ADVANCED SOLID MECHANICS

Also from Macmillan

Engineering Thermodynamics
Theory, worked examples and problems
G. Boxer

Essential Solid Mechanics

Theory, worked examples and problems
B.YJ. Young

Fluid Mechanics
Theory, worked examples and problems
H.C. Lowe

Introduction to Engineering Fluid Mechanics

Second Edition
J. Fox

Principles of Engineering Thermodynamics

E.M. Goodger
ADVANCED SOLID MECHANICS
Theory, worked examples and problems

P.R. Lancaster
D. Mitchell
Department of Mechanics/ Engineering,
University of Bradford

M
) P.R. Lancaster and D. Mitchell 1980

All rights reserved. No part of this publication may be reproduced

or transmitted, in any form or by any means, without permission.

First published 1980 by

THE MACMILLAN PRESS LTD
LOndon and Basingstoke
Associated companies in Delhi Dublin
Hong Kong Johannesburg Lagos Melbourne
New York Singapore and Tbkyo

ISBN 978-0-333-24013-7 ISBN 978-1-349-03906-7 (eBook)

DOI 10.1007/978-1-349-03906-7

This book is sold subject to the standard conditions of the Net

Book Agreement.

The paperback edition of this book is sold subject to the condition

that it shall not, by way of trade or otherwise, be lent, resold,
hired out, or otherwise circulated without the publisher's prior
consent in any form of binding or cover other than that in which it
is published and without a similar condition including this con-
dition being imposed on the subsequent purchaser.
CONTENTS

Preface vii
1 EQUATIONS OF ELASTICITY

1.1 Equilibrium Equations 1

1.2 Hooke's Law - Stress Strain Relationships for
Linear Elastic Materials 3
1.3 Equilibrium Equations for Two Dimensions 4
1.4 Plane Stress and Strain 5
1.5 Polar Coordinates 5
1.6 Stress at a Point 7
1.7 Principal Stresses 11
1.8 Stationary Values of Shear Stress and
Associated Planes 13
1.9 Strains and Displacements 18
1.10 Summary of Plane Stress Equations 27
1.11 Transformation Equations 28
1.12 Problems 30

2 STRESS FUNCTIONS
2.1 Stress Function 32
2.2 Cantilever Plate 35
2.3 Polar Coordinates 46
2.4 Solid of Revolution with Axially Symmetrical
Loading 49
2 .5 Curved Bars 53
2.6 Hole in Plate Subjected to Pure Tension 56
2.7 Concentrated Forces 62
2.8 Problems 71

3 TORSION
3.1 Displacements, Strains and Stresses 74
3.2 The Equilibrium Equations 76
3.3 The Boundary Conditions 76
3.4 Analogies to the Torsion Problem 82
3.5 Torsion of Thin Walled Sections 85
 3.6 Stress Concentration in Thin Walled Sections 90
3.7 Problems 94

4 ENERGY METHODS
4.1 Potential Energy 96
4.2 Virtual Work 115
4.3 Castigliano's Second Theorem and the Principle
of Virtual Work 118
4.4 Problems 120

5 NUMERICAL METHODS
5.1 Finite Difference Analysis 121
5.2 Finite Element Method 142
5.3 Problems 169

6 PROBLEMS IN BENDING OF BARS AND PLATES

6.1 Beams on Elastic Foundations 172
6.2 Bending of Curved Bars 183
6.3 Bending of Thin Circular Plates 198
6.4 Problems 208

7 DEFORMATION BEYOND THE ELASTIC LIMIT

7.1 Yield Criteria 212
7.2 Thick Walled Cylinders 217
7.3 Residual Stresses 221
7.4 Rotating Discs 223
7.5 Bending of Beams 227
7.6 Problems 235

References 237

Index 23S
PREFACE

This book is intended for use by more advanced students studying

mechanics of materials at first degree level. Although the various
topics are treated in such a way that they are complete and stand
by themselves, it is assumed that the student has already some
knowledge of the subject, as might be gained from following a lower
level text, for example, Essential Solid Mechanics by B.W. Young
(Macmillan, 1976).

The contents show how the theory of elasticity may be used to

extend the range and complexity of problems covered by the general
title 'Mechanics of Materials', including elastic failure.

In addition there are chapters on energy methods to rcnind the

reader that these techniques are returning once more to prominence
and an introduction to numerical methods of solution, including
finite element methods.

Worked examples are included in each chapter as well as problems

with answers for solution by the student. These problems have been
drawn from the authors' own notes and given out to students over
many years so that it is impossible to be specific about sources.
We offer general ack11owledgement here to those sources.

The authors wish to acknowledge the comments of help from a

number of people, particularly Mr D.R.J. Mudge of Queen Mary
College, during the preparation of this book.
We also wish to thank Mrs M.J. Lancaster and Mrs S.M. Beaumont
for their help and patience in typing the manuscript.

University of Bradford P.R. Lancaster

D. Mitchell
1 EQUATIONS OF ELASTICITY

As in all branches of Newtonian mechanics, the equations of motion

apply to deforming materials. It is assumed that the deforming
materials are continuous, homogeneous, isotropic and elastic. These
assumptions appear to ignore the possibility of material failure but
the equations developed do show how analysis for stress and strain
enable design to be assessed and if necessary modified. A study of
mechanical failure alongside a stress and strain analysis is indeed
to be encouraged.
The meaning of 'elastic' is depicted in the stress versus strain
curves of figures l.la and l.lb. In figure l.la the material is
linear elastic, while figure l.lb depicts the behaviour of non-
linear elastic material. For both materials the important thing is
that when the forces are removed from the body the deformation is
also removed, that is the graphs of stress versus strain follow the
same line whether loading or unloading.

-I)

en
en
--
I)

en
en
Q)
.... ....
Q)

(a) (b)

st rai11 (e) strain(e)

Fig. 1.1
In a later section of this book, materials being deformed beyond
the limits of elasticity (plastically deforming) are considered.
However, whatever the material properties the equations of equilib-
rium are the same. The way in which materials behave when stressed
is defined by the stress-strain relationships.

1.1 Equilibrium Equations

Figure 1.2 shows an element of material on which normal stresses (a)

and shear stresses (T) are acting. Using a cartesian frame of
reference (xyz), let the element have side lengths of dx, dy and dz.
Since opposite faces have equal areas then equilibrium of forces
(stress x area) is satisfied in the x, y and z-directions.
By taking moments about the centre of the rectangle in a coord-
inate plane (say zx-plane - figure 1.2) then to obtain rotational
equilibrium

1
 TXZ (dz dy)d2X + (T XZ + d:XZ
oX
dx
dx)dy dz T- TZX(dx dy)~
2

which gives

dT h
+--
xz dx = + ~ dz
ax T
zx az
Z (force/Vel.) !
Taz +rz
oaz d
z Y (force/Vol.)
- Y
OTyzd
+- /
Tyz
oy
ay + ~d
oy y
OTyx d
Tyx + Oy Y

Oax d

~ X (force /Vol.)

Fig. 1. 2

As the size of the element tends to zero the second term of each
side of this expression vanishes. Thus at a point in the material
in equilibrium

T
xz = T zx (1.1)

It follows for the ideal elastic materials being considered here

that at a point within a deforming material, the shear stresses are
in equilibrium with their complementary shear stresses. Likewise,
couple equilibrium in the other two planes yield Txy = Tyx and Tyz
T The convention used here for the two subscripts is that the
yz
first refers to the normal to the plane on which the shear is acting
and the second to the direction of the shear stress. When the out-
ward normal acts in the positive coordinate direction, then the
positive shear stresses act in the positive directions of the other
two coordinates.

2
1.2 Hooke's Law, Stress-Strain Relationships for Linear Elastic
Materials

Robert Hooke stated for an elastic material 'Ut tensio sic vis'
which when freely translated says that 'the strain produced in an
elastic material is proporational to the applied stress'. That is,
stress strain is constant for a given material.

For a linear elastic material, this means that the principle of

superposition applies so that, for example, deformations for a
particular point in the material due to the action of a number of
forces may be determined independently and added vectorially. The
effect of the normal stress is to produce normal strains in the
direction of the stress. Thus the strain due to ox in the x-
direction is ox/E (extension) while in the other two coordinate
directions the stresses a y and a z will decrease the effective strain
in the x-direction by va y /E and -vo z/E. This is known as the
Poisson effect. Thus the total strain in the x-direction is

aX va va z
e
X =T - ___L-
E E

or e
X = i; [ox - v(o y + az )]

where E is Young's modulus of elasticity and vis Poisson's ratioihr

the material. These normal strains are unaffected by the shear
stresses but shear strains are produced, which cause distortion of
the element of material. The shear strain is simply

using Hooke's law, where G is the modulus of rigidjty for the

material. The complete stress-strain relationships are thus

1
e = [ax v(o + az)]
X y

1
e E [cry v(ox + oz)]
y

1
e [az - v(o X + ay )]
z E
(1. 2)
T
xy
Yxy T
T
zy
Yyz T
Tzx
Yzx T
3
1.3 Equilibrium Equations for Two Dimensions

Section 1.1 was concerned with rotational equilibrium conditions at

a particular point in a solid. To find the equations of equilibrium
for any point in the body, the same procedure of resolving forces is
used. In figure 1.2 the stresses acting on each coordinate plane
and their variations on opposite faces of a small cuboid are drawn.

For example, resolving forces in the x-direction and equating to

zero for equilibrium

a ax dT yx
aX dz dy - [a
X
+ ax dx)dz dy + T dz dx - [T
xy yx +
ay dy)dz dx

dT zx )
+ '1'
ZX
dx dy - (T
zx + ---az dz dx dy + X dx dy dz = 0

Resolving forces in the other two directions gives corresponding

equations. Thus

a aX ()T
--+ ~ zx
+ -- + X 0
ax ay az

()T aa ()T
~ + ____r + ~ + y 0 (1. 3)
ax ay az

3 Txz aa z
--+ ~ + -- + z = 0
ax ay az

These equations ensure that equilibrium of the material is

maintained. Because many engineering problems may be considered
static or not moving with acceleration, the stresses obtained are
for a steady state condition. Consider now two-dimensional systems
in which Tzx = Tzy = az = Z = 0. Then equations 1.3 reduce to

( 1. 3a)

for material lying in the xy-plane. X and Y represent body forces

per volume which may be inertia forces due to rotation, field effects,
for example magnetic field~ or simply the weight of the body itself.

4
1.4 Plane Stress and Strain

Equations 1.2 are applicable generally to three-dimensional problems.

If the stresses are variable in only two dimensions then the system
usually reduces either to plane stress or plane strain.

(a) Piane strain all the strain takes place in one plane, say the
xy-plane. Then ez 0 and from equations 1.2
(J
z

Thus the other two strains become

ex = ~[a
Ex
-__
1-vy
a)
v_
(1. 4)

e = ~[a v_ a)
-__
y E y 1-v X

and the shear strain y xy = Txy /G remains unaltered.

In plane strain the constants of elasticity may be considered to

be modified and the effective m~dulus of elasticity is E/(1 - v2)
while Poisson's ratio becomes v/(1 - v) and G remains unaltered.
{b) Plane stress all the stresses are confined to one plane, say
the xy-plane. Then az = 0 and

ez - 'L(a
E X + a )
y

while the other three strains are

(1. 5)

To achieve plane stress there must be 'thickness' strains and for

plane strain there must be an applied stress in the z-direction.

1.5 Polar Coordinates

It is often more convenient to use a polar coordinate system for

solving elasticity problems. The appropriate polar equations are
therefore derived. Figure 1.3 shows an element cut from a piece of
stressed material. By use of polar coordinates a set of equilibrium
equations is obtained, similar to equations 1. 3. Again accelerations
are assumed to be zero. The coordinate directions are taken to be
r e z and in two dimensions it is assumed that Tez Trz' az and Z
are zero.
5
 Fig. 1.3a
Resolving forces as before, in the radial direction

[a r de
+a 0 rdr)(r + dr)dz de - o r dz de - oe dz dr sin 2
ar r

- [cre + ae
aoe de ) dz dr sin 2de - Tre dz dr cos 2de + [Tre

ilTre
+~ de ) dz dr cos 2
de +
Rr dr de dz 0

That is

aor de ClTre
~ dr de + or dr de - 2oe dr 2 + ~ dr de + Rr dr de = 0

on neglecting, as is usual, second-order small quantities and using

sin oe = oe and cos oe = 1; since oe is small, we obtain

l
aor l(ClTreJ or- oe
--+--- + +R= 0
ar r ae r
and for the tangential direction (1. 6)

oTre 2Tre
!roe) + - - + - - +
r ae ar r
s = 0

where R and S are the body forces per volume in the radial and
tangential directions.

6
 z
+ azoa,
a,+
d
z

(b)

Fig. 1. 3b

The results for resolving forces in the rez-directions are given

in equations 1.7 (figure 1.3b).

aa r dT
rz a - as
r
-- + l[<lTrSJ + --+ + R 0
ar r ae az r

dT rz aa z Trz
--+ l(<lT ze) + -- + -- + z 0 (1. 7)
ar r ae az r

3Tre 3T ze 2Tre
--+
ar
_!_rae)
r ae
+--+ --+
az r
s 0

1.6 Stress at a Point

Referring to figure 1.4, in which a tetrahedron has edges in the

coordinate directions, so defining a plane abc, let N be the
normal to the plane abc and let the angles between the normal and
coordinate directions x, y and z be cos-1 ~. cos- 1 m and cos-1 n.
The tetrahedron is considered to be quite small so that the
resulting equations are not only valid at a point, but, of greater
importance, they are valid in a region where the stresses are
varying. Thus area oac is the projection of the area abc in the

7
 Fig. 1.4

zy-plane, or area oac = area abc x cosine of the angle between the
normal and the x-direction, that is

6oac 6abc x ~

6oab

6obc
6abc x m

6abc x n
} (1.8)

and ~. m and n are known as the direction cosines of the normal to

the plane with respect to the x, y and z-directions (shown positive
in figure 1.4). Figure 1.5 shows the applied stresses (direct) crx,
a and a and (shear) T , T and T acting on an element and p ,
y z xy yz zx x
py and pz are the components of stress, that is components of the
total force per unit area, on the plane abc, in the coordinate
directions x, y and z. To find p , p and p , resolve forces in the
X y Z
coordinate directions in turn; for example, in the x-direction

a
X
(6aoc) + T
YX
(6aob) + T
ZX
(6obc) = p X (6abc)
and, using equations 1.8, this reduces to

p
X
= aX~ + T
yx
m+ T
ZX
n

Similarly, resolving forces in the y and z-directions gives values

for p y and p z . Thus

I
Px a ~ + T m+ T n
X yx zx

Py T ~ + a m + T n (1.9)
xy y zy
T + T m + a 2n
Pz xz ~ yz
8
 Fig. 1.5
Having these components enables the normal and shear stresses on
face abc to be found. Equations 1.9 may be written in matrix form
T T
yx zx
T m (1.9a)
zy
T
XZ
T
yz
n

Eigenvalues and eigenvectors of the matrix can be found by apply-

ing the mathematical rules for matrices. The eigenvalues of the
above are the principal stresses and the eigenvectors the corres-
ponding direction cosines of the normals to the principal planes.
Example 1.1
Find the resultant stress, the normal stress and the resultant shear
stress on the plane abc in figure 1.5.
(a) Resultant stress The resultant stress on the plane is given by
the resultant of the forces acting on abc divided by the area of abc.
Thus it will be evident that
a = (p 2 + Py 2 + Pz 2)!
R X
(b) Normal stress To find the normal stress on the plane, resolve
forces parallel to the normal N. (Note that the area of abc is
common to all forces acting on this face.) Thus
aN = p X1 + p y m + p Zn

9
(c) Resultant shear stress The normal force and the resultant
force lie in one plane (figure 1.6). The shear force (TR x ~abc)
forms the closing side of the force triangle and hence
T
R = l(a R2 - aN2 )

c
Fig. 1.6
(Note that TR must act in the plane of abc.)
It will be appreciated that the magnitude of the shear stress on
plane abc will depend on its direction in the plane.
Example 1.2

Find the direction cosines of TR with respect to the given axes in

figure 1.7.

Fig. 1.7

Let ~1, m1 and n1 be the direction cosines of the direction of TR

with respect to xyz and resolve forces parallel to the x, y and z-
coo~dinates in turn.

Direction of x

10
Hence

Similarly
p - aNm
ml = -LY_ __
TR

and n 1

Px py and pz being given by equation 1.9.

A similar method can be used for any shear stress component.

The converse, to find a shear stress corresponding to a given set
of direction cosines, is to resolve forces due to p , p and p in
X y Z
the direction required. Thus, for any direction in the plane abc
Tad = pxtad + pymad + pznad

provided that tad 2 + mad 2 + nad 2 1.

1.7 Principal Stresses

A principal plane is defined as one on which the shear stress
is zero. The normal stress on this plane is known as the principal
stress and is denoted by p. From figure 1.5 orientate plane abc to
coincide with a principal plane (see figure 1.8). Let the
direction cosines of this plane be tp, rnp and np. Then resolving
forces on the plane in the coordinate directions gives

Px pR.
p
Py
Pz
pmp
pn
} (1.10)

p z
N

Fig. 1.8

11
 This set of equations does not in any way conflict with
equations 1.9 when ~. m and n are replaced by ~ , m and n Thus
p p p
equations 1.9 may be regarded as being obtained by external force
considerations and equations 1.10 by considering the internal force
system for a particular plane on which there is no shear stress
acting. Thus combining equations 1.9 and 1.10 gives

p)~ + T m + T n 0
p yx p zx p

T
xy ~ p + (cry - p)mp + T n
zy p
0
} (1.11)

xz ~ p + T yz mp + (cr z - p)np = 0
T

Equations 1.11 are homogeneous equations for , m and n ; ~

p p p
for non-zero values, the determinant of the coefficients of
~p' mp and np must be zero, then

T (cr - p) T 0 (1.12)
xy y zy

T T
XZ yz

from which the following cubic equation for p can be found

(1.13)

where

I1 (cr + cr + cr )
X y Z

I2 cr cr
X y
+

2
cr cr
y z
+ cr cr

2
Z X
- T
xy
2
2 - T
yz
2 - T
zx
2
) (1.14)

I3 = cr T + cr T + cr T - 2T T T - cr cr cr
x yz y zx z xy xy yz zx X y Z

Now I 1 , I 2 and I 3 are called the stress invariants, that is they

do not alter in value whatever the orientation of the axes x, y and
z, which have been chosen arbitrarily.

Equation 1.13 always has three real roots Pl P2 and P3 and

consequently three sets of direction cosines, one set for each p,
can be found using equation 1.11 combined with the geometrical
relationship ~2 + m2 + n 2 = 1.

Thus, for example, to find the direction cosines for P2 sub-

stitute p 2 in equations 1.11 and hence obtain ratios say ~2Im2,
n 2/m 2 . Substituting these in the geometric relationship
~2 2 + m2 2 + n 22 = 1 will give absolute values for ~2. m2 and n2.

12
It can be shown that the three principal planes are orthogonal and
hence also are the directions of the principal stresses.

1.8 Stationary Values of Shear Stress and Associated Planes

Shear planes contain the maximum shear stresses corresponding to
two of the principal stresses [cf. two dimensions where Tmax=
(p 1 - P2)/2]. Thus there will be three of these planes and their
corresponding stresses. The problem is to find values of , m and
n which define these planes.
It has been shown that
T 2 (J 2 (J 2
R R N

that is
T 2 = p 2 + p 2 + p 2 _ (p + p m + p n)2
R X y Z X y Z

This can be expanded making use of 2 + m2 + n2 = 1, and the

following can be obtained for any orientation of plane abc.
T 2 = (p m- p )2 + (p n - p m)2 + (p - p n)2 (1.15)
R X y y z z X

If the axes are re-orientated to coincide with the principal axes

(figure 1.9) then by resolving forces in the usual way gives

Px = P1, Py = P2m and Pz = p3n (1.16)

3 z) N

P,

p3
Fig. 1.9

Hence substituting 1.16 in equation 1.15 gives

(1.17)

13
Reduce the variables to two by using ~2 + m2 + n2 = 1 to eliminate
n, the

'R2 (Pl pz)2~2m2 + (pz - P3)2m2(1 t2 - m2)

+ (P3 - Pl)2t2(1 - t2 - m2)

and for stationary values of 'R put a,R;a~ 0 and a,R;am =0

so that

2~m2(Pl - Pz)2 - (pz - P3)22tm2

+ (p3 - pi) 2 (n - 4~ 3 - 2~m 2 ) = o (a)

and

2~ 2 m(pl - P2) 2 + CP2 - P3) 2 (2m - 2mt 2 - 4m 3)

- (P3 - Pl)22m~2 0 (b)

If t = 0, (a) is satisfied but (b) requires that (p 2 - p 3)2(2m

4m 3) = 0, that is

m = 0 (minimum <) or m = 1/12 (maximum <)

and then the corresponding value of n can be found.

Table 1.1 sets out the maximum and minimum values of' and the
direction cosines of the normals to the planes on which they act
with respect to the principal axes.
TABLE 1.1
T T
min max
~ 0 0 1 0 1/12 1/12
m 0 1 0 1/12 0 1/12

n 1 0 0 1/12 1/12 0

T 0 0 0 HP2 -p3) ! (p 1 - p 3) HPl - P2)

It should be noted that the principal shear planes bisect the

principal planes. Corresponding to each of the maximum shear
stresses there is a normal stress.

oN= Pl2 + P2m2 + P3n2

with say t = 0, m = 1/12 and n 1/12

P2 + P3
ON 2

14
Thus on a principal plane of shear there is a shear stress T
CP2 - P3)/2 together with a normal stress oN = CP2 + P3)/2.
Example 1.3

The stresses on the coordinate planes acting at a point in a piece

of material are ox = 4, cry = 7 and crz = -9

Txy T =4T =T =5T T

xz
0
Yx ' yz zy ' zx
Calculate

(1) The normal stress, resultant stress and the resul~ant shear
stress on a plane whose normal makes angles of 50 with the
positive x-axis and 600 with the positive y-axis.
(2) The direction cosines with resrect to the x, y and z-axes the
resultant shear stress.
(3) The principal stresses and the direction cosines of the normals
to the principal planes with respect to x, y and z.
(4) The maximum shear stress at the point in the material.

Using figure 1.5

Px cr R. + T m + Tzx n
X yx

Py T R, + cr m + T n
xy y zy

p z = Txz R, + T m + cr n
yz z

cos 50 0.64279 R,

cos 60 0.50000 = m

From R- 2 + m2 + n2 = 1, n = 0.58037; the positive value is taken

because cos- 1 R, and cos- 1 m are both acute, that is cos- 1 n is
54.528 o.
Hence
4.57116

Py 8.97301

Pz -2.72333

(1) normal stress oN p R. + p m + pzn

X y
cr = 5.84426
N ,
resultant stress oR (p 2 + p 2 + p 2)2
X y Z

10.43202

15
 resultant shear stress
1
T = (cr 2 - cr 2)
R R N

TR = 8.64128
(2) Let ~ 1 , mr, n 1 be the direction cosines of this shear stress
with respect to the x, y and z-axes. Then

p - cr ~
X N = 0.09426 (84.592)
TR

0.70023 (45.555)

-0.70767 (135.046)

(3) The principal stresses are given by the determinant (equation

1.12)

(4 - p) 4 0

4 (7 p) 5 0

0 5 -(9 + p)

which on expansion and simplification gives

p3 - 2p2 - 112p + 208 = 0

To solve this equation, Newton's approximation is used, that is if

p 1 is a root then a more accurate root is given by

P2 = Pl ful
- f' (p)

where f'(p) is the differential coefficient of f(p). Try p 1 1,

then
1 - 1 - 2 - 112 + 208 95
P2 3 - 4 - 112 1 + 113

95
P2 =1 + 113 = 1.8407
and repeating the process gives p = 1.8526 as a root. The other two
are 10.67 and - 10.52. The principal stresses are therefore 1.85

16
(tensile), 10.67 (tensile) and -10.52 (compressive).
In order to find the direction cosines of the principal planes
use equations 1.11 to give
(4 - p) + 4m =0
4 + (7 p)m + 5n 0

5m (9 + p)n =0
Now for a specific principal stress, say p = 1.85, substitute
in the above equations and let 1, m1 and n1 be the direction
cosines of the normal to the plane of this stress.

(4 - 1.85)1 + 4m1 = 0

or 4 - - - 1.8604
- -2.15 -

also 5m1 (9 + 1. 85)n1 0

therefore
n1 5
m1 = 10.85 = 0.4608

(-1.8604m1) 2 + m1 2 + (0.4608m1) 2 1

2 1
or m1 = '4.6734
and m1 0.4626 (62.44)

so 1 -0.8606 (149. 38)

and n1 = 0.2132 (77.69)

that is, the direction cosines of the 1.85 stress plane are
1 = -0.8606, m1 = 0.4626 and n1 = 0.2132. 2, m2 and n2, 3,
m3 and n 3 can be found in a similar way by subsitituting p 2 = 10.67
and P3 = -10.52 in turn which give 2, m2 and n2, 3, m3 and n3
from the above equations; the student should confirm these results.

(4) The maximum shear stress is the greatest of (p 1 - p 2)/2,

CP2 - P3)/2 or (p3 - P1)/2 and for this example [10.67 - (-10.52)]/2
= 10.59.
Note No units have been s~ecif!ed for stresses, but usually they are
quoted in newton (metre)- (Nm 2)(or pascal) or other recognised
dimensional units, for example N mm-2.

17
1.9 Strains and Displacements
A solution to an elasticity problem should provide complete
information on stress, displacement and strain. It is usual to
measure strain in some form to find stresses in experimental stress
analysis. The order of a large strain in steel is typically lQ-3
(note that the dimension of strain is L/L, that is a number only).
Displacements can be of high order, provided that the material
remains elastic and that the maximum permitted elastic strain for
the material is not exceeded. It is first necessary to establish
relationships between strain and displacement.

For example, for a simple beam the stresses and strains can be
obtained in terms of a simple displacement o in the direction of y
(figure 1.10). From the theory of simple bending

and a = Ee

In figure 1.11 A (x, y) is displaced to A' (x + u, y + v) where u

is the displacement in the x-direction and v is the displacement in
they-direction. The increase in length of AB to A'B' is dx x
(variation of u with respect to x), that is (au/ax)dx. Partial
derivatives are necessary since u is not independent of v. With the
usual definition and restrictions for strain ex = increase in length
of AB in direction of x/(original length of AB)

projection of A'B' - AB
AB
dx + au dx - dx
ax
dx

au
ax
and similarly in the y-direction
av
ey = ay
Because of shear, the two adjacent lines AB, AD rotate through
small angles e and A. From figure 1.11

~dx
tan e :: e ax
dx + au dx
ax

18
since (au/ax)dx is small compared with dx.

Similarly

tan ::: A = ~~

Shear strain is defined as the change in angle between two lines

originally at right angles, that is [8 - (-A)] (figure 1.11) or
av au
=-+-
ax ay
Thus for two dimensions in cartesian coordinates

l
au
-
ex =ax
av (1.18)
e y = ay
av au
Yxy = ax + ay

In what follows, deformations in the Xfl-plane only are

considered and subsequent modifications are made to take account
of plane stress or plane strain.

Fig. 1.10

1.9.1 Rectangular Coordinates

Figure 1.11 shows a rectangular element ABCD having side lengths

of dy and dx lying in the plane of a sheet. Stressing the sheet in
which the element lies produces a displacement of the element in
addition to straining. Direct strains are responsible for increases
in the lengths of the sides of the elements while shear strains

19
 y

~dy=**f====-1=1
v
+-_,__ _A
oDe I
-------~AI

dy
a
I
I

y
I I
I

l J.
I

.I
~--~L----------------~--------------.-x

X U

Fig. 1.11
produce rotation of the lines, that is a change of shape of the
element.

In the theory to be developed, it will be assumed that the

elements of lines being considered remain straight but are capable
of rotation. (This means that the material will be linear elastic.)

In three dimensions, w is the displacement in the z-direction,

then

au av aw
e X = ax , e = ay , e z =az-
y
( 1. 19)
au av av aw
Yxy =-+-
ay
-+
ax Yyz = az ay Yzx = az
'" + }
ax

It can be seen that u and v are not independent since yxy includes
strains which are different to the direct strains. The equation for
y may be regarded as the equation of compatibility for y and v and
xy
hence the strains e and e . This ensures that e , e and y will
X y X y xy
be so related that the stress-strain relationships are satisfied at
all points in the deforming plate. Generally speaking, compatibility
means that the material remains continuous and no voids can occur
within it. The stresses and strains at all points within the
deformed body can never in themselves produce discontinuities .. For
any particular point in the sheet of material, figures 1.12a and 1.12b
can be constructed from figure 1.11, so that A' of the distorted

20
triangle A'B'C' coincides with A of the undistorted triangle ABC.
From the geometry of figure 1.12 the strains at the point can be
found for
PC' = CQ cos a + C'Q sin a

[~~ dx + ~~ dyJ cos e + [;; dy + ~~ dx) sin a

and the strain

PC' PC'
ea AC = ds

therefore

ea = (;~) (!:) + [~~) [~~] cos a

+ (;;) [*) + [~:) [~:) sin a

au+ sin a cos a (au av) ~; sin 2 a (1. 20a)

cos 2 a -
ax ay
+
ax +

'Y xy
e X cos 2 a + e y sin 2 a + sin 2 a
2

PC C'Q cos a - CQ sin a

PC
$a = AC

cos 2 a av 2
=

av s1n
ay a cos a +
ax - sin a au
ax - sin a cos a au ay

and at right angles to this (a + ~/2)

$'a = - ~;sin a + sin2 a ~~ + ~~sin a cos a - cos 2 a ~~

and 'Ya $a - $'a

= 'Y xy cos 2a - (e x - e y ) sin 2a

Example 1.4 Strain Gauge Rosette

An application of the two-dimensional strain equation (equation

1. 20a) when the axes (x, y) are coincident with the principal axes
(1, 2), is to be found when using resistance strain gauge rosettes.
Rosettes usually have three strain gauges which measure the 'strain
at a point' for a single application of stress. Figure 1.13 shows
three strain gauges separated by angles a and S. Axes 1 and 2 are the
principal strain axes and a is a reference angle which defines the

21
 G~dy
y

~(0~--t-------+-----------x
X u (a)

Fig. 1.12a

Fig. 1.12b

22
location of gauge a with respect to the 1-axis.

1
Fig. 1.13

From equation 1.20a

e
9
= ex cos2 9 + e
y
sin 2 9 + y
xy
sin 29

when referred to principal axes 1 and 2 become

or ------2-- + cos 29

that is

e
e
a + a cos 29
where a (e 1 + e2)/2, and 8 = (el + e2)/2, and the unknowns,
to be found from strain measurements, are a, 8 and 8. There
are three measurements

e a + 8 cos 29 (i)
a

a + 8 cos 2(9 + A) (ii)

ec =a + 8 cos 2(9 + A+ ~) (iii)

For commercially bought gauges it is usual for a = 8. These are

the formulae necessary to find the principal strains at the point.
Hence the principal stresses can be found for

23
and

so

and p2 = _E__ (e2 + vel)

1 - v2

It is also possible to find other direct and shear strains for

the given point. Solution of equations (i), (ii) and (iii) can be
tedious and if many results are to be analysed a small desk computer
will be found useful. It will be appreciated that e1, e 2 , p 1 and p 2
are the strains and stresses at the surface of the material.

Given

eb 1 X 10- 4

ec = 3 X 10- 4 with A. = 1.1 60

2 X 10-4 a + B cos 29 (i)

1 X 10- 4 a + B cos 2(9 + 60) (ii)

3 x 10-4 =a + B cos 2(9 + 120) (iii)

Eliminate a between (i) and (ii) and (i) and (iii), that is

10- 4 B cos 29 - B cos 2(e + 60)

2 X 10-4 = B cos 2(9 + 120) - B cos 2(9 + 60)

giving

10- 4 = ~ cos 29 +
2
~~ sin 29 (iv)

and 2 X 10-4 1(3)8 sin 29 (v)

which yield 9 -30 and B = 1.3333 and hence

a = 0.6666

Thus a + B e1 2 (say)

a - B e2 0.6667

The signs are determined by the loading situation, for example,

on a beam the rosette can be on the tensile or compressive side.

24
1.9.2 Polar Coordinates

The displacements are, again, represented by u and v to correspond

respectively with radial and tangential directions.
From figure 1.14

x r cos a

y r sin a

for a plane element of material lying perpendicular to the z-axis.

An element abed will appear as shown in the figure 1.14a when
undeformed and in figure 1.14b the element is shown displaced by
(u, v) and at the same time has been deformed by imposed forces.

(a) (b)

Fig. 1.14
If ad is displaced a distance u then be is displaced u + du. So
in the radial direction
extension of de
er original length of de

that is

Again partial derivatives are required since two dimensions are

being considered.

The radial strain is assumed to be unaffexted by second-order

effects of the tangential strain. A line such as ad will be
affected in both rand a-directions. Thus from figure 1.14b there
will be two components making up the tangential strain ea.

25
(1) Due to radial displacement u only

ad = r de

therefore the strain at a layer u from this layer is

(r + u)de - r de = ~
r de r

(2) Due to displacement v along S-axis there is an elongation ov so

that the strain is ov/ad and since ad = r de the additional strain
is (1/r)(av;ae) so

e
e
= ~r + ..!:..(av)
r ae

The shear strain Yre is the change in the angle at say a and this
is

Yre = ~ + (S - a)

Now (ignoring strains of second order)

av
s - ar
v
a =-
r

and ~ = H;~)
Thus the total change in angle at a' is

Yre = _!_(au) + av _ ~
r ae ar r

Summarising for polar coordinates

x = r cos 8 (1. 2la)

y r sin 8 (1.2lb)

(1.2lc)

ee = -
u + ..!:..( av) (1.2ld)
r r ae

av v
Yre = _!_(au) +- (1. 2le)
r ae ar r

and u is not independent of v.

26
1.10 Summary of Equations (Plane Stress)
Cartesian Coordinates

(a) Stress-strain relationships

e = av = .!..(a - va )
y ay E Y X'

(b) Equilibrium

aa h
~+~ + X 0
ax ay
aT aa
_l5X_ + _J.. + y 0
ax ay
(c) At a point

Px aX R. + Tyxm

p
y = T yx R. + a m
y
POlar Coordinates

(a) Stress-strain relationships

e
r
= ar
au = .!..(a - va )
E r a

e9 = .!!.r + !(av)
r aa = .!.(a
E a - varl)

1 (au)
=r as av
Yra + ar- r V Tr9
=G
(b) Equilibrium

!(aa 9) aTra 2Tre

+ ---ar- + - r - + S 0
r aa J

27
1.11 Transformation Equations
To obtain o and o6 from values of o and o transform equations can
r x Y.
be derived. In figure l.lSa the stresses at a point are shown in
both rectangular and polar coordinates. From these stresses
elemental triangles can be set up as shown in figure l.lSb. From
this the Mohr circle can be drawn (figure 1.16).

(a)
Fig. 1.15

T~--------------~aL-----------~--~

Fig. 1.16

28
 To obtain the value of ar note that 6 is the angle measured anti-
clockwise from the x-axis to the normal to the plane ac. On the
Mohr circle this corresponds to measuring 26 anticlockwise from the
datum ab. Note that points in the Mohr circle correspond to planes
in the stress diagram. Thus

ar + a6
--:2,--- + R cos (26 - a)

where R is the radius of ~1ohr circle.

ar + a6
cry ----2--- + R cos 26 cos a + R sin 26 sin a

and R cos a = FD or AF

R sin a 'r6

ar + a6 a6 - a r
a = --2-- + cos 26 + T sin 26
y 2 r6

a a6
r
= -y(l - cos 26) + 2(1 + cos 26) + T
r6
sin 26

a a sin 2 6 + a 6 cos 2 6 + T sin 26 (1. 22a)

y r r6

a r + a6
a - R cos(26 - a)
X 2

a cos 2 6 + a 6 sin 2 6 - T sin 26 (1.22b)

r r6

T
xy
-R sin(26 - a)

a6 - ar
sin 26 + T cos 26 (1. 22c)
2 r6

Equations 1.22 are the transform equations for rectangular

coordinates to polar coordinates.

29
1.12 Problems

1.1 A circular shaft 50 mm diameter is subjected to a bending

moment M and is at the same time transmitting a torque T. A 45
equi-angular resistance strain gauge rosette is fixed to the shaft
surface to monitor the torque. When the shaft is in a position
which produces maximum bending stress

ea - 10 x 10-s, eb = 5 x 10-s and ec = 7 x 10-s

Calculate the values of M and T from the readings. Take E = 207 kN/
mm- 2 , v = 0.3
[M = 113 N ml

1.2 A 45 equi-angular rosette is fixed to a plate structure and

the following readings are taken
e = 2 x lo- 4
b

Gauge (c) cannot be read because it is damaged. It is known however

that at the point of attachement of the rosette there is a thickness
strain (ez) = - 3 x lo-~+. Determine the maximum shear stress at the
point and its direction in relation to gauge (a). (Use figure 1.13.)
Take E = 207 kN mm- 2 , v = 0.3.

1.3 Referring to figure 1.5, the stresses acting at the point are
(units N mm- 2)

4, ay is unknown, az 0, T
4' 2 T 3.
yz T ZX xy
Calculate
(1) the direction cosines of the plane on which there is no stress,
(2) the value of cry,
(3) the principal stresses and their direction cosines,
(4) the greatest shear stress.
[Pl = 0, P2 = 6. 708, P3 = -6. 708]

1.4 Using figure 1.5 show that the stationary values of shear stress
at the point are ~CP1 - P2), !CP2 - p3), ~CP3- Pl) and hence
calculate the normal stresses on these shear planes.

1.5 If ax= 3, ay = 0, a z = -2, T xy = 4, T yz = 2 and T zx = 5,

calculate for a plane N(~, m, n) where ~ = 0.4 and m = 0.6 the
resultant shear stress and its direction cosines in relation to
the given axes. What are the principal stresses and the direction
cosines of the greatest of these?
[Principal stresses are 8.47, - 2.41, -5.06]

30
1.6 A resistance strain gauge rosette has arms a, b and c.
The following readings are obtained

-3e
a
Show that the maximum shear stress at the point is given by

~[28)~
1 + \} 3

assuming that the measurements are taken on the surface of a thin

plate subjected to a two-dimensional stress ~stem. The angle
between a and b is 30 between b and c is 60 . Show on a
diagram the directions of the prineipal stresses in relation to the
position of the rosette.

1.7 Using figure 1.5 calculate the principal stresses at a point

where crx = 2, cry = 3, crz = -1, Tyx = ~ and Tzx = Tzy = 0. For a
plane whose normal has angles Nx = 400, Ny = 70 calculate the
resultant shear stress on the plane and the direction cosines of its
direction on the plane.
[Principal stresses are 7.53, -2.53, -1.00]

1.8 For a general stress system (figure 1.5) show that the octa-
hedral shear stresses are given by expressions of the type

T = .!.I[(Pl- P2) 2 + (P2- P3) 2 + (P3- pl)]

oct 3

and calculate the corresponding normal stress on these planes. (An

octahedral plane has a normal whose direction cosines are equal,
that is, R- = m = n.)

31
2 STRESS FUNCTION

In this chapter a standard method of solution for two-dimensional

elasticity problems is developed. The method was first proposed
by G B Airy (1862) and involves finding a single function ~(x y)
which satisfies the compatibility requirements and from which the
stresses are easily derived.

2.1 Stress Function

To solve problems in elasticity it is necessary to satisfy boundary
conditions for both stress and strain. However, it is often an easier
matter to satisfy stress boundary conditions initially and to satisfy
strain boundary conditions later. For two-dimensional problems in
cartesian coordinates, equations 1.18 apply. Differentiate twice
the strain displacement equations and substitute the results in the
differentiated shear strain equation thus

a2 e X
=~
ayz axay 2

a e_
2
_ Y
= a-v-
3

ax 2 ax 2 ay
and
az
___2z = ~ a 3v
+--
axay axay 2 ax 2 ay

that is

(2.1)

This compatibility equation for strain has to be satisfied at all

points in the deforming material. Now substitute the stress equiv-
alents of e , e and y xy (plane stress) remembering that E =
x y
2G/(l + v), therefore

2(1 + v) azT az 1 az
~ = .!. - (a - va ) + - - (a - va )
E axay E ayz x y E ax2 y x

Cancelling E and re-arranging the equation gives

32
 (2. la)

Now differentiate the first of the equilibrium equations (1.3a)

with respect to x and the second with respect to y.

(2.2)

Adding these equations to obtain a2T /axay which is then

xy
substituted in equation 2.la gives

so
az
- - (cr
ax2 x Y
az
+ a ) + - - (cr
ayz x
+ a )
Y
-(l+v) t ax+aY~
-
ax
-
ay

[ax aY]
or Cl + v) Lax + aiJ (2. 3)

Equation 2.3 represents the compatibility condition for stress.

If body forces are absent then equation 2.3 reduces to

[-az- az) (cr

+ -- + a ) =0 (2. 3a)
ax2 ayz x Y

For plane strain conditions substitute v/(1 - v) for v in equation

2.3 to give (chapter 1, section 1.4)

(-az- + -az-) (cr + a ) = - 1 ) [ax

[r-::-"V a'(!
ax2 ayz x Y Lax + aYJ (2 .4)

If there are no body forces equation 2.4 reduces to equation 2.3a.

The stress or strain compatibility conditions are not solvable

except for the simplest of problems. However Airy suggested a stress

33
function ~(x, y), defined as follows

a
X

(2.5)

'xy

Equation 2.3a can thus be written

(2.6)

which is a biharmonic differential equation representing the compat-

ability equation for stress. In its differentiated form equation
2.6 becomes

~ + 2 a'+~ (2.6a)
ax 4 ax 2ay 2
This equation is .also difficult to solve for any but the simplest
of problems. However, numerical methods can be used as will be seen
later for quite complex problems. The advantage, analytically, of
equation 2.6a is that functions of x and y can be devised which rep-
resent particular stressing conditions. This is known as a semi-
inverse method of analysis, whilst developing ~ from known boundary
conditions is the direct method. The examples to follow apply to
plates lying and loaded in the x-y plane. In general plane stress
solutions are given. The stresses will clearly be in-plane stresses.

Example 2.1

(a) Consider a function ~ = ay 2 where a is constant. By substi-

tution, the compatibility equation is satisfied hence the stresses
can be written down immediately.

Using equations 2.5 to give a = 2a and a =

x y xy = 0. Hence ~ =
ay2 will represent s'imple tension, the forces being applied in the
direction of x.

(b) Similarly ~ bx 2 , b constant represents simple tension in the

y direction and consequently ~ =
ay 2 + bx 2 represents biaxial loadng
(figure 2.1). (Note that the principle of superposition applies
also to stress functions.)

34
 y 2b

2a

2b
Fig. 2.1

(c) A function in the form ~ = cxy, (c constant) also satisfies the

compatibility equation 2.6a. Hence the stresses are a = a = 0 and
X y
T
xy
= -c which represents a constant shear stress applied to a plate.

(d) Thus ~ = ay 2 + cxy + bx 2 represents bi-axial stresses accompanied

by shear stresses and it will be evident that quite complex problems
can be examined using this semi-inverse method. It is imperative,
however, that the compatibility condition (equation 2.6) is always
satisfied.

2.2 Cantilever Plate

2.2.1 Stress Boundaries

To solve the problem of an end loaded cantilever plate the direct

approach will be given in detail

~
~ __,.
X
~
,.....
p
L
y
Fig. 2.2
The force P acts over the free end of the plate and Ox represents
the centroidal axis for the undeformed plate. Let the thickness of
material be unity and the other dimensions be as shown in figure 2.2.
Take origin of coordinates at 0. The following are the know.n
boundary conditions
At y a T 0 for all values of x
xy
and at y a a 0 for all values of x
y
and at X = 0 a 0 for all values of y, which also means
X
that the bending moment is zero at X = 0.

35
 Note that the shear force is constant along the plate length and
consequently there can be no variation of shear stress from one
section to another, that is T will vary with y but not with x, so
xy
(2. 7)

where f1(y) is a function of y only. At this stage the negative

sign has little meaning since it can be absorbed in constants of the
ensuing equations.
Integrate equation 2.7 with respect toy

(2. 7a)

where f 2 (x) is an arbitrary function of x only, arising because of

the partial integration.

Integrating equation 2.7a gives

(2. 7b)

f 3 (y) is another function of y only.

A stress function of this form is thus applicable to the plate.

Now
a2 <P df 2 (x)
cr = - - = - - - = 0 for y a and all values of x.
Y ax2 dx

On integration f2 (x) = A a simple constant. Also

dfl(Y) d2f3(y)
cr X = a2.p = x - - - +
ay2 dy dy2

and when X = 0, cr X = 0 and thus

which on integration gives

B and C constants. Now this is a trivial solution and will not

affect stresses cr since orders of y less than two are eliminated
X
on differentiation and f3(y) can thus be discarded. Similarly
f 2 (x) = A will not affect the stresses so that .P = x/fl(y)dy.

36
 Now from the compatibility equation

[ ]ax 2
+
ay 2
[a2q,
ax 2
+ a2q,l
ay 2
= [ ] [o
ax 2
+
ay 2
+ /f~ (y)l
Y
0

giving

(Total differential coefficients can be used because f 1 (y) is a

function of y only.) Integrating this equation gives

L, M and N being constants of integration. Hence

<P = xJfl(y)dy

Now the constants L, M and N have to be determined.

There can be no resultant force over the end of the plate so for
unit thickness

or Ja(Lxy + Mx)dy 0
-a

2Max 0

giving

M =0
and q, = ~xy3
6
+ Nxy (2. Sa)

37
 It is left to the student to show that this satisfies the bi-
harmonic equation, hence

a
X
-- ~ -- Lxy
ay2

a
y

_tl_ L
T
xy axay - zY 2 - N

from T
xy
putting y = a where T
xy
0 yields

L 2
za -N

At the end of the bar

- P = fa-a xy dy
T

(the minus sign arises because of the convention of signs for shear
forces and bending moments.) Thus

that is
3P
- L
2a 3
and the complete stress function is

cp = _ 3P 3 3P
xy + 4a xy
2a 3

If I is the second moment of area for the cross-sectio n I 2a 3; 3

Hence

and a
(2. 9a)
X

a 0
y (2. 9b)

T = - _E_(a2 - y2}
xy 21 (2.9c)

38
 Thus for the plate a suitable stress function is

from which the stresses in equations 2.9 can be found.

Note that ox = -(Px)y/I = -My/I where M is the bending moment at

a section x, and when y is negative, stresses are tensile as
expected. The results are the same as those found from elementary
beam theory. The shear stress distribution is parabolic and is
independent of x. This means that the shear force P must be
applied in such a way that the stress distribution given by equation
2.9c is maintained at x = 0. In practice this cannot usually be
achieved but by making use of the principle of St. Venant, as long
as the stresses to be calculated are not in the immediate vicinity
of the applied forces, the solution will be the same as that given
by equations 2.9. This means that at whatever point or planes the
forces are applied there will be differences in the stress distrib-
utions in the immediate vicinity of these points. As will be seen,
localised stresses can also be dealt with using equations of
elasticity.

2.2.2 Displacements

The displacements of the cantilever plate shown in figure 2.2 can

be calculated, for plane stress conditions, as follows.

From equations given in 1.18

ex = au = .!..(o
E x
- vo )
y
= .!..(-
E
Pxy)
I
ax

Integrating with respect to x

(2 .lOa)

where fl(Y) is a function of y only.

e = av = .!..(o _ vo ) = vPxy
y <ly E y x EI

Integrating with respect to y

(2.10b)

where f 2 (x) is a function of x only.

Equations 2.10a and 2.10b are, as they stand, independent but as
with stresses, they must be compatible with each other so that the
strain distribution throughout the plate is continuous. The com-
patibility condition for strain is from 1.18.

39
that is
-Pa2 p 2 -Px2 dfl(Y) vPy2 dfz(x)
2GI + ~ = 2ET + ~ + 2EI + --ax- (2 .lOc)

(Note - the differential coefficients of f 1 (y) and f 2 (x) are total

because they are functions of only one variable.)

Equation 2.10c is an identity and the variables can be separated

following the usual mathematical techniques in the solution of
partial differential equations. It will be seen that on the left
hand side of equation 2.10c that a constant -Pa 2/2GI occurs. It is
not known how this constant is divided between the integrated
expressions for u and v (equations 2.10a and 2.10b). It is usual in
such instances to let A+ B = -Pa2/2GI. This does not increase the
constants of integration, which decide the boundary conditions
because if A is known, B can be found. Separating the variables for
X
-Px2 dfz(x)
A=--+---
2EI dx
giving
Px 3
f 2 (x) = 6EI + Ax + F

and for y
~ - vPy2 dfl (y)
B + 2GI - 2E I + ~

giving
_ ~ vPy 3
fl (y) - 6GI - 6EI + By + H

F and H both constants of integration to be determined from the

boundary conditions.
The complete expressions for u and v are

p 3 vPy3 + By + H - Px 2y (2 .lOd)
u =~- 6EI 2EI

Px 3 vPxy 2
v = 6EI + Ax + F + """"2'EE (2. De)

Both these equations give compatible results for both strain and
displacement.

40
 The sim~le theory of bending gives only part of equation 2.10e
(from -Eld v/dx 2 = M). The complete solution gives, in addition to
the vertical displacements (v), the horizontal displacements Cul
Thus, the resultant displacement of a point (x, y) is (u2 + v2),.
The presence of the modulus of rigidity G in equations 2.10d and
2.10e indicate deformation effects due to shear which are not dis-
covered using the elementary theory of bending.

The strain along the axis of the plane is given by au;ax when
y = 0. Thus equation 2.10d shows that there is no extension of the
centre line of the plate and this, therefore, is the neutral axis
(that is a strain free and hence stress free plane). Engineers are
often more interested in the vertical displacements. Using equation
2.10e putting y = 0 to give the displacements of the centre line of
the plate

2 2 3 Botmdary Conditions

(a) Let the slope of the centroidal axis at the point x = L and y
0 remain horizontal and also at that point, let u = v = 0, that is
there is no movement of the point.

The first condition requires av;ax 0 at x L, y 0, that is

and
PL 3
0 = 6 EI + AL + F

Hence

and

Thus for any point (x, y) in the plate, the displacement

(2 .lOf)

It is left as an exercise to find the corresponding expressions

for u.

41
(b) A second set of boundary conditions is to let the vertical
line at x = L, y = 0 remain vertical and again do not allow the
point to move. Thus

u v =0
at x L, y = 0.

au
and ay =0
at x = L, y 0

From au = 0
ay

PL 2
or B = 2GI

and hence

and using the expression for v

and

and hence
Px3 (Pa 2 PL 2) Pa2L PL 3 (2 .lOg)
v = 6EI - 2GI + 2EI X + 2GI + 3EI

Expressions 2.10f and 2.10g are very different in appearance

because of the choice of boundary conditions. In these expressions
only points defined by (x, y) in the plate can be used and state-
ments regarding how the boundaries are fixed have to be carefully
considered and analysed. For example, comparing equations 2.10f
and 2.10g when x = 0
PL 3
vf = 3EI

the usual formula derived from elementary theory.

PL 3 Pa 2 L PL 3 ,-1 3 E(a)Zl
vg = 3EI + 2GI = 3EI _ + 2 GL J

42
 Thus if a is small compared with L then the expressions vf
and v g are the same. However as the ratio (a/L) that is (depth/
length) increases, the plate becomes short and stocky and the extra
vertical deflection is brought about by the shear stresses within
the plate. Thus vf ignores shear effects because the plate has a
small (depth/length) as the elementary theory requires and v takes
g
into account shear effects when the plate is not necessarily small
in depth. It is evident that v is the more accurate of the two
g
formulae for deflection, but (a/L) must reach values of about a/L =
0.17 to increase the value of PL 3/3EI by 10% and when a/L = 0.25
the deflection due to shear amounts to 24% of the total. It is left
as an exercise for the reader to plot the u-values at x = 0 and
-a ~ y ~ a for this and deep plates which would give an indication
of whether or not plane sections remain plane, a usual assumption in
elementary theory.

Example 2.2

Given the stresses, find the strains for the plate shown in figure
2. 3.
w;unit length

y
Fig. 2.3

The stresses are given by

a
y

T
xy

where I is the second moment of the area of the cross-section about

the centroidal axis mm. The additional terms in the expressions for
a and a , not found by the simple theory of bending, allow for the
X y
normal loading on the top surface of the bar, which produces
compressive stresses (a ) and also compensation to correct for the
y

43
condition of zero bending moment at the ends of the plate (o ). The
X
stress function is found from a consideration of a general equation
~(x, y) and solving the constants from the boundary conditions as
indicated in section 2.2.1.

Now

ex= ~~ = i [ax - very]

and on substituting for a and a and integrating gives

X y

Similarly
w IL.- c2y2 ~ + (L- c2y2}
v = - 2EI [12 3 + 3 v 6 5

Now for compatibility of strain

w 2c 2x ( 2 x 3)
= 2EI [2xy2 - -5-+ R, X - 3

aul(Y) av 1 (x)
+ vx(y 2 - c2) + vxy2] + - - - + _a_x_
ay

which is true for all values of x andy. Let B and C be two

constants such as B + C = 0. Then

avl (x)
_a_x_ = - wc
2x w
2GI - 2EI [- 2c 2x
-5- + R.2x
x2
-3- vxc 2] + B

and
2 2
v 1 (x) = - wc x - 2EI
4GI
w
[- c2x2
5
R.2x2
+ -2--
x~+
TI

44
so that

+ ~ (c2x2 - R.2x2 + x~)

2EI 5 2 12 + vo + Bx

When x and y are zero, v = o, therefore v0 o and

w IY~ - c2y2 + 2c 3y + v rY~ - c2y2

v = - 2EI [12 2 3 l' 6 5

using E/2G = 1 + v. Now since av/ax = 0 when x = y = 0, B 0 C.

Also au 1 (y)/ay is a function of y only. Therefore

aul (y)
- -ay- = c

since C = 0. Thus

vx c:r.=.
2
w [[2x3y3 - 2c 5xy
u = 2EI + ( R.2x - 3x3) y + 3 - c2y + 2c3)]
-3- + A

Longitudinal strain

e au=~
x
=
ax 2EI l[2y3
3
- 2c2y
5
+ (R-2 - x2)y + v(L:.33 - c2y + 2c33)]

and on the central axis y = 0.

Therefore
vwc 3 vw
ex = 3EI = 2E

Thus the central axis is not the neutral axis. At the ends x = R.,
y = 0, v = 0 so that finally

45
 6 = 5wR.
24EI
4[1 + g
5
c2 (i
R.2 5
+ ~)]
2

Again, for deep plates the second term in the bracket, which is
due to shear, will be significant.

2.3 Polar Coordinates

Equation 2.5 is the stress compatibility equation in rectangular
coordinates. It is often more convenient to use a system of polar
coordinates. In what follows, the polar equivalents of direct and
shear stresses will be derived along with the stress compatibility
equation. An example will follow in which the stress/strain
equations are used with their displacements derived in chapter 1.
The r - 6 coordinate system is shown in figure 2.4 (r positive when
measured from origin, 6 positive measured anti-clockwise from the
x-axis.)

X
Fig. 2.4
No . the two independent relationships which can be obtained from
figute 2. 6 are
x2 + y2 = r2 (2.lla)

tan 6 = I.X (2.llb)

From equations 2.11

2r e~r) = 2x
ax

or ar X
- =- = cos 6
ax r

also sec 2 6 (a6) - L

ax x2

or
a6
-=- L x2 =- L sin 6
----
ax x2 r2 r2 r

46
 To find a transform for a 2 ~/ax 2 remember there are two variables,
so

Hence, using the above results for ar/ax and a~;ax

2.1 = cos e .1 - sin e 2.1

ax ar r ae

and the operator a;ax is

axa = cos
8 ~
ar
_ sin e a
- r - ar (2. 12)

and similarly

a = sin 8 ~ + cos e ~
ay ar r ar
Hence by differentiation

::: = aax(~ = (cos 8 aar _ si~ e aar]( cos 8 ~! _si~ e *)

Carrying out this operation leads to

+ s1n [-
. 2e a 2 ,~,
1 .::......:t.
r2 ae2
+-
r ar
=
1 a,~,)
(2.13)

It can be similarly shown that

a
x
= B
ay2
= sin 2e B -
ar2
2 sin e cos e [_!_ l1 - .!. a 2 ~)
r 2 ae r arae

(2.14)

Also a a2o~.
-- ~
Y ax 2

and

47
 Refer now to the transform equations for obtaining ax ay and
Txy in terms of ar a6 and Tre (see section 1.11).

ay = a9 cos 2e + ar sin 2e + Tre sin2e

(1.22a, b, & c)

Comparing these with equations 2.13 and 2.14 it may be deduced

that

(2 .lSa)

(2.15b)

(2.15c)

Adding equations 2.13 and 2.14

32 41-
ax2
+ tl
ay2
= (cos 2e + sin 2e) ( 32
ar2
f)

that is

ti+ti=.tl+.!.~+...!...tl (2.16)
ax2 ay2 ar2 r ar r2 ae2

and constant (2 .16a)

Finally, the compatibility equation, neglecting body forces is

(2 .17)

This is the stress compatibility equation when using polar

coordinates and 41 is a function of r and e. This equation although
more complicated in appearance than its equivalent rectangular
equation is generally more useful in the results which may be
obtained from it. The transforms from a a and T to a a 9 and
x y xy r
Tre should be remembered if conversions from one to the other are
required.
48
2.4 Solid of Revolution with Axially Symmetrical Loading

The stresses or and a 8 are independent of e and Tre must vanish

for this particular configuration of stress and geometry, since
elements on neighbouring radial planes are subjected to the same
stress at radius r. If there is no change of stress on these
neighbouring elements, then there is no shear.
y

Fig. 2.5

The argument applies to the whole body (figure 2.5). Hence,

where there is no dependence on e, the compatibility condition
reduces to

0 (2 .18)

when evaluated, the resulting differential equation is

(total differential coefficients because there is only one variable.)

The solution of this equation is
~ = Ainr + Br2inr + Cr 2 + D (2 .19)

2C + B(3 + 2inr) (2. 20a)

49
 a = .!._ l1 = ~ + 2C + B(1 + ZR.nr) (2.20b)
r r ar r2

Tr9 = o (2. 20c)

From the polar equations 1.21

giving

u = ~[- ~( 1 + v) + 2 Cr ( 1 - v) + B [ (1 - 3v) r

+ 2 ( 1 - v) ( r Q. nr - r) ] J + f 1 ( 9) (2.2la)

(f 1 (9) is a function of 9 only.)

Also ~
()9
= re 9 - u which after substitution of stresses and inte-
grating gives

v 4Br9
= - E-- I
f 1 (9)d9 + f 2 (r) (2.2lb)

(f 2 is a function of r only.)

For compatibility of strains

Tr9 oU oV
G = Y r9 = 0 = r1 ae + ar - r
~

giving on separating the variables

0 (2.2lc)

0 (2.2ld)

Differentiate equation 2.2lc and then integrate the resulting

differential equation

f1(8) = Mcos9 + Nsin9 (2.2le)

Integrate equation 2.2ld directly

f 2 (r) = Lr (2. 2lf)

(L. M and N are constants of integration.)

50
Finally

u = i[- ~(1 - v) + 2Cr(l - v) + B[(l - 3v)r

+ 2r(l- v)(~nr- 1)]] + Mcose + Nsine (2.22a)

4Bre
v = --E-- - Msine + Ncose + Lr (2.22b)

Equations 2.22a and 2.22b give the displacements for any radius r
at any inclination e. The following however, should be noted:

Equation 2.2le is independent of r. M and N can therefore only

describe translatory movements of the whole body. Consequently M and
N have no effect on the strain and can be eliminated from the dis-
placement equation.

From equation 2.22b, B must be zero otherwise shearing would occur

on radial planes due to the tangential displacement of a point given
by (ra, e = 0, 2~, 4~, etc). L must also be zero since this would
again imply shearing on tangential, and hence radial, planes (for
example, v will be different at r and (r + dr) implying a tangential
movement.)

Thus, for a cylindrical body where the stresses are distributed

symmetrically about an axis there is only a radial displacement
possible.

The relevant equations for axial symmetry are thus

0 (Equilibrium)

(Stress/strain)

y = 0
re

With these equations it is possible to solve the class ~f problems

where the loading and geometry exhibit axial symmetry.

Example 2.3

(a) Thick walled plate

do
r 0
dr +

51
 er du
= dr

ee -ur
eliminate crr and a 9 .

(b) Rotating discs

R now has a magnitude due to inertia forces and

(c) Temperature stresses

Let a be the coefficient of linear expansion of the material. The

equations necessary to solve this problem are

1
er - aT = E( 0r - va e)

1
ee - aT = EC 0 e - vcrr)

du
e r = dr

and u
ee =-
r
where er and e 9 are the total strains and therefore (er - aT)
(e 9 - aT) are strains due to stresses set up in the plate. T is a
function of r only, to maintain axi-symmetrical conditions.

Plane stress or plane strain solutions can be found for (a), (b),
and (c). With (a) the plane stress and the plane strain solutions
work out to be the same but different results will be found for (b)
and (c) where the plane stress solution has been used. (See Chapter
1, section 1.4.)

52
 For plane stress

+ ve - (1 + v)aT]
r

0
r

and for the plane strain problem (applicable to long tubes with re-
strained ends), the stress/strain relationships are

er - aT = t[crr - v(o 6 + oz)]

e6 - aT = t[oe - v(or + oz)]

e z - aT = t[0 z - v(o 6 + or)]

with ez = 0, so that oz

2. 5 Curved Bars

A curved bar subjected to end couples (fig. 2.6) can also be con-
sidered as a problem where there is symmetry about the axis 0 as far
as stresses are concerned, because at all sections the bending moment
is constant.

Fig. 2.6

Hence
A (2.20b)
or = - + 2C + B(1 + 2R.nr)
r2

53
 ~ + 2C + B(3 + 2~nr) (2. 20a)
r2

Tre =o (2.20c)

Also u = % [-~(1 + v) + 2Cr(l - v) + B[Cl - 3v)r

+ 2r(l - v) (~nr- 1~ + Mcose + Nsine (2. 22a)

4Bre
v = -E- - Msine + Ncose + Lr (2. 22b)

However, now B, L, M and N are not necessarily zero since the bar
will be displaced by the couples M.

The boundary conditions are:

Radial stresses are zero at the inside and outside surfaces

a = 0 for r = a and r = b (a)

r

There can be no resultant force over the end of the bar

(b)

By taking moments about the axis 0

(c)

From (a)

A 2Cr + B(l +
-+ 2~na) 0
a2

A
-+ 2C + B(l + 2~nb) 0
b2

r
(b) gives

Ja 6dr = 1
_cp(dr)
d2
dr2
= ~d~
dr b

and this is already satisfied by (a).

54
(c) gives

(a 9rdr Jr(::;)dr
b
or M = - [4>]
a

since the first part of the integral is satisfied by (a) and is zero.
So

A
- + 2C + B(l + 2R.na) 0
a2

A
- + 2C + B(l + 2R.nb) 0
b2

and D in the stress function (equation 2.19) vanishes.

These equations for A, B and C have to be solved for any partic-

ular geometry.

Example 2.4

If a = 100 mm, b 160 mm then

(1 + 2R.nl00) 2 A 0

(1 + 2R.nl60) 2 B 0

160
-R.nlOO M

from which A = 28.199M, B = 1.83 x 10-3M, C = 1.074 x 10-2M and the

stresses can now be found.

For a less stiff bar a = 280 mm, b 300 mm, then

1
(1 + 2R.n280) 2 A 0
280 2

1
(1 + 2R.n300) 2 B 0
300 2

300
-R.n280 -(300 2R.n300 - 280 2R.n280) - (300 2 - 2802) c M

A -212.52M, B = 7.7 x 10-sM, c=7 X 10-~+M.

55
 Having solved for the stress boundaries, it remains to say how
the bar is supported and what boundary conditions can be used to
describe the supports. Using equations 2.22a and 2.22b, A, B and C
are known and L, M and N are not necessarily zero for this problem.
Let the origin of coordinates be e = 0 and r = rm = (a+ b)/2.
Suppose that at e = 0 and r = rm av;ar = 0 and u = v = 0. These
conditions are sufficient to find L, M and N. Alternatively let
(1/r)(au;ae) = 0 and u = v = 0 at this point, then the radial tangent
would remain unaltered.

Further beam problems are discussed in chapter 6.

2.6 Hole in Plate Subjected to Pure Tension

In a tensile field in which there is no hole a 0, T 0 and

y xy
= ;ay 2 and hence

aX =tl=a

a a
X
/ \A}
Y

ar
a

'tre
J

Fig. 2.7 Fig. 2.8

This condition can be applied to the plate shown in figure 2.7
where the effect of the hole will be negligible at a large distance
'b' from it. Assume the distance is defined by a circle radius 'b'
and let the radius of the hole be 'a'.

In polar coordinates (y rsine)

(2. 23)

56
Then the stresses by equation 2.16

a
a2 cp a
= - - = -(1 + cos2e)
r ar2 2

= %(1 - cos2e)

= - ~(sin2e)
2

So for the point defined by ghj at the perimeter of the circle

radius 'b' the stresses are as shown in figure 2.8.

Note that a has two terms

r

or = %+ [%) cos2e
that is, at the boundary of the circle r = b, the stresses consist
of a constant radial stress(cr/2) applied to a thick walled plate
(axial symmetry) and a part which varies with cos2e. For this a
stress function has to be found. The stresses due to both parts for
a~ r ~ b can be superimposed to solve this problem of a plate having
a hole and subjected to tension. For the second part, assume
cp = f(r) cos2e (2.24)
where f(r) is a function of r only.

(Note that or and cr 6 vary with cos2e and Tre will vary as sin2e
using equations 2.16. Thus the assumption that cp = f(r)cos2e is
quite reasonable.)

Use the compatibility equation 2.18 to find f(r)

~
ar2
+ (!)r [lt)
ar
(..!.)
+
r2 ae2
(~) = d2f(r) (cos2e)
dr2

+ (i)(d~~)Jcos2e - (4 fr~))cos2e
Hence solve

57
 2
[:r 2 + (f:)( d~) _ r42][ d::~r) [f:)( dfd~))
+ _ 4:;r)] 0 (2.25)

since cos28 'f 0

Solving by any standard mathematical technique

D
f(r) = A + Br 2 + Cr 4 +- (2.26)
r2

so that
= (A + Br 2 + Cr4 + Q )cos28 ( 2. 2 7)
r2

is a satisfactory stress function since it fulfills all the necess-

ary conditions.
The stresses are, consequently

__ a2~ __ 6D
cr 8 ~ (2B + 12Cr 2 + -) cos28
ar 2 r4

[ 4A + 2B + 60 )cos28
r2 r4

= (- 2A + 2B + 6Cr 2 - 60)sin28
r2 r4

To determine A, B, C and D put in boundary conditions at r b

(equation 2.24)

or [%) cos28
for all values of e

T re = - (%) sin2e
for all values of e

58
at r = a

(J 0
r

Tre = 0 (unloaded boundary)

Hence for a
r
(J
2 = - 2B + 6D)
b'-~

0 = _ (4A + 2B + :~J and for Tre

ra2
-% = (- ~~ + 2B + 6Cb2 - :~)

0 = [ - -2A + 2B + 6Ca 2 - -6D)

a2 a'-~

In matrix form

4
-2 6 A
(J
0
b2 b'"'
2

4 6
2 0 B 0
a2 a'"'

2 6
2 6b 2 c (J
-z-
b2 b'-~

2 6
2 6a 2 D 0
a2 a'-~

Now make the plate very large by making b + oo. Then

(J cra'-+
B - 4 D - -4-

c cra 2
0 A = -2-

Hence

59
 ae = - %( 1 a4) cos26
+ 3=--
r4

a2
a
r = %[ 1 r2
a4) cos2e
- 4=--- + 3=--
r4
(2. 28)

Tre
= f ~- 2a2 - l)sin2e
2l r4 r2

To these are added the stresses due to the constant radial stress
applied at r = b and putting a/b = 0 by making b + =, for example
from previous theory for thick walled plates (see also Young's
Essential Solid Mechanics, page 214)

a A
r

so that with the given boundary conditions at r 0 and

r = b, or = a/2 then

A=Q.
2

and

when a/b + 0.

Thus or = %[ 1 - a:J
r

and a6 = %( 1 + a2)
r2

when added to equations 2.28 give for the total stresses

a2
a
r = %[ 1 - a2) +
r2
%[ 1 - r2
a4) cos2e
4-+ 3=---
r2

(2.29)

T = !!..(~- 2a 2 - l)sin2e
re 2 r4 r2

60
 Equations 2.29 show how local the disturbance is due to a hole
in an otherwise uniform tensile field (figure 2.9).
The greatest effect will be at r = a, then

and aemax = a(l - 2cos2e)

It will be evident that when e w/2

a
emax = 3a

Fig. 2.9
that is there is a stress concentration of three times the nominal
stress, aemax/a = 3 is known as the stress concentration factor
(S.C.F.).

r/a 1 1.1 1.2 1.4 1.5 1.8 2.0 2.5 3.0 3.5 4.0
a Ia 3 2.44 2.07 1.65 1.52 1.30 1.22 1.12 1.07 1.05 1.04
emax'
Table 2.1
Table 2.1 shows the variation of a6 as a/r is increased. It
max
can be seen that for a value of a/r = 3 the stress a 6 is almost
. d
equal to the nominally appl1e max
stress.
Example 2.5

Find the stress concentration factor for an infinite plate with a

hole which is subjected to biaxial stress and pure shear.

max simply put a = e + w/2 in the

To discover the effect of a2 on a 6
expression for aemax then at the periphery of the hole

w
aemax = a1(l - 2cos2e) + a2(l - 2cos2(e + 2 )

61
for biaxial tensile forces.

1. If a1 az = a, then S.C.F. is 2
2. If a1 -az = Ia! a pure shear stress, then
aemax = -4acos2e which gives an S.C.F. of 4 at e = ~/2.
The avoidance of producing stress concentration in engineering
components is clearly desirable since the stress levels can locally
be very high. Stress concentrations are to be found not only at
holes but also at fillet radii where the section of a shaft changes
and the value of the stress concentration factors can vary depending
on the stressing conditions, for example whether in tension or bend-
ing or torsion or a combination of these. Many experimental results
have been obtained, see for example reference ( 6 ) . Stress con-
centration factors can reach.high values and with ductHe materials
the stress concentrations produce yielding of the material and
consequently some spreading of the stresses.
2. 7 Concentrated Forces

The problem to be solved is that of a concentrated force acting on

the edge of a large (semi~infinite) plate. [The problem was first
studied by Boussinesq for a three dimensional body and the two dimen-
sional problem was solved by Flamant (France) and Michell (England).]
Referring to figure 2.10, the concentrated force P acts at a point
on the surface rnm and Flamant and Michell showed that the only stress
at radius r was a radial one provided that the axis of x was coinci~
ent with the line of action of P with the y axis orthogonal to this
to make a right-handed set of axes.

p
m y m m

ur I
X/
Fig. 2.10 Fig. 2.11
This applies even when the load is not perpendicular to mrn (fi~e
2.11). The angle e is measured anticlockwise positive from the x-
axis. The stress function describing the problem is

41 = A resinS (2. 30)

where A is a constant.

62
 This equation can be shown to satisfy the compatibility equation
2.17 and consequently the stresses are (equation 2.15)

ae =ti 0
ar 2

a
r [~) [*) +
( 1 Jfti)
r2 lae2
= ~ose
r

Tre =- 2[ (!)(.?.!))
<lr r ae
=0

Thus, only or has a value and to find the constant A, integration

of the forces around the semi-circle of radius r must equal the
applied force P. The equation for or denotes that as r decreases
the value of or increases so that at the point of contact of the
force the stresses are infinite. This is not possible and it is
assumed that there is a small semi-circle of yielded material in the
vicinity of the force around which a pressure is distributed and
equal to the radial stresses at that radius. Referring to figure
2.12, if t is the thickness of the plate, then resolving forces
vertically
p

Fig. 2.12

(compressive force)

and substituting for or

therefore

63
 It is usual to let P be the load per unit length in the z-direc-
tion., then
p
t = 1 and A =
1T

cp (Pre) . 6
- - s1n (2. 31)
1T

and CJ - (;~)cos a
}
r
(2. 32)
oe = T = 0
re

In this class of problem, the transform equations

CJ o cos 2 e + o6 sin 2 e - Tr 6sin2e

X r

oy o sin 2 e + o 6 cos 2 e + Tr 6 sin2e

r

- CJ 6 )
Txy cr 2 sin2e + Tr6 cos26
(2. 33)
and CJ o cos 2 e + o sin 2 e + T sin2e
r X y xy

oe o sin 2 e + o cos 2 e - T sin2e

X y xy

are often useful.

Referring to equation 2.32 and figure 2.13

d
D
Stress
rrajectory

X
AB=R

Fig. 2.13

64
and at point C where r d and e 0,

At point B1 on the circle of diameter d, r dcose and again

Thus the stress a is constant around the circle of diameter d and

r
acts radially towards the point A. In figure 2.13 the stress traj-
ectories are radial lines and concentric semi-circles and the lines
of constant stress difference (the iso-chromatics) are circles of
diameter d as indicated in figure 2.13.

Thus for a line mm distant D from the edge of the plate the
stresses a x , ay and T
xy can be readily found using the transform
equations 2.33 which reduce to

a a cos 2 e
X r

a a sin2e
y r

T
xy C 2r] sin2e

Now a
r
= - (2P) cos a
11R

and Rcose = D (where R AB fig. 2.13),

therefore

a
r (;~) cos 2e

ax (;~) cos4e

a
y (!~) cos 2esin2e

For a tangential load, figure 2.14, eis measured from line of

action of P which is coincident with the x-axis.

65
 p

y
Fig. 2.14

Then once again

o =- ( 2P)cose
r r'ITr

Again the integral of the resultant of the internal forces along

the semi-circle shown in figure 2.14 gives

J'ITorcoserde = 1T
2P J'IT cos 2 ede = 1T
2P( ~(e + !sin2e) ]'IT p
0 0 0

that is the internal forces balance the applied force and as before

T
re = 0e = 0

For an inclined force the results of the two previous examples can
be used.

I
I
Fig. 2.15

66
 The stress at point m (r, 6), figure 2.15 is

or=- ~~[cosacose +sinacos(I +a)] =- 2P(cosca + e))

That is the angle (a + 8) is still measured from the line of action

of the force P.

X
Fig. 2.16

In figure 2.16 for a couple applied to the boundary the stress

function will be different. The stress function for the forces at
(1) and (2) can be superimposed to give

c~Jr = cl>1 + ciJz

and since the forces are on a plane the stress function can be more
conveniently obtained as a function of x andy. Conversion to r - 8
can then be carried out. Then

cjl
r
= -cjl(x, y + a) + cjl(x, y)

If a is small then the first of the terms can be expanded using

Taylor's theorem.

cp(x, y + a) = cp(x, y) + a[~)

omitting higher orders of differential coefficients. Therefore

*
c~Jr = -cp(x, y) - a(*) + cp(x, y) = - a(~~)

Now = [~!) n;J + (tt) (~~)

and ~ = l. = sine (see figure 2.4)
ay r

ae sine
and ay = -r-

Hence
67
 * = [*)sine + (co:e) [*)
Now ~ = - (P:)esine

*
for a single force P acting on a plane and

=- (P11e)sin2e- [corse)(P;)cecose +sine)

[P11e) sin2e - (p11e) cos 2e - (~) sinecose

p
- nee + !sin2e)

and since

- a..!
ay

then ~r = P:ce + !sin2e)

and with a sufficiently small but keeping P finite, Pa M. Hence

for a concentrated couple M

~ = ~(e
1T
+ ;sin2e)

and e is measured from the vertical from the point of application of

Mas indicated in figure 2.16.

[~)[~) (- 2sin2e)
r

l
- ( ZM sin2e
1Tr2

with the usual restrictions applying to the point of application of

M.

In the foregoing, the coordinate axes have been determined by the

force and not the plane. Simple radial stress therefore applies to
the wedge shown in figure 2.17.

As before cre = Tre = 0 but the radial stress will have the form

68
 or=- (a:)cose

where a is a constant depending on the direction of P.

l'r
x._~~--~-t+~~~~~P- X

y
y
Fig. 2.17 Fig. 2.18

Equating internal and external components of force in the direction

of P gives

1 = a(S + !sin2S)

1
a = S + !sin2S

and 0 = _ Pcose
r r(S +!sin2S)
For P acting at right angles, figure 2.18, replace e by (270 + e)
in ar

Equating internal and external forces gives

P = JS a rdesine = -aP JS cos[ 3; + e)sinede

-S r -S
1
or a = S - ! sin2S

and a
r

Psine
a
r r(S - !sin2S)

and e is now measured as indicated in figure 2.18.

69
 With this equation for a , the stresses a and T can be found.
r y xy
The solution will be applicable to a triangle shaped beam.

2. 7.1 Displacements

For all the foregoing examples the usual equations to determine u

and v are used, that is

Yre = r
au) + av
(aB Clr
_~r = 0

and after integration the resulting constants would have to be

evaluated.

To find these constants the boundary conditions must be carefully

considered. For example, for a point force acting normal to the
edge of a semi-infinite plate, if it is assumed that there is no
tangential movement along the x-axis then when e = 0, v = 0. Note
that any termcontaining r only must vanish since the condition along
the x-axis must be true for all values of 0 < r < oo. Another con-
dition might then be that on the x-axis the radial displacement u = 0
at some distanced (figure 2.18). This will result in having a small
constant depending on the value of d.

2.7.2 UniformLoad

Uniformly distributed load - figure 2.19. The stress function for

this loading is given by

giving

and T 9 = - B

X
Fig. 2.19 Fig. 2.20

70
 The function a = B(r 2 s - r 1 2 s 1 ) applies to the plate shown in
figure 2.20, or

for compressive loading. (These are found by moving the origin of

coordinates and changing the sign of .)

The applications of the foregoing theory are to be found in many

areas of engineering. The problem of applying a concentrated force
is physically difficult as well as theoretically difficult to solve,
but nevertheless in design, corrections can be made to an otherwise
known distribution of stress when concentrated forces are encount-
ered. Herz's problem of contact stress is another important problem
which, in engineering is as important as the contact stresses just
studied, for here applications are to be found for ball and roller
bearings and gear teeth, as practical examples.

2.8 Problems

1. A curved beam has cross-section dimensions 2.5 mm thick x 25 mm

deep. The inner and outer radii of curvature are 100 mm and 125 mm.
The beam is subjected to a bending moment M and a suggested stress
function is

A, B, C and Dare constants. Show that the stress is suitable and

calculate values of a and a 8 in terms of M. Calculate also the
r
displacement equations taking the conditions at one end of the bar to
be u = v = 0 and av/ar = 0 at r = mean radius.

[A= -0.206M, B = -3.3 x 10- 5 M, c = 1.04 x 10- 4 M]

2. Show that = axy 3 + Bxy is a valid stress function. The function

applies to a rectangular plate shown in figure 2.21 lying in the x y-
plane. Calculate the displacements u and v for points A, B, C and D
given that at 0 u = v = au;ay = 0.

[A(u v) [ad 3(i - ~) , 0) ]

y

_dtt-A____s.....F...,.x
d D C
F------~

Fig. 2.21

71
3. The stress function ; = Cr 2 (cos 29 - cos 2a) applies to the wedge
shown in figure 2.22. Derive equations for a 9 , ar Tr 9 u and v
given Tr9 = T and a 9 = 0 at 9 = a. What is the value of C?
[C = _T_]
2sin2a

Fig. 2.22
4. A cantilever plate of length t as shown in figure 2.23 is sub-
jected to an end force P. Calculate using simple theory of bending
the bending and shear stress for a point A(xy). Hence using these
calculations the displacements u and v.

l
v =- 2EI -

Atl.,y)
'
~ X

~
Fig. 2.23 '
5. ~ = Ar2[lsin29- cos29tana + (a- 9)] is a suggested stress func-
tion for the plate shown in figure 2.24. Show that it satisfies the
compatibility equations and if so calculate the stresses at P(r 9)
and find the value of A. If a = 30 sketch a graph for ax at section
aa and compare the results l#ith elementary bending theory.

[A = - 2(a- tana)' a = 18.6q(O.S2 - 9 - !sin29]

x
qjunit len th

y
L
Fig. 2.24

6. Find a relationship between K and L to make ; = Kx 4 + Lx 2y2 a

valid strezs function. Hence find an expression for u and v and
sketch the boundary stresses.

[u Constant - 6EKl~3 (1+2v) - vxy2]]

72
7. The stresses at a point Q(x, y) in the cantilever shown in figure
2.25 are

cr 0
y

xy

Calculate for the point Q the displacement equations u and v making

any reasonable assumptions to determine the constants of inte-
gration. Hence, determine the strain on the central axis and the
displacement of the point 0, on the central axis, in the y-direction.
p
t'-' [x = 0, y = 0, ex = 2c]
0 ~ X
Q(X,Y)
~
I:'
L j
y I

Fig. 2.25

73
3 TORSION

Most students will already be familiar with the simple theory of

torsion of circular shafts, which is based on the simplifying
assumption that plane cross-sections of the shaft remain plane and
parallel to one another during straining. In turn, this means that
radial straight lines drawn on the cross-section remain straight
during straining, and that the shear strain is proportional to
radius.

A little thought will show that the above assumption simply is

not true for shafts of non-circular cross-section. For example, a
square prism of rubber when twisted in the hands would have its
originally plane end faces warped, showing that out of plane
displacements had resulted from the twisting. Torsion theory for
a three-dimensional body which allows for 'warping' of the cross-
section is developed here. A set of displacements is assumed which
means that compatibility conditions need not be considered, and
after satisfying the equilibrium equation and boundary conditions a
full solution is obtained. The original solution is due to St.
Venant.

3.1 Displacements, Strains and Stresses

y
y

Fig. 3.1

The shaft under consideration is assumed to be prismatic. The

origin of coordinates is at the fixed end and a couple T is applied
to the free end in the sense illustrated in figure 3.la. The angle
of twist per unit length is a so that the total twist for a shaft
of length i is ia. Consider a cross-section of the shaft at z from
the origin so that the angular displacement of a line OA drawn on
the cross-section is az, see figure 3.lb. The displacement of A .is
to A' where AA' is raz and the x component of displacement is

u = -AA'sine = -azy
74
since

sine .L
r

Similarly the y component of AA' is

v = AA'cose = azx
since
X
cose
r

The displacement w in the z direction is less easy to specify.

However, it seems reasonable to assume that w will be independent
of z but will depend on position within the cross-section. It is
also reasonable to assume that w will depend directly on the angle
of twist.

Therefore, the full set of displacements within the boundary of

the shaft are assumed to be

u = -azy
v = azy

w = al/J(x, y)
) (3.1)

where 1/J = 1/!(x, y) is some function of x andy (to be determined) and

is called the warping function. Using these displacements and the
definitions of chapter 1, the strains are easily deduced as

au 0, e av = 0 e aw = 0
e ay , z az
X ax y

au :lv
Yxy -+ 0
ay ax
aw= a(:ll/J - y)
:lu+ -

)
Yxz = -
:lz ax :lx
(3. 2)
= :lv + :lw = a(~ + x)
Yyz az :ly :ly

and the non-zero stresses are simply

Gy XZ Ga(~ - y)

}
T
xz ax
(3. 3)

T Gyyz Ga(~
:ly
+ x)
yz

75
3.2 The Equilibrium Equations
Assuming that 'body' or ~cceleration forces are zero, the
equilibrium equations in three dimensions are

and, remembering that only T and T are non-zero and independent

xz yz
of z these reduce to

This equation can be satisfied by introducing a stress function of

~ ~(x,y) such that

ll-
ax - -T
zy (3.4)

Also, from equations 3.3, the unknown function w(x, y) can be elim-
inated by performing the operation

that is

tl + tl = -2Ga (3. 5)
ay 2 ax 2
Thus, if a function ~ can be found to satisfy equation 3.5 the
equilibrium equations are also satisfied and the stress distribut-
ion can be found from equation 3.4. However the solution will not
be complete, or correct unless the boundary conditions are satis-
fied.

3.3 The Boundary Conditions

3.3.1 The Parallel Surface of a Bar

The direction cosines of the outward normal N, to an element of

surface ds are

76
 dy dx
R. = ds' m - ds' n =0

Fig. 3.2

The negative sign arises because x decreases as.s increases when

proceeding around the periphery of the cross-section in the
positive (anticlockwise) direction. n = 0 because the angle
between N and the z axis is zero.
As there are no external forces on the parallel surface of the
bar, the boundary conditions (equations 1. 9) reduce to the single
equation
R.-r
xz
+ m-r
yx
=0

or T
xz ds)
(2.r.1 _ T yz (dx1=O
ds)
Using equations 3.4 to substitute for -r xz and Tyx , this becomes

2.1.~ + .!!.. dx
ay ds ax ds
=.!!.!.
ds
=0

that is

(3.6)

That is, the value of~ must be constant around a boundary. where
the shaft has no holes, the cross-section has only one boundary,
and the value of ~ at the boundary can be chosen arbitrarily. It is
usually conveniently made equal to zero. However, where the shaft
has holes, and therefore more than one boundary, equation 3.6
applies to all boundaries but the constant value of ~ can be chosen
arbitrarily for only one of them. It follows that an extra equation
is needed for each extra boundary in order to fully define ~-
The remaining boundary equation described above indicates that
resolution of forces normal to the boundary gives a zero result.
Resolution at right angles (that is parallel to the boundary) must
give the total shear stress, see figure 3.3.

77
 Fig. 3. 3

Thus, resolving forces parallel to the boundary

T = T yz ~ - T
xz
m

using equations 3.3.

Now form the integral

Since the integral is taken around the boundary under consider-

ation and the function w(x, y) or w = aw(x, y) must be single
valued on the boundary the first two terms in brackets in the
above expression must be zero. Also J xdy = A and -~ydx = A where
A is the area enclosed within the boundary. Thus

JTds = 2GaA (3. 7)

For each boundary (other than the outside, where d~/ds = n,

~ = 0)
an equation like 3.7 can be written to find the constant
value of~ for that particular boundary.

3.3.2 The End Surfaces of the Bar

If the theory outlined above is correct and complete, summation of

forces at the ends of the bar should result in simply a torque,
with no resultant force. The magnitude of this couple will be
equal to the applied torque T.

At the ends of the bar the direction cosines are

~ 0

m 0

n = 1

78
 y

Fig. 3.4

Considering the end at z = t, then n = 1 and the force in the x

direction

x = fdx = JJ< xz dxdy

since ~ has the same value at y = y 1 and y2 (~ is constant around

the boundary).

Similarly
y =0
By moments about the origin 0, figure 3.4

T = JJ T
yz x dxdy - JJ T
xz y dxdy

where

JJ 'yzx dxdy = - Jf (*)x dxdy = - J[~x - f ~dx]dy

by parts, and similarly

-ff 'xzy dydx = -/[~y- f ~dy]dx

The first term in each of the above expressions must be zero
since ~ = 0 at the outer boundary. The expression forT then
reduces to

79
 T = 2 ff 4> dxdy (3.8)

If the cross-section of the shaft lies in the x - y plane and 4>

is thought of as a series of ordinates erected within the boundaries
of the cross-section then the tips of the ordinates will form a mound
or hill and the integral J/ 4> dxdy is clearly the volume within the
mound. That is, the torque is equal to twice the volume under the
'torsion hill'.

This visual idea of the distribution of 4> as a 'torsion hill' is

very useful, since the slope of the hill in' either direction (a4>/ax
or a4>jay) gives the shear stresses according to equations 3.4. More-
over, points of maximum slope (maximum shear stress) can often be
easily located using only experience and intuition.

Example 3.1

Fig. 3.5

The equation of the boundary of the shaft of circular cross-section

shown in figure 3.5 is x 2 + y2 = a 2 and choosing a trial stress
function

ensures that on the boundary, d4>/ds = 0 and 4> = 0. m is some constan~

still to be evaluated. Substitution of this expression for 4> into
equation 3.5 gives
Ga
m =- 2

and into equation 3.8 gives

T = Ga JJ (x 2 + y 2 - a 2)dxdy

Ga(lyy + lXX - Aa 2 ]

80
or T = + GaJ

where J = na~/2. the polar second moment of area of the shaft cross-
section. and

T = ~ = - Gay
XZ ay

T ~-
ay - Gax
yz
The t0tal shear stress at any point P (x. y) within the boundary
where x2 + y 2 = r2 is

T sine + T cose Gar

zx zy

Tr = Tzx cose + Tzy sine 0

as expected.

The distortion of the cross-sections can be found from equations

3.3, written in terms of w, the displacement in the z direction.
Thus

Txz - Gay = G[~: - ay]

T
yz
Gax G[~; + ax)

whence aw;ax = aw;ay = o. w = constant.

This seemingly trivial example of the circular shaft has been

chosen to illustrate the point that w is constant for all cross-
sections. The assumption made in the elementary theory is there-
fore justified - cross-sections of the shaft remain plane after
straining. The circular cross-section is however unique in this
respect - all other shapes of cross-section exhibit warping.
The student will by now appreciate that all that needs to be
done in order to solve the torsion problem is to write down an
equation which respresents the boundary of the cross-section in
the form f(x. y) = 0 and then choose a trial function~= mf(x. y).
This procedure ensures that d~/ds = ~ = 0 at the boundary and the
constant m can be found by substitution in equation 3.5. For
example, for an elliptical shaft a suitable function would be

81
see figure 3.6 and it is left as an exercise for the reader to
complete the solution

Fig. 3.6
However the number of useful solutions that can be found in this
way is limited to a few simple shapes. More complicated shapes can
be dealt with by superposition of simple solutions but the modern
trend is to use numerical methods, which are dealt with in chapter
5. There is one other class of problem that can be solved analyt-
ically by the use of analogues, namely the torsion of thin walled
sections. Analogy methods are dealt with in the next section.
3.4 Analogies to the Torsion Problem

3.4.1 The Hydrodynamic Analogy

The equation of continuity for the flow of an 'ideal' fluid (that is,
constant density and zero viscosity) in two dimensions is

(3.9)

where u and v are the components of velocity in the x and y direction

respectively and if it is assumed that the rotation or vorticity is
constant a second equation connecting the velocities is

av - -au = constant
-ax ay (3.10)

(Constant vorticity in this context has the physical meaning that a

straight radial line drawn in the surface of the fluid, remains
straight. That is, the angular velocity of a fluid element is
independent of its position.)

Writing u = a~/ay and v= -<l~/ax satisfies equation 3.9 and

equation 3.10 becomes

~a2"' + a2"' -- - constant

~
ax 2 ay 2
which is identical to equation 3.5. The gradient of ~ in a partic-
ular direction gives the velocity in a direction at right angles.
At the boundary the velocity must be parallel to the boundary and
the component of velocity normal to the boundary must be zero.

82
Clearly then, d~/ds = 0 and there is a perfect analogy between ~
as defined above and ~ as defined by equations 3.4.

Much of the value of the hydrodynamic analogy comes from visual-

ising the flow pattern and translating it into shear stress distr-
ibution. The velocity distribution in the hydrodynamic situation
is identical to the shear stress distribution for a shaft of the
same shape.

Fig. 3.7

For example, for a shaft with a keyway, figure 3.7, if one

imagines the simple fluid rotating within the cross-section, then
points A will become 'backwaters' with low fluid velocity (low
shear stress) whilst in the sharp corners (points B) the velocity
and hence the shear stress tends towards infinity.

The material of the shaft will always yield therefore at the

sharp internal corner of a keyway when the shaft is subjected to
torsion. Simiiarly, for a shaft of triangular cross-section it is
apparent that the corners will be areas of low shear stress and the
maximum shear stresses (velocities) will occur at the mid points of
the sides.

3.4.2 The Membrane Analogy

Consider a uniform elastic membrane sup?orted around the periphery

of a hole which has the same shape as the cross-section of the
shaft subjected to torsion. The membrane is stretched by a
uniform tension P/unit length and the supporting edges are assumed
frictionless so that the tension in the membrane within the bound-
ary is approximately P/unit length also. A pressure p is then
applied to one side of the membrane so that it is deflected in the
positive z direction (see figure 3.8) but the deflections are
limited so that the slope of the membrane in any direction is small
and the curvatures are given by
1 a2w
R=
xz ax 2
(3.11)

83
where Rxz is the radius of curvature in the xz plane and
similarly for R
yz

Fig. 3.8
For vertical equilibrium of the element dy dx in figure 3.8b
the upward component of force due to P in the xz plane, figure
3.8c is

Pdy sin(e + de) - Pdy sine ~ P dy de

if e and de are small.

Now de = Rdx = - (a 2w) dx

xz ax 2
A similar expression for the yz plane gives the total upward force
as
- p dydx (a2w + a2w)
ax 2 ay 2
and this must be equal to the downward force due to the pressure on
the element p dydx, whence

a2 w + a2w = _ ~ (3.12)
ax 2 ay 2
w and dw/ds are both zero at the boundary so that the deflection of
the membrane is exactly analogous to ~ in the torsion problem.

The membrane analogy lends added appeal to the 'torsion hill'

idea mentioned earlier and it is perfectly feasible for experi-
ments to be carried out using suitable membranes (for example,

84
a soap film) and measuring the shape of the hill. However the
membrane analogy can also be used in an analytical way to solve
the problem of the torsion of thin walled sections.

3.5 Torsion of Thin Walled Sections

'Thin walled' sections are those where, typically, the wall thick-
ness is less than one-tenth of the smallest other dimension of the
cross-sectio n. They may be of 'open' section, that is having a
single boundary, or 'box' section, that is having more than one
boundary. The fundamental assumptions about the shape of the
membrane which lead to the solution are different in each case.

3.5.1 'Open' Sections

Consider the shaft of narrow rectangular cross-sectio n shown in

figure 3. 9.
a
p

z
y
Fig. 3.9

It is assumed that the membrane deflects to the shape shown in

figure 3.9b and that the shape remains constant even at the ends.
The assumption is reasonable if a << b because then the proportion
of the volume under the membrane which is influenced by the change
of shape at the ends will be small.
An equation for the shape of the membrane can be deduced by
considering the equilibrium of a portion x wide and unit length in

per
they direction- figure 3.10.
H

Fig. 3.10

Whence
tane = - dw = ~
dx H

where H is the tension/uni t length at x = 0. Equilibrium of the

half membrane when x = a/2 requires that

85
that is

H ~ p

if a is small. Then

dw _e_ ( 3.13)
dx = p

Integrating gives

px2
w =- ~ + constant

and using w =0 at x = a/2 results in

Maximum deflection at x 0 is pa 2 /8P and the volume under the

membrane is

pba 3
12P

To obtain the results for torsion simply replace p/P by 2Ga in

the above equations, whence

2Ga X (3.13a)

and 'max Ga a at the centre of the long sides

(3.14a)

and the torque =2 x volume under the hill is given by

a3
T=Gab3 (3.15)

The above results can be extended to apply to the more complic-

ated shapes of cross-section shown in figure 3.11

For figure 3.lla simply replace b in equation 3.15 by nd to get

a3
T = Ga nd 3

86
 (a) (h) (c)

Fig. 3.11

and as before Tmax = Ga a

For the section of figure 3.llb

Ga(b 1a1 3 + b2a2 3)

T
3

and assuming that a2 > al

and similarly, for the I section, figure 3.llc

T
3

These expressions give results which are sufficiently accurate

for most engineering purposes but the expressions for Tmax are
approximately true only at points which are distant from the ends
of the cross-section. In addition, the sharp re-entrant corners
in the sections of figures 3.llb and c give rise to stress concen-
trations which may result in a value for the shear stress several
times greater than Tmax calculated from the above expressions. The
appropriate stress concentration factor will be derived in section
3.6.1.

3.5.2 'Box' Sections

Again the wall thickness is assumed to be small but box sections

have two or more boundaries. The torsion equations require that ~
is constant around each boundary and to make the membrane analogy
comply with this requirement imagine that a rigid plate, which is

87
the same shape as the boundary, is attached to the membrane at the
appropriate position and is free to move in the z direction but is
constrained to remain parallel to the xy plane.

(a)

(h)
rigid plate

Fig. 3.12

A hollow shaft (two boundaries) of small but variable wall thick-

ness t is shown in terms of the membrane analogy in figure 3.12.
The rigid plate ensures constancy of w around the hole and the
fundamental assumtion in the application of the analogy to this type
of cross-section is that the slope of the membrane is constant
across the wall thickness at any point.
Thus
h
T =- (3.16)
t

The true shape of the membrane is more likely to be as shown

dotted in figure 3.12b but so long as h is small and the wall
thickness is also small the approximation is not seriously in error.
The student should note the fundamental difference between this
approach and that used for the open cross-sections of the previous
section 3.5.1.

The volume under the membrane is very nearly Aph where Ap is

the area within the mean perimeter. Thus

T = 2A h (3.17)
p

For a complete solution a further equation is required. Using

equation 3.7 for shafts with holes gives

/T ds = 2Ga ~ (3.18)

where Ah is the area within the boundary of the hole. An example

will make the application of equations 3.16, 3.17, and 3.18 clear.

88
Example 3.2

A thin walled box section is shown in figure 3.13a and figure 3.13b
shows the assumed displacement of the membrane.

(a)

.....~.-+----r---+-+----:T--+J~~section
through BB rb)

Fig. 3.13
The thickness t of the wall is constant throughout the section,
so that applying equation 3.16, the shear stress in the part QRLM
of the wall is

hl
TI = t-

in MNPQ
h2
T2 =t- (a)

and in QM
hl - h2
T3 = -.....,t-- = T! - 'r2

These equations also reflect the hydrodynamic analogy - fluid

flowing around QRLM in the direction shown by the arrows with
velocity v1 must divide at M to flow along MNPQ with velocity V2
and along MQ with velocity v3. As the channels are all the same
width, continuity requires that

that is velocity is analogous to shear stress. The direction of

'flow' and hence the direction in which the shear stress acts is
important when applying equation 3.18. In general when applying

89
the hydrodynamic analogy to deduce directions it is the quantity
Vt which is constant.

If the wall thickness is small the sum of the areas of the holes
will be approximately equal to the area enclosed by the mean peri-
meter. Equation 3.17 therefore gives the torque

T (b)

There are three possible boundaries to which equation 3.18 can

be applied. Proceeding in the anticlockwise (positive) direction
around each boundary gives

for QRLMQ

for MNPQM

2Ga A2 (c)

and for QRLMNPQ

(Note that T3 is negative if proceeding around boundary MNPQM in

the anticlockwise direction.)

Although three equations have been written, only two are indepen-
dent - it can be seen that adding the first two equations gives the
third, which is to be expected. The usual procedure in a numerical
problem would be to write equations (c) in terms of h 1 and h 2 using
equations (a) and solve for h 1 and h 2 using any two equations (c).
The result is then substituted in equation (b) to find the torque in
terms of the angle of twist.

Of course, the sharp internal corners will give rise to stress

concentrations which means that the shear stress at the corners will
be greater than the calculated values of T1, T2 and T3.

To summarise, for box sections the relevant equations to be

satisfied are

h
T = t
T = 2 x volume under membrane

and jTds = 2Ga A

3.6 Stress Concentration in Thin Walled Sections

At a sharp internal corner in a thin walled section such as that

illustrated in figure 3.14 it would be reasonable to assume that
the shape of the deflected membrane was a solid of revolution, at

90
least over a small region subtending an angle 9 as shown.

Fig. 3.14

Transforming equation 3.5 into polar coordinates gives

and if the surface represented by the ordinate ~ (the surface of the

analogous membrane) is a solid of revolution, this reduces to

and since - d~/dr = Tez then this becomes

dTer Tez
--" + -- = +
dr r
2Ga (3.19)

whence, on integrating
T
ez
= ~r + Ga r (3.20)

The calculation of the constant C depends upon whether the shaft

cross-section is of the 'box' or 'open' type because the assumed
shape of the membrane is different in each case.

3.6.1 Open Sections

At some distance from the corner under consideration the maximum

shear stress in the wall will be approximately the same as that
found for a rectangular shaft, that is

T Ga t
0

so that equation 3.20 may be written

91
 (3.21)

and the assumed shape of the membrane along section ZZ is

approximately as shown in figure 3.15

r t .,

~
Section AA Section ZZ
Fig. 3.15

The exact position at which d~/dr = 0 cannot be easily deter-

mined, but as only an approximate result is required, it is
sufficient to assume that this occurs at r = a + t/2. Then
T ( )2
C = - ; a + ~)

and equation 3.21 becomes

This gives for the stress concentration factor, S.C.F., at r =a

S.C.F. = -Tazl
T
I
r=a
= 1 +t-
0
4a
(3.22)

The student should notice that as a tends to zero, the S.C.F.

becomes very large.

3.6.2 Box sections

For the case of box sections, the shear stress at some distance
from the corner is given by equation 3. 7, that is given by equation
3.7, that is

2GAa

where A is the area enclosed by the boundary of perimeter s. Using

this in equation 3.20 gives
C TOS
T6z = r + 2A r (3.23)

92
and the assumed shape of the membrane across section zz is as shown
in figure 3. 16.

Section AA Section ZZ
Fig. 3.16

In the hydrodynamic analogy shear stress is analogous to velocity

and continuity requires that the flow through section AA is equal
to that through section zz, therefore

and substituting from equation 3.23 into this equation and inte-
grating gives

For a thin walled section the quantities s(2ao + o2 )/4At and sr/2A
will be very small and can be neglected, whereupon on substituting
for C, equation 3.23 becomes
T t
0
Tez

Again, as only an approximate result is required o~ t and the

stress concentration factor is given at r = a as

Tez t
S.C.F. =-=--...:;.__ _ (3.]4)
T
0
a tn( 1 + ~)
Equations 3.22 and 3.24 give the approximate stress concentration
factors at re-entrant corners in thin walled open and box sections
respectively. These formulae are sufficiently accurate for prelim-
inary design work but more accurate values, usually determined by
numerical methods, can be obtained from appropriate reference works.

93
3. 7 Problems

1. Show that the function

<P = m[9a(x 2 + y2 ) + 6/3(3xy 2 - x 3) - a3]

satisfies the torsion equation 3.5 and the boundary conditions

equation 3.6 for a shaft of equilateral triangular cross-section
of side a, as shown in the figure.

Hence calculate the value of the constant m and plot the

distribution of shear stress along the line AX.
y

2. Show that the stress function

(due to Prandtl) is a solution of the torsion equation for the

shaft of cross-section shown below. Find the value of m and sketch
the shape of the 'torsion hill' giving key values.

r- 2acoso

3. Complete the solution for an elliptical cross-section shaft,

suggested in the text, and find an expression for the displacement
w in the z direction.

4. For the box section shown below calculate the maximum shear
stress in the material remote from the corners and the twist per
metre length of shaft when the applied torque is 3000 Nm. Calculate
also the minimum radii r1 and r 2 required to limit the maximum
shear stress at a corner to twice the maximum value already

94
calculated. Take G 0 82 GN/m 2

.024 radians/metre

83 N/mm 2 around area A2

1.15 mm

0.8 mm

5. Show that the central web in a symmetrical box section carries

no shear stress. What may be its purpose?

6. A tube of cross-section shown below is subjected to a torque of

2500 Nm. Calculate the shear stresses in the tube walls and the;
angle of twist per unit length. Take G = 82 GN/m 2 .

n
a = .028 radians/m

Tl 95.5 N/Jillll2

..
5 5
0
II)
T2 159.2 N/mm2
T
T3 35.3 N/mm 2

100 R 25 Tft 100.1 N/mm 2

95
4 ENERGY METHODS

The concepts of work and energy are fundamental to the methods of

analysis of many problems in the general field of applied mechanics.
In the more specialist field of mechanics of materials, these methods
are given less attention than they merit because the results they
give are often 'approximate' rather than 'exact' solutions. However,
it should be remembered that the approximate nature of the results
arises from the assumptions made when applying the concepts, rather
than from any failing in the concepts themselves.

The advent of the digital computer and the finite element approach
has made possible the application of energy methods to continuum
problems as well as the more complex structural problems. The object
of the chapter is to review the principles of minimum potential en-
ergy and virtual work and to prepare the ground for the finite
element method detailed in chapter 5.

4.1 Potential Energy

4.1.1 Conservative Forces

The student will already be familiar with the concept of potential

energy as work which a body is capable of doing by virtue of its
position in a gravitational field - a body of mass m at a vertical
height h above some datum in the earth's gravitational field has
potential energy mgh. The definition of conservative force enables
a wider and more general meaning to be attached to potential energy.
If a particle moves from A to B under the action of a force F which
has components X, Y and Z then the work done by F is
B
w J (Xdx + Ydy + Zdz) (4 .1)
A

X
Fig. 4.1

If the result of the integral in equation 4.1 depends only on the

coordinates of A and B then the work done is independent of the
path followed by the particle between A and B. The force F is then
called a conservative force. Moreover if the particle is now ret-
urned to A by any path, including the reverse of the path already
take from A to B, the net work done by F will be zero. Thus, points

96
like A and B can be associated with definite values of W; for any
position of the particle P the force F has a particular capacity
for doing work which is called its potential energy.
Potential energy is then defined by

or, if A is taken as the datum for measurement of potential energy

vb = -W =- I:(Xdx + Ydy + Zdz) (4.2)

One example of a conservative force has already been mentioned,

that due to gravity. In solid mechanics conservative forces arise
from the deformation of elastic (not necessarily linear) bodies. If
the material of the bodies is wholly or partially plastic or if
there is internal friction giving rise to hysteresis then the forces
are non-conservative.

4.1.2 Potential Energy and Elastic Bodies

In the previous section, potential energy was defined in terms of a

single force acting on a single particle. It is possible and more
useful to have a definition of potential energy for an elastic body
under the action of several forces. From now on the term 'force'
will generally include 'couple'.
Consider the simple spring shown in figure 4.2,initially un-
stretched under the action of forces F1 and F2 which may vary
with time. In a certain period of time end 1 is assumed to
move from A to C whilst end 2 moves from B to D. Then using
equation 4.1

where ~ is the change in length [(D - C) - (B - A)] and~

d1, whence

97
 A D
fL._J. T

Fig. 4. 2

The student will recognise the last term of the above equation
as the energy stored within the spring, or its strain energy. In
general the total potential energy of a loaded elastic body is
defined as

V = the strain energy of the body - the work done

by the external forces (4. 3)

Notice that, in this section nothing has been said about equil-
ibrium or the linearity of the spring. In fact the above definition
holds whether or not the body is in equilibrium and whether or not
the spring is linear. It is only necessary that the spring be
elastic.

4.1.3 The Principle of Minimum Potential Energ~

Suppose that the components of the resultant force acting on a body

are X, Y and Z (and also the resultant couple, if any, can be
resolved into components about the X, Y and Z axis). Differentiation
of equation 4.2 gives

1_Y = X
ax

av = Y etc
ay

and if the body is in equilibrium (that is X Y z, etc. 0) it

follows that

av = 0

}
ax
( 4 .4)
av = 0
ay etc. at equilibrium
That is, for a body in equilibrium, the potential energy has a
stationary value. It might be concluded on purely intuitive grounds
that a position of maximum potential energy corresponded to a
position of unstable equilibrium, a position of minimum potential
energy corresponded to a position of stable equilibrium whilst a
point of inflection on the potential energy curve indicated a
position of neutral equilibrium (figure 4. 3).

98
 Stable Unstable Neutral

Types of Equilibrium

Fig. 4.3

(a) Degrees of freedom It has already been shown that potential

energy of a particle acted upon by conservativ e forces depends only
on its position. However many systems are completely specified
kinematical ly by the use of only one coordinate, for example a
simple pendulum is completely defined when the angle of inclination
8 is specified (figure 4.4a) whereas the rigid bar in figure 4.4b
requires two coordinates to completely define its position, either
~1 and ~2 or ~1 and 8.

(a) ~)

Fig. 4.4

The degree of freedom of a body or system of bodies is therefore

defined as the minimum number of coordinates needed to completely
specify the configurati on of the body or system. These coordinates
are sometimes called generalized coordinates .

(b) Conservatio n of Energy From Newton's Laws of Motion it follows

that the work done on a body or system of bodies is equal to the
increase of kinetic energy.

dW = d(KE)

but from equation 4.2

dV = -dW = -d(KE)
or d(V + KE) = 0 (4.5)

99
which is a statement of the well known principle of conservation of
energy. The kinetic energy of a system of particles can be written
as
KE = ~(Em)VG2 + !E(mV2)

where VG is the velocity of the centre of the mass of the system

and V is the velocity of each particle relative to the centre of
the mass. If such a system could be represented by one coordinate
u (single degree of freedom) then

The potential energy V = V(u)

The kinetic energy KE = ~M(u)u 2

where M is the generalized mass referred to the coordinate u and
the energy equation becomes

V(u) + ~M(u)u 2 = constant (4.6)

(c) Stability of equilibrium - one degree of freedom. Different-

iating equation 4.6 with respect to time gives

v (u)u + M(u)uli + !M' (u)u 3 = o

or, dividing by li

V' (u) + UM(u) + ~'(u) 0 (4. 7)

where
d(V(u))
V' du

and M' = d(M(u))

du

It has already been shown in section (a) above that dV/du = 0

represents a position of equilibrium. In fact there may be several
solutions to the equation dV/du = 0 each representing a different
equilibrium position. Assume that one such position is given by
u = u1, so that V'(ul) = 0. Consider now a small displacement z1
from the equilibrium position so that u = u 1 + z1 or z1 = u - u1
and write equation 4.7 in terms of z1 and u 1, using Taylor's Series,
then

+ ~dM(ul) + z1M' Cu1) + [~1 )~" (ul) + }

(~l)~M'(ul) + z1M"(u1) + r~ 2 JM'"(ul) + } 0

100
since

Now since z1 is small, ZJ and ~1 are also small and also V1 (u 1 )=

0, so this reduces to (neglecting second order terms)

(4. 8)

Moreover, for the given position u = u1, V''(ul) and M(ul) are
constants. Thus, there are three solutions to equation 4.8 as
follows.

When V''(ul) is positive the solution has the form

V' '(Ul)
z1 = Acospt + Bsinpt with p2 = ~~~~
M(ul)

This implies that V'(u 1) = 0 represents a m1n1mum value of the

potential energy; the equilibrium position is stable because a
small displacement z1 from the equilibrium position u results in
an oscillation of amplitude z1 with simple harmonic motion about
the equilibrium position.

When V''(u 1) = 0 the solution is z 1 = 0, and there is therefore

no force tending either to restore or destroy equilibrium. This
indicates a position of neutral equilibrium. When V''(ul) is
negative, the solution has the form
Zl = Aept + Be-pt

Here the potential energy is a maximum and the equilibrium is

unstable.
The principle of minimum potential energy can now be stated.

Of all the assumed positions which a body or system of bodies may

occupy the position which makes the potential energy a minimum
represents a position of stable equilibrium.

The word 'may' in the above sentence is important. The body or

system of bodies may only assume positions which are compatible
with material continuity and the boundary conditions or external
constraints on the system.

Example 4.1

Simple tie bar - one degree of freedom, one equilibrium position.

The tie bar is fixed at one end and stretched by a force F so

that the configuration is completely determined by the displacement
x of the free end. From equation 4.3

101
 V = EA x2 - Fx
2R.

where the strain energy has been written in terms of the extension
X.

~~======~=
1111
c.s.areaA
1 x... ,
==~=~~
...
Mat'IA
a

Fig. 4.5 Fig. 4.6

Di fferenti at ing

dV = O EAx F
dx -R.--

FR.
or X = AE

gives the equilibrium position, and differentiatin g again

EA
T (positive)

shows that the equilibrium is stable.

Example 4.2

Light rigid bar on two wires initially of equal length - two degr-
ees of freedom.

In this example there are two degrees of freedom, since only two
from XI, x 2 and 8 are needed to completely specify the position
of the rigid bar.

Choose xi and x 2 as the coordinates whence the displacement of W

is (xib + x 2a)/(a +b) and equation 4.3 gives
EAAAxi2 EBABx22
v 2R.
+
2R.
- wciba + :2a)
+

3V Wb R.
dXI = 0 gives XI (a + b) EAAA

and~= 0 gives xz + Wa R.
ax 2 (a + b) EBAB

102
and again, a second differentiation reveals that this equilibrium

l
position is stable. The solution to this problem could be written
in matrix form as

EAAA
-r;: 0

! x1 !(a: b)) 0 (2)

EBAB
0
t;- x2 (a + b)

The terms in the square bracket represent the stiffness of the

wires and the terms Wb/(a + b) and Wa/(a + b) are the loads on each
wire or 'generalised' loads.
The above equation is a stiffness equation and the square bracket
is a stiffness matrix.

Notice that the introduction of a third wire (making the problem

statically indeterminate) does not add to the difficulty. There are
still two degrees of freedom and the solution is found in exactly
the same way as above. A different choice of coordinates would, of
course, lead to a different stiffness equation for the same situa-
tion. For example, on choosing x, and e to specify the displace-
ments, equation 4.3 gives

and the solution becomes, in stiffness equation form

This gives the same equilibrium position as the previous solution

but in a different form. Notice however that in both cases the
stiffness matrix is symmetrical.

Example 4.3

Ladder retained by spring - one degree of freedom, two equilibrium

positions.

Consider the uniform ladder shown in figure 4.7 which is re-

strained by the horizontal spring of stiffness K. The spring is
unstretched when the ladder is vertical. Assume that the reactions
at floor and wall are frictionless.

103
 Fig. 4. 7
Equation 4.3 can be written in terms of the single degree of
freedom e

V = K( 6sine) 2 - 4W(l - cose) (1)

2

dV
de =0 gives 0 36Ksinecose - 4Wsine

that is

sine = o, e o
or cos e = 9wK
Thus the two possible equilibrium positions are given by e 0
and cose = W/9K.
Differentiating 1 again gives

36Kcos2e - 4Wcose

putting e = 0 in this gives

36K - 4W

which is positive. (36K must be greater than 4W or the second

solution in meaningless.) e = 0 is therefore a position of stable
equilibrium.

104
 Putting cose = W/9K gives

36K(2cos 2 e 1) - 4Wcose

This time d 2 V/de 2 is negative and the equilibrium position is uns-

table.

4.1.4 Continuum Problems

After working through example 4.2 above, it will be clear that, if

there are n degrees of freedom (that is, n coordinates are needed
to completely describe the system) there can be derived n equations

which can then be solved for the n unknown equilibrium displacements.

However in continuum problems the number of degrees of freedom is

infinite, corresponding to the infinite choice of values for x, y
and z within the body or system of bodies being studied. The problem
may be reduced to one having a finite number of degrees of freedom
by making approximations. In general, the greater the number of
degrees of freedom included in the approximation the greater will be
the accuracy of the solution. Before discussing this method it is
necessary to generalise the expressions already obtained for work
done and strain energy to three dimensions.
(a) General expressions for strain energy and potential energy.
For a linear elastic material the six independent components of
stress are
{a} = {a a a T T T }
x y z xy yz zx

and the corresponding strains are

The strains are defined in terms of the displacements

{f} = {u, v, w}
au av av au
(ex= ax' ey = oy , " . Yx = ax+ oy " . etc.)

The stress strain relation written in matrix form is then

{cr} = [E] {e} (4.9)

105
where [E] is a six by six symmetrical matrix of material constants,
and the general expression for strain energy/unit volume becomes
U = ~{e}T {a} = ~{a}T {e} (see note below)
Substituting for {a} from equation 4.9 gives
U = ~{e}T [E] {e}

and for the whole body

U = JJJ~{e}T [E) {e} dxdydz ( 4 .10)

Note that { } represents a column vector althoufh the components

are written horizontally, for convenience. { } therefore repres-
ents a row vector. By the rules of vector multiplication if
{a} = [E] {e} (4.9)
then {a}T = {e}T [E]T ( 4. 9a)

and [E)T = [E] for a symmetrical matrix.

In general there will be body forces F per unit volume, with
components X, Y and Z acting upon the element dx, dy, dz and the
surface forces p per unit area are p , p and p .
X y Z
In terms of the displacements {f} = {u, v, w} equation 4.3
becomes

v =u - w

Ivol
~{e}T [E] {e} dv

-I {f}T [FJ dv (4.11)

vol

-I {f}T {p} dA - Ew.6.

surface 1 1

The second and third terms in equation 4.11 represent the potent-
ial energy lost (or work done) by the body forces and the surface
forces respectively. The last term represents the work done by
point loads. 6. is the displacement of the point of application of
1
P.1 measured in the direction of P ..
1

(b) The Rayleigh-Ritz method The procedure for reducing the number
of degrees of freedom to a manageable level in continuum problems is
to 'guess' the displacement field in terms of a finite number of

106
constants. The constants then take the place of the unknown dis-
placements and a solution for then is found. There are, of course,
conditions which the guessed displacement field must fulfill. The
functions chosen, usually polynomials, and a sufficient number of
derivatives must be continuous and must satisfy the boundary con-
ditions of the problems to be analysed. The method is originally
due to Rayleigh, who used it in connection with vibration problems,
and was later refined by Ritz.

Example 4.4

Axial member with uniformly distributed tensile load.

...
X

Fig. 4.8

Figure 4.8 represents an axial member with end loads P1 and P2

at ends 1 and 2 and also a constant tensile load of q/unit length.
The displacements at ends 1 and 2 are u 1 and u2 in the x direction.
A suitable displacement field would be

The boundary conditions to be satisfied are

x = 0, u = u1 gives a 0 u1

and x = L, u = u2 gives a1

and then
(1)

du u2 - ul
and e = dx = L

The reason for choosing only the first two terms of the poly-
nomial to represent the displacement field is now clear. In this
example unknown constants aa and al can be replaced by the dis-
placements at the ends of the bar. (These displacements are
called nodal degrees of freedom.) It is a requirement of the
method that the number of unknown constants in the polynomial which
represents the displacement is equal to the degrees of freedom.
Now instead of differentiating potential energy with respect to th~
a's it is differentiated with respect to the nodal degrees of free-
dom which have replaced the a's.

107
 The potential energy of the bar is

(2)

substituting for u and integrating gives

V = AJ[u 12 - 2u1u2 + u2 2) - ~L(ul + u2) - P1u1 - P2u2

Differentiate with respect to each u and equate to zero

av
-- = 0 gives
aul '

av
-- = 0 gives
au 2 '

or in matrix form

l :r~}
pl + = Et [ 1 -lllull (3)
p2 + -1 1 U2

The above equation cannot be solved for u1 and u2 if values of P1,

P2 and q are given. This is because the determinant of the coeff-
icients inside the square bracket is zero. In physical terms, it is
because the bar has not been fixed in space. To achieve this one
displacement in this example must be specified relative to the
reference axis.

For example, make u1 zero, then

u2 = (P2 + Sf)E~ = - [P1 +

qL) L
:f EA

is the solution

Example 4.5

Beam element

In this example there are two displacements at each end of the

beam and the displacements are different in kind, v is a trans-
lation, positive in the positive y direction and e = dv/dx is a
rotation or change od slope, positive in the anti-clockwise sense.
108
 This mixture of displacements makes no difference to the procedure.
As the total nodal degrees of freedom is four, a polynomial con-
taining four constants will be suitable.
v = a + aix + a2x 2 + a3x 3 1.
0
As in the previous example, now write the displacement field in
terms of the nodal displacements using the conditions
dv _
X = 0, V = VI, dx - 6I

dv
X = L, v = vz, dx = 62

1 I. L
Fig. 4.9
'tv; ...
12
X

whence

VI 1 0 0 0 a
0

6I 0 1 0 0 aI

v2 1 L L2 L3 a2

62 0 1 2L 3L 2 a3

and, by inversion, or direct solution of the above equations

a 1 0 0 0 vl
0

aI 0 1 0 0 9I

3 2 3 1 2.
a2 -T v2
L2 L2 T
2 1 2 1
a3 92
L3 L2 L3 L2

The potential energy of the beam can now be written

3.

and after substituting for v and integrating

109
 Now, as before, differentiate V with respect to each end
displacement in turn and equate to zero.

Substituting for a2, a3, aa2/av1 and aa3/av1 from equations 2

leaves

P1 = [[~~)v1 + [L62)el- (~~)v2 (L~)62]EI

+ 4.

and av 0
a6 1 =

gives

whence

5.

and by differentiating with respect to v 2 and e 2 in turn gives

similar equations to 4 and 5 for P2 and M2. The student should
satisfy himself that the following is the full solution in matrix
form
pl 12 6L -12 6L vl

Ml 6L 4L 2 -6L 21.2 61
EI
6.
L3
p2 -12 -6L 12 -6L v2

M2 6L 2L 2 -6L 4L 2 62

Notice again that the stiffness matrix is symmetrical and that

its determinant is zero. Imposing the constraints v1 = 61 = 0 and
making P2 = W, M2 0 leaves

110
 0

whence
w -WL, p -W

which is the familiar result for a cantilever with an end load W.

Example 4.6

Cantilever with point and uniformly distributed loads

~~1-....!::::..L--t;f2..,..~-=--L----+tt~
~ t tt 11 t f t t 3
W/unit length
Fig. 4.10

In the previous example the rigid body constraints could have

been imposed on the assumed displacement field at the outset by
making a 0 = a 1 = 0, to make v ~ dv/dx = 0 at x = 0. In addition
the solution to both examples 4 and 5 is exact because the 'correct'
displacement field has been chosen. In this example, which has
been chosen to demonstrate the Rayleigh-Ritz procedure in a purer
form, the solution will be an approximation.

Assume

1.

so that v and dv/dx are both zero at x = 0.

Potential energy is

If only the first term in equation 1 is retained (v a 2x2),

equation 2 becomes
8wL 3a 2
V = 4EILa 2 2 - P2a 2 L2 - 4P 3azL 2 - - -
3-

and dv/da 2 = 0 gives

111
 The solution for v is therefore

3.

If two terms of equation 1 are retained (v = a 2x2 + a 3x3), then

equation 2 becomes

Now avjaa2 = 0 and av;aa 3 = 0 yields values of a 2 and a3 which

on substitution in equation 1 gives a better approximation to the
displacement field than equation 3.
The inclusion of more and more terms greatly increases the
labour as well as the accuracy of the solution. If the exact dis-
placement field is indeed a polynomial, as assumed, the inclusion
of more higher order terms than necessary will result in values of
zero for the corresponding a's. If the polynomial is 'incomplete',
that is one or more of the lower order terms are omitted the
solution will never be exact however many higher terms are included.

Example 4.7

Triangular plate

~.u1 xu
~--------------------~~
Fig. 4.11

This example extends the method to the two dimensional stress

problem shown in figure 4.11. The triangle is loaded at each of
its three corners, as shown, and the problem is to discover the
displacements at each corner.

The loads on each corner are

{p} = {pl, P2> P3 Ps}

112
so that Pl> P3 and Ps are loads in the positive x direction at
corners.l, 2 and 3 and P2> P4 and PG are loads in the positive y
direction at corners 1, 2 and 3 respectively.
Similarly for the displacements

For the displacement field assume

1.

Following examples 4 and 5 the a's are now replaced by the

nodal displacements u1, u 2 ... etc. Evaluating u at each node

1 xl Yl

!::l .
2.

!::!
1 x2 Y2

1 x3 Y3

and solving for the a's gives

X2Y3 - X3Y2 X3Y1 - XIY3 X1Y2 - X2Yl

!::!'
1

!::!
= 2A Y2 - Y3 Y3 - Yl Yl - Y2

X3 - X2 XI - X3 x2 - xl

where A is the area of the triangle with an identical relationship

between a3, a4, as and u 2 , U~t I!Jld UG
The strains, by definition, are
au av - au av}
{e} = {ex, e y' y xy } = {ax'
ay' ay+ ax

Note that the strains (and therefore stresses) are constant

throughout the triangle, as a consequence of the assumed dis-
placement field, equation 1. For this reason this example is often
called the 'constant strain triangle'. Substituting for a 1 and a 2
from equation 3 and a4 and as from the corresponding equation in
u2, u4 and us gives, ~n re-arranging

113
 rl' r,-y,
ey

Yxy
=-
2A
0

X3-x2
0

x3-x2

Y2-Y3
Y3-Yl

Xl-X3
0

xl-x3

Y3-Yl
YI-Y2

x2-xl
x,:xJl~l
YI-Y u6

1
or {e} = ZA [B]{u} 4a.

The stress strain relationship (equation 4. 9) gives

[E]
E
(1 - v2) [ v
1

0
OJ
l~v
5.

for plane stress or

E
[E] 6.
(1 + v)(l - v)

for plane strain.

Use of equations 5 and 6 will depend upon the actual problem.
Using equation 4.11 with body forces and surface tractions zero
gives

V = II~{e}T [E) {e} t dxdy - I 61 p.u.

1 1
7.

where t is the thickness of the triangle in the z direction.

SUbstituting for {e}T and {e} from equations 4 and 4a and noting
that all the terms in the expressions for {e} are constants

V = {u} T [B] T [E) [B]{u} 8tA -

T
{u} {p} 8.

and taking av;aul = 0, av;au2 =0 ... av;au6 0 results in the

set of equilibrium equations.

t T
0 = 4A [B] [E][B]{u} - {p} 9.

or 0 = [K){u} - {p}

114
where [K] = (t/4A) [B]T[E][B] is a six by six symmetric matrix
called the stiffness matrix.

4.2 Virtual Work

4.2.1 Virtual Displacements

Before defining virtual work or the Principle of Virtual Work it is

necessary to explain what is meant by a 'virtual displacement'.
Simply, a virtual displacement is one during which the applied for-
ces remain constant in magnitude and direction.

Fig. 4.12

Thus, in figure 4.12 any force Xi undergoing a virtual displace-

ment ~xi will do virtual work Xi~xi' and the total virtual work for
the body undergoing virtual displacements ~xi and ~yi (i = 1 ... n)
is

WV.1rt = L1n(X.~x.
1 1
+ Y.~y.)
1 1
4.12

The student will have realised that real work is not the same as
virtual work and that the concept of virtual work is therefore an
artifice. The real work done by force Xi undergoing displacement
~x. is J~xi X.~x .. In examples where the virtual displacements are
1 0 1 1
very small the virtual work may be approximately equal to the real
work.

4.2.2 The Principle of Virtual Work

The Principle of Virtual Work simply states that if a body is in

equilibrium then the virtual work done in any virtual displacement
is zero.

For a rigid body in equilibrium the equations of equilibrium are

M = 0 (resultant moment about OX = O)

yz

115
 0 Mzx 0 (resultant moment about OY 0)

M 0 (resultant moment about OZ = O)

xy

Any displacement of the body in figure 4.12 (virtual or otherwis~

is equivalent to translation of 0 (~x, ~y, ~z) plus rotation about
each of the coordinate axes (~9 , ~e , ~e ), whereupon the virtual
X y Z
work would be

Wv 1rt = Ix.~ + IY.~ + Iz.~ +

1 X 1 y 1 Z
Myz ~e X + MZX~e y + Mxy~e Z

Since the displacements are arbitrary

Ix.~
1 X
= ~
X
Ix.
1
= o, etc

and hence

Example 4.8

Simple tie bar - see figure 4.5

Consider the tie bar shown in figure 4.5, fixed at one end and in
equilibrium when subjected to the tensile force F. A further
increase in displacement ~x will cause the internal force in the
bar to do work - (xAE/~)~x, negative because the force acts in the
opposite direction to the assumed virtual displacement. The total
virtual work is then
F~x - AE xdx 0
~

whence
~F
x = AE
Example 4.9

Rigid bar on two wires - see figure 4.6.

Assuming virtual displacements ~x 1 and ~x 2 from the equilibrium

position and applying the principle of virtual work gives

W Virt 0

[ Wb
or (a + b) 0

116
Since 6x 1 and 6x2 are arbitrary and independent, each bracket in
the above expression must be zero, whence

and x2

Example 4.10

Ladder retained by spring- see figure 4.7

Assuming that the ladder is in equilibrium in the position shown,
give the ladder a virtual displacement 68. The height of W above
the ground is

y = 4 cose

therefore

6y = -4sin8M

that is, the virtual displacement of W is 4sin868 downwards, and

similarly for the spring

x = 6sin8
6x = 6cos8M

The expression for virtual work is

WVirt = 4Wsin969 - (6Ksin9)6cos969 0

whence
sine = 0
4W
or cose = 36 K

4.2.3 Comparison of Virtual Work and Potential Energy

The above examples 8, 9 and 10 are simply examples 1, 2 and 3

re-worked using the Principle of Virtual work. It can be seen that
virtual work, although perhaps simpler to write down, yields less
information than potential energy (which also describes the nature
of equilibrium) and is conceptually more difficult. In both cases
the assumed displacements are an artifice, in the sence that the
external forces remain constant in magnitude and direction during
such displacements.

117
 The student should realise that the equation for v. the potential
energy, has no particular physical meaning, it is the values of
displacement which give V a stationary value to yield the equilibrium
equations which are important. The Principle of Minimum Potential
Energy is just as much a mathematical device as is the Principle of
Virtual Work. An example in which the conservative force is due to
gravity illustrates this point.

Equilib'm pos'n

Fig. 4,13

In figure 4.13 the unstretched spring of stiffness K has a mass

m'at its lower end which is released from rest. The reader knows
that under these conditions the mass will oscillate about the
equilibrium position with amplitude a and frequency l/2w~K{m, and
that, in the absence of damping, the mass m will never come to rest.
The force on the spring is constant for all x and the expression
for potential energy is therefore
Kx 2
V = - 2- - mgx

at the equilibrium position x = a and dV/dx 0, whence

a = 1!!K
K

Thus the Principle of Minimum Potential Energy is valid and yields

the equilibrium position even though the physical system is such
that the mass m would never be in equilibrium in the absence of
damping.
4.3 Castigliano's Second. Theorem and The Principle of Virtual Work

It is instructive to use the Principle of Virtual Work to derive

Castigliano's Second Theorem.

Consider a body in equilibrium under the action of external loads

(including the reactions) Pi, i = 1 ... n. The stress {a}= {ax
a a T } at any point within the body (x, y, z) will be a
y z zx
function of all the loads, and so will the strain {e} = {e ... y }.
X ZX

118
 Let the deflections of the body, at the points of application of
the loads, be o1 ... On in the direction of the loads.

Now suppose that the force, say Pk changes to Pk + dPk when dPk
is small. In order to maintain equilibrium the stress at all points
within the body must change by {dcr} where

{dcr} =
a
ar- {cr} dPk ( 4 .13)
k

Since equilibrium is maintained, the force dPk' the stresses {dcr}

and the reactions dRi (where dRi = a;aPk(Ri)dPk) must, independently,
be a system of forces in equilibrium. If the external displacements
are now taken to be the actual deflections oi' and the actual strains
{e}, within the body, the Principle of Virtual Work applied to the
incremental force system gives, since the reactions dRi do no work,

and use of equation 4.13 leads to

a T
ok = J {~} {e} dv ( 4.14)
vol. aPk

The integral on the right hand side of equation 4.14 is recognised

as the partial differential coefficient, with respect to Pk' of the
total complementary energy of the body, that is

(4.15)

This result is, in fact, Engesser's theorem which is more general

than Castigliano's theorem and applies to any elastic body, whether
linear or not.

e
Fig. 4.14

119
 If, and only if, the cr versus e graph is a straight line (that
is, the material is linear as well as being elastic) the complement-
ary energy equals the strain energy, see figure 4.14 and

(4.16)

which is Castigliano's second theorem.

4.4 Problems

1. For a beam, simply supported at each end, 'acceptable' displace-

ment functions are
1TX
v = as1n-y;- (a)

and v = ax(l - -[) (b)

Use the Principle of Minimum Potential Energy to deduce the def-

lection curve for a simply supported beam carrying a central load
W, for each of the above assumed shapes.

2. Using the second of the suggested displacement functions in

question one above, deduce the deflection curve for a simply supp-
orted beam carrying a uniformly distributed load. Calculate the
central deflection and compare it with the accurate result.
4
[ WL SWL 4)
96EI' 384EI

3. A simply supported beam of length L carries a load W at c from

the left hand end and b from the right hand end, (c + b = L). Use
Castigliano's theorem and the Principle of Minimum Potential Energy
to calculate the deflection at the load.

4. A cantilever carries a load W at its mid point and is propped

at its free end by an elastic prop of stiffness K. Assume the
deflection equation v = ax 2, which ensured v = dv/dx = 0 at x = 0,
and calculate the value of a using the Minimum Potential Energy
Method.

120
5 NUMERICAL METHODS

A complete analytical solution for the displacements, stresses or

strains in a body carrying loads can be found in only a limited
number of examples. Indeed, so few solutions exist that they are
usually known by the name of the discoverer, for example St Venant's
solution for torsion (see chapter 3) or the Boussinesq stress
function for a point load (see chapter 2). The difficulty in
finding an analytical solution usually lies in the complexity of the
boundary conditions - either the boundaries themselves have discon-
tinuities or the loading on the boundaries has irregularities or
discontinuities which are difficult or impossible to describe anal-
ytically.
However, these difficulties do not prevent an approximate
solution using numerical methods. In both the numerical methods
discussed below the governing differential equations are replaced
by a set of algebraic simultaneous equations which represent the
value of the unknown variable (displacement, stress, etc.) at
discrete points, usually called 'nodes'. Usually, the greater the
number of nodes the more accurate the result but, at the same time,
increasing the number of nodes increases the number of equations
to be solved.
There are many specialist texts covering both topics presented
below; in addition, entire books devoted to the mathematical
techniques required are available. It is also clear that the
solution of large numbers of simultaneous equations requires the
use of digital computers and a knowledge of programming. The reader
will understand that the following can be no more than an intro-
duction to Finite Difference and Finite Element methods of analysis.
5.1 Finite Difference Analysis
5.1.1 Difference Formulae for Partial Differential Coefficients

When learning the differential calculus the students become familiar

with the definition of a differential coefficient,

df(x) _ Lt f(x + t.x) - f(x)

""""dX- t.t
t.x-+0

or in terms of the notation given in the diagram, figure 5.1.

~ % Yn+l - Yn
(5 .la)
dxn t.x

(S.lb)

121
 (5 .lc)

Cs.ld)

~~~~~~~.-x
n-1 n n+t
Fig. 5.1
All the formula e in equatio ns 5.1 will give differe nt values for
(dy/dx) n but all will be 'approx imately ' the same and as 6x + 0,
all
will converg e on the 'true' value for (dy/dx) n. The express ions
in
equatio ns 5.1 are called 'differe nce formula e'; 5.la is a 'forward
differe nce', 5.lb is a 'backwa rd differe nce' and 5.lc and 5.ld are
'centra l differe nce' formula e, for obvious reasons .

Higher order differe ntial coeffic ients can be derived in differ-

ence form in just the same way. Using equatio n 5.lc

dy) - [~)
[ d2y) "'_i_[~)
d 2 "' dx dx "'
"' [dx n+l dx n-~
x n n 6x

Again using equatio n 5.lc gives

(dy) = Yn+ 1 - Yn
dx n+;1 6x

and similar ly for (dy/dx) 1 . Using these in the express ion for
n-;
(d 2y/dx 2) gives
n

The equatio ns which commonly occur in problem s in elastic ity

are partial differe ntial equatio ns.

Figure 5.2 shows the convent ional way of numbering the node
points in order to obtain differen ce formula e for point 0. The
four closest points are numbered 1 - 4 in anticloc kwise order and
the next 'circui t' is numbered 5 - 12 with point 5 next to point
on the X axis through 0. 1

122
 tp=tp{X,

Fig. 5.2
Extending the formulae above in terms of ~ ~(x, y) gives

(5. 3a)

(5. 3b)

and, taking the argument a step further

which becomes, on using equation 5.3a

(5.4a)

similarly

(5.4b)

and again using equation 5.2

[~) + (a2~) [~) 0
2 ar2
[ a~~ ) ~a2 (a2~)] ali ar2 3
ax2 ay 2 } --;2 ay 2 o::; h2

that is
~6 + ~8 + ~lo + ~12 - 2(~1 + ~2 + ~3 + ~~) + 4~0
[ a4~ )
ax 2ay 2
0
:t h~

(5.5)
123
 The reader should check that, using the alternative formulation
for equation 5.5 gives the same result

that is

5.1.2 Differential Equations Replaced by Difference Equations

It should now be clear that a partial differential coefficient can

be replaced by its approxima~e difference equivalent. The result-
ing equation refers to specific values of the unknown function at
node points within the boundaries of the problem. A difference
equation can be written for each node point, the boundary
conditions applied and the resulting set of simultaneous equations
solved for the discrete values of the unknown, ~.

For example the torsion equation (see Chapter 3)

~ + ~ = -2Ga
ax 2 ay 2
becomes
(5. 6)

in difference form, after using equations 5.3.

The boundary condition for torsion is d~/ds 0, that is ~ 0

for a single boundary.

Fig. 5. 3

5.1.3 Methods of Solution

In order to discuss methods of solution, it is useful to refer to

a specific example. Consider the torsion of a shaft of cross-
section shown in figure 5.4

124
 y
Jt
/ l'. h..-f{l =0
v
jE

c
ID

8
"<
D
w
X
A
" E

I
Fig. 5.4
:/

Only one quarter of the hexagonal cross-section is shown on a

square mesh of side h. The problem is to find ~ at the node points
(A, B, C, etc.) where ~ is given by

tl + tl = -2Ga
ax 2 ay 2
or, in difference form

(see figure 5.3 and equation 5.6), and~= 0 at all points on the
boundary of the cross-section. In fact, because of the symmetry of
the problem, only one eigth of the cross-section need be considered
and the problem has only five unknown values of ~ at nodes A, B, C,
D and E. For point A equation 5.6 becomes
(5.7a)

Similarly for point B, equation 5.6 gives

(5.7b)

and for point C

(5. 7c)

for point D

~E + ~B - 4~ 0 = -2Ga h 2 (~ = 0 at boundary) (5. 7d)

and point E
(5. 7e)

125
 Equations 5.7a- 5.7e comprise a set of five equations with five
unknowns and can therefore be solved for the values of ~ at the
node points. Three methods of solution will now be discussed.

(a) Gaussian elimination - straight forward solution of

simultaneous equations.

Equations 5.7 can be written in matrix form, as follows

-4 0 4 0 -2

0 -4 2 2 -2

1 2 -4 0 (5. 8)

0 1 0 -4 1 -2

0 0 1 2 -4 -2

To solve for ~ the procedure is, beginning with N 1,

(i) The Nth equation is divided by its Nth term.

(ii) The Nth equation is multiplied by the Nth term of the (N + l)th
equation and subtracted from the (N + l)th equation.

This step is then repeated on the (N + 2)th equation, and so on

until equation N has been appropriately multiplied and subtracted
from each subsequent equation in turn.
(iii) Steps (i) and (ii) are then carried out for the (N + l)th
equation, and so on until every equation has been dealt with.

The effect of these operations is to reduce the matrix of

coefficients to upper triangular form with each diagonal coefficient
equal to unity. The last equation then has the form ~N = X and the
remaining unknowns can be found by back substitution.

The work can often be reduced by re-ordering the equations to

take advantage of already existing values of unity which can be made
to fall on the diagonal. Re-arranging equations 5.8 gives

1 2 -4 0 1 ~A -2 a

0 1 0 -4 1 ~B -2 b

0 0 1 2 -4 X
~c -2 Ga h 2 c (5. 9)

0 4 0 0 ~D -2 d

0 -4 2 2 0 ~E -2 e

126
where the first three equations are already in the form required.
Step (ii) above requires that equation (a) should be multiplied
by -4 and subtracted from equation (d); to complete the first
column of zeros

1 2 -4 0 1 --2

0 1 0 -4 1 --2

0 0 1

0 8 -12 0 4 -10

0 -4 2 2 0 -2

and carrying out step (ii) on equations (b) and (c) leaves

1 2 -4 0 1 -2

0 1 0 -4 1 -2

0 0 1

0 0 0 56 -52 -18

0 0 0 -18 12 -6

Carrying out steps (i) and (ii) on the remaining two equations
(Cc) and (d)) results in

1 2 -4 0 1 -2 a

0 1 0 -4 1 -2 b

0 0 1 2 -4 x 4>c c (5.10)

13
0 0 0 1 -rr 9
-28 d

5
0 0 0 0 1 e
2

From equation 5.10e it is now clear that

Equation 5.10d reads

13 9 2
<Pn - 14 <PE = - 28 Ga h

and substituting for <P gives

127
 It is left to the reader to continue the back substitution
process and show that

cf>c = 4Ga h 2

cf>B =a h2

and

Having found the values for 4> at each of the node points,
approximate values for ac~>/ax = Tyz and ac~>/ay = TXZ can be found
from the appropriate difference formulae. Also the applied couple
can be found from
j=N
T 2 JJ 4> dxdy ~ 2 2 (5.11)
j=l

where N is the total number of nodes. Of course, the relatively

coarse mesh used in the above example cannot be expected to yield
accurate values for applied torque and shear stress. For example,
a cross-section through the Y axis of the torsion hill for the
shaft cross-section of figure 5.4 would be as shown in figure 5.5.

A!--~c~-=e=--~F~ Y
Fig. 5.5

An approximate value for the shear stress at F is then

( T xz) F r
= [ JF ~
(cf>E - O)
--h- =2
5
Ga h (5 .12)

This is clearly an underestimate of the true value forT. A

better approximation may be made by fitting a curve to the values
of cf>A, cf>c and cf>E already obtained, and then differentiating to
find the slope at 0. A suitable polynomial would be

4> =a + by 2 + cy~ + dy6

using even powers of y because of the symmetry of the curve for cf>.
Substituting the conditions

128
 y 0, <PA = 29 Ga h 2

y h, <Pc =4 Ga h 2
5
y 2h, <PE =2 Ga h2

y 3h, <PF 0

gives

<P = ! [9 ~/2) Ga h 2

and <l<P/Cly = -y Ga, whence at y 3h the shear stress is

('rxz)o = 3Ga h (5. 13)

which should be compared with equation 5.12. This final approxi-

mation for the shear stress provides a v~lue which is quite accept-
able from an engineering point of view.

This section has provided only a brief outline of the elimination

method of solution of simultaneous equations. The pitfalls of the
method and the numerous other methods available are the subjects of
specialised texts.
(b) Iteration

In this process a set of values for <P is guessed and each value is
then subsequently 'improved' in turn. When the value at every
node point has been 'improved', one iteration has been completed
and the process begins again, and it is continued until the
improvement between successive iterations is negligible or falls
within the limits of accuracy required. The method will be
illustrated by using the same example as for section 5.1.3a.

First of all the difference equation 5.6 is transformed using

the substitution

(5.14)

so that equation 5.6 becomes

(5 .15)

This transformation makes the solution purely a numerical matter

and the value of N can be chosen to reflect the accuracy of the
solution. For example if N = 100 and the solution is sufficiently
accurate if successive iterations result in a change of value of

129
less than 1, then computation can be carried out in integers.

On this basis a set of values as shown in figure 5.6 will serve

as initial 'guesses'.

Fig. 5.6

Equation 5.15 for point A 11ow becomes

N + 41)Jc

so that a better estimate for point A is obtained simply by

substituting the guessed values into the above expression for ljJA.

Thus, an improved value for ljJA is

100 + 800 225

ljJA :i; 4

The original guess for ljJA is crossed out and the 'improved'
value substituted. This improved value is used in all subsequent
operations.

Following the same procedure, an improved value for point B is

given by

100 + 400 + 200

175
4

Similarly for point C

100 + 225 + 150 + 350

206
4

The process is continued until all the values have been amended,
thus completing an iteration. Further iterations are carried out
until a sufficiently accurate solution is reached.

The iteration procedure as outlined above is 'foolproof' in the

sense that, no matter how bad the initial guesses are, the procedure
will eventually converge on the correct result. The disadvantage

130
is that, unless special accelerating techniques are used, the rate
of convergence is often unacceptably slow.

(c) Relaxation

In the relaxation process the difference equation is again recast

in the form of equation 5.15 and a set of guessed nodal values
forms the starting point. The next step, however, is to calculate
the value of the 'residual' at each node. The residual is defined
as the amount by which each node is in error. From equation 5.15
R. = N + ,,, I + ,,, I + ,,, I + ,,, I - 4' I (5.16)
J 'I' j +1 'I' j +2 'I' j +3 'I' j +4 'I' J

where R. is the residual at node j and ljl'i is a guessed value, see

figure ~. 7.

i+2

i+3 j i+1

i+4

Fig. 5. 7

Subsequent steps in the process are designed to reduce the

residuals at each node to zero, or to an acceptably small value.
The value chosen for N plays the same role as in the iteration
process, section 5.1.3b.

Using the same example and guessed values as for the previous
se.ction the residuals are calculated below, using equation 5.16.

RA 100 + (4 X 200) (4 X 250) -100

RB 100 + (2 X 200) + (2 X 100) (4 X 180) -20

Rc 100 + 250 + (2 X 180) + 150 (4 X 200) 60

RD 100 + 150 + 180 - (4 X 100) = 30

RE = 100 + 200 + (2 X 100) - (4 X 150) = -100

These values are then entered to the right of the guessed starting
values as shown in figure 5.8.

The next step is to begin the process of reducing the residuals

to zero. An experienced operator can often see short-cuts in this
process which reduces the total number of operations. However a
systematic approach is always to operate on the residual having the
high~st numerical value, in this case either RA or ~

131
 Fig. 5.8

Choosing, arbitrarily, RA and examining the equation for RA above,

it can be seen that a change in the value of ~A of +1 changes the
residual at each of the four surrounding points by +1. This state-
ment is summarised by figure 5.9 which is referred to as a 'relax-
ation molecule'.

Fig. 5.9

In order to reduce the residual at A to zero therefore (-25)

must be added to the value at A. This has the effect of changing
the residual at each surrounding point by (-25). The effect of
applying this step is shown in figure 5.10. Again figures are
updated by crossing out rather than erasing so that a check can
always be made.

-25 0
250-100
A c E

Fig. 5.10

132
 Proceeding now to node E, the residual of -100 is made zero by
adding (-25) to the value at E. This causes a change of (-25) in
the residuals at D and C and these are also shown in figure 5.10.
The next highest residual is at point B and is reduced to zero
by the addition of (-5) to the value at B. The necessary amend-
ments to the residuals at the points surrounding B are then made
and it should be remembered when carrying out this operation that,
because of the symmetry of the problem, each point C receives a
contribution from two points B to its residual. The residuals at
C due to change in value of (-5) at B must be amended by 2 x (-5).
The result of this last step is also shown in figure 5.10.

The fortuitous choice of starting values for this problem has

resulted in a 'correct' solution after very few adjustments. All
the residuals are now zero and it remains only to sum the figures
in the value column in order to find the values of WA etc, for
example

WE = 150 - 25 = 125
Usually the relaxation process must be carried out many times
in order to achieve an acceptable result and as has been already
pointed out, experience and skill can greatly shorten the process.
5.1.4 Boundaries Not Coincident with Uesh

Generally the boundaries of the cross-section or component being

analysed will not conveniently pass through the node points of the
mesh and the difference equation must therefore be modified to
allow for this.

j+4
Fig. 5.11

In figure 5.11, B is a point on the boundary which does coincide

with the mesh, mh is the distance between points j and B in the x
direction (m < 1) and j + 1 is an 'imaginary' point on the mesh
outside the boundary.

Using the ideas of section 5.1.1 the slope in the x direction

at j is given by

133
 (forward difference)

and hence the value of ~ at the imaginary mesh point j + 1 can be

extrapolated as

h(M.)
ax . = (..!.)~
m B+ [1 - ..!.)~.
m J
J

into equation 5.3a gives

(S .17)

Using this expression for (a 2 ~/ax 2 ) rather than equation 5.3a

for points near the boundary, the rest of the analysis proceeds in
the usual way.

5.1.5 Alternative Shapes of Mesh

An alternative approach to the above problem is to use a different

shape of mesh. For example, the boundary of a particular problem
may be fitted more conveniently to a mesh of triangular or hexagonal
shape. Before proceeding to these shapes the reader should first
consider the relationship between the square mesh already discussed
and the stretched membrane used in the membrane analogy of chapter 3.

h
y
2
-- --1
I 0 I
3 h I I
1
__I
~-

(a) (b)
Fig. 5.12

If the membrane is replaced by a mesh of elastic strings then

(see figure 5.12a)

(i) The load on each node point is the pressure p on the membrane
multiplied by the area covered by each node. In this case where

134
the mesh is square the area is h 2 and the nodal load is ph2.
(ii) The tension in each string must equal the tension/unit length
of the membrane (P) multiplied by the length of the side of the
nodal area (h), thus the string tension is Ph for the square mesh.
From figure 5.12b, the equilibrium equation for node 0 is then

Ph[w1 ~ W0 ) + Ph r-2 h- w0 ) + Ph r-3 h- W0 )

+ Ph F4 h- wo) + ph 2 =0

or w1 + w2 + w3 + w 4 - 4w 0
ph2
p (5 .18)

which is exactly the same as equation 5.6.

Proceeding to the triangular mesh of figure 5.13 shows that the
area of influence of each node is hexagonal of side h/13 and area
13h 2/2. The nodal load is therefore pl3h 2/2, and the the tension
in each string is Ph/13 and the equilibrium equation for a typical
node is therefore

or w1 + wz ... + ws - 6w 0 + i(p~ 2 ) = 0 (5 .19)

Fig. 5.13

Similarly, for the hexagonal mesh of side h, the area associated

with each node is triangular, of side h/:3 and area 313h 2/4, figure
5.14.
The equilibrium equation is

0 (5.20)

135
5.1.6 Applications of the Finite Difference Method

Example 5.1 Torsion

The example chosen to illustrate the methods of finite difference

analysis was the torsion of a hexagonal shaft, figure 5.4. In
order to extend the application of the methods to cross-sections
with more than one boundary, the same shaft will be used but with
a square hole, of side 2h, as shown in figure 5.15.

/_ ~ h

i'.
~ B 1~,;

rpA_ ......._ A A c .rf{J 0

A B

A
/
Fig. 5.15
The general boundary condition d~/ds = 0 holds for both boundari~
but the value of ~ can be arbitrarily chosen for only one of them.
Usually the outer boundary is chosen to have ~ = 0 and the value of
~around the hole is designated~= ~A (constant). The second
boundary condition required is
/Tds = 2Ga A
where A is the area enclosed within the boundary.
In difference form, remembering that T is the slope of the hill
normal to the boundary ca~;an), this becomes

~=N(~A
1..
- ~i)h N
}: (~A -
)
~i = 2Ga A (5.21)
i=l h 1

where N is the total number of mesh connections to the boundary under

consideration, figure 5.16.

~
I
II

A 'A-
A

Fig. 5.16

136
 In the above example, equation 5.21 becomes, since the area of
the hole A = 4h2,

4(~A - 98 ) + S(~A - ~C) = 2Ga(4h2)

or 12~A - 4~ 8 - s~c = 4(2Ga h2) 1.

Again, because of symmetry, only one eighth of the cross-section

need be considered so that values at only two internal points remain
to be found. Proceeding as in section 5.1.3 - for point B
2.

and for point C

3.

Solution of equations 1, 2 and 3 above by the methods outlined

in section 5.1.3 results in

~A = [~)Ga h2

"'s
"'
[24)G h2
l3J a

~c [i~)Ga h2

Example 5.2 Laplace's equation

The compatibility condition, in terms of stress for two dimensions,

leads to Laplace's equation, see equation 2.3a when there are no
body forces acting.

(5.22)

Writing ~ (a X +a)
y
= (al + a2) this becomes, in-difference
form

~1 + ~2 + ~3 + ~~ - 4~0 0

~1 + ~2 + ~3 + ~~
(5.23)
4

That is the value of ~ at any point is the average of the four

surrounding values.

137
 The solution of equations 5.22 and 5.23 is most useful as an
adjunct to a photoelastic investigation.

The monochromatic fringe pattern gives the value of (a1 - a2)

at each point within the boundary and usually the separate values
of a1 and a2 are known, or can be deduced on the boundary and
hence (a1 + a2) = <fl is known at the botmdary also. With this
knowledge, equation 5.23 can be solved within the boundary to
yield the value of (<Ill + <fl2) at each mesh point within the
boundary. Thus, at each point a 1 - a2 is known from the photo-
elastic results and a1 + a2 is known from the numerical results
and the separate values a1 and a2 can be fotmd.

Example 5.3 The Bi-harmonic equation

Substituting for a and a in Laplace's equation 5.22 in terms of

X y
the Airy stress function discussed in Chapter 2 yields

~+~+ a4<fl = o (5.24)

ax 4 ax 2ay 2 ay4

which is known as the Bi-harmonic equation. A solution for <fl

can be found using the finite difference method and hence the
values of ax, ay and Txy can also be approximated. In difference
form, equation 5.24 becomes, using equations 5.4 and 5.5

+ <l>s +<1>7 +cf>g +cf>u = o (5. 25)

As before an equation such as 5.25 can be written for each point

on the mesh and the resulting set of equations solved for <fl at the
node points. However the values of <fl around the boundary have first
to be calculated.

Recalling the boundary conditions in two dimensions

aR.+T m p
X xy X

T
yx R. + ay m = Py

and substituting for a , a and T Tyx in terms of <fl and

x y xy
using R. = dy/ds and m = -dx/ds gives

P _
X -
~(~)
ds ay2
+ dx[ 32<fl)
ds 3x3y
_J!_[li)
- ds 3y
(5. 26a)

138
 d
- ds
(a'a:xJ1 (5.26b)

whence on integrating

it=
ay
J px ds (5.27)

ax - J py ds
it=

Using these equations the values of 41 can be calculated for all

boundary points, as demonstrated below.

Consider the rectangular plate, of uniform thickness t, shown

in figure 5.17.

Stress u

Fig. 5.17

The plate is loaded as shown and the problem is to evaluate 41

for each portion of the boundary.
(i) The bottom edge, y = 0, -a/2 ~ x::; +a/2

ds = +dx

Applying equations 5.27

*
it=
ay
f(O)dx =A

= - J [%) dx = - a; + B

and integrating the second of these with respect to x gives

ax 2
- - 2- + Bx + C

139
where A, B and C are constants of integration. However, the values
of A, B and C are not important because the essential problem is to
determine the stresses ax ay and Txy which depend on the second
derivatives of ' Thus A, B and C can be chosen arbitrarily and the
obvious thing to do is to make them all zero, so that, along the
bottom edge

ll: 0
ay
ax
l l : -2
ax

ax 2
and
' = -4
Notice that these equations give a symmetrical shape for ' which
agrees with the symmetry of the problem. At x = +a/2

.!!.- 0
ay

aa
l l : -4
ax

and these values will be needed for the next part of the boundary.
(ii) The right hand edge, x = a/2, 0 ~ y ~ b
0

Equations 5.27 give

lt_ = j(O)dy = J
ay

lt_
ax
= -j(O)dy = K
and integrating again

' = Jy + L
This time the constants of integration must be evaluated from the
values obtained for the corner x = a/2, y = 0, so that

140
 ~ = 0
ay

~ _ cra
ax- 4

cra 2
' = - "T6
(iii) The top edge, y b, a/4 ::; x ~ a/2

pX = py = 0
ds = -dx
Equations 5.27 lead to

~= 0
ay

~- cra
ax -4

crax cra 2
cp - - --+
4 "T6
(iv) The top edge, y b, -a/4 ::; x ::; a/4

pX = 0, Py = -cr
ds = -dx
Equations 5.27 give

~- M
ay

~=
ax -f -cr(dx) - crx + N

crx 2
' = - --2-- + Nx + P

and using values already obtained at x = a/4, y b, allows M, N and

P to be evaluated as
cra 2
M = 0, N = 0, P = ~

Although the value of cp is now known for all points on the

boundary a difficulty remains because, for points adjacent to the
boundary, equation 5.25 demands values of cp actually outside the
boundary. For example, a point like 0 on the mesh shown in figure
5.18 requires values of cp at points 5 and 11.

141
 7
boundary
8 2 6 :_;?
f+-
9 3 0 1 5

10 4 12

11

Fig. 5.18

The required values are easily found by extrapolation from

boundary values just derived.

Thus
h (acp j
ax] 1

5.2 Finite Element Method

5.2.1 The Element Stiffness Matrix

The basis of the finite element method is the element stiffness

matrix. Any structure or component may be broken down into a
finite number of relatively simple elements whose load-displacement
characteristics are known, can easily be derived or approximated
using well tried methods in the theory of elasticity. The structure
or component is then re-assembled using the chosen elements and the
overall or structure stiffness equations thus derived. These
equations, linking the structure loads and structure displacements
are approximate, the degree of approximation depending on how well
the assenbly elements accommodate the loading and boundary conditions
and obey the rules of compatibility. The element stiffness matrix
is thus the basic building block of the method.

Consider an arbitrarily shaped element with n 'nodes' as shown

in figure 5.19. A 'node' is a point of connection between one ele-
ment and another. Each node is, in general, free to move in several
directions (degrees of freedom) and thus each node will have a
number of element displacements (u) associated with it and with
each displacement will be associated an element load (p).

Each element load pi will depend upon every element displacement

(u1, uz un) and for a linear elastic solid a typical equation
would be

142
 y

Fig. 5.19

(5. 28)

where i could take any value between 1 and n. The coefficients

K.. in equation 5.28 are called stiffness coefficients and the
lJ
whole set of n equations would be written

or in shorter matrix form

I 5.29

{p} = [K] {u} (5. 30)

where {p} is the column vector of loads, {u} is the column vector
of displacements and [K] is the element stiffness matrix. (The
notation {}Twill be used to represent a row vector.) Some obser-
vations can be made immediately.
1. The stiffness matrix will be square, of size n x n where n is
the number of nodes (m) multiplied by the number of degrees of
freedom at each node.
2. A particular element stiffness coefficient K.. will be found
lJ
in the ith row of the jth column in the stiffness matrix.

3. The matrix will be symmetrical, that is

K . :: K .
lJ Jl

For example, the element stiffness coefficient which multiplies

u7 in the expression for pz(K 27 ) will be identical to the element
stiffness coefficient which multiplies uz in the expression for
p 7 (K7z). This is simply a statement of the Maxwell reciprocal
theorem.

143
4. The set of equations 5.29 or 5.30 is singular. That is the
equations cannot be solved for the u's given an arbitrary set of
loads {p}. This is because no physical restriction has been
imposed on any of the displacements (boundary conditions) so that
the element is free to move as a rigid body.
The reader will by now have realised that the connection between
degrees of freedom, displacement and node number needs to be rigidly
defined. In this book the following system will be used. Suppose,
for example, that there are three degrees of freedom at each node
and that they are (x displacement), (y displacement) and (rotation
about the Z axis) as might be found in a rigid jointed two dimen-
sional framework - see figure 5.20.

Y~3k-1

~"""'T.......- .... x
J3k-2

Fig. 5.20
The u 1, u2 and u3 are, respectively, the X, Y and e 'displace-
ments' at node 1, and similarly for the 'loads' Pl pz and P3
Following figure 5.20 the reader will appreciate that, for example,
u7 is the X displacement at node 3 of the element. Notice that
the 'displacements' need not be of the same type. In the example
above, u3 is a rotation (radians) whereas u1 and uz are trans-
lations and similarly the 1 loads 1 will be a mixture of couples and
forces. Of course, the units of the element stiffness coefficients
will take care of these differences. Notice also that all displace-
ments are given the symbol ui, whereas generally, for example, the
symbols u, v and w are used to.represent translations in the X, Y
and Z directions respectively. So long as the remarks above are
kept in mind, this should present no difficulty.

In this book, only three elements will be discussed, which can

then be used singly or in combination. Although simple the elements
can be used to solve quite complicated problems, and the main
methods of derivation of element stiffness matrices will have been
covered.
(a) The end load element (direct method}

Although this book is not intended to cover the mechanics of

structures, this element is included because of its simplicity and
because it may be used in combination with the simple bending element
(section 5.2.lb).
The end load element is simply what its name suggests - a com-
ponent which can carry and transmit only loads directed along its
length as might be found in a simple pin jointed frame.

144
This element is a simple beam with a node at each end each having
two degrees of freedom, viz. a displacement in the Y direction
and a rotation about the Z axis positive in the sense shown in
figure 5.22. The method used here to derive the element stiffness
matrix is called the unit displacement method.
In equations 5.29, if u 1 is made equal to unity and all the other
displacements are made zero the equations reduce to
p1 K 11 xl+O

P2 1 + 0

that is, the expressions for the loads simply become the first
column of the stiffness matrix. Similarly making u 2 equal to
unity, all other displacements zero, yields the second column of
the stiffness matrix, and so on.
For the beam shown in figure 5.22 the bending moment equation
becomes

EI(~)2 {a)
dx
(remember the 'load' p 2 is a couple at node 1)

PlX2
then EI[*) -2-- pzx + A (b)

PIX 3 pzx 2
El y = - 6 - - - - Ax + B
2+
{c)

To obtain column 1 of the element stiffness matrix make

Uz 0, u3 = 0 and u~ = 0 and solve for p 1, p 2 , p 3 and p~.

uz 0, X = 0 in {b) gives A =0
U3 Yz 0, X L in (c) gives B = pzL 2/2 - P1L 3/6

Ul Yl 1 at x =0 in (c) gives EI = pzL 2/2 - p 1L3/6

and u~ 0, X = L in (b) gives 0 = p 1L2/2 - pzL

Hence
12EI
P1
L3
 The coordinate system is shown in figure 5.21 - there is only
one degree of freedom (in the X direction) and there are two nodes
so the element stiffness matrix will be 2 x 2.

P. .u1 areaA,Youngs MOdE

..
Fig. 5.21
The element stiffness equations can be directly derived, as
follows

For equilibrium of the element Pl + P2 0 and by definition

and combining these in matrix form gives

fPl} = Ac [ 1 -1] x ful} (5. 31)

1P2 -1 1 1u2

or {p} [K] {u} (5. 32)

where

[K] =AE [l -l] (5. 33)

L -1 1

the element stiffness matrix.

The reader should examine the above stiffness matrix with

particular attention to observations 1 to 4 in section 5.2.1.

(b) The simple bending element (unit displacement method)

Lx ~,u,
p2u2 1
El

L
E?i:JP,U,

Fig. 5.22
 6EI
P2 = -
L2

For equilibrium of the element (see figure 5.22) resolving

forces in the Y direction

P1 + P3 =0
and taking moments about node 1

P2 + P~ + P3L =0
gives

P3

p~

T~e first column (and row, by symmetry) of the element stiff-

ness matrix can thus be filled in
12 6L -12 6L

EI 6L X X X
L3 -12 X X X

6L X X X

It is left as an exercise for the reader to fill in the rest of

the coefficients. The final result is
12 6L -12 6L
6L 4L 2 -6L 2L 2
EI (5.34)
[K] =-
L3 -12 -6L 12 -6L
6L 2L 2 -6L 4L 2
(c) 'l'he constant strain triangle

This element is a triangular plane element of thickness t and is

so called because the assumed displacement field gives rise to
constant strains (and stresses) within each element. Although the
thickness (t) of each element must be constant the assembly of
elements which represents a particular problem need not all be of
the same thickness.

147
 In the previous elements an exact solution for the stiffness
matrix has been found whereas to 'assume' a displacement field
(as will be done for this element) clearly implies an approximation.
The justification in this case is that, if the element is small
enough, it is reasonable to assume no variation of stress or strain
within its boundaries.
Although this is a relatively unsophisticated element the reader
will find it extensively used to solve quite difficult problems and
it provides an excellent introduction to the techniques of stress
analysis using the finite element method.
The element is shown in figure 5.23 and the assumed displacements
are

uX al + a2x + a3y (x direction)

(5.35)
uy aa. + asx + a6Y (y direction)

Fig. 5.23
Notice that the element is numbered in an anticlockwise directio~
Quantities which will be needed subsequently are defined as follows

al X3 X2 bl Y2 Y3

a2 Xl X3 b2 Y3 - Yl (5.36)

a3 = X2 - Xl b3 = Yl - Y2
Evaluating ux at each node (corner) gives

[!-
1 xl Yl

!::!
1 X2 Y2

1 X3 Y3

148
and solving for the a's

all l [X2Y3-Y2X3 X2Yl-Y3Xl

{ a2 = 2A Y2-Y3 Y3-Yl
a3 x3-x2 x1-x3
where A is the area of the triangle.

The y displacements at each node u2, u4 and u 6 will be connected

to a4, as and a5 in an identical way so that the full solution for
the a's is written, using equations 5.36
al X2Y3-Y2X3 0 X3Yl-Y3Xl 0 XlY2-YlX2 0 Ut

a2 bl 0 b2 0 b3 0 u2

a3 a1 0 a2 0 a3 0 U3
1
= 2A 0 XIY2-Y1X2 u4
a4 0 X2Y3-Y2X3 0 X3Y1-Y3X1

as 0 bl 0 b2 0 b3 us

a5 0 a1 0 a2 0 a3 u6

(5. 37)
Recalling the definitions of strain in two dimensions ex aux/~.

e auy/ay, Yxy = au/ax + aux/ay leads to

y

~[:
0 b2 0 b3 ul

{e) {:: l a1 0 a2 0 OJ
a3 u2

Yxy 1 bl a2 b2 a3 b3 U3
u4

us

or {e} = 2~ [B) {u} (5. 38)

The stress-strain relationships in two dimensions are

149
 l
where for plane strain,

[E] E
(1-v) (l-2v) r:v
0
l~v
0 l-2v
: (5. 39)

-2-

and for plane stress

,;j
v
E

[:
[E] 1 (5 .40)
l-v 2
0

and E is Young's Modulus.

The forces acting at the mid points of each side of the triangle
may be written in terms of the nodal coordinates, as shown in figure
5.24.

Fig. 5.24

For example, the force acting at the mid point of the side joining
nodes 1 and 3 is

in the negative X direction (using equations 5. 36). Each of the

mid-side forces is now divided equally between its adjacent nodes
so that

Pl b3+b2 0 a3+a2

P2 0 a3+a2 b3+b2

(:~l
P3 bl+b3 0 al+a3
{p} - [%) 0 bl+b3
P4 al+a3

Ps bl+b2 0 al+a2

P6 0 al+a2 bl.bl

150
and again noting that b3 + b2 -b1 etc. from equations 5.36, this
becomes

bl 0 al
0 al bl

b2 0 a2
{p} [I) 0 a2 b2
{a}

b3 0 a3

or (see equation 5.38)

{p} = [t) [B]T {a} (5.41)

Using the appropriate form for [E] (equations 5.39 and 5.40) and
substituting for {a} = [E]{e} from equation 5.38 gives

{p} = ( 4~) [B] T [E] [B] {u}

and the element stiffness matrix is therefore

[K] = [ 4~) [B] T [E] [B] (5. 42)

The reader should compare this method with the energy method
used to derive the same results in example 4. 7.
The method ~ed in chapter four is of course more general and
can be applied to an element of any shape. Also, more sophisticated
and accurate displacement fields can be used. The difficulties
presented by the integral in the expression for potential energy
are overcome using numerical techniques.
5.2.2 Assembly of Structure Stiffness Matrix

Consider the 'structure' shown in figure 5.25 which is to be

simulated by two plane triangular elements of the type just
discussed.
The stiffness equations for each element can be written in the
form
{p} [K] {u}

.151
 4....,~7
0
'zll1
1
Fig. 5.25

or Pl Ku K12 K13 K14 K1s K16 ul

P2 K21 K22 K23 K24 K2s K26 u2

P3 K31 K32 K33 K34 K3s K35 U3

P4 K41 K42 K43 K44 K~ts K45 u4

Ps Ks1 Ks2 Ks3 Ks4 Kss Ks6 us

P6 K61 K52 K53 K54 K5s K66 U6

using the notation of the previous section and figure 5. 25.

The stiffness equations for the whole problem will include
eight loads and eight displacements. The overall stiffness matrix
therefore is 8 x 8 and the problem is to combine the coefficients
of the element stiffness matrices so that they produce the 8 x 8
stiffness matrix which represents the overall behaviour of the
structure. A set of 'equilibrium' equations can be written as
follows - see figure 5.25.

Ps 1Ps + 2Psi P6 = 1P6 + 2P6i P7 = 2P3

Pa = 2P4 (5.43)

The reader should study these equations carefully and note that
the conventions used in deriving the element stiffness matrix are
carried through here. The prefix refers to the element number.
Thus, P5, the X load at node 3 of the structure is composed of the
X load at node 3 of element (1) ( 1ps)and the X load at node 3 of
element (2) (2Ps). Notice that node 4 of the structure is node 2
of element (2) because the nodes of the element must be taken in
anticlockwise order. Thus

P7 = (the X load at node 2 of eleaent (2)) 2P3

152
 The equation for Ps can be further expanded as

Ps = 1Ps + 2Ps 1Ks1 1u1 + 1Ks2 JU2 1Ks6 lu6

+ 2Ks1 2u1 + 2Ks2 2u2 2Ks6 2u6
and similarly

P7 = 2P3 2K31 2u1 + 2K32 2u2 2K36 2U6 etc

Another look at figure 5.25 will show that

ul 1u1; u2 = 1u2; u3 = JU3 = 2u1

u4 1u4 2u2; Us }US = 2us;

u6 lu6 2u6; u7 2u3; Us = 2u4 (5.44)

and these are sometimes referred to as compatibility equations.
They enable element displacements in the equations for the structure
loads (above) to be replaced by structure displacements. The
equation for Ps then becomes

Ps 1Ks1U1 + 1Ks2U2 + C1Ks3 + 2Ks1)U3 + C1Ks4 + 2Ks2)U4

+ C1Kss + 2Kss)Us + C1Ks6 + 2Ks6)U6 + 2Ks3 U7 + 2Ks4 Us

and the complete set of equations is

0 0

0 0 U2
P3 1K31 1K32 1K31+2K11 1K34+2K12 1K3s+2K15 1K35+2K16 2K13 2K14 U3
P4 1K41 1K42 1K43+2K21 1K44+2K22 1K4s+2K2s 1K46+2K26 2K23 2K24 U4
Ps 1Ks1 1Ks2 1Ks3+zKs1 1Ks4+zKsz 1Kss+zKss 1Ks6+2Ks6 zKs3 zKs4 Us
P6 1K61 1K62 1K63+2K61 1K64+2K62 1K6s+zK6s 1K66+2K66 zK63 2K64 U6
0 0

0 0 2K43 2K44 Us
(5.45)

Alternatively, an individual stiffness coefficient, for example

2K43 may be located in the overall stiffness matrix by reasoning
as follows:- 2K43 is associated with 2P4 and 2u3 of element (2).
From the equilibrium equation 5.43, 2p4 is part of Ps and from the
compatibility equations 5.44, 2u3 = U7. Therefore zK~t3 is to be
found in row 8, column 7 of the overall stiffness matrix.

153
 A numerical example will also help to make the preceding sections
more clear. In figure 5.25 the coordinates of the nodes are as
shown below

Node X coordinate y coordinate

1 0 0
2 1 0
3 0 2
4 1 2

and so for element ( 1) , using equations 5. 36 and 5.38 the matrix

[B] is

dB]

[:
-1
0

-1

-2
2

0
0

2
0

1
]
~ ~
1~2 ~1
Fig. 5.26 Fig. 5.27

and similarly for element (2)

0 2 0 -2
2 [B] [ :
-1 0 1 0

-1 0 1 2 0
_:J
The area of each element is 1 and, assuming plane stress and
v = 0.3 (equation 5.40), the stiffness matrix [K] for each element
is calculated as (using equation 5.42)
4. 35 1.3 -4 -.7 -. 35 -.6

2.4 -.6 -1.11 -.7 -1

4 0 0 .6
I(K] l.l[t;)
symmetrical 1.4 .7 0

. 35 0

154
and . 35 0 -. 35 -.7 0 . 7
1 -.6 -1 .6 0
4.35 1.3 -4 -.7
I(K) l.l[t:)
symmetrical 2.4 -.6 -1.4

4 0

1.4
Note that the coefficients in the expressions for 1 [K] and 2 [K]
are identical but that they appear in different places. This is
to be expected because.elements (1) and (2) are the same size and
shape but are differently orientated. The expression for 2 [K)
could have been obtained from the expression for 1 [K] by a
suitable rearrangement, or transformation. This will be further
discussed in section 5.2.4.

Values may now be substituted into equation 5.45 to give the

overall stiffness matrix.

4. 35 1. 3 -4.0 -.7 -.35 -.6 0 0

2.4 -.6 -1.4 -.7 -1.0 0 0

4.35 0 0 1.3 -. 35 -.7

2.4 1.3 0 -.6 -1.0
[K)
symmetrical 4.35 0 -4.0 -.6

2.4 -.7-1.4
4. 35 1.3
2.4

5.2.3 Nodal Constraints and Solution of Equations

Given an arbitrary set of loads [P) the equations 5.45 could not
be solved for the unknown displacements [U] for the reasons already
given in section 5.2.1. The equations are singular because no
restriction has been placed on any displacement and the structure
is therefore free to move as a rigid body. In order to reduce the
equations to solvable form a sufficient number of displacements
must be specified so that rigid body motion is prevented. If only
this number of displacements is specified, the problem is statically
determinate. If more than this minimum number of constraints is
specified the problem becomes statically indeterminate because there
are redundant constraints. In either case, the method of solution
is identical. For example, in figure 5.25 assume that u1, U2 and u4

155
are all zero and that loads are applied to nodes 3 and 4. This
means that the x and y displacements at node 1 are zero and the y
displacement at node 2 is zero and the problem is statically
determinate, that is the reactions in the x and y directions at
node 1 and the reaction in the y direction at node 2 can be found
from the equations for statical equilibrium.

The procedure is now to remove the rows and columns relating

to displacements ul, u2 and u4 from equations 5.46. This leaves
4.35 0 1.3 -.35 -.7 u3

[::1 4. 35 0 -4.0 -.6 Us

2.4 -.7 -1.4 l.l[t;) UG (5. 4 7)

EI
4. 35 1.3 u7
2.4 Us

n
where the rows which have been removed are
ul

r-~
1.3 -4.0 -.7 -. 35 -.6 0 0

p2

p4
= l.l[t4E) 1.3

-.7
2.4

-1.4
-.6

0
-1.4
2.4
-.7

1.3
-1.0
0
0

-.6
_,:J u2
U3
0

u4 0

Us
UG

u7
ua
(5. 48)

The 5 x 5 set of equations 5.47 can now be solved for the five
unknown displacements U3, Us, UG, U7 and Ua. Notice that the
reduced matrix is still symmetrical, but is no longer singular.
Having found the unknown displacements the reactions P1, P2 and
P 3 can be found by substitution in equations 5.48, remembering
that u 1 = u 2 = u4 = o.

In the above example, if the applied loads are P3 Ps Pa 0,

PG = P7 = 1, equations 5.47 becomes

156
 0 4.35 0 1.3 -.35 -.7 U3
0 4.35 0 -4.0 -.6 Us
1 1.1[~) 2.4 -.7 -1.4 Us
1 4. 35 1.3 u7
0 symmetrical 2.4 Us

which on solving gives

-.1679
u,1
Us 2.6971

"'u7J . ,\ [,~] 1. 2127

2.9749

Us -.2787

and on substituting these values for the U's in equations 5.48 the
reactions are calculated as - see figure 5.28.

pl} {-1.000005}
{ P2 = -2.9993

plf 1.99999

~-3 ~2
-~...a"'_-1
_ _~..,. P:fO

Fig. 5.28

As the problem is statically determinate, the above reactions

could have been calculated by application of the conditions of
equilibrium for co-planar forces.

An example of a statically indeterminate problem is obtained

by making UJ = Uz = U3 = U'+ = 0.
 Equations 5.45 now become

0 4. 35 0 -4 -.6 Us
1 2.4 -.7 -1.4 Us
1.1 [E4t)
1 4.35 1.3 u7
0 synunetrical 2.4 Ua

and the 'removed' equations are

pl 4. 35 1.3 -4 -.7 -.35 -.6 0

p2 1.3 2.4 -.6 -1.4 -.7 -1 0
l.l[E4t)
p3 -4 -.6 4.35 0 0 1.3 -.35 -.7 u3 =0
plf -.7 -1.4 0 2.4 1.3 0 -.6 -1 Ulf =0
Us

Us

u7

Ua
and the solution is

Us 2.6902 pl -1.61171

Us 1.1169 p2 -3.00004

u7 2.9681 (E~) /1 and

p3 0.61173

Ua -0.2837 plf 2.00010

Again, examination of the reactions shows that the force system

is in equilibrium.

Having found the full set of displacements the three components

of strain (and hence stress) for any particular element can be
calculated from equation 5.38. If it is required to display a
stress distribution for a particular component, the stresses are
usually taken to act at the centroid of each element.
5.2.4 Transformation of coordinates

The end load element (section 5. 2 .la) and the simple bending element
(section 5.2.lb) will generally be employed in structures where
theoverall coodinate system does not coincide with the element
coordinate system used to derive the element stiffness matrices.
Consequently, the element stiffness matrix must be transformed
from element coordinates to overall or structure coordinates before
being assembled into the overall stiffness matrix.
158
 Fig. 5.29

Figure 5.29 illustrates the problem for a simple end load element
with nodes at 1 and 2. In 'element' coordinates the stiffness
equations for the element are (equation 5.31)

Pl'] = Et [ 1 -l]lul'] (5. 31)

l
P2' _-1 1 u2'
where primes refer to the element system.
In the global system, loads and displacements are to be measured
in the X and Y directions so that the element stiffness equations
need to have the form

P1 Ku K12 K13 K14 Ul

P2 K21 K22 K23 K24 u2

(5.49)
P3 K31 K32 K33 K34 u3

P4 K41 K42 K43 K44 u4

where u1 is the X displacement at node 1 and so on.

The problem does not arise for the constant strain triangular
element of section 5.2.lc because the element stiffness matrix is
derived directly in terms of global coordinates. However, 'diffi-
cult' or unusual nodal constraints are sometimes encountered which
make a transformation of coordinates desirable. For example at the
support shown in figure 5.30 theY' displacement is zero whilst
displacement in the X' direction is allowed. This would be
difficult or inconvenient to describe in the global XY coordinates.

Returning to figure 5.29 the forces acting at node 1 are related

by the expressions

Pl = Pl 1 cosa P1'~

159
 P2 =PI' cos(~/2- a) = P1'm
or in matrix form

c:} = 1:1 PI' (5. SO)

with an identical relationship at node 2.

Similarly for the displacements

UI' = UI cosa + u2 cos(~/2- a) uit + u2m
or in matrix form

UI 1 = [t m]fl} (5.51)

' ll2
with an identical relationship at node 2. Equations 5.50 and 5.51
are more generally written
{p} = [T]{p I} (S .52)

and { u' } = [T] T{u} (5.53

where [T] is a matrix of direction cosines appropriate to the trans-
formation required and [T]T is the transpose of [T]. Using these
expressions in the element stiffness equation 5.30 gives
{p} = [T]{p'} = [T][K']{u'}
therefore
{p} [T] [K' ] [T] T{u} (5.54)

and [K] [T][K'][T]T (5. 55)

where [K] is the element stiffness matrix in global coordinates ~nd
[K'] is the element stiffness matrix in local or element coordinates.
In the example shown in figure 5.30 only the elements containing
node j need be treated and only the loads and displacements at node
j need be transformed. The transformation (for the whole element)
is then

Fig. 5.31

160
 PI 1 0 0 0 0 0 PI

P2 0 1 0 0 0 0 P2

P3 1 0 0 R. m 0 0 P3
Plf' 0 0 -m R. 0 0 Plf

Ps 0 0 0 0 1 0 Ps

P6 0 0 0 0 0 1 P6

or {pI} [T] {p}

T
and {u} = [T] {u'}

The element stiffness matrix is now transformed as described

above, then added to the overall stiffness matrix in the usual
way. On specifying the nodal constraints it is then only necessary
to make Uy (the Y displacement at node j which has been transformed
to the Y' direction) equal to zero. On solving for the reactions
the value obtained for Py will be the reaction in the direction Y'.

5.2.5 Applications of the Finite Element Nethod

Example 5.4 Pin-jointed 3D framework

Fig. 5.31

Pin-jointed is taken to mean that no bending moment can be

transmitted by the connection in any direction, so that the six
elements in figure 5.31 are end load elements. They are the same
length R. and are assumed to be identical.

There are four nodes, each with three degrees of freedom so that
the overall or structure stiffness matrix will be 12 x 12 in size.
If the end load element has direction cosines R., m and n with
reference to the global coordinates, the required transformations
are

161
 Pl !1. 0

P2 m 0

P3 n 0
t'l} or [p] [T] [p'] for loads
0 !1. P'2
P~+

Ps 0 m

P6 0 n

and ul

u2

[:
mn
f'l} 0 0 0] u3 [T) T [u)
or [u']
u'2 0 0 !1. m n Uq

us

UG

for displacements.

Using these expressions and equations 5.31 gives the stiffness

matrix for an end load element in global coordinates

[K) I(-C)
(C)
(-C) I X EA
(C) L

where

j:: :: ::)
[C]

n!l. nm n2

The direction cosines for each element are shown below so that
it is a simple matter to complete the stiffness matrix for each
element using the above expressions.

162
 Member/Direction R. m n
Cosines
1 1 0 0

1
2 -2 ~2 0

3
1
2
112 0

4
1
2 273
1
I!3
5 -2
1
273
1 II3

6 0 ;r3 I!3
The overall stiffness matrix is then assembled according to the
procedure of section 5.2.2 and is shown in symbolic form in figure
5.32. Substitution of the appropriate values for the coefficients.
gives the stiffness matrix as
1
13 1 1
3
2 73
1 1
76
-1 0 0
4
1
-4 0 -4-m -76
13 1
5 1
6 372 0 0 0 --;r
3
-4 0
1
-m -rr1 -372
2 1 1 2
3 0 0 0 0 0 0 -76 -372 -3
3 1 1 1 13 1 1 1
2 -73 -m -4 4 0
4 473 76
13
5
6
1
76 4
3
-4 0
1
473 -rr1 -312
1

2 1 1 2
Symmetrical 3 0 0 0
76 -372 -3
EA
k =T 1 0 0 0 0
0
2
11 12 1 12
6 3
0 -3 -3
2 12 2
0 -3
3 3
1
0 0
2
1 4
2 372
2

163
 IK11 lK12 lK13
3K 11 3K 12 3K 13 1K 14 1K 15 1K16 3K 14 3K 15 3K 16 4K 14 4K1 s 4K 16
4K 11 4K 12 4K 13
1K 22 1K 23
3K22 3K 23 1K 24 lKzs lKzs 3Kz 4 3Kzs 3Kzs 4Kz 4 4Kzs 4Kzs
4K 22 4Kz3
1K33
3K33 1K34 1K3s 1K36 3K34 3K3s 3K36 4K34 4K3s 4K36
4K33
1K44 1K4s 1K46
2K 11 2K12 2K13 2K 14 2K 15 2K 16 SK 14 SK 15 SK 16
SKu SK12 SK13
lKss lKss
2Kzz 2Kz3 2K 24 2K 25 2K 26 SK 24 SK 25 SK 26
5Kzz 5Kz3
lKss
2K33 2K34 2K3s 2K36 SK 34 SK 35 SK 36
SK33
2K44 2K4s 2K46
3K44 3K45 3K46 6K 14 6K 15 6K 16
6Kn 6Kl2 6K13
2Kss 2Kss
3Kss 3Kss 6K 24 6K 25 6K 26
6Kzz 6Kz3
2Kss
3Kss 6K34 6K35 6K36
6K33

4K44 4K45 4K46

SK44 SK4s SK46
6K44 6K45 6K46
4Kss 4Kss
SKss SKss
6Kss 6Kss
4Kss
SK66
6K66

Fig. 5. 32
Overall Stiffness Matrix for Structure of Fig. 5.33

Suppose now that the structure is fixed at the three bottom

corners and therefore the only possible displacements are u 10 , u1 1
and u 12 . Assume that the node 4 carries a unit load in each
coordinate direction, that is P1 0 = P11 = P12 = 1.

164
 The equations to be solved reduce to

l:::j E~ l l
whence

= 9.5:47
u12 -3.98528

Now that the entire displacement vector is known, the reactions

can be found by substitution into the overall equation. This is
left as an exercise for the student who should also check that the
entire structure is in equilibrium.

Example 5.5 Statically indeterminate beam

~:
(!)

:fc
@
L1
tw
Fig. 5. 33
L2

Matrix methods for the solution of structural problems have been in

use for many years and are particularly useful for the type of prob-
lem shown in figure 5.33 where there is a redundant constraint and
a shaft of suddently varying cross-section.
Using the results of section S.2.lb the overall stiffness equat-
ions are easily assembled as

12 E1I 1 _12 E1I1 6E1I 1 U1

L13 ~
L1 L13 L12
0 0

4 E111 -~ 2E111 0 0 u2
L1 L2 1 L1
12 [~+ E2I2) _6 (E1I1 _ E2I2) _ 12 E2I2 6Ez..!z.. u
r:;z
L1 3 """"(2T L22 LzT L22 3
L1
+
4 [Ell 1 E2I2)
L2
-6E212
L22
2E2I2
L2
u
4

Ps 12 E2I2 _6 E2I2 u
L23 L22 5

4E2I2 uG
L2
165
 The nodal constraints are U1 = u2 U5 0 and the load vector is
T
{P} = {P 1, P2 - W, 0, P5, - C}
The three equations to be solved are for U3, U~ and UG.

Example 5.6 Plane stress problem with skew support

The problem shown in figure 5.25 is again considered but this time
with a skew support, that is, node 4 is constrained to move in the
X' direction (Fig. 5.34). The nodal constaint at node 4 is clearly
that the Y' displacement must be zero. Only element (2) is affected
and the necessary transformation matrix is (referring to section
5.2.4)

1 0 0 0 0 0

0 1 0 0 0 0
1 1
0 0 +72 -72 0 0
[T] 1 1
0 0 0 0
Tz Tz
0 0 0 0 1 0

0 0 0 0 0 1

since R, cos- 1 ( -45) and m = cos - 1 (90 ( -45))

I
y

Fig. 5. 34

The transformed stiffness matrix for element (2) is then

[K'] = [T] [K] [T] T

where [K] is given in section 5.2.2. Following the rules of matrix
multiplication gives

166
 0.35 0 0. 247 -. 742 0 0.7

1 o. 282 -1.131 0.6 0

2.575 0.975 -2.404 0.495

2 [K'] l.l(E4t)
4.675 -3.25 -1.485

4 0

synunetrical 1.4

Assembling the overall stiffness matrix in the usual way gives

4. 35 1.3 -4 -0.7 -0.35 -0.6 0 0

2.4 -0.6 -1.4 -0.7 -1 0 0

4. 35 0 0 1.3 0.247 -0.742

2.4 1.3 0 0.283 -1.131

[K] 1.1 [E4t)
4. 35 0 -2.404 -3.253

2.4 0.495 -1.485

2.075 0.975

synunetrical 4.675

Notice that this result could have been obtained by applying the
transformation

1 0 0 0 0 0 0 0

1 0 0 0 0 0 0

1 0 0 0 0 0

1 0 0 0 0
[T]
1 0 0 0

1 0 0
1 1
72-72
1 1
zero 72 72

to the overall stiffness matrix given in equation 5.46.

167
 Applying the nodal constraints U1 = U2 = U3 U4 u 8 0 and
noting that the load vector is

{P} = {P 1, P2 , P 3, P4 , 1, 1, 0, P'a}T

leaves the equations to be solved as

2.4
- 2 404]
.495 UG
{usl
.495 2.075 U7'

l
whence

5
u ]
UG =
{.527]
306 [ E~) l \
U7' .537

and substituting back for the reactions gives

pl -. 368

Pz -.675

p3 .530

p4 .837

Pa' -1.644

Once again it is left as an exercise for the reader to cl1eck

that these reactions are in equilibrium with the applied loads at
node 3- see figure 5.35.

Pe;=1 e8 .. -1-644
I

R=1
I
~=0
7

~--'675 ~=837

P.=-368 Pa'530
1

Fig. 5.35

168
5. 3 Problems
1. Use the finite difference method to deduce the applied torque T
in terms of the twist per unit length a for a shaft of square cross-
section of side a. Adopt a square mesh of size 6 x 6 and note that
only one eighth of the cross-section need be considered due to symm-
etry - see figure.
/
!Y'D
~8
A a

a
I

2. C~lculate the value of ~ at the node points for the hexagonal

shaft with square hole shown in the figure. Estimate the value of
the maximum shear stress and suggest ways of improving the accuracy
of the estimate without increasing the number of nodes.
24 2
TI Gah
v ~
IB ~
~A=

/ ~B = 20 Gah2
~ ~ 13
~ A
~c = 30 Gah2
~ v 13
~ v Tmax = [*] D = ~j Gah
A better approximation can be made by fitting a polynomial to ~C'
~A and ~D (= O) and differentiating.

3. Deduce the Airy stress function ~ and a~/ax and a~/ay around
the boundary of the thin plate loaded as shown in the figure and
hence deduce the expression for ~ within the plate.
( -a ~ x ~ a, y = 0) "' = klx3 l l = klx2 l l = 0
- - "' 6 ' ax 2 ' ay

_~ k1a 3 l l l l = k1a 2
(0 ~ Y ~ b' x = a) ~ - 2 + 6 ' ay = kzy' ax 2

uy.. ty
(-a ~ x ::; a, y r---r~-L..._, UJ!'O

whence

169
4. Six bars, all exactly alike, are connected at four joints to
form a regular tetrahedron in such a way that the bars carry axial
load only - see figure. A load P with direction cosines (2/7, 6/7,
3/7) is applied at joint 4. Describe how you would set up the
stiffness matrix for this structure. State why the matrix, as it
stands, is singular and how you would reduce it to a solvable form.

(Joint 1 is fixed in the XYZ direction

Joint 2 is fixed in the X and Z direction
Joint 3 is fixed in the Z direction only)

Illustrate your answer by using (4) with ends at joints 1 and 4 as

an example.

z
4

(For member (4), direction cosines are = 1/2, m = 1/213 and n =

12/3 and the stiffness matrix for member (4) in structure coordin-
ates becomes

1 1 1
1
4 473 76
1 1 12
2
473 TI IT
1 12
3
76 TI
EA
L
10

+ 11

12

1 2 3 10 11 12

Each term is then placed in the structure stiffness matrix

according to the structure displacement numbers alongside the
above matrix.

170
5. A rectangular plate of uniform thickness t is to be simulated
by two constant strain triangular elements of identical proportions
as shown in the figure. The element stiffness matrix for element
number 1 with corners at nodes 1, 2 and 3 is

4. 32 1. 30 -4.00 -0.70 -0.35 -0.60

2.40 -0.60 -1.40 -0.70 -1.00

4.00 0.00 0.00 0.60

K l.lOEt
=--4-
1.40 0. 70 0.00

0.35 0.00

symmetrical 1.00

where E is Young's Modulus and Poisson's ratio has been assumed to

have a value of 0.3 (plane stress).

Write down, by inspection, the stiffness matrix for element 2

and hence deduce the overall stiffness matrix for the assembly. If
nodes 1, 2 and 4 are fixed and a load of 1 unit is applied in the
y direction at node 3, calculate the unknown displacements and the
reactions at each node.

Comment of the accuracy of the values you have obtained. What

difference would it have made if the plate had been divided along
the line 1 - 4 instead of 2 - 3?
4 P=1
us = 0, u6 = 2.64Et
-.7
Plf 0, P7 2."4 2 units
-1 -1.4
P2 = 2.4' Ps 0, Ps =2T

1.3
P3 = 2.4' PG 1

171
6 PROBLEMS IN BENDING OF BARS AND PLATES

In this chapter a number of problems are studied in which bending is

responsible for the main stresses. The method of approach is not
purely from elasticity considerations, but rather ad hoc, which is
often called 'strength of materials'. It should be appreciated that
actual conditions should comply with the assumptions made for the
results of calculations to be accurate enough for design purposes.

6.1 Beams on Elastic Foundations

Consider a bar supported on a foundation (figure 6.1) which cannot

be regarded as being rigid, but responds to forces acting on the bar
by deforming itself. The support in this problem is one which will
itself cause the foundation to deflect in the same manner as the
beam. This is equivalent to fastening the bar to the foundation.
In what follows the foundation is considered to be elastic so that

1
the deflection (or deformation) at any point along the bar will be
proportional to the reaction (p) per unit length at the point,
Arbitrary loading including weight
of beam

fiY(~ X

Fig. 6.1

that is
p (J -y
therefore
p = -ky

Usually the bar will be loaded by forces acting on it, setting up

the reaction p, but in the equation which follows the general curve
is first considered and any distributed forces acting on the bar can
be taken into account by including a particular integral to the
solution. Other forces can be taken into account in the constants
of integration.

Thus from figure 6.1, using elementary beam theory

EI(d'+y1 = p = -ky
dx4 J

or ~+ (kJ
Ef y = 0 (6 .1)
dx 4

172
k is the foundation constant (units - force/unit length/ unit
deflection) for a bar of unit width. (Note that k must be modified
if the bar has a width which is different to the unit chosen for
determining k.) Implied in the foregoing are the usual assumptions
made in elementary beam theory.
The solution to equation 6.1 is

-AX AX DsinAx)
y =e (ACOSAX + BsinAX) + e (CCOSAX + (6. 2)

where A4 = k/4EI and A, B, C and D are constants of integration to

*
be determined from the boundary conditions. (Units of A = 1/length.)
From elementary beam theory, the following formulae apply

= the slope (6.3a)

-EI (d2y) sagging bending moment (M) (6. 3b)

dx 2

-EI[~)3 shear force (F) (6. 3c)

dx

EI[d 4y) =loading (w) (6.3d)

dx 4

In a loaded beam supported on an elastic foundation, sufficient

end conditions may be known and the values of A, B, C and D found.
However, the procedures for dealing with a complicated loading
involves a considerable amount of work.

The problem is solvable in a simpler way if beams of unlimited

length are considered and the principle of superposition applied.
In this way a number of results can be obtained but in particular
some useful mechanical engineering results can be obtained without
resorting to the more general theory of shells.

6.1.1 Semi-infinite and Infinite Beams

6.1.1a Concentrated force on semi-infinite bar

In figure 6.2 a semi-infinite beam is loaded by a force P at its end,

as shown.

The equation
-AX AX
y = e (AcosAx + BsinAx) + e (CcosAx + DsinAX)

applies and the constants are to be evaluated for the known end
conditions in the usual way, viz:-

173
 X

Fig. 6.2

When x ~ oo, y must ~ 0 and since eAX ~ ~, therefore C and D must

be zero, hence

y = e -AX (ACOSAX +
.
BslnAX ) (6. 4)

and this equation tends to zero as x tends to infinity, as would be

expected. This means simply that the effects of P, the disturbance
becoming less the greater the distance from P. Using equations 6.3,

*
for the conditions that at x = 0, M = 0 and F = -P, then different-
iate equation 6.4

= Ae-AX((B - A)cosh + (-B - A)sinAx}

d2v __ AX
~ A2e- (-2BcOSAX + 2AsinAx) (6.5)

d 3Y = A3e-AX(2(A + B)COSAX + 2(B - A)sin x)

dx 3
Hence, bending moment is

M =- El[~) =- EIA 2e-AX(-2BcoSAX + 2AsinAX)

dx 2

Shear force is
F= - EI [d 3y)
dx 3

There>fore

0 2B

or B 0

and

or A =-p-
2EIA 3

174
therefore
-AX
Y =~ COSAX
2EIA 3

Pe-AX
M = - -A- sinAx (6. 6)

-AX (COS Ax SinA X)

F Pe -

Note that AX has units length/length, that is an absolute number

(radians), and A- 1 is sometimes known as the characteristic length.
(This is analogous to wavelength in vibration problems.) It can be
seen from equation 6.6 that y, M and F are all oscillatory in
character.

6.l.lb Concentrated force on infinite bar (figure 6.3)

p

y
Fig. 6.3 Fig. 6.4
Again, take the origin through the line of action of the ~orce P and
x along the undeformed line of the foundation. Since the length of
the bar is infinite the loading is symmetrical and only half need be
considered (semi-infinite), figure 6.4.

Because of symmetry there must be a reactive couple M0 applied at

the load point. The end conditions are x = 0, M = M0 , F = -P/2 and
dy/dx = 0.

Now, as before, y = e-AX(AcosAX + BsinAx) applies and using

equations 6.5 gives, and hence in terms of the known force P
p
- 2 = -EIA 34A
p
or A = B =--
8EIA 3

Hence
Pe-AX PAe-AX
y = --(COSAX + sinh) = zr(cosh + sinAx)
8EH 3

175
and M
0
=~
4A.

6.l.lc Couple acting on an infinite bar (figure 6.5)

The equation for a single force on an infinite beam has the general
form

y = ke -A.x( cosA.x + s1nA.x

. )

where k = PA./Zk, that is y = kA(A.x) for the force P at 0. For the

force -P at 01

Y1 = -kA(A.(x - a))
since P is reversed, so at any station x

y = -k[A(A.(x- a)) - A(A.x)]

(This analytically is the same as -k[f(x- a) - f(x)] and can be

expanded using Taylor's theorem.

Thus

f(x - a) - f(x) = f(x) - a[:!) - f(x)

no further terms being required if a is small.

Or in the problem

so y = - PZA.k2 .2ae-A.xsinA.x = - Pa(A. 2 e:A.x)sinA.x

and the limit

Lt(Pa) M
a .... o

y (MA. Je -A.x Sl.nA.x

= - l-r-
2 . (6. 7)

the deflection for a concentrated couple (figure 6.6)

Fig. 6.5 Fi!!. 6.6

176
6.l.ld Uniform loading on semi-infinite 6. 7)
w/unit length x

a b
y
L
Fig. 6.7

The deflection at A produced by a small force (wdx) is given by

y = ~~(COSAx + sinh)

where P = wdx, therefore the deflection at A for a length of loading

L

b wdxAe-AX fa d - -AX
y-- f 2k (COSAX + sinAX) + 2kx (Ae
W )(COSAX + sinA)
0 0

=
wAk fb (COSAX
2 +
.
S1nAX)e-AXdx +
WAk fa (COSAX
2 + sinAx)e -AX dx
0 0

=
2k t
WA -e -AXCOSAX-~b
A
-0
+ 2k
t
WA -e -AXCOSAX a
A 1 0

Y = 2wk [ 2 - e-Ab COSAb - e -Aa COSAa J (6. 8)

If a and b are large enough, then

w
y=k

so that at sections a good distance from the ends of the load, bend-
ing of the beam can be neglected (figure 6.7). There are many other
structural problems which can now be solved using the above results,
(see references). It should be noted that the following classifi-
cation applies to beams on elastic foundations.

Short beam is one in which Ai ~ n/4

Medium length beams n; 4 ~ Ai ~ n

Infinite (long) beams Ai ~ n

177
6.1.2 Discontinuity Stresses

A number of problems to determine the stresses in thin tubes utilise

the foregoing theory. A thin tube is one where the ratio of the
inner radius to wall thickness is greater than ten although some
authors do quote twenty. Consider a thin tube (figure 6.8) which is

Deformed position

Undeformed position
Fig. 6. 8 Fig. 6.9

acted on by a uniformly distributed force around the circumference

at a particular section. The force will produce a deflection, y sa~
which will vary along the length of the tube. Instead of consider-
ing the whole tube, consider a length AC of unit width and thickness
t which is small compared with r the radius of the tube.

At a section B (figure 6.9) the tube and hence B on AC will have

been displaced an amount 'y' as shown, therefore the circumferential
strain

211(r - y) - 2u
211r

e
e =-r.r
indicating compressive stresses in the wall due to the distributed
force. The circumferential stress is Ee 9

- Ey
r

and the force/unit length due to this is stress x area where

the area is 1 x t (figure

Fig. 6.10

that is
p
unit length

178
the resultant of the forces P balances the applied distributed force
and since the width of the element is 1, and re = o, then

2P sin[}) = Resisting force per unit length

Approximately,

. [e)2 2e
s1n +

and resisting force Pe = ~r

Hence

Resisting force
(unit length)

that is

the force a y (6.9)

and so the result satisfies the assumption made for a beam on an

elastic foundation that

p = ky

and k = Eto
r2

o is an arbitrary arc length which difficulty can be overcome by

making its value unity which implies that r must be large in order
that 8 is small. In addition r must be much larger than t in order
that the assumption that cr 8 across the wall is constant. So

where
(EI) I D (6 .10)

that is
12Et (1 - "11 2 )
(6.11)
4r 2 .Et 3

(that is A-1 still represents the characteristc length and A has the
units (length)- 1 ).

179
 Typically r/t ~ 20 (r the minimum radius) for a thin walled tube,
therefore

-1 t
and ). = 3.479

for a steel tube and if r/t = 10 then ).-1 = t/2.5. Theseexpressthe

fact that the effects of the force considered are local to the zone
of application.

Example 6.1

Reinforcing ring - figure 6.11

y
Fig. 6.11

Assume the ring is rigid and that the tube is long and subjected to
an internal pressure p. Then at a large distance from the ring the
tube deformation can be considered to be due to the internal
pressure p.

If the tube can be considered 'open ended', then

and 0

therefore

re 6 f[J'f) ~=
tE 0 (6.12)

or 6 = t
where k = Et
r2

If the tube is closed at the ends, then

180
 pr2 (6 .13)
and 0 = 2tE(2 - v)

Generally, unless a~ is very much greater than zero then equation

6.12 holds and equation 6.13 can be diregarded.

If the origin of the coordinates are taken through the deformed

line of the tube as shown in figure 6.12, then when x = 0, o = pr2/Et
(open ended).
y = e-AX(AcOSAX + BsinAx)

where ~~ = 0 at x = 0.

Substituting these values

y nr2 ( e- AX (COS AX
= "1ft + sinAX) }

Hence the restraining force due to the ring

-D(d3y) = - ~
dx3 2

p
2 pr 2t 2A3
or 0
3 (1 - \12)

M
pr 2 t~
6(1 - v 2)
0

since
A4 3(1 - v 2)
r2t2

6M
0
and a
max t2

The stress in the tube due to the presence of the ring is

_ pr 13(1 - v 2 )
ao - T 1 - v2

181
 = 1.816- (with v = 0.3)

that is 82% larger than the circumferential stress in the tube

before application of the ring.

Example 6.2

Flange on a long tube - figure 6.12

y
Fig. 6.12

Again the theory for a beam on an elastic foundation can be used.

The same system of coordinates as in the previous problem is used.
Then again depending on whether the tube has open ends or closed
ends, when x = 0, y = 6 = pr 2 /Et (open) or 6 = pr 2 /2tE(2 - v)
(closed).

If the flange can be considered to be a simple support then

there will be no reactive couple acting on the tube, figure 6.13a,
and then at x = 0, M = 0. If however there is a 'built-in' effect
then there will be n8 change of slope in the tube, that is at x = O,
dy/dx = 0. The results are shown under each figure.
~

y Fig. 6.13a

at x = 0, y =6 and X = 0, M
0
=0
y = 6e -AXCOSAX

M -2DA 26e -;l.xsin).x

36e -).x (COSAX

F -20~ - sin;l.x)

Hence
p -2DA 36
0

182
 -----+--ct
y
Fig. 6.13b

at x = 0, y =6 and x = 0, dy/dx = 0

-AX
y 6e (COSAX + sinAx)

M - 2DA 2 6e -AX (sinh - COS AX)

When X = 0, M = 2DA 2 6
0

F -4DA 3 6e-AXCOSAX

When A 0, p = -4DA 3 6 (compressive)

0

The foregoing examples illustrate the elementary use of the

theory of beams on an elastic foundation. To extend the application
further a number of references are given at the end of the book.

It should be noted that the more accurate theory of shells can

also be used instead of this theory. However, the setting up of the
equations for shells is not within the scope of this book.

6.2 Bending of Curved Bars

This section will deal with curved beams which may be considered
(a) thin, where only stresses due to bending or torsion are
significant and energy due to shear and axial stresses is insig-
nificant (say R/t >>10) and (b) thick bars which may be considered
as an addendum to chapter 2 where stress functions are used to
determine stresses in problems of plane bending.

6.2.1 Thin Curved Bars

To solve problems involving thin curved bars use is made of

Castigliano's Second Theorem, viz: aU/aPl = 61. The theorem has
been developed in chapter 4 (section 4.3) where P 1 is one of the
forces acting on a linear elastic body and 6 1 is the corresponding
displacement, that is the displacement in the direction of the force.
Now U is the total elastic strain energy of the stressed body. The

183
 expression for strain energy is quite general and evidently contains
P1, P2 ..... Pr' etc. Use can be made of the second theorem to find
unknown reactions or general deflection equations, when necessary,
by using the concept of 'dummy forces'. Analogous with these
expressions are

where M. and T. are bending and twisting couples and B. andy.

l. l. l. l.
represent the correspondi ng slope or rotation. It is usual in
problems involving thin bars to ignore stresses due to tension or
compression forces and only to include torsional or bending strain
energy.

A few examples will illustrate the application of Castigliano 's

second theorem.
Example 6.3 Restricted movement

A curved bar is constrained to move in a given direction by the

application of a force P. As a consequence a reaction is set up.
p
p
A Q

Fig. 6.14a Fig. 6.14b

In figure 6.14a the point A of the bar is constrained to move

vertically in direction of P. Since A would, if free, have two
orthogonal components of deflection then a force Q must be exp-
erienced by the guide. An equal and opposite force acts on the bar.
By Castigliano 's second theorem aU/aQ = 0 and aU/aP = ~ the def-
lection along the guide. Assume the guide to be frictionles s, from
figure 6.14b the bending moment at C

e = PRsine - QR(1 - cose)

M 0 ~ e ~ Tr/2 (6.14)

au aM 9
Now au
aq = aMe aq (6.15)

and u = r2:1
2 ds (6 .16)

184
where s defines the curve.
If the radius of curvature is constant, then s R8 and ds Rd8.
Also

au JMa
<lMa = Ef Rd8

so

R3 f7f/2
or 0 = Ef [Psine - Q(1 - cos a) ] (- ( 1 - cos a)) de
0

Hence
Q= ___Z_P___ = 1.4037P
37f - 8
To find the movement of P

au
ail= 0

aMa .
and aP = sme

so 0 ~: C 2
[P>ine ~ Q(l ~ oo.e) ]>inede

PR 3
o = o.oB35 ET

on substituting for Q.

Example 6.4 out of Plane Bending

Figure 6.15 shows a point force P applied at the free end of a

cantilever bar of fixed radius R. The force is applied at right

185
angles to the plane of curvature of the bar. The bar will be sub-
jected to bending (Me) and twisting (Te) at any section.

At C

Me PRsine

Te PR(l - cose)

Once again the corresponding deflection at P is ~p = au;aP but,

the equation now has two terms to account for the bending and
twisting, that is
au _ au aMe au aTe
a'P- aMe. ai' +aTe aP (6 .17)

or - cos e) de

Substitution of Me and Te gives

p
Fig. 6.15 Fig. 6.16
If the amount of twist or the slope is required at C then 'dummy'
couples can be used. The integration then h.>.s to be carried out in
two parts (figure 6.16), MandT represent the vectors of these
couples.

I
0 ~ 0 ~ a

Me PRsine (6.19a)

Te PR (1 - cos e) (as before)

186
 l
a ~ e ~ Tr/2

Me PRsine + M (6.19b)

Te = PR(l - cose) + T

for the directions shown and

au
aM = t3

au
and aT = y

Now all the strain energy due to bending and torsion has to be
taken into account but only equations 6.19b will be necessary in
the final equation since equations 6.19a do not contain M and T.

Thus for example

au au (6. 20)
aM = aMe

and aMe/aM and are/aM are non-zero only for equations 6.19b.

Hence

t3
au
aM
I PR~~ne.l.Rde
Tfh

a
(6.21)

which can be integrated to give the required slope at any angle a.

An expression can be found for y in a similar way.
Example 6.5 Unknown Constraints due to Slip

A curved bar slips vertically by a small amount o at one end.

Determine the reactions for the following conditions.

(i) Assume one end pinned- figure 6.17. Since A is pinned, there
cannot be a bending moment there. At C

Me = PRsine - QR(l - cose) (6.22)

au
Now -- = 0
aQO

au
and ail= 0
0

187
and proceeding as in example 6.3 gives the required results for P0
and Q0

Fig. 6.17 Fig. 6.18

(ii) Bar built-in at both ends- figure 6.18. A is built-in as
well as B, then fixing moments would be required at both A and B.
Suppose now end A slips an amount 6 in the direction of P only, so
that no movement occurs in the direction of Q

Now taking moments from end A

and here

;~ = 0 (assuming no change of slope due to slip)

0

au
~=0
0

and au = 6
aP 0

and the resulting equations which give P0 , Q0 and M0 are

2P R
M =_o__
0 TT

(6. 23)

Q R p RTT
-M __...... + _o_
0 2 4

188
Example 6.6 Deflection curve (vertical component for a quarter
ring subjected to an end force) - figure 6.19
p

Fig. 6.19

The end A is free to move and the end G is built-in so that the
tangent is vertical there. Vertical components given by au;aq for
various a's are found as follows

Between A and C

PRsine (6. 24)

Between C and G

PRsin6 + QR(sin6 - sina) (l :; 6 :; Tr/2 (6.25)

and the limits are important.

Now the vertical movement of Q is

a~
-- +
au
aQ aM0

7fh M
0 + J EfD . P2 (sin6 - sina)de
(l

= !._.
3 J7fh Psin6(sin6 - sina)d6
EI
(l

and since Q is a dummy force it can now be made zero

Tr/z
PR 3 cosesina-~
oQ = Er [He - !sin26) +
-a

PR 3
Er I}- -r- !sin2a]
(l
(6.26)

189
 Putting a = 0 gives

Example 6.7 Uniformly loaded bar built-in at both ends- figures

6.20a and b

Built-in

(a)
Uniform load
w unit length

Fig. 6.20

In the diagram the reaction and couples at each end are shown.
M and T are shown as vectors for convenience of resolution about
a point on the bar. The loading per unit length is w.
Take moments about C (figure 6.20b)

M0 MA'"'' - TAinO - f'RdoRsin(O-o) VARsinO

0

and T0 MA'ino TA'"'" - f'RdoR(l-oo(O-oJ) VAR(l-ooO)

0

Hence

Me MAcose TAsine + wR2 (l-cos6) + VARsine

T
a = MAsine + TAcose - wR 2 (6-sin6) + VAR(l-cos6) } (6 .2 7)

Having obtained the general expressions for M6 and T6 it remains

to find MA, TA and VA, MB T8 and v8 .

190
 The usual statics equations apply

and a moment equation is obtainable by putting e = n in the equation

for M8 (equation 6.27) and similarly for T6 and the following cond-
itions apply at A if the end if built-in

No movement of the bar in the direction of VA' so

No rotation of the bar in the direction of TA, so

au = 0 (6.28)
a~

No flexure of the bar in the direction of MA, so

au 0
a~

These are all the equations necessary to obtain a complete

solution of this problem.

6.2.2 Bending of Thick Curved Bars with Large Curvature

The elasticity solution for a curved bar subjected to bending is

given by the equations 2.20 in which A, B and C have to be deter-
mined from the boundary conditions. There is no restriction on
the ratio of a/b except the obvious when a = b and when a = 0 when
the stresses would be zero or infinite (see figure 6.21). Thus for
pure bending the exact solution can be obtained and only at the
ends of the bar will the stresses be different from those calcul-
ated. An appeal to St. Venant shows that away from the ends, the
stresses would be as calculated. Two other situations can be
evaluated using a stress function and these are summarised below.

Fig. 6.21

191
 In figure 6.21 shear forces are applied at the free ends of a
curved bar. This is not now a problem of ?~ial symmetry but a
stress function can be deduced if it is assumed that bending
moments produce only direct stresses. Then the bending moment due
to P at e (figure 6.21) is proportional to sine and the direct
stresses will thus be proportional to sine. Thus assume a stress
function of the from ~ = f(r)sin9. Substituting this equation into
the compatibility condition (equation 2.17) yields

~ = (A 1r3 + B: + c1r + D1r~nr)sin8

and the stresses and displacements can be found from equations 2.15
and 1.21.

If P was applied to the curved bar as shown in figure 6.22a then

a similar procedure can be used to obtain ~ = f(r)cose but the
solution now includes the effect of a couple also applied at the end
(figure 6.22b). For the shear forces (figure 6.21) and couples
(chapter 2) applied to curved bars there is complete equilibrium.
However as shown in figure 6.22a a reaction of 2P must be applied as
indicated to balance the applied forces. This, as shown in figure
6.22b, results in an unresolved couple at the mid section which must
be reacted at the surfaces and thus, from the stresses obtained
from this function must be subtracted the stresses due to a couple
acting on the ends to obtain a solution for the bar subjected only
to forces applied as shown in figure 6.22a.

Fig. 6.22

With these results inclined forces can be dealt with and the
stresses found using the principle of superposition.

Whilst the foregoing give results which are exact for both
stresses and deformation there are a number of engineering problems
for which only the stress distribution is required. The following
simplified theory also allows for a variation in the cross-section
of the bar. (Figure 6.23)

It is assumed that transverse sections of bar originally plane,

remain so after bending. Due to bending a plane such as pp rotates
relative to a plane mm by an amount dS to position p'p'.

192
 b
+Y
a

Fig. 6.23

Let

R radius of curvature of centroidal axis

r = radius of curvature of neutral axis

e = distance between R and r

that is

e = R - r

Also take y to be positive when measured towards centre of

curvature and M is assumed positive when it produces tension on
outer fibres, that is when it decreases the radius of curvature.

At a distance y from neutral surface an element is extended an

amount yde and the strain in the element is

Extension of ss
Original length of ss
y
-_ r-:y- e
[de) (6.29)

and consequently the stress in the element when there is no lateral

pressure between layers is

E x strain

that is

a = --.2'_ [dee) (6.30)

r - Y

This distribution of stress is not linear, as with thin bars, and

the neutral axis will be displaced towards the centre of curvature.

193
 In the above expression r and de are unknown. To find them use
equations of statics

L(stress x area) = 0 (6. 31)

over the cross-section which express the fact that there is no

resultant force and this combined with

l:Cstress x area)y = -M (6.32)

gives the applied bending moment.

IA
adA = o- Ede
e
J ydA
A r - Y
(6. 33)

and the second of these equations gives

IA
aydA = -M = Ede
e
J y2dA
r- y
(6. 34)

Now the integral

Jy2dA =-fey- ___!L_)dA =- J ydA + JrydA

r-y r-y r-y

The first integral represents the first moment of are about the
neutral axis and the second integral from equation 6.33 is zero,
since r is fixed for any particular section.

Let Jy 2dA/(r - y) be denoted by Ae, that is - fydA Ae where A

is the cross-sectional area.

From equation 6.34

Ede M
e - Ae
and thus using equation 6.30

()" = -
My (6.35)
Ae(r - y)
The main shortcoming of the foregoing theory is that the bending
stress a only has been taken into account and radial stresses have
been ignored. A similar theory to that developed here for radial
stresses shows that for solid sections the radial stress is
generally small. However if the sections are discontinuous, for
example with I and T sections, then the radial stress can be
appreciable,particularly if the web is small in thickness compared
to the overall dimensions. Care has therefore to be exercised for
this type of section.

194
 If a shear force acts across the section then the elementary
theory used for shear stresses in beams can be used and this theory
can be found in most strength of materials text books. Similarly
a tensile force applied to the section may be assumed to produce a
tensile stress which is uniformly distributed across the section.
It must be remembered that the section must be some distance from
the point of application of the force for the above to apply.

For any cross-section (figure 6.24)

R r v R r
v

0 0 --.4.--Q

Fig. 6.24 Fig. 6.25

From equation 6.33

J~=O
r - y

and from figure 6.24

v = r - y

Hence on substituting for y

v~A = IdA
A
or r
fdA
v

(a) Hence for the rectangle (figure 6.25) since dA Bdv

195
and e =R- r
(b) Circular cross-section (figure 6.26)

n
R dv

-+--+-0
Fig. 6.26

Again
A
r =
R+afdA
R-a v

Now dA = bdv
1
and b = 2[a 2 - (R - v)2]'
'lfa2
So r =
1
r+a 2 [a 2 - (R -v)2]2 dv
R-a v

This is an awkward integral but yields:-

Another version of this formula (see references) gives

1 1
r = ~[(R+a) 2 + (R-a) 2 ]2

(c) Tee section (figure 6.27)

196
 (d) Trapezoidal section (figure 6.28)

~ b2 ~ b2
r---- I~ ~"I
dv~
I \
d2
- f - - - ~I I
h
c
n-- f - - - n'
v J
I
I -
-
~~n d h

d1 R r b1
k ,...;.- g
g b1 ~

...... - - ~
'V

Fig. 6.27 Fig. 6.28

b b2+[bl~b 2 ](h-v)
Hence

so that

r =

and finally

r = ____________ (~~-l~~-b-2~)~d_____________
r1 ~ b 2]Mn(~)- (b1 - b2)

Note that if b 2 = 0 then this equation for r reduces to that for

a triangular section.

197
6.3 Bending of Thin Circular Plates

In this section the plates which will be dealt with are those in
which the deflections of the middle surface are small compared to
the dimensions of the plate. Thus the slope at any position of the
plate is small and an analogy may be drawn from the theory of the
simple bending of beams in which the curvature can be written in
terms of the second differential coefficient only since squares of
the slope can be ignored. There are a number of references given in
which thick plates and thin plates with large curvatures are dealt
with.

The assumptions are

(1) Deflection of the middle surface of the plate is small com-

pared to the thickness of the plate.

(2) The middle surface of the plate remains unstressed during

bending,

(3) Points in the plate lying initially on a normal to the

middle plane of the plate remain on the normal to the middle
plane after bending.

(4) Normal stresses across the thickness of the plate can be

ignored.

(5) The plate thickness is small compared to any other dimension

The theory will be developed first using cartesian coordinates

and then using polar coordinates using the theory already outlined
in chapter 2.

If a long plate is bent to a cylindrical surface then the problem

is that of bending a wide beam and a number of useful results can be
obtained with regard to plates. The notation is given in figure
6.29.

If the plate is lone, then a unit width of strip can be cut from
the plate so long as the distance from the ends is large to comply
with the principle of St. Venant. Then for a two-dimensional
element

e
X
= .!.(a
E X
- va y ) (6.36a)

e
y
= 0 = .!.(a
E y
- va X) (6.36b)

since at. a large distance from the ends there will be no strain in
the y-direction.

198
 y

Fig. 6.29 Fig. 6.30

Substitute (b) in (a) then

- v 2cr X} = -1 -E-v- cr X
1 2
e = -(a
X E X

and transposing

E
cr =---e (6. 37)
X 1 - V2 X

and as for a simple beam ex z/r where (1/r ) is the curvature

XZ XZ
of the plate.

Hence
=-E_z_[l)
crx 1 - v 2 rxz

and consequently (figure 6.30)

M =
X
(rdz J (6.38)
xz

that is

M
X [r:J
The term Eh 3/12(1 - v2 ) is called the flexural rigidity of the
plate and is to be found in all thin plate problems. It corresponds
to EI for a beam of unit width but there is added stiffness since
anticlastic curvature is prevented. The accepted notation for the
flexural rigidity is D. Thus

199
 D
M
X r
xz

y y
My
Moments per Radii of
unit length curvature

Fig. 6. 31 Fig. 6.32

If a plate is subjected to moments Mx/unit length and My/unit

length, the planes of the moment being parallel zx and zy planes
respectively, then in any layer

z
e = --
x r xz

and e
z
y r
yz

where ryz is the radius of curvature in plane y - z and rxz is the

radius of curvature in plane x - z.

Also approximately

1 a 2w
and - - =-
r ay2
XZ

I
and a __E_(e + ve ) = _E_z_ (--1- +
X 1 - \) 2 X y 1 - v 2 rxz r;J
(6. 39)
a
y
= _E_(e
1 - \)2 y
+ ve ) = _E_z
x
-[-1-
1 - v2 ryz
+ ~)
rxz

200
also MX
t 2 crxz.dz
_hi
2
rh lz
I
) _hh
[ 1- + -v-)
Ez2- -
--
1 - v 2 rxz ryz
dz

M
X
oir- xz r~j
1 +

(6.40)
M - 0r-a2w
--+va2wl
-
X
_ax 2 ay 2_

and M
y D~l +
r:j
- yz

6.3.1 Circular Plate

For a circular plate x rcose and y rsine

Fig. 6.33

By analogy with the development of equation 2.16

a2w + a 2w = a 2w + .!.(aw] + _!_[a 2w)

ax2 ay2 ar2 r ar r2 aa2

and for a plate loaded symmetrically about its axis, there is no

dependence on e and consequently the sum of the principal curv-
atures reduces to

a 2w a 2w d 2w .!_(dw] (6. 41)

--+--=--+
r dr
ax 2 ay 2 dr 2

and are total differential coefficients since w is a function of r

only.

201
 0

z
Fig. 6. 34

Figure 6.34 shows a section through the axis of a deformed plate.

Take axes through OC and let C be on the undeformed line of the
plate. Then at radius r the deflection is w. The radius OA (r)
n
cuts the plate at A normal to the middle surface and is a principal
radius of curvature. So neglecting squares of the slope at A (that
is <j> is small)

1 _ d2w = _ ~(dw) = d<j> (6. 42)

r
n dr2 dr dr dr

since the slope at A is <j> = -dw/dr (that is w decreases as r

increases from the origin C).

The other principal radius of curvature is rt and is in the

tangential direction and lies in a plane at right angles to the
plane containing rn. This radius is OA and

for small angles such as <j>.

Hence

...!...
rt
= f
r
= - _rl(ddwr) (6."-3)

So the sum of the curvatures can be explained geometrically.

FromM =- D[a 2w/ax 2 + v(a 2w;ay 2)], the bending moments per unit
X
length of radius and arc are thus given by

I
Mr o[!:: + ~(~;)] = o(* + (~) <~>]
(6.44)

Me D [H~;J + v r:::n = D [~ + v( ~:n

202
6.3.2 Moment Equilibrium {figure 6.35)

There will be a variation of shear force Q/unit length and Mr/unit

length with r. There will be no variation of Me with r since the
plate is loaded symmetrically about 0. Since the plate is thin only
stresses due to bending will be taken into account.
O+~dr

Fig. 6. 35

Take moments about the right hand end of the element on mm and
remembering that Mr and Me are moments/unit length

[
u
'"'r + dM
drr dr l(
r + dr ) d e - Mrr d e - 2~es1''2""""r
. -de ...

+ (Q + :; d1 (r + dr)dedr 0

therefore

dM
r d: drde + Mrdrde - Medrde + Qrdrde

+ terms of second order smallness =0

Hence

dM
r
-- + (6.45)
dr

and on substituting for Mr and Me

203
 ~r 0 [d<P
dr Ldr
+ .::._
r
cpl)
J + Q[[d<P
r dr
+ ~
r
<t>) _ (ir + vd<t>)]
dr
+ Q O

that is

.!. d<P + ....::....cp - ...i. - \) d<P Q

r dr r2 r2 r dr =- 'D

or 2
D

which may be written

~r.!. ~Cr<P)J = _ g_ (6.46)

drlr dr D

In terms of the displacement w, dw/dr - cp, then

d[ld dw] Q (6. 4 7)

drlr dr (r dr) =D
which is the necessary relationship between shear force/unit length
at the radius with the deflection w at that radius. This with the
moments equations Mr and M8 are all that are necessary for the
solution of thin plates which are loaded normally but symmetrically
about an axis. Note that the theory allows for plates with a con-
centric hole but eccentric holes require modification of the theory.
Note also that as with one-dimensional beams, the theory for plates
gives stresses and strains in terms of the deflections.
Example 6.7 Uniform pressure applied to one surface of a circular
plate

Fig. 6.36

Take origin through the axis of the plate, then for equilibrium of
the section of the plate of radius r,

2rrrQ = prrr2

204
or Q = pr
2

d
dr
[1
[! d [r ddwr)]
dr

and after integrating three times. t!1e deflection surface is

n-r 4 Ar 2
w= ~ + --- + Bnr + C
64D 4

and the slope equation for any radius is

dw pr 3
--=---+-+-
Ar B
dr 16D 2 r

The boundary conditions for a simple support around the periphery

are

dw
r = 0 dr = 0 (i)

when r = a w = 0 (ii)

when r = a Mr 0 (iii)

that is no applied radial moments. (Nothing can be said of the

tangential moment/unit length unless one is applied.)

(i) gives B = 0 otherwise infinite deflections and hence stresses

would be obtained at r = 0.

(ii) gives

and from (iii)

M
r
0 = _ 0 f~
[16D
+ ~
2
+ (P.C 2A)] r=a
v 16D +

pr 2 A
or 160 (3 + v) = - zCl + v)

or A=- pa2[~)
8D 1 + v

205
and C = _pa'+ + pa'+(~)
64D 32D 1 + v

=~[~)
64D 1 + v
Hence

w
= ~ _ pa 2r 2 (3
64D 32D 1 + v
+ v) + pa'+[~)
64D 1 + v

when r = 0 gives the result for the maximum deflection of the plate.

~[5 + \))
wmax = 64D "'l'+'V

The bending moments at any radius are

M = p(3 + v)( 2- r2)

r 16 a

and Me = -fu[a 2 (3 + v) - r2(1 + 3v)J

Note when r = a, M = 0 and Me = pa 2/8(1 - v) and when r = 0, Mr =

r
Me = pa 2 (3 + v)/16 and the maximum stress therefore occurs at the
centre of the plate. Now from equations 6.43 and 6.39 (polar form)

0
r
= _E_z
1 _
-[.!!!
v2 dr
+ v~)
r

= _ E z (Mr)
1 - v2 D

and substituting for D

12M z
r
(J ---
r

and since surface stresses are of major interest where z h/2

6M
(J
r
r h2

6Me
and oe
h2

for any radius r.

206
 The boundary conditions for a built-in periphery are

when r dw
0, dr = 0

when r dw
a, dr - 0

when r a, w = 0

Hence

B 0, A=-~ and c - pa4

- 640

- pr4 pa2r2 + pa4

and w - 640 - 320 640

and at the centre w

max
= pa 4/640 which is less than that for a simple
support in the ratio (5 + v)/(1 + v).

Figure 6.37 gives diagrams of standard situations and boundary

conditions from which more complicated problems may be solved.

tp
t t
[1]

tp
~ ~ [2]

M(l I,M [3]

~ [4]
WfUnlt length

Me [5]

(See Table 1 )
Fig. 6.37
207
 TABLE 6.1

(See Fig. 6. 37)

Shear Force per Boundary

Plate
Unit Length Q Conditions

p r 0 dw/dr 0
1 =a =0
2nr r w =0 Mr

2
p r 0 r =a dw/dr =0
2nr r =a w =0

3 0 r a Mr M
r =0 w =0 dw/dr =0

b > r > a r b w =0
4 Total load p = 2naw
Q = wa/r r =a r =b 1\lr = 0

at r =a or r b w= 0
5 0 at r a Mr Ma
r b Mr Mb

6.4 Problems
1. A thin tube of radius rand thickness t is to have a flange made
from the same material fitted to its end. The cross-section
dimensions of the ring are small compared to the radius r and the
tube is subjected to a pressure p. It may be assumed that there is
no change in slope at the end of the tube due to the ring and that
it is open ended.

Calculate in terms ofp, rand t and the elastic constants for the
material the value of the bending moment on the end of the tube
assuming the ring is rigid. If the bending moment on the end of the
tube is to be limited to M and the ring is considered to be elastic,
0
calculate the force/unit length between the ring and the tube.
[Ans. P
0
= 4DoA3 and M
0
2DoA 2 ]

2. A horizontal semi-circular bar of circular cross-section is

built-in at its ends and loaded as shown in the diagram. Show that
the twisting moments at the fixed ends have a value PR (n - 2)/2n
Calculate also, the bending moments at the fixed ends and show
diagrams of the bending and twisting moments for the bar.
[Ans. M
o
PR and T
2 o
= PR (n 2n- 2)]

208
 Fig. 6.38

3. A curved bar has a rectangular cross section b mm wide and d mm

deep. If R is the radius of the centroidal axis, show to a first
approximation that the distance between the neutral axis and the
centroidal axis is

A curved bar of rectangular cross-section 25 mm wide by SO mm

deep, carries a bending moment of 1250 Nm which tends to decrease
the radius of curvature. The bar is semi-cricular and the radius of
the centroidal axis is 300 mm. Calculate the value of the maximum
stress in the bar stating where it occurs. What is the error in the
stress when the above approximation for e is used.

4. A thin flat circular plate is loaded by a uniform pressure

applied entirely over one side. A uniform radial couple M0 /unit
length is also applied at the periphery in such a way as to make the
central deflection of the plate zero. Assuming that there is no
deflection of the periphery calculate the value of M0 and hence show
that the ratio of the radial stress/ tangential stress is given by

(5 + v)
1 + Sv

where v is Poisson's Ratio for the material.

5. A beam of section modulus EI rests on an elastic foundation. A

force P is applied to the beam as shown in the figure. Calculate
the equations for deflection and bending moment at a distance x
along the beam and find the ratio MX /M0 where MX is the bending
moment at a distance x from 0 and M0 is the bending moment at 0.
Hence if A= 0.15 (EI Ncm 2) find x when M /M = 0.1. Use without
X 0
proof may be made of

-AX (AcosAx e
AX (CcosAX DsinAx)
y = e + BsinAx) + +

209
 k
and A = 414EI

where k is constant, and the following table

x(cm) 2 3 4 5

cos Ax 0.9554 0.9005 0.8255 0. 7319

sinAx 0.2954 0.4348 0.5644 0.6514

-AX 0. 7408 0.6376 0.5488 0.4724

e

Fig. 6. 39
[Ans. 43 m]

6. In the part structure shown in the figure, a thin flat plate is

to be suspended from a rod and carry a total load P distributed
uniformly around its periphery.

Calculate the displacement at the edges of the plate relative to

its centre assuming that the diameter of the rod is insignificant
compared to the diameter of the plate. Hence calculate the thick-
ness of the plate if this displacement is not to exceed 5 mm for an
edge loading of one Newton/mm.

Take E = 210 kN/mm 2 , v = 0.3 and a 250 mm.

P [Ans. 5.87 mm]

Fig. 6.40

210
7. A thin ring is simply supported on frictionless pins at the
opposite ends of a horizontal diameter and loaded by a vertical
force of 2W at its lowest point. Show that the bending moment at
the point of application of the load is WR/2 and determine the
vertical and horizontal reactions in the ring at this point. Hence
calculate the bending moments at one of the supports and the highest
point in the ring. Sketch the bending moment diagram for the ring
showing principal values.
[Ans. Mat highest point = 0.36WR]

Fig. 6.41

211
7 DEFORMATION BEYOND THE ELASTIC LIMIT

It is well known that metallic materials exhibit P.lastic as well as

elastic behaviour. Here, the worq plastic is used to denote that
the yield stress of the material has been reached after its elastic
limit has been exceeded. It also clenotes that in the plastic state
flow of the material can proceed. Flow of material involves very
large strains of the order SO% - 90% and this branch of plasticity
is to be found in production processes. In this chapter plastic
strains of the same order as the elastic strains only are consid-
ered. It should be noted that other materials flow but the laws
governing their flow are different to the one reserved for metals.

In what follows.it will be assumed that the material being dealt

with is steel for which there is a sharp yield point. ~'lhen the
yield point has been reached it will be assumed that no further
mcrease of force is necessary to produce deformation. The stress/
strain characteristic can therefore be either of figure 7.la or
figure 7.lb in which the usually small elastic component of defor-
mation can be ignored compared to the plastic component. Once
deformation has been extended into the plastic range, then on
unloading there will be residual deformations Er (figure 7.la and
b), which in certain structural members can be an advantage.
u u
y Yt--__.---r-_

(a) (b)

Fig. 7.1
In this section generally, only small plastic deformations will
be considered and large deformations such as those found in metal
working operations are not considered. The problems to be studied
are all statically determinate. It is, therefore, reasonable to
ignore work hardening and the use of a constant yield stress (Y
simple tension = simple compression = 2k, where k is the yield
shear stress) is quite justified. (It should be noted that if the
structure is subjected to reversals of stress in the plastic range,
then the assumption that yield stress of the material in tension is
the same as that in compression is not normally justified unless it
can be demonstrated that Bauschinger effects are negligible for
that rna terial.)

7.1 Yield Criteria

Yield criteria have been thoroughly investigated and can be obtain-

ed from considerations of the modes in which yielding can take

212
place. From experimental and analytical approaches two criteria
emerge which are useful for most ductile materials.

7.1.1 Tresca Yield Criterion (Maximum shear stress)

Yielding takes place when the greatest shear stress in a piece of

material reaches a critical shear stress k. In terms of the prin-
cipal stresses, yielding occurs when the greatest of

\cPl- P2)/2,, I(P2- P3)/2,, or I(P3- Pl)/21 = k (7.1)

(the yield shear stress). For simple tension, wher.e the yield
stress is Y, then, the Tresca yield condition gives

Pl = Y, P2 = P3 = 0 andY = 2k
7.1.2 Maxwell or von Mises Yield Criterion (Maximum shear strain
energy)

It can be shown (see for example Essential Solid Mechanics, B W

Young) that the shear strain energy (or energy of distortion is
given by

where Pl P2 and P3 are the principal stresses.

Now E = 2G(l + v), so

Us= l~G[(Pl - P2) 2 + CP2- P3) 2 + (p3- Pl) 2]

Yielding occurs when the shear strain energy (T 2/2G) per volume
reaches a certain critical value, equivalent to that found in a
simple tension test.

or (7 .2)

where k is the yield shear stress of the material.

Again in simple tension, when the yield stress Y is reached

Pl = Y and P2 = P3 = 0, or 2Y 2 = 6k 2 or Y = 13 k

213
 Thus for the two theories the relationship between yield shear
stress (k) and tensile (compressive) yield stress are

Tresca Y = 2k
Haxwell Y = 13k
Example 7.1 Thin walled tube subjected to tension and torsion
(See Fig. 7 .?)
y

Fig. 7. 2
An element in the wall of the cylinder will be subjected to a two-
dimensional stress system (Figure 7.3) o, due to the tension P
(assumed to produce a uniform stress o) and a shear stress T. These
in turn produce principal stresses in the plane of the material.

and P2 = ~ - [(~) 2 + T2

where o = P/A = P/2uat and

r T Ta/J T/2nta 2 . (The polar second
moment of area J = 2nta 3).

(a) According to the Tresca criterion, yielding occurs when the

greatest of (p 1 - p 2)/2, P2/2, p 1/2 reaches a value of Y/2 = k.

Evidently (Pl - p 2)/2 is the greatest, so

that is
(v) 2 4(y/ 1
+ (7. 3)

214
(b) According to ~axwell (von Mises) criterion, since p 3 0
equation 7.2 reduces to

which on substitution for p 1 and p 2 gives

0)2 T)2
(y- + 3(y-J = 1 (7.4)

The theories thus agree when shear is absent (see figure 7.4).

Fig. 7.4

Example 7.2 Thin walled cylinder subject to internal pressure and

torsion (closed ends)

A thin walled cylinder is subjected to an internal pressure and

subsequently to torsion. An element opqr in the wall of the tube
now has stresses o, crt and T where o = pr/2t, crt = pr/t and as
before T = T/2nta 2 . The principal stresses are:-

Pl
+ ;teat ~
P2
;feat
The radial stresses are small compared to either o or crt and for
the purpose of analysis can be ignored.

and p1

215
 P2
3crR.
-2-- ~cr2R.( + T2

where crR. pa/2t and T = T/2TTta2

It is left as an exercise to show that yielding will occur

according t~ the equations given below.

1. Tresca

T 2
[crv.t( + 4(y-) 1 (7.5)

2. Maxwell or von Mises (T'l 3 O)

cr 2
[YR.) + (f) 2 = ~ (7.6)

The foregoing are examples of in-plane stresses in thin walled

members.

Example 7. 3 Torsion and Bending (Figure 7. 5)

Fig. 7.5

A circular shaft is subjected to a bending moment Manda torque T.

An element on the surface of the shaft will have the following
stresses acting on it, crb due to bending and T due to torsion and
the principal stresses at the surface are:-

P2 =
0
: - ;fcr:f + T2

and according to the Tresca criterion yielding occurs when

216
or [ (Jyb) 2 + 4 (f-) 2 1

where

Td
and T = 4f

where I is the second moment of area about the shaft diameter.

The Maxwell (von Mises) criterion gives (cr 3 = 0)

( (Jyb J 12 + 3 (-f)
2
= 1

7.2 Thick Walled Cylinders

7.2.1 Thick Walled Open Ended Cylinder

(CJ 2 = 0) Sub_ject to .Tnternal Pressure

Open ended cylinders are to be found in a number of engineering

applications. For example the problem of a hub shrunk onto a shaft,
or where movable pistons are fitted into the ends of the cylinder
so as not to produce longitudinal stresses in the cylinder wall.
Obviously care must be taken in the examples quoted to keep friction
to a minimum.

For this problem only the Tresca yield criterion will be used
when the greatest of (cr 8 - crr)/2, a 8 /2 or cr/2 becomes equal to
k = Y/2. Now in thick walled cylinders subjected to an internal
pressure crr is always compressive and hence (cr 8 - crr) is the great-
est stress difference which is required, that is cr 8 - crr = Y.

(a) Initial yielding - the following equations apply to the

cylinder shown in figure 7.6.

Fig. 7.6 Fig. 7.7

(J
r

217
 (7. 7)

Now T
max

and this has its maximum when r is least and yielding thus commences
at the bore of the cylinder.

According to Tresca's yield criterion

the pressure for yielding to just commence at r a, that is

The symbol p will denote this quantity throughout this section.

0

(b) Partial yielding - once yielding has begun further increases

in internal pressure produce more yielding in the material. To
find the equation for the yielded material use the equilibrium
equation
do r
r dr = 06 - 0r

and in the yielded zone o 6 - or Y, therefore

o Y R.n r + a ( 7. 9a)
r

and at r = b, or 0, therefore

r
or = Y R-n b (7.9b)

Let the yielding of the material proceed to a radius r c,

figure 7. 7.

Then for a ~ r ~ c

o = Y R-n !.
r b

218
 For the elastic zone c ~ r ~ b the elastic equations 7.7 apply.
Using these in the general form

Now at r = c, the elastic/plastic boundary

Hence for c ~ r ~ b

(7.10a)

(7.10b)

Now, in order to determine the radius c for a given applied pressure

p, at the boundary r = c, a is continuous. Equation 7.9 is there-
fore r

ar = Y R.n r + A

and at r = c

ac=YR.nc+A

From equation 7.8 at the boundary (p a )

0 c

a
c
= ::!:.
2
[1 - b2c2)
Hence at r = c, ar = -ac' that is

219
 - ~ (1 - ::J = Y ~n c + A

and A can be found. Substituting in equation 7.9a then for a~ r ~ c

cr
r
= - Y ~n . -
r
~2 [1 - b2c 2) (7. lla)

cr 6 = Y - Y ~n f - ~ [1 - ::) (7.llb)

Equations 7.10 and 7.11 give the complete stress distribution for
a partially yielded cylinder and graph 7.1 shows the tangential
stress distribution

@
10 10
Eqs. 710b & 711b
a 0-6
y 0-8
06
c
0'4
b
06
02 04
o;---~~--~---r--~----~
~3 0~ 0~ 0~ 0~ 0-8 0~ 1~
r /b

Graph 7.1

Put or = - p at r = a in equation 7.lla, then

( 7. 12)

which is the relationship betweenthe pressure p and the radius to

which yielding progresses.

(c) Full yielding - the pressure required to produce a fully

yielded wall is found from equation 7.9a with the following con-
ditions

cr
r
when r =a

cr
r
0 when r =b

220
These give

Pult = Y in ..
a ( 7 .13)

The foregoing is the basis of autofrettage theory where a.cylin-

der is subjected to sufficient pressure to cause partial yielding.
On subsequently being put into service the cylinder has a residual
compressive hoop stress in the bore due to the permanent enlargement
caused by the initial pressure. The combination of service hoop
stress (tensile) and residual hoop stress (compressive) results in
a smaller hoop stress at the bore than would otherwise have occurred.
In practice it must be assumed that the applied pressure produces
uniform expansion of the cylinder, and this is confirmed by detect-
ing the movement of the material at its outside diameter.

7.3 Residual Stresses

In order to find the residual stresses, which remain in the

cylinder after unloading apply p (equation 7.12) as a tension at the
bore of the cylinder on the assumption that the residual stresses
are elastic. This has the effect of reducing the applied pressure
to zero. The resulting stresses due top are added to equations 7.10
and 7.11. It is necessary to assume that residual stresses will be
elastic since generally when the body is unloaded no further plastic
deformation occurs and there is only elastic recovery. However since
part of the material has been yielded and part has remained elastic
the residual stresses will be complex. In this problem it is nece-
ssary to establish the condition that yielding will not continue
after unloading.

Due to p

a
r
[E_~
E
- 1) (1 - ::J
a2

ae
E
(1 + ::)
[b2 -
a2
1)
and using equation 7.8

a
-y-r = - ____ (a2 a2)
2po b2 r2
(7.14)
ae - ____ (a2 a2)
y-- +
2po b2 r2

221
Hence the residual stresses are (after some algebra)

or
(7.15)

a ~ r ~ c ( 7. 16)

Table 7.1 shows the pressures p 0 and pult for various (a/b)
ratios and graph 7.1 shows variations of c for a partially yielded
cylinder.
TABLE 7.1

a Just yielding Full yielding

b at Bore (p 0 /Y) (pul/Y)

0.1 0.50 2.30

0.2 0.48 1.61
0.3 0.46 1.20
0.4 0.42 0.92
0.5 0.38 0.69
0.6 0.32 0.51
0.7 0.26 0.36
0.8 0.18 0.22
0.9 0.10 0.10
1.0 0 0

222
 From these two sets of equations

a :; r :; c

c :; r :: b

Now as pressure is released (a8 - ar)/Y decreases until it

becomes zero, or 0 1- (p/p )(a 2/r 2) giving a radius r = p/p
0 0
after which (a 8 - ar)/Y becomes negative and will be greatest when
it is equal to -1 (that is to just to start yielding again at the
borer= a).
Using this value and putting r = a

or

Thus if yielding is to be prevented on tmloading the cylinder

7.4 Rotating Discs

The elastic formulae for rotating can be derived from eauations given
in section 2.4.

- 3-+8- v pw 2r 2
B
ar = A (7 .17a)
r2

ae = A+...!._ 1
- - 8- - pw 2r 2
+ 3u
(7.17b)
r2

A and B are constants and depend on the boundary conditions and w

is the uniform angular velocity of the disc.

Putting

pw 2 (3 + v) = a
8
and B

a
r
A

223
 Since these equations are due to rotation both crr and cr 8 remain
tensile. The third stress cr = 0 for a thin disc.
z

and hence the Tresca yield condition is

( 7 .18)

Now cr 8 is greater when r = a, the inside radius of a hollow disc,

or when r = 0 for a solid disc.

Again the elastic stresses, equations 7.17 apply in the elastic

region from the boundary of the elastic/plastic zone.

7.4.1 Solid Discs (figure 7.8)

Fig. 7.8

Elastically crr = 0 at r a and when r = 0, B 0 since there cannot

be infinite stresses at the centre, hence

(7 .19a)

and (7.19b)

when r =0
(7.20)

that is

Y = pw/( 3 ; v)a2 (7.21)

where w is the angular velocity to just produce yielding at the

y
centre of the disc, or

224
 w
y [-c-3-+-=-8:.:....)-p-a-,r

As the angular velocity is increased, yielding will continue from

the centre to some radius r . From r to a in the elastic zone
c c
B
CJ A- nr 2 (7.22a)
re r2

cre A+.!_- 6r 2 (7.22b)

e r2

and from r = 0 to r rc in the plastic zone

y (7. 23a)
cre
p

CJ = n(a 2 - r 2) Y- nr2 (7.23b)

rp

and r can be found in a way similar to that for thick walled

c
cylinders in addition to A and B, for when r rc, cr cr and
= Y, r =a and or= 0.
cr 8 re rp

7.4.2 Hollow Disc

V/7/J bJ///J
wrad/sec j a b J
Fig. 7.9

The elastic equations are used to find the constants A and B. Assume
or 0 when r = a and r = b (figure 7.9)

0 A-.!_- na 2
a2

B
0 A- nb 2
b2

therefore

B na 2b 2

and A nb 2 + na 2 n(a 2 + b2 )

225
therefore
2 2
a
r
a(a 2 + b2) -aa
- b- (7. 24~
r2
2 2
ae a(a 2 + b 2) +aa- b- - Br2 (7 .24b)
r2

a = 0
z
Of course these will be modified if an internal or external
pressure is applied to the disc.

Again, the yield condition is still a6 = Y as already explained.

So for a hollow disc, yielding will commence at some angular
velocity w given by
y

on substituting r = a in equation 7.24b and remembering that

a = pw 2 (3 + v)/8 and B = pw 2 (1 + 3v)/8. Following the procedure
y y
for a solid disc, if the angular velocity is increased beyond w
y
then the material will yield to a radius r = rc.

Now for the plastic zone where

ae =y (7 .25~
y

a ~ r ~ r
c
aa 2b 2 ar 2
+ Ba2 - ab 2 - -
ar =y --
r2
(7 .2Sb)
y

and in the elastic zone where

ar A-_!_- ar 2 (7. 26~

e r2

rc ~ r ~ b

ae A+_!_- Br2 (7.26b)

e r2

with the conditions a

r =0 at r a and r = b, and a
re
r = r c and with w > wy

226
7.5 Bending of Beams

As with all structural members subjected to forces sufficient to

cause yielding of the material, removal of them produces elastic
recovery of the material accompanied by residual stresses. These
deformations and stresses can be useful in beams as in other
structures but nevertheless careful design is required to use beams
in the plastic range for as with other structures, it is the unknown
residual stresses which cannot always be reckoned with and allowed
for in design calculations.

7.5.1 Stresses

In what follows the yield stress is taken to be Y and to have the

same value in tension as in compression. Shear effects will not be
taken into account but an indication of their influence on the
theory will be given.

The elastic stress distribution of a beam subjected to bending

is shown in figure 7.10b.

Fig. 7.10

This form of distribution persists up to the point where

yielding of the outer fibres commences due, say to a non-symmet-
rical section after which the distribution is modified to accommo-
date further increases in the applied forces (figure 7.10c). It is
at this stage that for non-regular shapes the neutral axis changes
its position. When the section becomes fully plastic (Figure 7.11)
then since, in general, there can be no resultant force over the
cross-section
y
P.
N

~ 1--
y
Fig. 7.11

227
 pl - p2 =0 (7 .27a)

Pl y f bldY (7 .27b)

p2 =y f b2dy (7.27c)

Hence the neutral axis will be positioned so that the areas on

either side of NA are equal, that is

Jb1dy = fb2dy

The idea of a real material yielding right up to the axis is not

quite true since there is always an elastic component and hence an
elastic core in the beam.

7.5.2 Partial Xielding (figure 7.12)

N A

{a) (b)
Fig. 7.12
The internal force system is in equilibrium, hence

Now from figure 7.12b

y (J

di" = d2
therefore

(J = y [:~)
and hence

228
 pl y
('"
dl
b dy

p2 y
[l:Jb dy (7.28)

and the applied moment at the section will be found by taking moments
about NA.

(7. 29)

Equation 7. 29 gives the value of c for a given applied ~~.

The moment to just produce yielding in the outer layers is evid-

ently given by equation 7.29 when c = 0. For full yielding A1 = A2
and the applied moment can easily be found.

If the cross-sec tion of the beam is rectangul ar or circular, the

yielding spreads from the outer fibres by eq~al amounts.

Example 7.4

For a rectangul ar cross-sec tion (figure 7.13) find the ratio ~y/~ult
where My is the moment to just produce yielding and ~ult is the
moment to fully yield the material.

B y
C1tl P.
D y
P.
_{_
y
Fig. 7.13
D;
Jcrb dy J 2
2Y
ifY
B d _ 2YB[D 2 )
Y-204 =-4-
BDY

229
(P1 = average stress x area.)

Therefore
2 D = BD2 y
~y = pl X - X (7. Yl)
3 6

(Note from

M o M YI
T = y = D/2

y BD
2

therefore
so 2
Mult = -4- Y (7. 31)

My/Mult = 2/3, that is the moment to produce full yielding is SO%

greater than that to just produce yielding in the outer layers.

7.5.3 Residual Stresses

Consider a bar acted on by couples of magnitude Mult so that the

whole of the bar is yielded, after which the couples are removed.
On removal of Mult residual stresses will be set up. Again it is
assumed that the residual stresses will be elastic and in order to
find their magnitude apply equal and opposite couples to the bar ano
add the results nf this couple to those of the couple producing full
yielding.

since on what was the tension side has now a compressive elastic
component whilst on the original compressive side, a tensile stress
will be added. This is shown diagrammatically in figure 7.14.
y

'---~(a)
y
+ 15 y
(c)

Fig. 7.14

230
 It should be noted that figure 7.14c would be modified because
of an elastic core and a more representative diagram for the resid-
ual stresses is as shown by the dashed line on figure 7.14c.
A similar procedure is followed for a partially yielded beam.
7.5.4 Plastic Hinges

It will be evident that to load a beam by transverse forces may

produce partial or full yielding of the material. Consider a bar
loaded by a single force (figure 7.15) which is sufficiently large
to produce full yielding at the section where it is applied.

Fig. 7.15

Then at distances away from this section the bending moment is

less than Mult and only partial yielding will be produced. At
some s.ection the bending moment is reduced to My At this section
PLl BD2
M=-2-=My=TY

for a rectangular cross-section. Hence

so that the extent of yielding along the beam in the region of the
force can be calculated. If P was less than that required to
produce Mul t .the same procedure for finding the appropriate L1 is
used. However when the force P produces Mult then a so called
plastic hinge is formed and it can be assumed that further
increases of P produce rigid body rotation about the formed plastic
zone where all the deformation now takes place. No further increase
in the bending moment can now occur. For all statically determinate
beams the ratio of ultimate load to yield load is always Mult/MY.

231
 The application of the idea of plastic hinges is to be found
mainly in structures but in mechanical engineering structures the
theory is useful since it goes some way towards finding the values
of limiting loads.

Fig. 7.16 Fig. 7.17

~fuen, for example, the beam is statically indeterminate (figure

7.16) and with a uniformly distributed load applied the bending
moment diagram consists of the sum of the fixing moment diagram
and the force moment diagram. The numerically greatest bending
moments are at the fixings but there is also a maximum halfway along
the beam. w can be increased until plastic hinges form at A and B
but this alone will not cause collapse of the beam. This will be
brought about only if another hinge can form and this occurs at C
halfway along the beam.

The mode of collapse is shown in figure 7.17 and the three

moments have the same value, and no further increase in w is poss-
ible because of the excessive deformations which would result.
Because A, B and C are hinges the beam is now statically determined
and for half the beam (figure 7.18) taking moments about A

wl
2
Fig. 7.18

wL L
-z- 4

232
or limiting

w =---

Using the principle of virtual work (figure 7.19)

Fig. 7.19

if the bars rotate through a small angle a then work done by each
half of the bar is

2lw2L(L: )J
and the work done by the moments is

(at A) (at B) (at C)

therefore

wL2a = 4M a
4 ult

or w = ---

as before.

For the cantilever shown in figure 7.20

and for AB also taking moments at A

For CB and taking moments at C

-M + R L2 = 0
ult B

233
 M IElastic Collapse
(a) (b)

Fig. 7.20

therefore from these

M lt
~=~

and the maximum value of P

and if L1 = L2 = 2L

6Mult
p =-L-

Estimates of deflection of beams under yielding conditions can be

made if required from a knowledge of the radius of curvature of the
elastic portion. However it should be remembered that until a
plastic hinge is formed, when the beam will have large deflections,
the deflections will be small and may be estimated from the elastic
zones (using M/1 = E/R). The effect of shear stress in practical
problems is usually small but approximations may be made to take it
into account and the reader is referred to works on structural anal-
ysis for a complete discussion.

234
7.6 Problems

1. A thick walled cylinder, 50 mm inside diameter and 150 mm out-

side diameter, is subjected to an internal pressure such that the
cylinder yields to a diameter of 60 mm. Sketch the stress distri-
bution in the cylinder wall for this condition.

If the internal pressure is now released, calculate the permanent

change in the inside diameter.

Assume perfectly elastic/plastic material with a yield stress of

200 MN/m 2 and that the cylinder ~ields according to the Tresca
theory. Take E = 200 x 10 3 MN/m and v = 0.3.

[Ans. 6r =5 x 10- 3 mm]

2. Derive experssions, using the Maxwell (von Mises) and Tresca

yield conditions, for the plastic yielding of a thin walled tube
subjected to combined tension and torsion. Assume the tube to be
thin enough for the stresses to be uniformly distributed through the
wall thickness. Hence from your result show that for the given tube

where a is a constant depending on the choice of yield criterion,a

and T are the stresses set up in the tube due to tension and torsion
and Y is the yield stress of the material in simple tension. Sketch
the results as curves of a/Y against T/Y for each of the two yield
criteria.

3. A rectangular cross-section beam, 50 mm wide and 100 mm deep, is

subjected to bending produced by equal and opposite couples of
magnitude 11/12 Mult where Mult is the couple needed to fully yield
the beam cross-section. Calculate and sketch the bending stress
distribution across the beam.

Assume that the material is perfectly elastic/plastic with a

yield stress a = 200 MN/rn 2 in tension and compression and that
y
unloading takes place elastically. (Maximum residual stress
3o/8.)

4. A flat turbine disc of uniform thickness is 889 mm outside dia-

meter and 178 mm inside diameter. The tensile radial rim loading
due to blades and shrouding is 6.205 Nmm- 2 at 3,000 r.p.m. Assuming
that the inside periphery is unloaded, and that the disc is free
from stress at zero speed, calculate the speeds at which (a) yield
first occurs at the inside edge, and (b) the disc just becomes
plastic throughout.

Assume that yielding takes place at a constant stress difference

of 275 Nmm- 2 Take density of steel to be p = 7833 Kgm- 3 and v =
0.3.

235
5. A thick walled circular cylinder has an internal radius 'a'
and an external radius '2a 1 The cylinder is subjected to an
internal pressure sufficient to cause yielding throughout half
the wall thickness. The pressure is then removed. Determine the
residual stress components at the inside and outside surfaces of
the cylinder.

Assume that the cylinder material is elastic perfectly plastic

with a yield stress Y (assume that the axial stress is zero).

6. Show for an open ended thick walled cylinder subjected to an

internal pressure, that the equation of an element at radius r is

do
r--.. = a - a
dr 8 r

or and cr 8 being the radial and tangential stresses at radius r.

If the yield stress of the material is 308 Nmm- 2 and can be

assumed constant during yielding, calculate, using the Tresca yield
condition, the thickness of t of the yielded portion of the tube
for an internal pressure of 93 Nmm- 2 . The bore of the tube is 12.7
mm and the outside diameter is 19 mm. What is the maximum internal
pressure which may be applied to produce yielding of the material
throughout the wall thickness.

[Ans. t = 6.6 mm, p = 125 Nmm-2]

7. A pressure PI is applied at the bore of the compound tube of

materials A &B.If TA and TB are the maximum shear stresses (corr~
esponding to the yield points of the materials) show that the
minimum external radius is given by

2ahATB

Calculate also the pressure between the two cylinders when Pl is

applied at the bore of the compound cylinder, in terms of Pl> TA
and T 8 .

8. A beam of rectangular cross-section is subjected to a bending

moment M which is greater than ~ but less than Mult' so that the
outer fibres are yielded to a depth (d/2 - e) where d is the depth
of the cross-section. If the material of the beam is elastic
perfectly plastic show that

where b is the width of the section and Y is the yield stress of

the material.
236
 REFERENCES

1. TIMOSHENKO, s p Strength of Materials Van Nostrand

2. HETENYI, M Beams of Elastic Ann Arbor

Foundations Univ. Michigan
Press

3. Ed. FLUGGE, s Encyclopedia of Physics Springer Verlag

Berlin

4. HALL, SA et al Machine Design Schatun Pub. Co. ,

New York

5. ROARK, R J & Formulas for Stress McGraw-Hill

YOUNG, W C and Stra.i.n Kogakusha Ltd.

6. Engineering Sciences
Data Unit - Stress
and Strength

7. SHIGLEY, J E Mechanical Engineering McGraw-Hill

Design Kogakusha Ltd.
8. TOUCHERT, T R Energy Principles in McGraw-Hill
Structural Mechanics Kogakusha Ltd

9. RICHARDS, T H Energy Methods in Ellis Horwood

Stress Analysis Series in Eng.
Sciences

10. HILL, R Mathematical Theory of Oxford

Plasticity

11. HORNE, M R Plastic Theory of Nelson

Structures
12. JOHNSON, W & Engineering Plasticity Van Nostrand
MELLOR, P B Reinhold

13. TIMOSHENKO, s p Mechanics of Materials Van Nostrand

& GERE, J M Reinhold

14. YOUNG, B W Essential Solid Macmillan Press

Mechanics

237
INDEX

Airy - stress function 32

Beams
- elastic foundation 172
- yielding of 227
- residual stresses in 230
- curved 53, 183, 191
Bending of thin circular plates 198
Bi-Harmonic Equation 34, 138
Boundary conditions 35, 41, 76
Castigliano 119, 183
Compatibility 32, 34, 48
Concentrated force 62
Conservative forces 96
Constant strain 147
Continuum 105
Constraints (nodal) 155
Coordinates
- rectangular 1' 19
- polar 5, 25, 46
Curved bars 53, 183
Cylinders
- thick 51, 217
- thin 178
Degrees of freedom 99
Direction cosines 8
Discs
- rotating 52, 223
- temperature stresses 52
Displacements 18, 39, 74
Eigenvalue 9
Eigenvector 9
Elasticity 1
Energy methods 96
Energy
- potential 97
- minimum potential 98
- conservation 99
- complementary 119
Equilibrium 76
Equilibrium equations 1, 4
Finite difference 121
Finite elements 142
Flange on tube 182
Flexural rigidity 179
Force on plate 207

238
Gaussian elimination 126

Hooke's Law 3
Hole in plate 56

Iteration 129

Neutral axis 45
Numerical methods 121

Plastic hinge 231

Plate
- built-in 207
- cantilever 35
- circular 198
- curved 53
- simple support 207
- tension in 56
Principle of virtual work 115

Rayleigh-Ritz 106
Reimforcing ring 180
Relaxation 131
Residual 132
Residual stress 221
Rosette 21
Rotating discs 52

Solids of revolution 49
Stability 100
Statically indeterminate 165
Stiffness matrix 142, 151
Strain 18
- biaxial 35
- gauge rosette 21
Stress
- at a point 7
- axial symmetry 49
- bending 39
- biaxial 35, 61
- boundary 35
- complementary shear 2
- concentration 61, 90
- discontinuity 178
- function 32
- invariant 12
- plane 5
- principal 11
- resultant 9
shear distribution in plates 39
- stationary shear 13
- strain relationships 3
- temperature 52
- trajectory 64
Superposition 3

239
Temperature stres.s 52
Theories of elastic failure 213
Thin plates 198
Torsion 74' 136
- circular shaft 80
- elliptical shaft 81
- hill 80
- hydrodynamic analogy 82
- membrane analogy 83
- thin walled sections 85
Transformation 158
Transform equations 28

Uniform loading 70

Virtual displacements 115

Virtual work 115

Yield criteria 212

- maximum shear strain energy (von Mises) 213
- maximum shear stress (Tresca) 213

240