of the matrix in

eqn (25). But this determinant is given by the expression on the left-hand side of eqn (29),

and we note also that e= 0, since CT~, CT~ and (J) are all zero. Hence, using eqns (29) and

(34)-(37), we have

where (i, j, k) =

(1, 2,

3) or (2,

3,

1) or (3,

2

12

~= _ h

K

[1 _ h2{(C 11 _ 2C13)a2 + (C 22

C 44

10

C 55

C 33

+

h4 (a 2

120

-

C 4 4

13 2

+ -

C 55

)

a 2 f32[(C12 + 2c 66 )2 -

_

2C 23 )f3 2 }](a 2 '111 + 132'122)

C 33

C 11 C2 2 J + O(h 6 )

(54)

where '111 and '122 are defined by eqns (44) and K by eqn (33). Consider now the lateral deflection of each of the two surfaces z =

shall denote by W. sin ax sin f3y. From the third of eqns (20) with K; = 0 we find that

± !-h, which we

W. =

3 L Kicosh!-hqi'

i= 1

452

Hence, from (53)

W. H. WITTRICK

ow. = (P/h)'L*[(X(hqr - jjqJ) + f3(e J .qJ -

eiqr) + (eijj -

ejh)qrqJJ

where the meaning of the summation 'L* is the same as in eqns (18) and (19). This equation

can be written more concisely as

ow. = (P/h)[ -(XH~II) - H~21) + H~31)]

(55)

where the H~! quantities are as defined in eqns (32). But from eqn (39) we have

H( -1) =

~

[H(2) _ Q H(1) + Q H(O)]/Q

~

1

~

2

~

3

and hence, using Table 1, we find that

H~Il) = H~21) = 0 and H~3l) = -K.

Equation (55) therefore becomes

KP

.- --

he'

(56)

Finally, on substituting eqn (54) and using the binomial theorem we find that

w.

W.I

=

1 + h2[(C l1 _

10

C ss

2C 13 )(X2 + (e 22 _

C33

C44

2C 23 )f3 2

C33

+ (X2f32(~+ ~) {(e l2

C44

c ss

+ 2C66)2 -

ellen}] + O(h4)

«(X2'111 + 13 2 '122)

(57)

where W.I is the deflection according to classical plate theory, defined in eqn (51). Let the lateral deflection of the middle plane be denoted by Wo sin (Xx sin f3y. From the third of eqns (20) with K; = 0 we have

Hence, from eqn (53)

Wo =

3

'L

i= 1

Ki =

3

'L sech!hqi{Kicosh!hqJ

i= 1

=

3 (h2

i~1

1 -

sqr +

)

Kicoshlhqi

h2

3

= W. - -

'L qr Kicoshlhqi + 8 i=1

Wo = w. - ~; 'L* [(Xq~(hqr - h.qJ) + f3q~(eJ.qJ - eiqr) + Q3(e i jj - ejh)] +

After some cyclic modification of the suffixes i, j, k this becomes

Wo -

_

W. - 80 L. [(Xqi qj(jj - h) + I'qi qj(ei - e j ) + Q3(e i jj - ejJi

2

2

a 2

2

Ph ~*

1')]

+

Analytical. three-dimensional elasticity solutions to some plate problems

which is more concisely written as

u?

rro =

W

• -

Ph [

88

ex

H(O) + H(O)

32

31

+

H(O)]

33

+

453

Hence, using Table 1, and eqns (13) and (56), we obtain

Wo = w.[ 1 + h2(C13 iX 8 2 C :

C 23 P2) + O(h 4 )].

(58)

Finally, on substituting eqn (57) for w., we find that

Wo

~I

=

1 + h2[(C 11 _

10

C 55

3C13)ex2 + (C 22 _ 3C 23 )f3 2

4c 33

C 44

4c 33

+ iX2f32(~ + ~) {(e 12 ~2C 66 )2 ~C II C22}] + O(h4).

C44

C5 5

(ex "11

+ f3

"22)

In the special case of an isotropic plate, we have

C II

= C22 = (C 1 2 + 2C66) = 2G/(1

-

v).

Equations (57)-(59) are then much simpler

w w.

cJ

= 1 + (12/5)A + O(A2)

Wo

W.

3v

= 1 + 2(1 _

v) A + O(A 2)

(59)

(60)

(61)

(62)

Wo

-

~I

3(8 - = 1 + 10(1 -

3v) A + O(A2)

v)

(63)

where A is defined in eqn (46).

The difference between Wo and W. is not a trivial one. It is primarily due to the strains

(1 x and (1,. This is easily

demonstrated, as follows. According to classical plate theory the bending strains ex and

are za 2 ~I and zf32 ~I times (sin ax sin f3y), respectively. Hence from eqns (4), assuming that

(1% is negligibly small, we find that

e z arising from the Poisson's ratio effect

of the bending stresses

e y

e% = -Z(C 13 ex 2 + c23f32)(~Jcdsinaxsinf3y

and on integrating this with respect to z between 0 and !h we obtain

Wo -

w. = h2(C 13 ex 2 + C23f3 2 )

8c 33

~I

which agrees with eqn (58) to first order. The magnitude of the difference is most easily appreciated if the plate is isotropic, in which case eqns (61)-·(63) show that, to first order

Wo -

W.

W. - ~I

5v

=--

8(1

-

v)

454

W. H. WITTRICK

This is equal to 0.3125 if v = 1/3 say, so that the change in the half-thickness is by no means negligible compared to the deflection due to shear. Now many higher order plate theories assume that w is constant through the thickness, and if so the question immediately arises as to whether one should compare it with W. or W o , or maybe some average value between the two. There appear to be at least two logical reasons for choosing W. for this purpose. First the work done by the lateral pressure p is directly associated with w., so that if a higher order plate theory agrees with W. it should correctly predict the total strain energy in the plate. Second, the middle plane deflection W o has no obvious physical significance and any experimental verification would inevitably involve measurement of

w An additional, though perhaps minor, advantage of W. is that the first-order correction

to w.:I in eqn (61) is independent of v. To conclude this section we consider the bending strains Bx and By at the two surfaces, which are of the form

{Bx;e y} =

It can be shown that

±{ex;ey} sin ax sin py

atz = ±!h.

(64)

ex =

(P/2(})[a 2 H\Ol- (h 2 /12){a 2 HW + apHW +

aH\l~} + O(h4)]

e y = (P/2(})[P2H~Od - (h 2 /12){apHW + p2HW +

PH~lj} + O(h4)].

On using Table expressions

1, eqns (13) with

e= 0, and eqns (51) and (54) we finally obtain the

e =

x

!hW a 2

1

c

[

2

1 + _

10

h2 {a

2

P '122

(a '111 + p2'122)

(

'122

C 44

_!I!!

C 55

)

+ ~

2 (-

6

C 55

~

7)

C33

+

p2(C 12 + 2C 66 _

6

C 55

t, = !hJ-Y.IP2 [1 + h 2 {

10

(a

7C 23 )} + O(h 4 )]

C 33

2

2

a '111 2

'111 + P '122)

(!I!!_'122) + {32(C 22 _ 7C 23 )

C 5 5

C44

6

C44

C33

+ a 2 (C 1 2 + 2C 66 _ 7C13)} + O(h~)]

6

C 44

C 33

(65)

where '111 and '122 are defined by eqns (44). If the plate is isotropic the equations are much simpler and it is not difficult to obtain higher order approximations by assuming isotropy from the outset. The result for the fourth approximation is as follows

ex

a 2

= fy

p2

= lhW

Z

cI

[1

(2- 7v) A

v)

+ 5(1 _

~A2 _ ~A3

+ 175

875

O(A4)]

+

.

(66)

3. SOLUTIONS USING MINDLIN THEORY AND COMPARISONS WITH THE THREE-DIMENSIONAL ELASTICITY SOLUTIONS

In this section we shall solve the eigenvalue problem and the static loading problem using Mindlin's plate theory[2]. We shall compare the results with the solutions obtained in Sections 2.3 and 2.4, enabling us to draw certain conclusions about the Mindlin theory, and especially about the so-called "effective shear rigidities", that appear to be new. In order to make a proper comparison with the three-dimensional theory, we need a version of Mindlin's plate equations which, in addition to transverse shear deformations,

Analytical, three-dimensional elasticity solutions to some plate problems

also incorporates:

455

(a)

the general orthotropic constitutive equations, eqns (4),

(b)

the effects of rotational inertia, and

(c)

the complete quadratic expressions (3) for the second-order strains e~ and e; when

calculating the change of potential energy of the static, in-plane, compressive stresses CT~

and CT~.

Despite the extensive literature on Mindlin's plate theory the author has been unable to find a version in the published literature that satisfies all these requirements, and for that reason we list the relevant equations in the next section.

3.1. The equations of Mindlin's plate theory for an orthotropic material

The central assumptions of Mindlin's theory are that the lateral displacement w is independent of z, and that points on a normal to the middle plane before bending remain on a straight line, though not a normal to the middle surface, after bending. Thus the assumed displacement field is

u = -z'Px(x,y,t),

v = -z'Py(x,y,t),

w = w(x,y,t)

(67)

where 'Px and 'Pyare rotations of the normal. In relating the stresses CTx and CT y to the strains it is also assumed that the plate is in a state of plane stress, so that CT z = O. This enables e z to be related to ex and e y and then eliminated from the equations for CT x and CT y in eqns (4). The result is

CT x

=

clle x + c 12 e y ,

CT y = c 12 e x

+ c 22 e y

(68)

where C ll , c 12 and C 22 are defined by eqns (38). Also eqns (67) imply that the strains Yxz and Yyz are independent of z; to allow for the fact that this is not strictly correct the actual shear moduli C 44 and C 55 are replaced by empirical effective moduli Ct4 and C~5' Hence the shear stress-strain equations become

'yz = Ct4Y Y Z'

'xz = C~5Yxz'

'xy = C66Yxy'

(69)

The derivation of the governing differential equations and natural boundary conditions from Hamilton's principle now proceeds along standard lines, with the following results. The stress resultants in the plate are given by

-

h 3 {_

12

C ll

h 3 {_

o'Px

ox

o'Px

o'Py}

_

Mx

=

+ C 12 oy

O'Py}

_ My = - 12 C 1 2 oX + C 22 oy

M

_

xy -

h 3 - 12 C 66

{o'Px

oy

+

O'Py}

ox

Qx = hC~5g:- 'Px}

Qy = hCt4 {~;-

'P y} .

(70)

456

W. H. WITTRICK

The equations of motion are

_ OQx _ OQy + h!e(w} = oy

ox

p

h

3

oMx + oMxy _ Qx + -!e('I'J = 0

ox

oy

12

oMy

Ty +

oMxy _ Q + -2('I'y} = 0

ox

y

12

h