Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Int. .I.

Solids Structures,

1972, Vol. 8, pp. 69 to 91. Pergamon

Press. Printed in Great

Britain

VIBRATION AND STABILITY OF CRACKED


RECTANGULAR
PLATES?
B.

STAHL~

Northwestern

and L. M.
University,

KEERD

Evanston,

Illinois

Abstract-The
present paper deals with eigenvalue problems ofcracked rectangular plates. Vibration and buckling
problems are solved for a plate with a crack emanating from one edge and for a plate with a centrally located
internal crack. The problems are formulated
as dual series equations and reduced to homogeneous
Fredholm
integral equations ofthe second kind. The singularity of the solution in each case is isolated and treated analytically.
Numerical results for the natural frequencies and moment distributions
are compared with the work of other
investigators.
Vibration and buckling mode shapes are also illustrated for a cracked plate.

INTRODUCTION
THE objective of the present paper is to develop and demonstrate a method for determining
the natural frequencies and buckling loads of rectangular plates with mixed boundary
conditions arising from cracks. The plates are of uniform thickness and are simply supported
along the four outer edges. Two cracked configurations are considered: a stationary crack
penetrating the plate from one edge, and a centrally located internal crack.
Recently, Keer and Sve [l] have given a rigorous treatment of the symmetric bending of a
cracked rectangular plate due to a static, uniform, transverse load. They formulated the
problem as dual series equations and used a modification ofthe technique by Westmann and
Yang [2] to obtain a solution in terms of a Fredholm integral equation of the second kind.
Three cases were studied in their paper : two collinear external cracks of equal length, an
internal crack centrally located and a single external crack. The dual series equation solution
techniques presented by Keer and Sve are applied here to corresponding vibration and
buckling problems. In addition, solutions to the more difficult problems associated with
antisymmetric bending are given here.
A vibration analysisofa cracked rectangular plate has been made by Lynn and Kumbasar
[3], who used a Greens function approach to obtain a Fredholm integral equation of the
first kind. The subdomain method [4] was employed to satisfy the boundary conditions.
No attempt was made to account for the singularity of the solution. A comparison of their
numerical work with results obtained by the present method is made in this paper.
Numerical treatments of plates with mixed boundary conditions are given by Chen and
Pickett [5,6]. A survey of numerical methods for solving plate bending problems has been
made by Leissa et al. [7]. It was concluded there that discontinuous boundary conditions
are poorly approximated without the use of appropriate singularity functions. The work
in this paper tends to confirm this conclusion.

t Presented at the Third Canadian Congress of Applied Mechanics, the University of Calgary
1 Currently Senior Research Engineer, AMOCO Production
Company, Tulsa, Oklahoma.
$ Professor of Civil Engineering.
69

(1971).

70

B. STAHL and L. M. KEER

A large amount of literature is available concerning


mixed boundary value problems
of infinite plates. Although the present analyses deal with finite plates, the solution of infinite
plate problems provides the correct singularity to be used in the vicinity of the mixed condition. The local behavior at the point of discontinuity
of the boundary conditions must be
the same for both finite and infinite plates. For a plate containing
a crack, the nature of the
singularity may be deduced from the paper by Goodier [8] by considering
the crack as a
limiting case ofan elliptical hole. The complex variable method as illustrated by Sih et ul. [9]
in conjunction
with the paper by Yu [lo] may also be used to deduce the nature of the
singularity. In addition, the Fadle eigenfunction
expansion technique employed by Williams
[ll, 121 may be used. From these methods, it is found that the moment stress resultant is
singular as the inverse square root of the distance from the base of the crack. It will be
shown that these singularities
are indeed properly incorporated.
The present work is restricted to elementary plate theory, with the notation given by
Timoshenko
and Woinowsky-Krieger
[13]. Since the plates are finite, the problems are
formulated by means of Fourier series. The mixed boundary conditions lead to dual series
equations with a weight function. The dual series equations are then converted by means of
certain integral representations
to a homogeneous
Fredholm
integral equation
of the
second kind from which the natural frequencies, the buckling parameter and the auxiliary
function 0(p) are determined. The eigenfunctions
(mode shapes) may be computed by use of
O(P).

PLATE

WITH

CRACK

EMANATING

FROM

ONE

In this section problems involving the vibration


and stability
emanating from one edge will be investigated. The various problems
as internal crack problems with an appropriate
continuation
of
differential equation which governs the free transverse motion w,(x,
in Fig. l(a)? is given by
v4w

+-II a2wo
DF=.

EDGE

of plates with a crack


can also be considered
the plate. The partial
y, t) of a plate as shown

()

Letting
w,(x, y, r) = w(x, y) exp(iQr),
the time dependence

can be eliminated

so as to obtain

v4w+w

= 0.

Utilizing the Levy-Nadai


approach
[13] for plates
edges, substitution
of the partial expansion
w=

(2)

m= 1,2,...

Y,(Y)

(3)
simply

supported

on two opposite

sinbx)

f The coordinates
and dimensions shown on the plate geometry figures are scaled by the factor x/Z, where ii
is the actual plate length. Actual (barredj coordinates
and dimensions are obtained by use of X = 6x/n, j = cly/n,
15= iiblrr,T = 2/n and C, = 2,/n.

71

Vibration and stability of cracked rectangular plates

t--T-

(a)

FIG. 1. Geometry of cracked plates.

into (3) leads to an ordinary

Y,(y) = ;[A,,,

differential

sinh(r,y)

+B,

equation

cosh(r,y)+

for Y, which has as its general solution

C, sinh(r,y)+D,

cosh(r,y)]

(5)

where
rl = [kQ+m2]+,
r2 = [-kO+m]+,
k = [p/D]+.

(6)
I

The buckling and vibration problems are very similar because the governing differential
equations are almost identical and the same boundary conditions are applicable. By applying a uniform compressive load Ton the edges perpendicular
to the crack [see Fig. l(a)],
the governing partial differential equation may be written as

(7)
By use of a partial expansion such as that given by equation
equation can be developed for Y, which has the solution

Y,(Y) = ;[A,

sinh(p,y)

+ B, cosh(p,y)+

C, sinh(p,y)

(4), an ordinary

+ D, cosh(p,y)]

differential

(8)

72

B. STAHL

and L. M.

KEER

where
y = x/2b
Pl = [mpy +

m21f

(9)

p2 = [-mpy+m2]+
PY = JVID).

By comparing equations (S), (9) with (5), (6), it is evident that the buckling problem can be
obtained from the vibration problem by redefining rI and r2 according to (9) and by replacing kR with mpy.
As is evident from Fig. l(a) for all problems considered in this section, the following
boundary conditions hold :
w,

$ =0, lyl

5 b,

w,$=0, lyl = b,

x=0,71

(10)

OIx<7l.

(11)

Boundary conditions (lo), (11) express the fact that the edges x = 0, rc and y = + b are
simply supported. The remaining boundary conditions for each of the problems to be considered are obtained by focusing attention on the line y = 0.
(a) Symmetric vibration of plate with crack from one edge
The boundary conditions appropriate to this case of mode shapes symmetric about
y = 0 are given by (lo), (11) and the following :

a a2w

v, - -

.+(2-v)$
aY [ aY

2
-?"+"ti=(),

Boundary
(14) in 0 s
By use
Boundary

Y = 0,

= 0,

aw
ay= 0,
MY,

ay2 ax2

Y = 0,

Olxlrc

C<Xln:

Y = 0,

OIX<c.

(12)
(13)

(14)

conditions (12), (13) m c < x I 71follow from symmetry conditions and (12),
x < c arise from the stress free condition on the face of the crack.
of the partial expansion given in (4), it is seen that (10) is automatically satisfied.
conditions (11) and (12) lead to the relations
A, = r,[kR+(l-

v)m2]C,,JrI[kR--(1

- v)m2]

B, = -r,[kQ+(l-v)m2]tanh(r,b)C,/r,[kQ-(l-v)m2]
D, = -tanh(r,b)C,.

(15)
(16)

(17)

73

Vibration and stability of cracked rectangular plates

The problem is therefore reduced to the determination of the constants C,. The remaining
boundary conditions (13), (14) are mixed with respect to the slope and moment, and they
are written as the dual series equations
f

P, sin(mx) = 0,

C<XllI

(18)

In= l,Z,...

m[l + F,(fz)]P, sin+)

,=!2

= 0,

0 5 x < c

(19)

1 1..

where P, and F&2) are given by


P, = 2r,kn[k;R-(l-v)m2]-C,,

l+F,(R)

(20)
v)m2]2

= [(l-v)(v+3)mkR]-i
-

(21)

As m --f co, the weight function F&2) approaches zero as mS4. The constant 1 in the bracketed term of (19) serves to isolate the singularity associated with the static problem of a
cracked plate strip infinite in the y-coordinate. This is readily seen to be the case since
F,,,(R) -+ 0 by letting b --f co and then letting rR + 0.
The dual series equations (18), (19) may be reduced to a single integral equation by
representing the unknown coefficients P,,, by a finite Hankel transform
c
P, =

s0

tvWl(mt)

dt,

(22)

where Jl(mt) is the Bessel function of the first kind and first order. By substituting P,
into the first of the dual series equations, interchanging the order of summation and integration, and using the identity
. J,(mt)sin(mx)
, I..

= xt-(t2-x2)-fH(t-x),

,=Fz

(23)

where H(t - x) is the Heaviside function, it is easily seen that (18) is automatically satisfied.
Equation (23) and all of the integral representations of series involving Bessel functions
used throughout this paper may be derived by appropriate contour integrations around the
first quadrant of the complex plane [l].
Integrating (19) once with respect to x and substituting (22) yields, after an interchange
in the order of summation and integration,
c
[l + F,,,(Q)]Jl(mt) cos(mx) dt = 0,
OIX<C.
tcp(t) f
(24)
s0
In= 1,2,...
By utilizing the integral representation
2
m=

tacos

= t--xt-(x2-tZ)-iH(x-t)

1,2,...

(25)

rm

-2

J0

[exp(2rrs)- l]-Ii

cosh(xs) ds,

B.

74

equation

STAHL

L.

and

KEER

M.

(24) is put into the form of Abels integral

equation

x2 - t2)- f dt = h(x),

(26)

where

xh(x)(t2-x)-~ dx.

(27)

I l(ts) in (25) is the modified Bessel function of the first kind and first order. With the help
of certain identities found in [14], the final result becomes
I
Q(P) +

K(p,

s0

where O(p) = cp(cp). The kernel of the integral


K(p,

r) = c2r

0,

r)O(r) dr =

OlpSl,

equation

(28)

is written

as

2
F,(Cl)mJ,(mcr)J,(mpc)
i m= l,Z,...
cc

s[exp(27ts)-

+2

l]- ll,(rsc)l,(psc)

ds

s0

(29)
I

It should be remarked
that (24) should include an arbitrary constant of integration
as a
result of integrating (19) with respect to x. In the process of solving Abels integral equation,
however, the arbitrary constant has no effect and may be ignored.
The singularity of M,, at the crack tip may be verified by considering the second dual
equation (22)
series equation (19), which represents M,, on y = 0, and by substituting
written in the form

s
c

Pm =

-m-ccp(c)J,(mc)+m-

It may be shown that in the vicinity

[w(t)1
dt.

JoW&

(30)

of the crack tip,


(31)

and thus the moment is square root singular near the crack tip which is in complete agreement with the references cited in the Introduction.
For the purpose of calculating the moment distribution
along the untracked
segment,
it is convenient to write equation (19) in the form

Myyly=oBy substituting
may be written

-& m-_f

Pmcos(mx) +

(22) into (32) and using the integral


as

mF,(R)P,

sin(mx).

(32)

of (25), equation

(32)

m= 1,2,...

l.Z....

representation

M,,(x, 0) - c2

K(x,

P)%P) dp +

f
ln= 1,2,...

mF,(Q)P,

sin(mx), x > c,

(33)

Vibration and stability of cracked rectangular plates

75

where
K(x, p) = -cp2[X2-&?-f
+2P

,,a [exp(2zs) - l] - I ,(cps)s sinh(xs) ds

and

(34)

Pm = c2

P~-YP)J,(~cP)
dp.

(35)

The moment field at points other than y = 0 can also be calculated but with a little more
difficulty. An expression for M,, on y > 0 can be written in the form

M!JY-

m[exp( - my) + F,(Q, y)]P, sin(mx).


f
RI=1,2,...

(36)

For large y, the above series can be summed as it stands. For y << 6, the series converges
too slowly because of the moment singularity. Therefore it is necessary to isolate the series
corresponding
to the first term in the brackets and to substitute appropriate
integral representations.
A procedure
for integral equations
analogous
to the one required here for
series equations
was used by Sve and Keer [15] in calculating
punch isochromatics.
vibration qf plate with crack from one edge

(b) Antisymmetric

The shape function is given by (4) and the coordinate


system is shown in Fig. l(a).
Boundary conditions
(lo), (11) again hold and the boundary conditions
on the line y = 0
are given by

_ti++yzzw=o
= 0,

ax2

MY, ay2

w2-a2w_o
ax

v,m-

OIxSn

ax2

a azw
,+(2-&J
ay [ay

Y =

0,

= 0,

c<x17c

Y = 0,

OIx<c.

It should be noted that the above boundary


conditions
are also applicable
which is partly free and partly simply supported along the line y = 0.
By applying (11) and (37), one is led to the relations
A, = - coth(r,b)B,

v)m2]B,/[ - kQ+(l-

- kfl+( 1 - u)m]

for a plate

(41)
v)m2].

The mixed boundary conditions


(38) and (39) lead to dual series equations
and (19), but where P, and F,(R) are given by
F,,,= 2kRm2B,/[

(39)

(40)

C, = coth(r,b)B,[kn+(l-v)m2]/[-kfi+(l-v)m2]
D, = - [kQ+(l-

(37)

(42)
given by (18)

(43)

B. STAHL and L. M. KEER

16

and
l+F,,,(R) = [m3kR(v+3)(v-1)]-1{r,coth(r,b)[kS1-(l-v)m2]2
- r2 coth(r,b) [/&I+ (1 - v)m]}.

(44)

The first of the dual series equations for this case represents the condition a2w/~x2 = 0 on
y = 0 and c < x i n, which facilitates the introduction of the appropriate singularity at
the tip of the crack, but is not sufficient to insure that w = awlax = 0 on y = 0, c < x I /t.
Hence, the auxiliary condition that aw/ax = 0 on y = 0 and x = rr is imposed.
The appropriate representation for P, which results in the proper singularity and
at the same time permits the auxiliary condition to be satisfied, is
E
m = 1,2,...
P, = E.Z,(mc)+
(45)
tdt)J,(mt)
dt,
s0
The constant E is determined by imposing
$7c, 0) =

i:

m= 1,2,...

m-P,

cos m7c = 0.

(46)

which leads to
c

E=

J
0

tdt)k(t)

dt,

(47)

where
k(t) = -

me .Zl(mt)( - 1)

fn=1,2,...

m- .Zl(mc)( - 1).

f
i m=1,2

,...

(48)

It can be shown that


m-lJl(mt)(-l)m
f
m=1,2,...

OIt<n.

= -at,

(49)

Hence,
k(t) = - tJc,

(50)

and substituting E into equation (45) results in

J [
E

P, =

tcp(t) .Z,(mt)-iJ,(mc)

dt,

m = 1,2,...

(51)

It is easily verified that the first of the dual series equations is automatically satisfied. By
the procedure indicated in the previous case, the second of the dual series equations leads to
(28) where
K(p, I) = c2r
+2

m
J
m=1,2,...

F,(Q)[Jl(mcr)

rJl(mc)lmJI(w4

[exp(2rrs)- l]- [Z,(scr)-rZ,(sc)]sZ,(scp)

ds

(52)

17

Vibration and stability of cracked rectangular plates

The singularity of M,, at the root of the crack may be verified in a manner analogous
to the symmetric case. It is easily shown that

Mxxly=O
-

P, sin(mx).

if

(53)

m=l,Z,...

Substituting (45) into (53) results in


c
Mxxly=~ -E

J,(mc) sin(mx)+

s0

In=1,2,...

tcp(t)

.Z,(mt) sin(mx) dt.

(54)

nt= 1,2,...

The first term in (54) results in


w E(2c)-%++O(E+)
MX_X

(55)

in the vicinity of the root of the crack along y = 0.


(c) Symmetric buckling of plate with crack from one edge and uniform load on edges
perpendicular to crack

It is evident that the buckling problem can be obtained from the vibration problem by
redefinition of rl and r2 according to equations (9) and by replacing kR with mpy. Thus the
values of A,, B, and D, as given by (15)-(17) hold subject to this redefinition. The governing
dual series equations may be immediately written as [see equations (18)-(21)]
f

P, sin(mx) = 0,

c<x<n:

(56)

fn=1,2,...
,j2

,,, 41

>9

0 I x < c,

+~,(dIP, sin(mx)= 0,

(57)

where
P, = 2p,mpy[mpy - (1 - v)m2]- i C,,

l+F,(p)

tanh(p

= [(l-v)(v+3)m2py]-
i

p1

(58)

1b)
[wv+U - 4m212

- tanW2b)
p2 [mm-0 - M212

(59)

Solving the above dual series equations in the same manner as the vibration case results in
(28) where the kernel is given by
K(p, r) = c2r

F
Up)mJl(mcr)Jl(mcp)
i In=1,2,...
m

+2

s[exp(2ns)- l]-Z,(rsc)Z,(psc)
s 0

ds

(60)
I

The difficulty with this result is that as m -+ co, F,,,(p) + 0 as m-2 and hence the series in
(60) converges slowly. Since the series must be evaluated for each trial value of the parameter
p, it seems desirable to improve the convergence. By a procedure similar to that used by

78

B. STAHL and L. M. KEER

Sih and Loeber


series equations

[16] in a problem involving


may be written as

a penny-shaped

crack, the second of the dual

(61)

OIx<c,

+ G,(p)]P,,, sin(mx) = 0,
where
h(v) =

5v2+2v+l

(62)

8( 1 - V)(3 + V)

and

y2p2h(v)

G,(P)= ~

m2

~I&).

(63)

It can be shown that as m --+ co, G,(p) --f 0 as rnm4. It seems reasonable
in the form
K(p, r) = c2r

f
G,(p)J,(mcr)mJ,(mcp)-y2p2h(v)
i nl= 1,2,...

to write the kernel

[m2 + 11-l

m= 1,2,...

(64)
x J,(mcr)mJ,(mcp)

+2

m [exp(2zs)-

l]-Z,(crs)sZ,(psc)

s0

ds
I

The second series in (64) coverges slowly. However, it need be calculated only once for
each crack length and therefore (64) provides a relatively efficient means of computing the
critical load parameter.
By means of a contour integration
of
exp(izw)J,(tw)

cos(xw)

(w2 + 1) sin(7rw)
around

the first quadrant,

,j21 I,,,[m+

the following

can be derived :

identity

l]- .Z,(mt) cos(mx) = 1:


+2

J1(tu~~~)du

m[exp(2nv)-

l]- Z,(W) cosh(xu) dv


.

v2-

(65)

By use of (65) and with the help of an identity found in [14, p. 6941, the kernel of the integral
equation may be written in the alternative form
K(p, r) = c2r

f
G,(dJ
i fn=l,Z,...

,(mcr)mJ l(mcd

+2

[exp(2xs)-

l]- I,(crs)sl,(pcs)

ds

- y2p2h(v)

+ 2y2p2h(v)

Jl(cr)Kl(cp),

p2r

Kl(crVl(cp),

r2p

om[exp(2ns)-

l]- [s2-

l]-Z,(crs)sZ,(cps)

ds .

Vibration

and stability

of cracked

rectangular

79

plates

The above kernel requires the evaluation of a principal value integral. Since there is no
apparent advantage in its use, equation (64) was used for the numerical calculations.

PLATE

WITH

INTERNAL

CRACK

In this section problems involving the vibration and stability of plates with an internal
crack will be considered, the geometry of which is given in Fig. l(b). Only the cases of vibration and buckling symmetric with respect to the y-axis will be considered since the cases
that are antisymmetric
about the y-axis have essentially been solved in the preceding section
if the plate dimension is reduced by half.
If the plate is simply supported on all four sides, the corresponding
boundary conditions
are expressed by
Z2W
W,,

w,
In view of the above
pansion

7171

ox

0,

IYI I b,

0,

lyl =b,

azw

ay

boundary

conditions,

w=

(66)

x = -22
+x$.

(67)

the shape function

is taken

Y,(y) cos(mx),
f
??I=1.3,...

as the partial

ex-

(68)

where Y,(y) is given by (5).


(a) Symmetric-symmetric

vibration of plate with internal crack

The first case to be considered is that in which the mode shapes are symmetric about
both coordinate axes. Considering the lower right-hand quadrant of the plate, the additional
boundary conditions resulting from the presence of the crack may be expressed by

a azw
v, - -

y+(2-v)$1 =0,

Y =

ay[ ay

aw

ay=0,
M,,

a$

ax2

Y = 0,

(69)

c<x<F

= 0,

_12w+!5Lo

OSX<;

0,

OIx<c.

(71)

By using equations
(67) and (69) three relations may be found between the unknowns
A,, B,, C, and D, and they are those given by (lSH17). The mixed boundary conditions
(70), (71) lead to the dual series equations

If! Pmcos(mx)

= 0,

C<X2

m= 1,3,...

j3 I 1,,,m[l +F,(Q)IP,cos(mx)
where P, and F,,,(R) are given by equations

= 0,

2
0 5 x < c

(20) and (21), respectively.

(72)

(73)

80

B. STAHL and L. M. KEER

The representation
crack tip, is given by

for P,,,, which introduces a square root moment singularity at the

P, =

E
J
0

Gil

(74)

dt,

where Jo(mt) is the Bessel function of the first kind and zeroth order. It can be easily verified
that (74) satisfies the first of the dual series equations. Integrating the second of the equations
once with respect to x between the limits of 0 and x, and substituting (74) yields, after an
interchange in the order of summation and integration,

Eq(t)
J 2
0

o<x<c.

[l -t F,,,(R)]Jo(mt) sin(mx) dt = 0,

(75)

nl= i,3,...

By use of an integral representation for the first term in (75), it can be put into the form of
Abels integral equation. With the help of identities found in [143, the final result becomes
(28). where
K(p, r) = 2c2p

?
1 In= 1,3,...

~,(fW0(mcrW0(mcp)

Om[exp(rrs)+

l]- s~~(rsc)~~(~sc)

(74)

(b) Symmetric-antisymmetric vibration of plate with internal crack


For this case, the boundary conditions which are applicable, in addition to those given
by (66) and (67), are
Y = 0,
MY,_c!TfViiZM=o,
ay2 ax2
a~

azw

ax

ax2

w=_=_--_o

f$--

Y=

a a2w
JY

i+(2-Y)g =0,

[ ay

0,

o-ix<;

(77)

C<X<E
2

y = 0,

o<x<c.

(79)

The above boundary conditions are derived from considerations of the antisymmetry of the
deflection function in the y-coordinate and the stress free state along the face of the crack.
The shape function is given by (68) and satisfies boundary conditions (66). Upon application of the boundary conditions (67) and (77), the three relations between the unknowns
A,, B,, C, and D, are found to be given by equations (40)-(42). The mixed boundary
conditions above lead to the dual series equations
P, sin(mx) = 0,
?
I!!=1,3....

ctx<n

m2[l + F,(Q)]P, cos(mx) = 0,


23....

(80)

2
OlX<C

(81)

81

Vibration and stability of cracked rectangular plates

where
P, = 2mmB,/[-&+(l-v)rr?]

(82)

and where F,(Q) is defined by (44). The first of the dual series equations above arises from
aw/ax = 0 on y = 0, c < x < 71/2.This condition is used rather than the other two in (78)
because it permits the appropriate singularity at the crack tip to be introduced. However, all
conditions of (78) are automatically satisfied since the shape function equals zero at x = 7c/2.
The solution to the above dual series equations proceeds by taking the finite Hankel
transform of (22) as the appropriate representation for P,,,. The first of the dual series equations will be automatically satisfied. The second of the dual series equations is integrated
twice with respect to x. The first integration takes place between the limits of zero and x.
A constant of integration which appears with the second integration may be ignored for
reasons given earlier. With an appropriate contour integration and by use ofcertain identities
foundin [14] thesecondofthedual
seriesequationsfinallyleadstoahomogeneousFredholm
integral equation of the second kind as given by (28). The kernel of this integral equation is
written as
K(p, r) = 2c2r

f
~,(NJ,(~cr)mJ,(~cP)
i m=1,3,...

-s
x

[exp@s) + l]- I,(rsc)sl,(psc)

(83)

(c) Symmetric-symmetric buckling of plate with internal crack and uniform load on edges
perpendicular to crack

The analogy between buckling and vibration has been pointed out earlier. The dual
series equations may be written down immediately as
Pmcos(mx) = 0,

m~+1+ %(P)IP, co44

c<x<;

= 0,

OlX<C

where h(v) and G,(p) are defined by (62) and (63). The solution to the dual series equations
may also be written by inspection. A homogeneous Fredholm integral equation of the
second kind is obtained as in (28) and the kernel is written as
K(P, r) = 2c2p

f
G,(p)Jo(mcr)mJo(mcp)
!?I=1,3,...

- y2p2h(v)

Om
s

(m2 + l)- JO(mcr)mJo(mcp)


f
In=1,3,...

[exp(rcs) + l]- sZ,(rsc)Z,(psc) ds

82

B. STAHL and L. M. KEER

NUMERICAL

PROCEDURE

AND RESULTS

Both the vibration and buckling problems have been reduced to the solution ofa homogeneous Fredholm integral equation of the second kind. The kernel of the integral equation
contains the unknown parameter being sought, e.g. the frequency of free vibration. Because
of the complexity of the kernel, the integral equation is treated numerically
by reducing
it to a system of homogeneous
algebraic equations using Simpsons rule. In matrix notation, the integral equation may be written as
[Z] (0) +
where [I] is the unit matrix
Defining

a nontrivial
determinant

[R&q] {H} = 0

and K(0) is the discretized

kernel

of the integral

equation.

solution of (88) is found by searching for that value of 51 which causes the
of the matrix A to vanish. The frequency factor,fis related to the frequency Q

by
I& =

.f(n/iq2(1 + y2).

In all calculations
the value of Poissons ratio was 0.3 (V = 0.3).
The search procedure for finding the frequency is a trial and error process. First, a
crack length is specified and the matrix A of(88) is calculated for several values offrequency.
An estimate of the trial frequencies can be obtained by calculating
the upper and lower
bounds, i.e. the untracked
and completely cracked cases, by use of standard techniques.
The determinant
of the matrix A corresponding
to each value of the frequency is then computed. If the estimates of the trial frequencies are reasonably good, a change in sign of the
determinants
would be observed and an approximate
value of the frequency can be obtained by locating the crossing. Further trials can be used to pinpoint the value of the
frequency.
A very efficient technique for obtaining the frequency is Mullers iterative method [17],
which is a quadratic interpolation
scheme for finding the roots. It has been used for all of
the analyses presented here. Usually five or six evaluations of the matrix A were found to be
sufficient for an accurate determination
of the frequency or buckling factors.
In the calculation of the matrix A of (88) the infinite series of the kernel were evaluated
to a relative error criterion of l/5000, i.e. the series evaluation
was terminated
when the
ratio of the absolute value of the last term calculated to the absolute value of the sum of all
previous terms became less than l/5000. Increasing the accuracy of the series evaluations
did not improve the final results. The integrand of the infinite integral in the kernel is a
monotonically
increasing function up to some maximum. After the maximum is reached,
the integrand decays exponentially.
Simpsons rule was used in the evaluation of the infinite
integrals and a sufficient number of intervals were used to ensure accurate results.
Another factor which affects the accuracy of the solution is the order of the matrix A.
The numerical results for the crack cases are tabulated in Tables l-8 and were obtained
from eleventh-order
matrices. For some crack lengths the fourth significant figure of the
frequency factor for the higher modes may not be accurate.

Vibration

TABLE

I. FREQUENCY

Mode

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
r Numbers

and stability

of cracked

rectangular

83

plates

FACTORS AND EIGENVALLJE RATIOS FOR PLATE WITH CRACK


VIBRATION MODES SYMMETRIC IN y. y = 2.0

Mode 2

Mode 3

i/n,

,?/A,

,qiL,

1.OOO(1)i

1.000
1.000
0.996
0.982
0.950
0.904
0.856
0.813
0.778
0.755
0.747

1.600(2)
1.598
1.578
1.528
1.485
1.475
1.471
1.439
1.382
1.329
1.308(2)

1.000
0.999
0.993
0.977
0.964
0.960
0.959
0.948
0.929
0.911
0.904

2.600(3)
2.596
2.565
2.527
2.522
2.501
2.459
2.449
2.436
2.395
2.368(3)

1.000
0.999
0.993
0.986
0.985
0.98 1
0.972
0.971
0.968
0.960
0.954

1,000
0.992
0.964
0.902
0.818
0,733
0.661
0.606
0.570
0.558(l)
in parentheses

indicate

the number

of semiwaves

in the x coordinate

TARI.C 2. FREQUENCY FACTORSAND EIGENVALUE RATIOS FOR PLATE WITH CRACK


VIBRATION MODES ANTISYMMETRIC IN .v.y = 2.0

Mode 1
J

I.jn,

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0

3.400( 1)t
3.400
3.386
2.917
2.028
1.492
1.165
0.951
0.801
0.692
0.558(l)

1.000
1.000
0.998
0.926
0.772
0.662
0.585
0.529
0.485
0.451
0.405

4.000(2)
3.991
3.932
3.545
3.521
3.416
2.870
2.319
1.915
1.620
1.308(2)

in parentheses

FROM ONE EDGE.

Mode 2

c/n

t Numbers

FROM ONE EDGE

indicate

the number

Mode 3

.iJi,
1.000
0.999
0.991
0.941
0.938
0.924
0.847
0.761
0.692
0.636
0.572

of semiwaves

in the x coordinate.

TABLE 3. BUCKLING FACTOR vs. CRACK


LENGTH FOR PLATE WITH CRACK FROM ONE
EDGE. 7 = 2.0

0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.oo
t Numbers in parentheses
of semiwaves in x coordinate.

2.000(2)7

1.998
1.964
1.861
1.744
1.652
1.583
I.528
1.473
1.428
1.394( 1)
indicate

number

J
5.000(3)
4.997
4.820
4.416
4.385
4.015
3.852
3.790
3.371
2.877
2.368(3)

i/i.,
1.ooo
1.000
0.982
0.940
0.937
0.896
0.878
0.871
0.821
0.759
0.688

B. STAHL and L. M. KEER

TABLE 4. FREQUENCY

FACTOR vs. CRACK

LENGTH FOR PLATE WITH INTERNAL CRACK.

VIBRATION MODES SYMMETRIC IN x AND SYMMETRIC IN y. y = I.0

Mode l(a)

Mode l(b)

Mode 3(a)

c1r.t

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50

l.OOO(l)f
0.994
0.978
0.954
0.926
0.897
0.871
0.848
0.831
0.821
0.817(l)

t Numbers

in parentheses

indicate

5.000(l)
4.934
4.760
4.540
4.332
4.162
4.032
3.936
3.869
3.828
3.814(l)
number

of semiwaves

5.000(3)
5.000
5.000
4.999
4.999
4.985
4.964
4.935
4.902
4.876
4.865(3)
in x coordinate

TABLE 5. FREQUENCY FACTOR vs. CRACK LENGTH FOR PLATE WITH INTERNAL CRACK.
VIBRATION MODES ANTISYMMETRIC IN x AND SYMMETRIC IN y. ;I= I.0

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
t Numbers

Mode 2(a)

Mode 2(b)

Mode 4

6.500(2)

2.500(2)?
2.500
2.499
2.494
2.484
2.467
2.443
2.417
2.394
2.377
2.368(2)
in parentheses

indicate

6.499
6.490
6.452
6.366
6.228
6.060
5.896
5.762
5.675
5625(2)
number

of semiwaves

8.500(4)
8.499
8.492
8.470
8.444
8.429
8.427
8.420
8.395
8.363
8.348(4)
in x coordinate.

TABLE 6. FREQUENCY FACTOR vs. CRACK LENGTH FOR PLATE WITH INTERNAL CRACK.
VIBRATION MODES SYMMETRIC IN x AND ANTISYMMETRIC IN y. y = I.0

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.49
0.50
t Numbers

Mode l(c)

Mode 3(b)

Mode l(d)

2.500(l)+
2499
2.491
2.456
2.362
2.180
1.924
1.652
1.407
1.192
1.007
0.817(l)

6.500(3)
6.498
6.463
6.288
5.656
4.827
4.326
4.085
3.971
3.909

8.500(l)
8.494
8.392
7-890
7.143
6.842
6.672
6.467
6.149
5.700
5.235
4.865(3)

in parentheses

indicate

3.814(l)
number

of semiwaves

in x coordinate

Vibration

and stability

of cracked

rectangular

plates

85

TABLE I. FREQUENCY FACTOR vs. CRACK


LENGTH FOR PLATE WITH INTERNAL CRACK.
VIBRATION MODES ANTISYMMETRIC IN x AND
ANTISYMMETRIC IN y. 7 = 1.0

cln

Mode 2(c)
/

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50

4.000(2yf
4.000
4.000
3.997
3.982
3.938
3.829
3626
3.330
2.983
2.368(2)

t Numbers in parentheses
of semiwaves in x coordinate.

indicate

number

TABLE 8. BUCKLING FACTOR vs. CRACK


LENGTH FOR PLATE WITH INTERNAL CRACK.
y = 1.0

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
t Numbers in parentheses
of semiwaves in x coordinate.

2.00ql)t
1.989
1.956
1.907
1.850
1.792
1.739
1.695
1.662
1.642
1.635(l)
indicate

number

The frequency factors for the plate with a crack from one edge are tabulated in Tables 1
and 2 as a function of crack length. The modes are numbered according to the number of
semiwaves in the x-coordinate. In addition to the frequency factors, eigenvalue ratios are
tabulated in order to facilitate a comparison with the figures given in Ref. [3]. The eigenvalue
1 is related to the frequency factor f by
*

=/(;)2[l+y21[;]+
= [p$].

(901

The value 1, for each mode equals ;1evaluated at c/z = 0.


The solid lines on Figs. 2 and 3 represent the above tabular data in graphical form. The
dashed lines are taken from Ref. [3]. The maximum difference between the solid and dashed
line in mode 1 of the antisymmetric modes is about 11 per cent. Considering that the method

86

B. STAHL

and L. M.

KEER

Vibration

and stability

of cracked

rectangular

plates

87

of Ref. [3] does not account for the singularity of the solution, the agreement between the
two methods with regard to the global quantities such as frequency is seen to be quite
reasonable.
For the symmetric vibration modes of the plate with a crack from one edge, relative
moment distributions were also calculated according to (33) and they are shown by the
solid lines on Figs. 4-6. Superimposed by the dashed lines are the moment distributions
taken from Ref. [3]. It is seen that good agreement exists away from the crack tip. Close to the
crack tip, however, the moment distributions differ quite drastically. This indicates that
even though global quantities such as frequency can be computed reasonably well by methods
which do not account for the singularity, it is extremely difficult to accurately compute
point-wise quantities such as the moment distribution in the vicinity of the point where the
boundary conditions are discontinuous. Finally, the contours of constant deflection for the
first three symmetric vibration modes are illustrated in Fig. 7.
The buckling factors for the plate with a crack from one edge are tabulated in Table 3.
Figure 8 illustrates the buckling mode shapes for three different crack lengths. It is interesting

1.0

=:
;

0.5

0
z
3
P

0.0

w
>
5
i

-0.5

-I .o

0.0
FIG. 4. Moment

distribution

along untracked
segment c, for plate with crack from one edge. Vibration
modes symmetric in y. c1 = 0.37~ y = 2.0.

0.5

MODE

0.0

- I.0

c,

l.OC,

FIG. 5. Moment distribution


atong untracked
segment c, for plate with crack
from one edge. Vibration modes symmetric in y. c1 = 0.5x, y = 2.0.

0.0

-1.0

0.5c,
FIG. 6. Moment distribution
along untracked
segment c1 for plate with crack
from one edge. Vibration modes symmetric in y. c, = 0477c,y = 2.0.

5
ii -0.5

5
d

-0.5

0.0

0.5

1.0

0.0

if
3
P

)r

ii
3

0.5

1.0

Vibration and stability of cracked rectangular plates

89

FIG. 7. Contours of constant deflection for symmetric vibration of plate with crack from one edge
c = 0.5~ y = 2.0. Upper: mode 1; middle : mode 2; lower : mode 3.

to observe that when the crack length is small, the buckled shape has two semiwaves in the
x-coordinates. As the crack length increases, the buckled shape goes into one semiwave.
Tables 4-7 summarize the numerical results for the vibration modes of the plate with
an internal crack. The modes are numbered according to the number of semiwaves that
occur in the x-coordinate when the plate is untracked. The letters behind the numbers
differentiate modes having different number of semiwaves in the y-coordinate. The buckling
factors for the plate with an internal crack are tabulated in Tabie 8.

90

B. STAHL and

FIG. 8. Contours

L. M. KEER

of constant deflection for buckling mode of plate with crack from one edge. Upper:
c = O.ln, : = 2.0; middle: c = 0.5~, y = 2.0; lower : c = 0.9~. y = 2.0.

Acknowledgements-The
first author is grateful to the National
for financial support during his tenure as a graduate student.

Science Foundation

and Northwestern

University

REFERENCES
[I] L. M. KEER and C. SVE, On the bending

of cracked plates. Int. J. Solids Sfruct. 6, 1545-1559 (1970).


[2] R. A. WESTMANNand W. H. YANG, Stress analysis of cracked rectangular
beams. J. nppl. Mech. 34, 693- 701
( 1967).
[3] P. P. LYNN and N. KUMBASAR, Free Vibrations of Thin Rectangular
Plates Having Narrow Cracks with
Simply Supported Edges, in Developments in Mechanics, Vol. 4, Proc. Tenth Midwestern
Mechanics Conference, Colorado State University, Fort Collins. Colorado, August 21-23. pp. 91 I-928 (1967).

Vibration

and stability

of cracked

rectangular

plates

91

(1956).
141 S. H. CRANDALL, Engineering Analysis. McGraw-Hill
[51 S. S-H. CHEN and G. PICKETT, Bending of plates ofany shape and with any variation

in boundary conditions.
J. appl. Mech. 34,217-218 (1967).
161 S. S-H. CHEN and G. PICKETT, Bending of Uniform Plates of Arbitrary Shapes and with Mixed Boundary
Conditions, in Developments in Mechanics. Vol. 4, Proc. Tenth Midwestern Mechanics Conference. Colorado
State University, Fort Collins, Colorado, August 21-23, pp. 411427 (1967).
[71 A. W. LEISSA, W. E. CLAUSEN, L. E. HULBERT and A. T. HOPPER, A comparison of approximate methods for
the solution of plate bending problems. AIAA Jnl. 7, 920-928 (1967).
PI J. N. GOODIER, The influence of circular and elliptical holes on the transverse flexureof elasticplates.Phil.
Mag. 22, 69-80 (1936).
factors for plane extension and plate
c91 G. C. SIH, P. C. PARIS and F. ERDOGAN, Crack-tip, stress-intensity
bending problems. J. appl. Mech. 29, 306-3 12 (1962).
[lOI Y. Y. Yu, The Influence of a Small Hole or Rigid Inclusion on the Transverse Flexure of Thin Plates, Proc.
Second U.S. National Congress of Applied Mechanics, pp. 38 l-387 (1954).
[I 11 M. L. WILLIAMS, The bending stress distribution at the base of a stationary crack. J. appl. Mech. 28, 78-82
(1961).
Resulting from Various Boundary
Conditions
in Angular
Cl21 M. L. WILLIAMS, Surface Stress Singularities
Corners of Plates Under Bending, U.S. National Congress of Applied Mechanics, Illinois Institute of Technology, Chicago, Illinois, pp. 325-329 (1952).
S. TIMOSHENKOand S. WOINOWSKY-KRIEGER. Theorv of Plates and Shells. 2nd edition. McGraw-Hill
(1959).
I. S. GRADSHTEYN and 1. M. RYZHIC, Tables qf Inteirak Series and Products, 4th edition, p. 427. Academic
Press (1965).
C. SVE and L. M. KEER, Indentations
of an elastic layer by moving punches. Int. J. Solids Struck 5.795-816
(1969).
G. C. SIH and J. F. LOEBER, Torsional
vibration of an elastic solidcontaining a penny-shaped crack. J.
acoust. Sot. Am. 44, 1237-1245 (1968).
S. D. CONTE, Elementary Numerical Analysis. McGraw-Hill
(1965).

(Received 11 December 1970; revised 3 June 1971)


A6CTpaKT-Pa6Ora

pacchIaTpwaaeT
aonpoc
3anas Ha co6crBeHHble 3HaqeHHa, Kacamuwxm
npnMoyro,,bHblX "JlaCTMHOK c TpeLI,MHaMM. PeLuaIo-rcs
33naqM KOfle6aHMA M ~CTO~~YMBO~TM,LUIR nnaCTMHKM
c TpeqHHoc7, pacnpocTpatiflmmeficfl
OT Kpan M .wln nnaCTMHKM
c KOHUeHTpMYeCKM
pacnonoxeHoR
BKyTpeHHefi ~PelWHOfi. 3aL,aW +OpMyflMpyWTCa
a @OpMe
ypaBHeHMti B IIapHblX pflL,nXCT CBOL,flTCR
K O,.,HO,,OflHblMMHTerpanbHblM
YPaBHeHMflM
4peUOJTbMa
~~oporo
pona. Ann
Kamfloro cnyqan
Bbl~eJ,S4eTC,3CMHryJ,,3pHOCTbpe~eHMIIM 3aTeM OnpeIVUReTCfl aHWIWWWeCKM.CpaBHMBa~~Cfl
',,,C,,~HHbIe
pe3yJIbTaTbI LUIR CO6C~BeHHblX '(BCTOT I, PaCllpeLleneHMe MOMeHTOB,
C pa6oToR
,QpyrMX l4CCJleJlOBaTenet+. MJ,,,lOCTpMpylOTCSl@OpMblKOne6aHMfi
M BblllyqMBaHl4fl
IWR ~TIIaCTMHKMC TpelllMHOii.

You might also like