Todd J. D., 1974, Structural Theory and Analysis

STRUCTURAL THEORY
AND ANALYSIS
Other engineering titles from Macmillan
J. G. A. Croll and A. C. Walker: Elements of Structural Stability
R. T. Fenner: Computing for Engineers
J. A. Fox: An Introduction to Engineering Fluid Mechanics
E. M. Wilson: Engineering Hydrology, Second Edition
Structural Theory
and Analysis
J.D. Todd
Fellow of St. Edmund Hall and
University Lecturer in Engineering Science
University of Oxford
M
© J.D. Todd 1974
All rights reserved. No part of this publication may be

reproduced or transmitted, in any form or by any means,
without permission
First published 1974 by

THE MACMILLAN PRESS LTD
London and Basingstoke
Associated companies inNew Y ark Dublin
Melbourne Johannesburg and Madras
SBN 333 18021 6 (hard cover)
333 15693 5 (paper cover)
ISBN 978-0-333-15693-3 ISBN 978-1-349-15554-5 (eBook)
DOI 10.1007/978-1-349-15554-5
The paperback edition of this book is sold subject to the condition that it shall
not, by way of trade or otherwise, be lent, re-sold, 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 condition being imposed on the subsequent purchaser.
CONTENTS
Preface ix
Plane statics 1
1.1 Introduction 1
1.2 Equations of equilibrium 2
1.3 Supports, reactions and free-body diagrams 2
1.4 Stability and determinacy of reactions 4
1.5 Calculation of reactions 5
1.6 Equation of condition 8
1. 7 Principle of virtual work 9
1.8 Shear force and bending moment 11
1.9 Relations between load, shear force and bending moment 15
1.10 Principle of superposition 16
1.11 The force polygon 20
1.12 The funicular polygon 21
1.13 Graphical determination of reactions 22
1.14 Graphical construction of bending-moment diagrams 23
1.15 Graphics applied to a three-pin arch 25
1.16 The differential equation for a vertical load funicular polygon 27
1.17 In traduction to influence lines 28
1.18 Influence lines for reactions, shear force and bending moment 29
1.19 Loading systems 30
1.20 Application of virtual-work methods to influence lines 32
1.21 Multi-load systems 35
1.22 Influence lines for girders with floor beams 39
1.23 Influence line for three-pin arch 40
2 Statically determinate structures 45

2.1 Simple plane trusses 45
2.2 Stability and determinacy 46
2.3 Resolution at joints 48
2.4 Graphical method 50
2.5 Method of sections 51
v
vi CONTENTS
2.6 Compound trusses 53

2. 7 Complex trusses 53
2.8 Virtual work 55
2.9 Statically determinate space structures 57
2.10 Tension coefficients 58
2.11 Rigid jointed structures 60
2.12 Influence lines for statically determinate trusses 66
3 Elementary elasticity, plasticity and bending of beams 73

3.1 Sign conventions 73
3.2 Stress 74
3.3 Strain 76
3.4 Relations between stress and strain 78
3.5 Composite rods with axial tension or compression 80
3.6 Principal stresses in two dimensions 83
3. 7 The Mohr circle diagram for stress 85
3.8 Principal strain in two dimensions 87
3.9 The Mohr circle diagram for strain 89
3.10 Relationships between elastic constants 92
3.11 Strain energy 95
3.12 Pure bending of beams 97
3.13 Bending of a composite beam 102
3.14 Bending of unsymmetrical sections 104
3.15 Strain energy due to bending 107
3.16 Combined bending and axial force 108
3.17 Plastic bending of beams 109
3.18 Combined plastic bending and axial force 114
4 Torsion and shear effects 118

4.1 Torsion of a circular cross-section 118
4.2 Strain energy due to torsion 121
4.3 Combined torsion, bending and axial force 122
4.4 Plastic torsion of circular rods 122
4.5 Shear stresses in beams 124
4.6 Shear-stress distribution in flanges 129
4. 7 Shear centre 131
4.8 Torsion of thin-walled tubes 132
4.9 Torsion of a thin rectangular section 134
5 Deflection of beams 138

5.1 Introduction 138
5.2 Deflection by direct integration 139
5.3 Deflections using singularity functions or the Macaulay method 141
5.4 Moment-area methods 146
5.5 Use of standard cases 151
5.6 Deflections due to shear 153
CONTENTS vii
6 Virtual work and energy methods 157
6.2 Strain energy and complementary energy 157
6.3 Flexibility coefficients 159
6.4 Maxwell's reciprocal theorem 161
6.5 The flexibility approach 162
6.6 Kinematic indeterminacy 165
6. 7 The stiffness approach 165
6.8 The method of real work 168
6.9 Method of virtual work 170
6.10 Virtual work applied to statically determinate systems 171
6.11 Energy theorems and their application to statically determinate
structures 177
6.12 Williot-Mohr displacement diagram 185
6.13 Virtual work applied to a statically indeterminate truss 188
6.14 Engesser's theorem of compatibility 192
6.15 Trusses with several redundants 195
6.16 The trussed beam 198
6.17 Virtual work and energy methods applied to frames 199
6.18 Ring and arch problems 203
6.19 Redundant trusses using the stiffness approach 209
7 Moment distribution and slope deflection 214
7.1 Moment distribution 214
7.2 Sign convention 215
7.3 Stiffness and carry-over factor 215
7.4 Distribution factor 217
7.5 Fixed end moments 217
7.6 Examples 220
7. 7 Modified stiffness 222
7.8 Deflection of supports and sidesway 227
7.9 Frames with inclined members 233
7.10 Rectangular multi-storey frames 235
7.11 The slope-deflection equation 236
8 Stiffness and flexibility methods 244
8.2 Outline of the stiffness method 244
8.3 Stiffness matrix for a space structure member 247
8.4 Pin-jointed plane trusses 250
8.5 Example of a pin-jointed redundant truss 253
8.6 The space truss 259
8.7 Continuous beams 261
8.8 The plane frame 264
8.9 Grillages 269
8.1 0 Special cases 270
8.11 Outline of the flexibility method 270
viii CONTENTS
8.12 Member flexibility matrix 274

8.13 Example of a pin-jointed redundant truss 275
8.14 Example of a plane frame 279
8.15 Choice of stiffness or flexibility approach 284
9 Influence lines for statically indeterminate beams 287

9.1 Beams with two spans 287
9.2 Betti's reciprocal theorem 289
9.3 Applications of Betti's theorem to influence lines 291
9.4 Multi-span beams 294
10 Stability of columns 298

10.2 Euler critical loads 299
10.3 Strut with initial deformation 304
10.4 Struts made from ideal elasto-plastic materials 305
10.5 Double-modulus theory 307
10.6 Tangent-modulus theory 309
10.7 Practical strut formulae 310
10.8 The Southwell method 311
10.9 Energy methods 313
II Plastic analysis of beams and frames 318

11.2 Collapse of redundant beams 3I9
11.3 Load factor 324
11.4 Basic theorems 325
11.5 Graphical analysis 325
11.6 Virtual-work approach 330
11.7 Combination of mechanisms 334
SuggestionsforjUrtherreading 344
Answers to problems 345
Index 350
PREFACE
This text has been written primarily for the benefit of undergraduates and I sup-
pose that the preface should be directed towards enlightening the reader about the
contents and the reasons for writing a book. I have often wondered if an under-
graduate reads a preface; he has in all probability been directed to a particular
text by his teachers in the subject, and is far more concerned about the number
of worked examples or whether the book appears to be too difficult or too easy
for his tastes. However there are a few remarks that I should like to make.
I have been concerned for a number of years with lecturing and demonstrating
topics in both structures and elasticity. In addition I have been extremely fortunate
In teaching undergraduates in pairs using a tutorial system. This has I hope enabled
me to discover the areas of structures that appear to be difficult and that are not
sufficiently well covered by existing texts. I am also very conscious of the funda-
mental errors that are made year after year by successive generations of students.
I hope that I shall perhaps eliminate some of these difficulties and also point out
the common pitfalls.
The book is intended to serve an undergraduate as a basic text in structures. It
is not possible to cover all the material required for the various degree courses in
a single volume and, by the time the final year of a course is reached, books of a
more specialised nature will no doubt be required. The first two chapters are con-
cerned with statically determinate structures; it is still essential to have a firm
grasp and feel for this subject, as it is one of the foundations of structures. The
other main foundation is elasticity; it would therefore not be possible to make a book
on structures self contained without including some elements of elasticity. In three
chapters I have attempted to cover all the basic elasticity theory, but I have not
dwelt on the more mathematical aspects. An attempt has been made to use a con-
sistent sign-convention throughout the text.
The emphasis in structural teaching has been changing over the past few years,
particularly as computers have become freely available. It is however still neces-
sary to have a real understanding of basic structural theory. In chapter 6 the
fundamental ideas of stiffness and flexibility are introduced. Also included are
examples using work and energy methods in the solution of redundant structures
and in the determination of displacements. A separate chapter covers the elements
of moment distribution and slope deflection-these are still important methods of
analysis that can give a rapid solution if a structure is not too complicated. A
ix
X PREFACE
knowledge of matrices is required in chapter 8-it is assumed that all undergraduate
engineers now receive instruction in this subject as part of their mathematics
course. This chapter discusses stiffness and flexibility methods, and although
these are often covered by a book devoted entirely to the subject, I have attempted
to provide sufficient information in one chapter working entirely from the member
approach. It may possibly be a little indigestible at a first reading, but the reader
will have secured sufficient basic knowledge from the previous chapters and with a
little perseverance will win through, particularly as there are several examples
worked out in detail using both stiffness and flexibility.
The final three chapters discuss topics that I believe should be covered
by a fundamental text. Unfortunately it has not been possible when discussing
stability to cover more than the behaviour of a strut in isolation. The last chapter
does I hope provide an adequate introduction to plastic collapse so that more
definitive works can be read without difficulty.
SI units have been used when any numerical values have been required. Prob-
lems for solution are provided at the end of each chapter; a number of these have
been taken from Oxford examination papers, by permission of the Clarendon
Press, Oxford, but I am responsible for the answers provided.
There can of course be no original material in a book of this nature and
I am indebted to the many different sources that have been used over the
years in forming my ideas in the presentation of the subject matter.
Oxford 1974 J.D. TODD

1 PLANE STATICS
1.1 Introduction
In the first place we need to define what is meant by the term structure. The
reader may well have a mental picture of a bridge, dam or large building; any of
these would most certainly be classed as a structure, but they severely limit the
use of a term which can be applied in a much wider sense. A car, aeroplane, milk
crate or even the chair that you are sitting on are all structures. They are all
designed to carry a particular form of applied loading and, what is most important,
in the majority of cases only deform by a small amount when the loading is
applied. All engineering structures or structural elements are subject to external
forces or loads. These will induce other external forces or reactions at the points
of support of the structure. Consider the chair: it is designed to carry an external
load, provided by your own mass; forces are transmitted through the various mem-
bers of the chair and the reactions act at the ends of the legs. In the case of an
aeroplane in level flight, there must be a vertical component of reaction acting on
the wings as a lift force to balance the mass of the plane. The structure of the
wing is so designed that it is capable of transmitting this force to the fuselage.
Certain types of structures, for example a dam, rely partly on their large mass
to resist the applied forces; we shall not concern ourselves with these types of
structure in this text. If a structure is stationary or moving with a constant
velocity, the resultant of all the applied loads and reactions will be zero, and the
structure is said to be in equilibrium. If however the velocity is not constant it is
necessary to consider inertia forces in addition to the applied forces. We shall
confine our attention to structures that are in static equilibrium. Very often it
will be found that we have to simplify and idealise a structure in order to obtain
a theoretical solution that will give us a fairly good idea of the loading in indivi-
dual members. When the member loads have been deduced we are able to proceed
with the design of each member in turn.
The loads that are applied to a structure can be divided into two categories,
dead and live loads. The dead loading arises solely from the mass of the structure
itself, and it is possible to calculate this to a fairly high degree of accuracy. The
live loading may arise from a variety of causes and can often only be estimated.
Examples of live loads are moving loads, wind and snow loading, impact and
earthquake shocks. In addition to these loadings, members of a structure can have
their loading altered by changes in geometry, movements of supports and changes
of temperature.
2 STRUCTURAL THEORY AND ANALYSIS
1.2 Equations of equilibrium

A system of forces is said to be in equilibrium when the resultant of all the forces
and the resultant of all the moments at one point are equal to zero. For a three-
dimensional system with a set of mutually perpendicular axes, x, y and z, six
conditions must be satisfied. These six conditions can be stated in mathematical
terms as follows:
2 Px = 2 Py = 2 Pz = 0 (1.1)
2 Mx = 2 My = L Mz = 0 (1.2)
where Px is the component of any force in the x direction and Mx is the moment
of a force P about the x axis.
The equations may be written in vector notation as
R =Rxi + Ryi + Rzk =0 (1.3)
MR = MRxi + MRyi + MRzk = 0 (1.4)
where R is the resultant of the system of forces and MR is the resultant moment.
It is obvious that Rx = 'l:,Px andMRx = 'l:,Mx, etc.
In fact it is not necessary for the axes to be orthogonal- -any three axes can be
chosen for the summation of the forces, and any three axes for the summation of
the moments. The case of parallel axes would of course be excepted.
If the system is coplanar, that is say Pz = 0 and Mx =My= 0, then there will be
only three conditions of equilibrium
'J.Px = L Py = 0 (1.5)
LMz=O (1.6)
Several special cases arise for a coplanar system

(1) If there are only two forces acting on a body that is in equilibrium,
then the forces must be equal and opposite.
(2) If there are only three forces acting on a body that is in equilibrium,
then the three forces must be concurrent.
(3) A set of coplanar forces not in equilibrium can be reduced to a single
resultant force or resultant moment.
1.3 Supports, reactions and free-body diagrams

The loads applied to a structure must be transmitted via the various members of
the structure to a number of specified points or supports. The resulting forces and
moments at these points or supports are called reactions. The supports can be
classified into three different types-roller, pinned or fixed. It must be emphasised
that these are all ideal support conditions, that is no friction exists in the roller or
pin, and there is no movement or rotation in the fixed support. In practice con-
ditions may well deviate from these ideal situations.
PLANE STATICS 3
~8
(a)
/p
t------~----,-,~-t~~
VA Ra Va
(b)
Figure 1.1
The simple beam in figure 1.1 a has an inclined force P applied to it. At end A
there is a roller support-this will allow rotation or horizontal displacement to
take place quite freely-thus there can only be a vertical reaction at A. End B has
a pinned support, vertical and horizontal displacement is therefore prevented at
this point, but rotation is allowed to take place freely. The reaction at B will be
inclined at an angle to the vertical, but may be resolved into two components,
one horizontal and the other vertical.
If a line drawing of the beam is now made and the equivalent reactions are
applied (figure 1.1 b) we have what is termed a free-body diagram. VA and VB
represent the vertical components of reaction at A and B respectively and H 8 the
horizontal component of the reaction at B. However it would have been possible
for VB and H B to have been combined and replaced by the single inclined reaction
R B. A horizontal reaction H A could have been added to the diagram, but this is of
course equal to zero.
At this stage it might be as well to state that it is always essential to draw a free-
body diagram for a structure. Care must be taken to ensure that all the applied
loads and reactions have been shown on the diagram.
In figure 1.2a, the same beam is shown with a fixed support at B, this is some-
times called an encastre or built-in support. In addition to the vertical and hori-
zontal components ofreaction V 8 andH8 , a further reaction or momentM8 is
required at B. If the support is rigid this moment will prevent any rotation from
taking place. The free-body diagram is shown in figure 1.2b.
/p
i (a)
~
/p Me
r-...
t~HB
tVA Va
(b)
Figure 1.2
1.4 Stability and determinacy of reactions

If we consider a two-dimensional system with known applied loads, the free-body
diagram can be drawn showing all the unknown reactions. With the system in
static equilibrium the three conditions of equilibrium can be applied, and will
result in three equations in terms of the unknown reactions. The equations can
then be solved simultaneously. For a complete solution there will in general be a
limitation of three unknowns.
If there are less than three unknown independent reactions, there will not be
sufficient unknowns to satisfy the three equations and the system will not be in
equilibrium. It is then termed statically unstable so far as the external supports
are concerned.
~
W/: (a) ~
'
~
(b)
/I',
/ I ..._
/ I ..._
/ ..........
/ .....
*~(e)
/ .....
/ I '......_
~ ~,,
(c)
Figure 1.3
If there are more than three unknowns, the equations cannot be completely
solved and the system will be statically indeterminate or redundant. For example
if there were five unknowns, two of them could be assigned any value, the remain-
ing three could then be found from the equations of equilibrium and would be
entirely dependent on the values chosen for the first two reactions. This does not
mean that a statically indeterminate system is insoluble, the name implies that the
system cannot be solved by the use of statics alone. Additional information will
be required about the manner in which the system deforms under the applied
loading. This type of structure is generally described as redundant and methods
of solution will be discussed later in this book.
In figure 1.3 several examples of different structures are shown. The beam, a,
has only two unknown vertical reactions and will therefore be an unstable system.
The beam, b, has four unknown reactions, one at the left-hand end and three at
the right-hand end, the beam is therefore statically indeterminate to the first
degree, that is there is one more reaction than can be determined by statics alone.
The portal frame, c, is also statically indeterminate to the first degree, as there will
PLANE STATICS 5
be a vertical and horizontal reaction at each support. The frame, d, is statically
determinate for reactions, but it will be seen later that it is statically indeterminate
so far as the forces in the members are concerned. It should be noted that in all
cases the question of statical determinacy is independent of the loading applied to
the system. Reverting to system a: if vertical external forces are applied it might
be tempting to say that there would only be vertical components of reaction at
each support, the support system shown is quite capable of sustaining these,
hence the system is stable. However an instability arises if a horizontal force,
however, small, is applied. It can be seen that the question of stability and deter-
minacy of the reactions must be independent of the actual loading that is applied
to the structure. In effect it would be no use designing a structure to withstand
only vertical loads, if it collapsed when a small lateral wind load was applied. In e
we have a case where the lines of action of three supports, shown dotted, all pass
through a single point. Any loading system that is applied to the beam would
have a moment about the point of intersection and this cannot be resisted by the
reactions, hence the beam would tend to rotate about the point. Based on this
fact a general statement can be made: the reactions must be capable of resisting
any small displacement or rotation that is applied to the structure.
In certain cases the number of unknown reactions can be reduced by what is
termed an equation of condition. This will be introduced in the examples which
follow and will be discussed at greater length in section 1.6.
1.5 Calculation of reactions

This can probably best be illustrated by one or two examples, but before dealing
with specific cases we ought to consider the different types of loading that may
be applied. A concentrated load is assumed to be acting at a point; in actual fact
this is an impossibility as there would have to be an infinite stress in the member
at the point directly under the load. In practice the material under the load will
deform and the load will then be spread over a small area hence reducing the stress
Figure 1.4
concentration. However, for the purpose of calculation it will be sufficiently

accurate to assume that the load is acting at a point. Another very common type
of load is referred to as a uniformly distributed load (U.D.L.). As the name
suggests, the load is distributed along the surface of the member with a constant
value per unit length, or per unit area.
Assume that the moment about point Cis required for the uniformly distri-
buted load p per unit length, extending over length b, see figure I .4. Consider an
6 STRUCTURAL TIIEORY AND ANALYSIS
element dx of the load distant x from C. The moment of this 'concentrated' load
about Cis px dx. The total moment is found by integrating this expression
Me = a f a+b
px dx = pb a +
(
2b) (I. 7)
It is clear from equation 1.7 that the moment can be found by replacing the U.D.L.
by a point load, equal in magnitude to its total value, and acting at its centroid.
If the load was not uniformly distributed but was some function of x, the
moment could be found in a similar manner. The load would be replaced by a
concentrated load equal to the total distributed load and acting at the centroid
of the load system.
E'·-r·t""
.
·
A I
40kNfm
B
(0) (b)
Figure 1.5
As a numerical example consider the beam in figure 1.5a. The free-body diagram
is first drawn as in b. By horizontal resolution it is seen at once that HA is zero.
Take moments about A for the complete system
(40 X 5 X 2·5) + 60(5 + 2·5) = 5VB
Vs = 190 kN
Resolving vertically
VA + VB = 200 + 60
therefore
VA=70kN
Often it is possible to obtain a check on the work by taking moments about
some other suitable point, for example :l: Ms
(60 X 2·5)= (40 X 5 X 2·5)- 5 VA
therefore
VA=70kN
This last equation is equivalent to saying that the sum of the moments about B is
zero. This is true so long as there is no externally applied moment at B.
In figure 1.6a the beam passes over three supports and the portion AD is con-
nected to DC by a pin. From the free-body diagram (figure 1.6b) it would at first
appear that the problem is statically indeterminate as there are four unknown re-
actions. However, as there is a pin at D, the bending moment there will be zero
and so we have an additional piece of information. This is known as an equation
of condition.
PLANE STATICS 7
30
20kN/m
(a)
20 kN/m
~~
t~ tva (b)
Figure 1.6
Resolving forces horizontally

HA = -30kN
The negative sign indicates that HA is in the opposite direction to that assumed in
the diagram. 1:M0 for DC gives
6Vc = 40 x 4
therefore
Vc = 26.7 kN
Note when taking moments at the pin, only the moments are included for the
forces on the beam DC.
1:MA gives
(20 x 9 2 )/2 + (40 x 13) = SVa + (26·7 xIS)
Va = I86 kN
VA= 220- 26·7- I86 = 7·3 kN
A check may be obtained by taking moments about B.
A further example involving the use of an equation of condition is the three-
pin arch (figure 1.7a). The moment at the pinE will be zero. Then
1: ME for portion ECD

6H0 = 3V0
1: ME for portion ABE
3VA =4HA +(30x I)
CrN
j30kN
T 2DkN
r-----'-----.~
6m
j
HA
A ~
t
--
~
Ho
0 ~
(a)
~
(b)
+
Vo
Figure 1.7
VA+ Vn = 30
Resolving horizontally
HA + 20=Hn
Thus there are four equations in terms of the four unknowns. Solving these we
obtain
VA= 2 kN HA = -6 kN Vn = 28 kN Hn = 14 kN
The negative sign for H A indicates that the direction of action of H A is opposite
to that shown in figure 1. 7b.
1.6 Equation of condition

This has already been introduced in some of the problems solved in the last
section. It can be seen that an equation of condition can reduce the degree of
indeterminacy of a structure.
(a)
PLANE STATICS 9
The symmetrical arch rib, figure 1.8a, is built-in at both ends and has loads P
symmetrically applied. From the free-body diagram, figure 1.8b, it can be seen
that there are six unknown reactions. Apparently the arch is indeterminate to the
third degree. However if use is made of symmetry, HA =Hs,MA =Ms and VA= Vs,
it can be shown by vertical resolution that the values of VA and Vs are each equal
toP. Thus the arch will only be indeterminate to the second degree. The beam in
figure 1.9a has two equal spans and the loading is symmetrical. It can be seen
from the free-body diagram figure 1.9b that the problem is indeterminate to the
t t
A' B' ic
(a)
tp t
+vA +va tVc
(b)
Figure 1.9
first degree. There is however an equation of condition which gives VA = Vc and it

is tempting to say that the problem is statically determinate. However the equation
of condition will not in this case reduce the degree of indeterminacy. All that the
equation states is that the sum of the moments about the centre is zero-this is of
course to be expected as there is no external moment applied at the centre support.
It should be emphasised that in general an equation of condition of this type
can only be used if there is complete symmetry both for the loading and for the
structure. In the example given, both of the beams would have to be of the same
length, same cross-section, and made of the same material, for complete symmetry.
1. 7 Principle of virtual work

This is an extremely important principle which will have a number of important
applications in the field of structural engineering.
If a force is given a displacement in a particular direction, the work done is
given by the dot product of the force and the displacement, that is, work is only
done when the force and the displacement are in the same direction. Thus it is
necessary to take the resolved part of the displacement in the direction of the
force. Now consider a small particle acted on by a number of forces Ph P 2 , Pn·
The forces are all concurrent and have a resultant PR· Suppose that the particle is
now displaced through a small distance A in a particular direction. If the forces do
not change magnitude the work done is
P1 . A +P2. A+··· +Pn .A =PR .A
Had the forces been in equilibrium, the resultant force PR would have been
zero, hence
(1.8)
This is the principle of virtual work which states: If a particle that is in equi-
librium is given a displacement in any direction, then the total work done by the
forces acting on the particle is zero. The only requirement is that the forces must
stay constant during the displacement. They can be generalised forces, that is,
they need not be linear forces but could be moments or torques. The displace-
ments could also be linear or rotational. It is sometimes referred to as the prin-
ciple of virtual displacements, since it is not necessary for the displacement to
take place-it could be imaginary. We are in fact stating the principle of equilib-
rium in a slightly different way.
The principle can easily be extended from a small particle to a structure that is
in equilibrium. The structure could be thought of as a large number of particles
each of which is in equilibrium-thus the principle will still apply.
o'
---..
-
s' _, ...- ....- ' ' ......
......
,...-
...... 30
......
50kN
20kN/m .....
HA-- 40 1 .....
D
VA Sm
Bh~ 4m 4m __J_2mJ Vc
(0)
, __ /--~-;
c'
--·
~-
w
tA tB D ct
(b)
Figure 1.10
We shall now make use of this principle to find the reaction VB in the problem
already solved in figure 1.6. The beam has been redrawn (figure 1.1 Oa). As the
reaction at B is required a virtual displacement is applied at point Bin the
direction of the required reaction that is, vertically. The beam will take up the
deformed position as shown in figure 1.1 Oa, all displacements being greatly
exaggerated. B has moved to a position B' while the beam pivots about the end
points A and C and hinges about D'.
If the vertical displacment BB' is v , the centroid of the U.D.L. will move
vertically through 9/1 Ov and the point of application of the 50 kN load has
moved vertically through 3/Sv.
PLANE STATICS 11
Applying the principle of virtual displacements and noting that only the verti-
cal component of the 50 kN load is required
-(180 x -fov)- (40 x ~v) + VBv = 0
therefore
VB= 186 kN
The negative signs are required because the direction of action of the forces is
opposite to that of the displacements.
If the reaction V c is required, a vertical displacement v 1 is applied at C. The
beam will now remain on its supports at A and Band pivot about D (figure 1.10b ).
Thus the work equation becomes
-(40 x jv 1) + Vcv 1 =0
therefore
Vc = 26·7 kN
1.8 Shear force and bending moment

The aim of all structural design is to produce a structure that is capable of carry-
ing the specified forces. Individual members of the structure may have to carry
axial and transverse loads, bending moments and torques, thus the design could
become extremely complicated.
For the time being the discussion will be limited to a simple straight beam,
the applied loads and reactions lie in a single plane which also contains the
centroidal axis of the beam.
I
X
t'? ~
iB
\
A' (a)
X
Q
tfT ~
\
!L t~F F4t
M
tVA tVa
Q
(b)
Figure 1.11
Such a beam is shown in figure 1.1la. Imagine the beam to be sectioned at XX.
Then the free-body diagram will be as shown in figure 1.11 b. To keep each portion
of the beam in equilibrium the following forces will be required: an axial force F,
a transverse force Q, known as the shear force, and a bending moment M For
equilibrium at the section the corresponding forces in each section must oppose
each other; for example if there is a clockwise moment M shown on the R.H.S.
this must be opposed by an anti-clockwise moment M on the L.H.S. The values
of these forces can be determined from the load applied to the beam and the
values of the reactions. A sign convention has to be adopted. In a later chapter
we have to use a sign convention for stresses and the convention used here will be
similar. A right-handed triad for x, y and z is adopted, the x axis points to the
right and they axis vertically upwards.
Axial force
The axial force is the resultant of all components of applied loads and reactions
acting parallel to the axis of the beam at the point considered. Tensile forces will
be taken as positive.
(b)
Figure 1.12
Shear force
If a beam is in equilibrium and we imagine it to be sectioned vertically, it will be
necessary to apply forces normal to the axis of the beam, lying in the section, to
maintain equilibrium. These forces Q are known as shear forces. If the resultant
of all the forces acting on the left-hand portion of the beam is downward then a
force Q would have to act upwards (figure 1.12a). For the right-hand portion of
the beam the resultant force and the shear force will each act in opposite senses.
Positive shear is that shown in the diagram, that is, the resultant forces would tend
to move the left-hand portion down and the right-hand portion up.
Bending moment
The bending moment is the sum of the moments taken about an axis normal to
the plane of the applied loads and reactions, and passing through the centroidal
axis of the beam at the point considered. The bending moment will be taken as
positive when it is 'sagging'. This means that the beam will be concave upwards-
the fibres at the top edge of the beam will be compressed, while those at the
bottom edge will be extended. A negative bending moment is termed 'hogging'.
Figure 1.12b shows a positive bending moment.
We are now in a position to calculate and sketch both bending-moment and
shear-force diagrams for a beam. It is convenient to sketch the beam and then
draw the shear-force and bending-moment diagrams underneath. It should be
PLANE STATICS 13
--x----fP
~f - - - - -
il2v=:==
j _ _ ___- _ve_ _ _ _~
PL pi~
PI~
re-
(a) (b )
p
~-o-r b
'I
t .t t-x-- +
• ve
-ve
(c) (d)
Figure 1.13
emphasised that in all the cases which follow, the beam has been considered as
light, that is, making no contribution to either the shear force or bending
moment.
Four simple cases are shown in figure 1.13. The first of these, a, is a cantilever
with a point load P applied at the free end. Considering a section at a point dis-
tant x from the free end, the value of the shear force is - P. So that the value of
the shear force remains constant along the length of the beam at -P. The value of
the bending moment is -Px as the beam is hogging. The bending moment will
have a maximum value at the fixed end of- PI and a value of zero at the free end
with a linear variation.
Figure 1.13b shows the same cantilever with a U.D.L. of p per unit length. At
x the shear force is -px and the bending moment is -px 2 /2. The shear force has
a linear variation from zero at the free end to -pi at the fixed end. The bending
moment increases in a parabolic manner from zero at the free end to - pf2/2 at
the fixed end.
The shear force and bending moment diagrams for a simply supported beam
with a U.D.L. are shown in figure 1.13c. At a point x from the left-hand support
the shear force is- p [(1/2)- x], that is, a linear distribution with a zero value at the
mid-point of the beam. The bending moment at the same point has a value of
p(lx - x2)f2, that is, a parabolic variation with a maximum value of p/ 2/8 at the
mid point.
A simply supported beam with a point force P applied is shown in figure 1.13d.
The reaction at the left-hand end is Pb/1 and that at the right-hand end Pa/1.
For 0 < x <a, the shear force will be constant at -Pb/1. When x =a, there is a
sudden change in value by an amount equal to the force P.
For a< x <I, the shear force is -P(b/1- 1) = Pa/1, another constant value.
For 0 < x <a, the bending moment is Pbx/1, a linearly increasing value that
will have a maximum of Pab/1 when x =a.
For a< x <I, the bending moment is
Pbx/1- P(x- a)= Pa(l- x)/1
A slightly more complicated case with numerical values is that of the simply
supported beam figure 1.14. All distances are in metres. It is first necessary to
determine the reactions
8V8 = 150 + 240 + (40 x 9)

Thus
V8 = 93·8 kN VA= 46·2 kN
r;---- 2 5 - - -c""
' t..,6_0_kN_ _ 3·5 _ _ _0...,.~ 40kN
.t-t---------8---------8.:1~
20kN/m
HA..._ • Va
E
53·8
13·8
46 ·2
/
40
Figure 1.14
PLANE STATICS 15
The shear force diagram can now be drawn. At A the value is equal to the
reaction VA, that is, -46·2 and it stays constant at this value from A to C. At C
there is a sudden change of +60 and the shear force stays constant at 13·8 from
C to D. At D there is a change of +40 and the value stays constant at 53·8 from
D to B. At B the shear force will change from 53·8 to -40 a total change of 93·8
equal to the reaction VB· From B toE there will be a linear decrease with a zero
value at E.
The bending moment at A is zero and will increase linearly to 46·2 x 2·5 =
115·5 kNm at C. There is a linear variation between C and D where the value is
(46·2 x 6)- (60 x 3·5) = 67·0 kNm. Again there is linear variation between D and
B where the value is -20 x 2 2 /2 = -40 kNm. From B to E there is a parabolic
variation with zero bending moment at E.
It can be seen that a point load causes a sudden change in the value of the
shear force equal in magnitude to the value of the point load. It also causes a dis-
continuity in the bending-moment curve. A couple applied to a point on a beam
will cause a sudden change in the value of the bending moment equal to the value
of the couple, but there will be no change in the value of the shear force at the
point of application of the couple.
1.9 Relations between load, shear force and bending moment

Figure 1.15 shows a small element dx cut from a beam. It is loaded by a force,
on the top surface, of average intensity p per unit length, acting upwards. To
keep the element in equilibrium it is necessary to add both shear forces and bend-
Figure 1.15
ing moments. These will change in value from one side of the element to the
other. The sign convention previously adopted has been used. As the load acts in
a positive y direction, it is considered positive.
Q +dQ= Q- p dx
thus
dQ
-=-p (1.9)
dx
Taking moments about the right-hand edge, omitting the term involving the
load as a second order quantity
M+dM+Qdx=M
thus
dM
-=-Q (1.10)
dx
Making use of equation 1.1 0, we can express equation 1.9 as
d 2M
-=p (1.11)
dx 2
Equation 1.10 shows that if the shear force is negative then the rate of change of
bending moment is positive, so that the value of the bending moment is increasing.
Also tb~ value of the bending moment will be a maximum or a minimum when the
shear force is zero. Exceptions to this would be when the maximum value of the
bending moment occurs at the end of a beam, or when there is a sudden change
in the value of a bending moment due to the application of a couple.
Equation 1.10 can be integrated such that
M2 -M1 = Jx,x, -Qdx

Thus the change in moment between any two points is given by the area of the
shear-force diagram between the two points. An exception to this would be if a
couple were applied between the two points considered.
It is essential to acquire the skill of sketching bending-moment diagrams both
rapidly and easily. The relations that have been derived can be of some assistance
in achieving this, so also can the principle that is introduced in the next section.
1.1 0 Principle of superposition

This principle states that, for a linear elastic system in which changes in geometry
are small, the effect m due to a cause M can be added to the effect n due to a
cause N. Thus the result will be the same as the effect m + n due to a cause M + N.
A simple example of this is a linear spring balance where the elongation of the
spring due to a load M +N could be found by summing the elongations resulting
from the loads M and N applied separately.
The principle is not valid if the material of the system is non-elastic or has
exceeded the yield point, or if the geometry of the structure changes appreciably
as the load is applied.
Obviously the geometry of all structures must change slightly as loads are
applied. Individual members will extend or contract resulting in points on the
structure deflecting. The changes in the majority of cases are small and can be
neglected in applying the principle.
An example of a structure in which the geometry is dependent upon the load
is shown in figure 1.16. The two uniform members of the same length are made
PLANE STATICS 17
of linear elastic material, they are each pinned at one end to a rigid support and
are joined together by a pin. A vertical load Pis applied at the pin, and the dotted
line shows the deflected form of the beams. It will be found that the deflection
of the mid-point will not be proportional to the load.
Figure 1.16
A small knowledge of elasticity is assumed in the following. Let the deflection

bed and the original length of each member be l.
If d ~ l, tan 8 ~ 8 = djl, the tensile force Fin each member is
P P Pl
--~-=-
2 sine 28 2d
The strain e is given by
Making use of Hooke's law

F
e=-
AE
where A is the cross-sectional area, therefore
d2 PI
2J2 = 2AEd
or
AEd 3
P=-~3-
So that when 8 is small the relation between the load and displacement is of a
cubic form.
The principle of superposition can sometimes be used to good effect when
shear-force and bending-moment diagrams are required. In every case it would be
perfectly possible to draw individual bending-moment diagrams for each load
applied separately to a beam, and then finally sum the diagrams to get the total
effect when all the loads are applied simultaneously. For the majority of cases
this would be a rather tedious approach, and in general it is necessary to be selec-
tive. One particular application is in the case of statically indeterminate systems,
where it is possible to get some idea of the shapes of the S.F. and B.M. diagrams.
It is, of course, not possible to assign final numerical values to the diagrams.
Consider a beam built-in at each end and carrying a U.D.L. (figure 1.17a).
Omitting any effects due to axial force the problem is statically indeterminate to
the second degree, but may be reduced to a single degree by an equation of con-
dition which enables us, from symmetry considerations, to state that the vertical
reactions at each end are pl/2. There will however be an unknown moment M
acting at each end. From symmetry it may be argued that the moments at each
end will be of equal magnitude.
Consider the problem as two separate cases: a simply supported beam carrying
a U.D.L. (figure 1.17b) and an unloaded simply supported beam with a moment
M applied at each end (figure 1.17e). The S.F. and B.M. diagrams for the first
case are shown in figures 1.17c and d. The B.M. diagram is parabolic with a maxi-
~? Zm. Jiiii!#P~ (a l
1----- I -----1
(b)
(c)
(d)
,..-....
.....;.M
.:___ _ _ _ __ Mc.:__---. (e)
t +
- ve (f)
v "'SI (g )
Figure 1.17
mum value of pl 2 /8; a standard case that should be remembered. (For a central
point load of P the maximum B.M. is at the centre and has a value of Pl/4.) Now
consider the case of the beam simply supported with terminal moments - M . The
sign is negative as these moments tend to hog the beam. The reactions at the ends
will be zero, thus the shear force is everywhere zero. The bending moment will be
constant at a value of- M (figure 1.17f). The two bending-moment diagrams can
now be combined to give the final result. The simplest way of doing this is to
redraw figure 1.17d, invert figure 1.17f, and superimpose these giving figure 1.17g.
The two diagrams now have to be subtracted, leaving the shaded area. From this it
may be seen that the maximum bending moment can either occur at the centre of
the beam, pl 2 /8 - M, or at the ends of the beam,- M The S.F. diagram does not
have to be redrawn and is identical to figure 1.17c.
PLANE STATICS 19
¥~'$< ' "'~ ?f'.&" -~!Y"&""&.:l 'IJD.' !!I (a)
v---
~
(b)
tel
•ve (dl
r--------___ (el
~I
~ (f)
~
(g)
- ve
Figure 1.18
A further example is that of a propped cantilever (figure 1.18a). The value of

the vertical reaction V cannot be determined by statics, Figures 1, 18b and c show
the S.F, and B.M, diagrams for the loaded cantilever without the prop, Figures
1.18d and e show the cantilever loaded with Vat the right-hand end, The com-
bined S,F, and B.M, diagrams are shown in figures 1.18fand g, the shaded area
once again giving the final values, Note that the point of maximum positive
bending-moment occurs when the shear force is zero. To find the value of V it is
necessary to add some further information to the problem, In this case it might
be stated that the prop is rigid, that is, the deflection at the end of the cantilever
is zero. Deflection of beams will be found in a later chapter,
f:!B
~ w
\ X .tV
(a)
•ve
Figure 1.19
It is possible in certain circumstances to obtain a complete solution if further
information or an equation of condition is given. Suppose that the height of the
support at A (figure l.l9a) is adjustable and that it is varied until the maximum
bending moment in the beam is as small as possible. Perhaps this is rather an arti-
ficial case but it will enable us to obtain the complete B.M. diagram.
The B.M. diagram (figure 1.19b) can be sketched and it is seen that the maxi-
mum moment could occur either at the root of the cantilever or somewhere in the
positive region. The optimum is obtained when the two values are equal in mag-
nitude but opposite in sign. (Give some thought to this statement and make sure
you know why it is correct.)
px2
M =--+Vx
X 2
This will be a maximum when dM/dx =0, that is, x = V/p. Actually this position
could have been obtained at once by saying that the bending moment is a maxi-
mum when the shear force is zero.
The bending moment at B is
pl2
vz--
2
therefore
px2 pl2
- - Vx= V l - -
2 2
since
V= px x 2 + 2lx -1 2 = 0 x = (.,j2 -1)1
Hence
V= (vf2- 1)pl
It is now possible to find the bending moment at any point in terms of p and
l alone.
1.11 The force polygon

If the direction and magnitude are known for each force that is acting on a body,
a polygon can be drawn to a suitable scale, the forces being represented by the
sides of the polygon. If the polygon closes, the system of forces will be in equilib-
rium. If the polygon does not close, the resultant force acting on the body can be
found, both in magnitude and direction, by the line required to close the polygon.
This can easily be verified by applying the triangle of forces.
In the free-body diagram (figure 1.20a) three forces are shown acting on a
body. The force polygon (figure 1.20b) is drawn by making AB, BC, etc., parallel
and scaled in value to P1o P2 , etc. For the case shown, the polygon does not close
thus indicating that the forces are not in equilibrium. We can however close the
polygon with a line DA. This will represent the magnitude and direction of the
PLANE STATICS 21
0
\
\
c
',II
' ', ~2 '
' ' '
~--- \
\
' '
A
B
Fi
(a) (b)
Figure 1.20
force P4 that has to be applied to the body to keep it in equilibrium. This can be
seen by applying the triangle of forces. The resultant of P 1 and P 2 is R 12 , which
when combined withP 3 givesP4 .
We must note at this stage that the construction gives the magnitude and
direction of the resultant, but will not give the point of application, unless it so
happens that all the forces pass through a single point. To find the point of
application a further construction is required known as the funicular or string
polygon.
1.12 The funicular polygon
~ c
'T
a
A
Fi
T
(a) (b)
Figure 1.21
The same free-body diagram and force polygon are shown in figures 1.21a and b.
Any point 0 is chosen on the force polygon diagram, this is known as the pole.
It need not lie inside the force polygon but a point inside it has been chosen for
convenience. Join OA, OB, etc.-these lines are termed rays. OAB now forms a
force triangle, thus P 1 can be resolved into two other forces: PA in direction AO
and PB in direction OB. Similarly P 2 is resolved into PB in direction BO and Pc
in direction OC. It is seen that by choosing a single point of intersection 0, PB,
for example, will represent a common component of forces P 1 and P 2 •
In the free-body diagram the rays must intersect the lines of action of the
various forces, thus a line parallel to BO must intersect P 1 and P 2 . Choose any
point Q on the line of action of PI> and draw a line parallel to OB to cut P2 at R.
From R draw a line parallel to OC to cut P3 at S, from S a line parallel to OD and
from Q a line parallel to OA. Let these last two lines intersect at T. QRST is known
as the funicular polygon and T is a point on the line of application of the equili-
brating force P4 •
This can be seen quite easily as we can replace the forces Pas follows in the
funicular polygon diagram.
PI by PA in direction AO and P8 in direction OB
P 2 by P8 in direction BO and Pc in direction OC
P 3 by Pc in direction CO and Po in direction OD.
The forces P 8 and Pc will cancel out in pairs leaving only forces PA and Po acting
at T. The resultant of PA and P0 is P 4 . So that T will be a point on the line of
application of P4·
Two conditions must be satisfied for a system of forces to be in equilibrium:
(i) the force polygon must close; (ii) the funicular polygon must close.
A particular case can arise when the first condition is satisfied. However when
the funicular polygon is drawn, the first and last lines are parallel, indicating that
the resultant on the system is a couple.
1.13 Graphical determination of reactions

The force and funicular polygons can be used to determine the reactions that
result when a beam is loaded.
I s
I
Ia)
Figure 1.22
The beam in figure 1.22a will have a vertical reaction at B but the direction of
the reaction at A is unknown. Draw lines on the force polygon (figure 1.22b) to
represent PI, P 2 and P3 . At this stage it will not be possible to close this diagram
as there are two separate closing lines representing R A and R 8 . It is convenient to
letter the spaces between the forces in figure 1.22a, this is known as Bow's nota-
tion. Thus Q is the space between RA and PI> R between PI and P 2 , etc. Finally
U is the space underneath the beam between the two reactions. This system of
lettering is transferred to the force polygon, that is, PI is the force qr.
PLANE STATICS 23
Next choose a suitable pole 0 and draw in the rays. The reason for the lettering
system can now be seen as a line on the funicular polygon, across space Q for
example, will be parallel to Oq. Using this system we are unlikely to make a mis-
take by drawing a line across the wrong space in a complicated system.
The funicular polygon can now be drawn. It is essential in this case to start the
funicular polygon at A, as it is known that the reaction at A passes through this
point. The various lines are shown in figure 1.22a. The reaction at B is known to
be vertical; the line across space T cuts a vertical through Bat C. AC is the closing
line on the funicular polygon. A line parallel to AC is drawn through 0 on the
force polygon, this is shown dotted. The force polygon is now closed by two lines
representing R A and R B. R B is known to be vertical and must pass through t.
Draw tu to cut the dotted ray at u. The two reactions are then represented by qu
and ut.
If we imagined a massless string fixed at A and C the funicular polygon would
represent the shape that the string would take up if loads P were hung from it,
hence the name string polygon. If the pole 0 had been chosen on the other side
of the force polygon the funicular polygon would be inverted. We could now
think of a series of massless rods, instead of strings, supporting the loads. Hence
the term line arch.
1.14 Graphical construction of bending-moment diagrams

Figure 1.23 shows a light string fixed at the two end points A and B. The string
carries two loadsP 1 andP2 • Since the loads are vertical,HA =H 8 and the hori-
zontal component of tension in the string will be constant at H, say.
Figure 1.23
For a flexible string the bending moment at any point on the string must be
zero. Then for 1.: Me
VAx =H(y + xld) + 'L mx (1.12)
1.: mx represents the moment about C of all the applied loads P to the left of C.
For l.:M8
(1.13)
~ m 8 represents the sum of the moments of the applied loads P about B.
Eliminating VA from equations 1.12 and 1.13
Hy = L ms · ~ - .L mx (1.14)
l
Now ~ m 8 fl would be equal to VA for the case when H = 0, thus the right-hand
side of the equation represents the bending moment at x for a simple beam of
length l with the same system of loads applied. Thus, considering the string, if we
multiply the horizontal force H by the vertical distance from a point on the string
to the line joining the two ends, we shall obtain a value for the bending moment
at the same point in a simple beam.
This analysis can be used for the graphical construction of bending-moment
diagrams for beams with applied vertical loads.
(a)
1---H
(b)
Figure 1.24
The beam AB in figure 1.24a is simply supported with vertical loads P 1 and P 2
applied. The force and funicular polygons are constructed. In this case it is not
necessary to start the funicular polygon at A. The direction of the reaction at A
is vertical and any point on the vertical through A will suffice. From the previous
discussion we can see that the funicular polygon will represent such a string fixed
at A' and B'. Thus the bending moment on the beam at X will be given by Hy.
The question that naturally arises is, what is the correct value of H? In the case of
the string, H was the value of the horizontal force at the supports. It can be seen
at once from figure 1.24b that this is represented by the horizontal distance from
0 to the vertical line representing the load.
It is of course necessary to take into account various scale factors when the
numerical value of the bending moment at a particular point is required in a
problem. If the force polygon has been drawn to a scale s2 then the true value of
the horizontal thrust is S2 H, and if the line diagram has been drawn to a scaleS t.
the value of the bending moment will be given by S 1 S 2 Hy.
PLANE STATICS 25
If some of the loads on the beam are not vertical the problem cannot be treated
directly, the reason for this is that the value· of H would not be constant. However,
if only the vertical components of the loads are used in the construction, the
resulting funicular polygon will represent the B.M. diagram, as the horizontal
components do not affect the value of the bending moment.
When part of the loading system in a beam is uniformly distributed, it is best
to consider it for the time being as a concentrated load placed at the centroid of
the distributed load, and equal to the total value of the distributed load. The
force and funicular polygon can be drawn . The portion of the funicular polygon
between A and C in figure 1.25 will be incorrect as the variation of bending
moment for a uniformly distributed load should be parabolic. The diagram can
however be corrected by drawing in a parabola between A' and C' such that A'D'
and D'C' are tangents to the parabola.
c 8
-a-
vo p
7
a'
A
I
-- ------
~ or c' -
Figure 1.25
A non-uniform load can be plotted in graphical form and divided into anum-
ber of vertical elements or strips. The area and centroid of each strip can be
found. The area will represent a small load which can be considered as acting
through the centroid. The force and funicular polygon can be drawn, the latter
will have to be slightly corrected by making the rays tangential to a curve, which
can be drawn freehand.
1.15 Graphics applied to a three-pin arch
The problem of finding the reactions for a three-pin arch has already been dealt
with theoretically in section 1.5. To solve the problem graphically it is necessary
to consider the loads on each half of the arch separately.
Figure 1.26a shows the arch with the complete loading, and figure 1.26b shows
it with the loads on the right-hand side removed. Considering this second case,
the direction of the reaction R B must pass through the points B and C as the
(c)
Figure 1.26
bending moment at both these points is zero and there is no other load on the
right-hand side. The force and funicular polygons are drawn. The ray fo will pass
between the lines of action of P2 and R 8 . The closing line from the funicular
polygon is transferred to the force polygon. As the direction of R 8 is known, a
line parallel to this can be drawn from f to cut the final ray at g. RA and R 8 will
be represented by dg and gf.
It will next be necessary to consider the arch with the loads on the right-hand
half removed. The reactions can again be found. Finally the two sets of reactions
can be combined to give the values when the arch is loaded on both sides.
The funicular polygons will not give a measure of the bending moment at any
point as the horizontal thrust at the abutments has to be taken into account.
However, once the reactions are known the moments can be found by statics.
Next consider a set of vertical loads applied to a three-pin arch (figure 1.27). It
is essential that the funicular polygon, which also represents the line of thrust,
A B
Figure 1.27
PLANE STATICS 27
should pass through the three pins, that is, the bending moment at each pin is
zero. This will mean that there is only one possible position for the pole on the
force diagram. The location of this point is left as an exercise for the reader. It
can be seen that the closing line on the funicular polygon must be horizontal.
We have already shown in section 1.14 that the horizontal distance H on the
force polygon represents the horizontal thrust at the abutments and, for a vertical
system of loads, will also represent the horizontal component of force at any
point in the arch.
Any vertical distance y' on the funicular polygon multiplied by the polar dis-
tance H on the force polygon gives the bending moment at a particular point on
a beam with the same vertical loading applied. So that in the case of the arch Hy'
is the bending moment at a point Don the arch due to the vertical applied loads
and the vertical component of the end reaction. To find the total moment the
effect of the horizontal thrust must be taken into account.
M 0 =Hy' - Hy =H(y'- y)
where y is the vertical coordinate of the arch at the point D. Note that y' - y is
the vertical distance between the arch and the funicular polygon.
If the arch had been designed such that its axis coincides with the funicular
polygon, then the bending moment at every point on the arch would be zero, and
the arch would only be subject to direct thrust. The funicular polygon represents
the line of thrust in the arch, and is sometimes referred to as the pressure line.
We now have a very simple way of determining the position of the line of
thrust. The moment at any point is divided by the value of the horizontal thrust,
giving the value of y'- y.
1.16 The differential equation for a vertical load funicular polygon
Figure 1.28a shows the diagram of a non-uniform vertical load applied to a simply
supported beam. To draw the force and funicular polygons (figures 1.28b and c)
the load is split into a number of elements. It will be noted that the position of the
pole 0 has been chosen such that the closing line on the funicular polygon is hori-
0
A
t VA
X ,C
: (a)
t B
Va
X
E
-H_____.j
(b) (c)
Figure 1.28
zontal. OF is a typical ray on the force polygon and is shown as the tangent JK to
the funicular polygon. The closing ray on the force polygon is OG, thus DG and
GE represent VA and V 8 .
The length DF is fK p dx where p, the load, is some function of x.
Thus FG represents VA - fK p dx. Now
tan L FOG = VA - fK p dx
H
and this is equal to the slope of JK on the funicular polygon. Thus
VA - fK p dx =(dy)
H dx c
and since VA is constant
(1.15)
This equation is known as the differential equation for a vertical load funicular
polygon.
As a particular application let us find the shape of an arch that is carrying a
load uniformly distributed over the span, such that the bending moment at any
point on the arch is zero.
As the bending moment is zero it is essential that the funicular polygon and
the centre line of the arch coincide.
In this case p is constant and equation 1.15 can be integrated twice to give
px2
y = - - +Ax+B
2H
Applying suitable end conditions y = 0 when x = 0 or l
This means that the arch would have to be of a parabolic form with a central rise
ofpl 2 /8H
1.17 Introduction to influence lines

The external loading systems introduced so far have been fixed in position. Many
structures are however subjected to moving loads, and it is necessary to investigate
the effect that may result from these loads. A simple example of this would be a
vehicle crossing a truss bridge; the forces in the members will of course vary with
the position of the load. The variation of force in a particular member, with respect
to the load position, can be displayed in a graphical form, known as an influence-
line diagram.
An influence line can be defmed as the graph of a particular function plotted
against the position of a unit load as the unit load transverses the structure. The
PLANE STATICS 29
function could be one of many different quantities, for example the shear force,
bending moment or deflection at a particular point, or perhaps the force in a
particular member.
It is absolutely essential from the outset to distinguish between, say, the S.F.
diagram for a beam and the influence line for shear force at a particular point on
the beam. The first diagram will give the value of the shear force at any point on
the beam for a particular set of applied loads; whereas the influence line will give
a diagram showing how the shear force varies at one particular point as a unit
load moves over the beam.
1.18 Influence lines for reactions, shear force and bending moment
.,
-------:~8
(a)
(b)
(c)
(d)
Figure 1.29
A simply supported beam is shown in figure 1.29a. The influence line for the
reaction at A is first required. If the unit load is placed a distance x from A the
value of VA can be found by taking moments about B
VA/=1(1-x)
1-x
VA=--
1
This value of VA can now be plotted against x to form the influence line (figure
1.29b). VA has a maximum value of unity when the applied load is at A and a
value of zero when the applied load is at B, the variation being linear.
The influence lines are now required for shear force and bending moment at C,
distant a from A.
Shear force
For x <a the shear force at C is equal to the value of V 8 which is xjl. The shear
force= xjl, adopting the usual sign convention.
For x >a the shear force at Cis equal to the value of- VA, that is,-(!- x)/l.
In both expressions there is a linear variation of shear force with position of the
applied load. When x =a there is a sudden change in value from ajl to -(l- a)/l,
a total change of unity.
The influence line is shown in figure 1.29c. Note that the inclined lines must
be parallel to one another and that there is a unit change at C.
Bending moment
For x <a the bending moment at Cis
X
VB (l - a) = - (l - a)
l
For x >a the bending moment at Cis
VAa= e~X )a
Both of the expressions are positive as the bending moment is sagging and
they also vary linearly with x.
At C whenx =a
a
M=-(l-a)
l
The influence line for the bending moment at C is shown in figure 1.29d.
1.19 Loading systems

At this stage it might be pertinent to enquire into the possible use of an influence
line. So long as the system is linearly elastic the principle of superposition can be
invoked, and the influence line used to find the value of the function when any
loading system is applied to the structure.
Let figure 1.30b represent the influence line at C for a particular function.
This has of course been drawn for a unit load, so that if a load P crosses the
beam, the value of the function will be given by Py, where y is the height of the
influence-line diagram"at the point under the load. The maximum value is of
course given by Pymax. For a series of concentrated loads the value of the function
will be given by ~ Py.
Next consider the case of a uniformly distributed load p. The load acting on a
small element dx of the beam is p dx; as dx -+ 0 the load can be considered as
acting at a point and the value of the function would be py dx. The total value of
the function due to the complete distributed load can be found by integrating
this expression. That is p I y dx since p is a constant. I y dx is the area of the
influence-line diagram underneath the uniformly distributed load.
PLANE STATICS 31
(b )
Figure 1.30
It is comparatively simple to find the maximum value of a function for a single

point load, but the case of a uniformly distributed load requires further discussion.
Referring back to the S.F. diagram shown in figure 1.29c. If the U.D.L. is
larger than the span of the beam and moving from left to right, the maximum
positive shear will occur when the front of the load has just reached C, and is
equal top times the area above the base line. The maximum negative shear will
occur when the rear of the load has reached C. If the load is shorter than the span,
no particular difficulty arises ; again there would be two cases to consider, the
front and the rear of the load at C.
Figure 1.31
There is no need to discuss the case of bending moment when the load is longer
than the span. If the length of the load is b and this is shorter than the span l an
area such as that shown in figure 1.31 will have to be a maximum. The influence
line for bending moment has been drawn for a point distant a from the left-hand
support, the maximum height of the diagram being c, with the rear of the load at
a distance x from the point considered.
The simplest approach is to find when the unshaded area is a minimum.
c(a - x ) 2 c(/ - a - b + x ) 2
Unshaded area A = + ......:....-----'--
2a 2(1- a)
when dA/dx =0
(a-x) (1-a-b+x)
---+ =0.
a 1- a
This simplifies to
x a
-=- (1.16)
b
Thus the point considered divides both the length of the load and the span in
the same ratio. Alternatively the two heights MN and OP are equal.
The influence lines considered so far have been constructed for very simple
cases. If use is made of the principle of virtual work, the influence line can often
be sketched very rapidly for more complicated cases.
'
1. 20 Application of virtual-work methods to influence lines
A c B
~~ ~
~
tp I I
(b)
_____, +Q
(e )
(c l
//~//1
(f)
~i (d)
Figure 1.32
Reactions
We now meet our first real application of virtual work. The influence line for the
vertical reaction at A is required. A virtual displacement Ll is applied to the beam
PLANE STATICS 33
at A. The corresponding displacement of Pis ·o (figure 1.32b ). Since the system is
in equilibrium
or
If we make both P and Ll equal to unity, then VA = o. Thus an influence line has
been constructed for the vertical reaction at A.
In general, to draw the influence line for a reaction, the point of application
of the reaction is given a unit displacement in the direction of the reaction. The
resulting beam displacement will give the influence line.
Shear force
The influence line for shear force at point Cis required. It is necessary to remove
resistance to shear force at this point, but the resistance to both bending movement
and direct force must remain intact. We could imagine the beam cut through at C
and a mechanism such as that shown in figure 1.32e introduced. When the shear
forces Q are applied this would deform as shown in figure 1.32f, the final position
being when both the links were vertical. The deformed beam is shown in figure
1.32c and the two portions are displaced through a relative vertical distance Ll at
C, the corresponding displacement of P being o.
By the principle of virtual work
QLI- Po= 0
Again put Ll and P equal to unity. Thus Q =o and the influence line for the shear
force at C has been constructed.
It is essential that the two displaced portions of the beam should remain parallel.
If not, there would be an extra term in the virtual-work equation involving the
bending moment at C and the relative rotation of the two parts of the beam.
Bending moment
The influence line for bending moment at point C is required. The resistance to
bending moment must be removed, but the beam must still be capable of carrying
both shear and axial forces. This can be achieved by inserting a pin at C. Bending
moments Min opposite directions are applied on each side of the pin (figure 1.32d),
and a relative virtual rotation e is imposed, e of course being a very small angle.
The corresponding displacement of Pis o.
By virtual work
Me- Po= o
Thus
Once again put P = 1, e = 1 then M = o and the influence line for bending moment
has been constructed.
As 8 is assumed small, the distances shown at the supports can be taken as afJ
and bfJ. Thus the height of the influence-line diagram at a distanced from the
left-hand support is
dbfJ db(J db
- - = - =-
a+b I
when 8 is unity. So that the maximum height of the diagram is abjl and this
occurs at C.
The application of these methods can probably best be illustrated by means of
an example.
A 8 R s c p D T U E F (a)
v I'-. t"-...
~ ...........
~
(b)
I/ I/~
[7
(c)
~ 17
w
'
I/~ (d)
~~1/c P """" 1.---
Figure 1.33
A bridge girder (figure 1.33a) has a total length of 100m. It is supported at

six points A, B, C, D, E and F, each span being 20m. The girder is continuous
over the supports, but has hinges at R and Sin span BC and at T and U in span
DE. BR = SC =DT = UE = 5 m.
Influence lines are required for the following: vertical reaction at C, and shear
force and bending moment at P (situated 5 m from C in span CD).
The influence lines have all been sketched in figure 1.33. Figure 1.33b shows
the reaction at C, c and d show the shear force and bending moment at P. In all
cases the deformed beam, that is, the influence line, must remain in contact with
the supports, except in b where unit displacement is applied at support C. The
beam will pivot about the hinges SandT.
PLANE STATICS 35
For the vertical reaction the height of the influence line at Cis unity. For the
shear force there is a unit change at P and in the deformed position SP is parallel
to PT. For bending moment WC =CPO, where 8 is the relative rotation of the two
parts of the beam at P. Now CP is 5 m and with 8 = 1, WC = 5 m.
The results can now be written down.
A-R s c p D T U-F
Vc 0 5/4 1 3/4 0 -1/4 0

shear force 0 -1/4 0 1/4 to -3/4 0 1/4 0
bending moment 0 -15/4 0 15/4 0 -5/4 0
1.21 Multi-load systems

So far we have only considered the use of the influence-line diagram for single
concentrated loads and uniformly distributed loads. The situation will often
arise however, when a series of concentrated loads roll across the structure, and
the maximum value of a particular function for a chosen point is required. Alter-
natively the absolute maximum value for a function for any point on the structure
might be required.
~B
(a)
(b)
Figure 1.34
Shear force
The train load P 1 - P6 (figure 1.34a) is travelling across the beam from left to
right and the maximum value of the shear force at C is required. First the
influence line for shear force at Cis sketched as shown in figure 1.34b. As the
loads move across the beam, the positive value of the shear force will steadily
increase until P 1 is just to the left of C. When P 1 crosses C there will be a sudden
decrease in shear force by an amountP 1 • With the loads continuing to move to
the right the value of the positive shear force will again increase while that of the
negative shear force due toP 1 will decrease. This state of affairs. will continue until
P 2 is just to the left of C, and the net shear force can be found. It can now be
seen that the maximum value of the positive shear will occur when one of the
point loads is just to the left of C. It also follows that the maximum value of
negative shear will occur when one of the loads is just to the right of C.
The maximum value can be found by trial and error, but will always occur
when one of the loads is just to the left or right of the point considered. Trial-and-
error methods are often rather tedious, and it is possible to reduce the calculation
involved.
Suppose that load Pn has just crossed C, the positive shear will have decreased
by Pn. As the loads advance there will be a gradual rise in the positive shear equal
to the change in value of the reaction VB. This will increase until Pn + 1 reaches
the point C. The total change in shear force dQ is given by
dQ = LIVB- Pn (1.17)
where Ll VB represents the difference in the value of VB when Pn and Pn + 1 are

placed in turn at point C. If equation 1.17 is positive, the shear force will have
been increased by moving the loads forward. The loads are thus advanced until
there is a change of sign in equation 1.17. Once the sign becomes negative the
maximum value of the shear force has been passed. The change in VB can be
found very simply if all the loads remain on the beam. Care must be exercised if
a load comes onto or leaves the beam.
The absolute maximum value of the shear force, that is, the maximum value
that can occur at any point of the beam, will always be at one of the supports for
a simply supported beam. Consideration of the shear-force envelope discussed
below will confirm this.
In figure 1.34 the influence line was drawn for point C. There is a unit change
in the value of the shear force at this point. The influence line for a point next to
the support A would be a triangle of unit negative value at A and zero at B.
Similarly for a point at B we should have a triangle with unit positive height at B.
If we draw both of these influence lines on the same diagram (shown dotted in
figure 1.34b ), we shall have two parallel lines, unit distance apart. These lines
form the envelope for the series of shear-force influence lines. If the influence line
is required for a specific point, it is only necessary to join the two parallel lines by
a line perpendicular to the base line, passing through the required point.
Bending moment
The train load (figure 1.35a) is crossing the simply supported beam from left to
right. The influence line for bending moment at point C has been drawn in figure
1.35b. The question arises, where should the loads be placed such that the
bending moment at Cis a maximum?
Let R 1 be the resultant of all the loads on the beam to the right of C, and R 2
the resultant of those on the left. The value of the bending moment at C will be
given by
PLANE STATICS 37
A~f':1-----a...;....:_ ~--+i -~~

_,,...j...l -b B
I (a) I
I I
I I
~
~~~
f-x 2 ~ ~x 1 ~
(b)
Figure 1.35
Let the loads now move a small distance dx to the right, such that dx = dx 1 =
dx 2 , but the values of R 1 and R 2 remain unchanged. There will be a change in the
value of the bending moment
dM=( R:d- R~d)dx

Thus, as the loads move further onto the beam, the value of M will continue
to increase as long as
(1.18)
Let us now assume that the inequality is true, and that P5 comes onto the beam
before P 3 has reached C. The value of R 2 will be increased while that of R 1 stays
constant. Thus the inequality will not be affected by a load coming onto the beam.
The same kind of argument will apply if P 1 leaves the beam.
If however P 3 reaches and crosses C the value of R 1 will increase suddenly and
that of R 2 will decrease and the inequality may no longer be true.
We can now see that the maximum bending moment will occur when one of
the loads is at the point C. The inequality (equation 1.18) may be used to deter-
mine which load will give the maximum value.
So far we have only considered a method for finding the maximum bending
moment that can occur at a particular point. Let us next consider the problem of
finding the absolute maximum bending moment as the train of loads crosses a
simply supported beam. From our previous cases, it is clear that the maximum
bending moment must occur underneath one of the loads.
Let us first find the position of P4 such that the maximum bending moment
occurs under P4 (figure 1.36).
Let ~ P be the total load on the beam, and let a be the distance of the centroid
of this load from P4.
Figure 1.36
If x is the distance of P4 from the left-hand support

(1-x-a)
VA='2P 1
The value of the bending moment under P4 is

VAx-P 5 b 5 =M
dM (1- 2x- a)
- =P = 0 for a maximum
dx 1
then
1- a
x=-- (l.19)
2
Thus the distance of P 4 from the support at A is(/- a)/2. The distance from
the position of l:: P to the support B is I- x- a= (l- a)/2.
Thus the maximum moment, under a particular concentrated load, will occur
when the mid-point of the beam bisects the distance between the load being con-
sidered and the centroid of all the loads on the beam.
To fmd the absolute maximum it is necessary to proceed by trial and error.
However in general the maximum bending moment will occur somewhere near
the centre of a simply supported beam. Assuming that it occurs at the middle, the
inequality 1.18 may first be used to find which load should be at the mid point,
then the method just described can be used to make final adjustments. Hence the
absolute maximum value can be found.
x - - - 1 - - - - - 1-x ----+-~
A c 8
Figure 1.37
PLANE STATICS 39
Before concluding this section let us determine the shape of the bending-
moment envelope. The maximum height of the influence line for the bending
moment at C (figure 1.37) has a value x(l- x)/l (section 1.20). Thus the bending-
moment envelope will be a parabola with maximum height l/4.
1.22 Influence lines for girders with floor beams

In the case of a fairly large span, the loading system is seldom applied directly to
the beam. The usual form of construction is shown in figure 1.38a. Large beams,
called girders, are first placed over the span, and then a number of floor beams
a-e, are placed on top of, and at right angles to, the girders. These floor beams
carry stringers, which are parallel to the girders and are simply supported on each
stringer. The decking system or roadway is placed on the top of the stringers. The
8
(a)
I --- --- T
1
l
Figure 1.38
parts of the girder between a and b etc. are termed panels, and a, b, etc., are
referred to as panel points. The effect of a load P placed on a panel will be trans-
mitted to the girder via the two panel points. The floor beams are considered as
being simply supported at the panel points.
The values of the reactions at A and B will not be affected by this type of con-
struction, thus the influence line for reactions is identical to that of a simply
supported beam.
For a load applied at any particular point on a stringer, the value of the shear
force will be constant in any particular panel. Thus we can refer to the shear force
in a particular panel rather than the shear force at a particular point on the girder.
We shall now construct the influence line for shear force in panel cd where the
length cd is/. When a unit load is applied to the left of c the shear force will be
equal to VB, and with the load applied to the right of d, the shear force will be
equal to VA. These two portions of the influence line can be constructed and are
identical to that of a simple beam. Consider the load applied between c and d at a
distance x from c, with panel lengths all equal to I. The reaction at c will be
(/- x)/1 and that at d will be xjl. The shear in the panel will be given by
where m and n are the heights of the influence line at c and d respectively. This is
the equation of a straight line, joining the ends of the two lines that have already
been constructed (figure 1.38b ).
The influence line for bending moment has to be drawn for a particular point
on the girder as the bending moment will not be constant in a panel. We shall draw
the influence line for point C on the girder. First the influence line for a simple
beam is drawn, this will be correct apart from the portion between the two panel
points c and d (figure 1.38c).
The bending moment between c and d can be expressed as
1-x) x
M= ( -~- p +lq
where p and q are the heights of the influence line at c and d respectively. Again
we have the equation of a straight line joining the two ends of the influence line
already constructed.
1.23 Influence line for three-pin arch

The symmetrical three-pin arch (figure 1.39a) has a rise h and a span /. The
influence lines at D (x, y) are required for bending moment, shear force and axial
force.
For a unit load to the left of D
Mo = VB(l-x)-Hy
For a unit load to the right of D
Mo = VAx-Hy
It can be seen that the influence line for bending moment can be obtained from
the influence lines for VA, VB and H. The influence lines for VA and VB are iden-
tical to those for a simply supported beam (figures 1.39b and c). A closer inspec-
tion of the first term in each of the expressions for Mo, reveals that this represents
the influence line for the bending moment at D for a simply supported beam,
that is, H = 0.
The influence line for His shown in figure 1.39d. The shape of this can be
determined by the method of virtual displacements; the maximum value is l/4h.
PLANE STATICS 41
c
H- .--x-t
t~
~------------ (------------~
VA~
(b)
Va
(c)
(d)
(f)
Figure 1.39
The influence line for the bending moment at D can now be drawn and con-
sists of the difference between the influence line for M 0 for a simply supported
beam and a constant y times the influence line for H (figure 1.39e ).
The influence line for the shear force at D can be found from the combination
of two effects; the vertical load acting on a simply supported beam and the hori-
zontal thrust. If 8 is the slope of the arch at D, the influence line for shear force
is given by
Influe~ce line for shear at D,} cos 8 _{Influence line for H} sine
{ for a Simply supported beam
This is sketched in figure 1.39f.
The construction of the influence line for axial force is left as an exercise for
the reader.
Problems
1.1 A beam ABC 12 m long is simply supported at A and B where AB = 8 m.
Point loads of 15 kN and 6 kN are applied at the mid-point of AB and at C
respectively. A uniformly distributed load of 4 kN/m extends for 6 m from A.
Determine the values of the reactions by resolving and taking moments. Check
your results by using virtual work. Sketch the shear-force and bending-moment
diagrams and find the values under the point loads.
1.2 A simply supported beam of length lis symmetrically loaded with a distri-
buted load that increases linearly from zero at the supports top per unit length
at the centre. Sketch the shear-force and bending-moment diagrams and find the
value of the equivalent uniformly distributed load that would give a maximum
bending moment equal in magnitude to the actual maximum bending moment.
A support is now introduced at the centre of the beam and its height is ad-
justed until the bending moment at the centre is zero. Find the position and
magnitude of the maximum bending moment.
(a) (b)
p 1.3
1.3 Two graphs are shown in P 1.3a and b. Take these first to be bending-moment
diagrams and sketch the corresponding shear and loading diagrams. Then take
them to be loading diagrams for beams simply supported at each end and sketch
the corresponding shear-force and bending-moment diagrams.
1.4 A beam is simply supported at A, B, C and D. AB =CD= 8 m, BC =12m.

Pins are inserted in span BC, 3 m from B and C respectively. AB has a uniformly
distributed load of 4 kN/m, and CD a uniformly distributed load of 6 kN/m, with
a point load of I 0 kN applied at the centre of span BC. Show that the beam is
statically determinate, and find the values of the reactions at the supports. Draw
the shear-force and bending-moment diagrams.
1.5 Sketch the shear-force and bending-moment diagrams for the four beams
shown in P 1.5. State whether each case is statically determinate or redundant.
PLANE STATICS 43
• t
p 1.5
1.6 Show that the bending-moment diagram for a statically determinate beam
with vertical loads applied is represented by a funicular polygon. Illustrate your
answer with reference to a simply supported beam of length l with vertical loads
of P/2 at the quarter points and a load P applied at the centre.
A symmetrical arch of varying cross-section is semicircular in form and has a
span of/. It is pin-jointed at the abutments. A loading system identical to that on
the beam is applied. The horizontal thrust is found to be P/2. Discuss how the
line of thrust might be found. Sketch the bending-moment diagram for the arch,
and determine the maximum value.
1.7 Show that the differential equation for a vertical load funicular polygon is
given by the expression d 2y/dx 2 = -p/H Discuss what is meant by H
An arch is pinned at each abutment and has a span of/. A third pin is placed
at mid-span where the rise is l/2. Define the shape of the arch for the following
load system if no bending action is permitted.
Horizontal distance from l I 31 7/ 3/

left-hand abutment 8 4 4 8 4 4
p
Load - P 2P P distributed load of
2 4P
l per unit horizontal length
1.8 For the beam described in problem 1.1 when no loading is present, sketch
the influence line for the reactions at A and B, and the influence lines for shear
force and bending moment at the mid-point of AB.
1.9 For the beam described in problem 1.4 when no loading is present, sketch
the influence lines for the reactions at A and B, and the influence line for shear
force and bending moment at a point 5 m from A. Determine the maximum
bending moment and shear force at this point when two loads of 6 kN and 3 kN
spaced 3 m apart traverse the beam.
1.10 A symmetrical parabolic arch rib is hinged at both abutments and also at
the crown. The span is 40 m and the rise 10 m. Draw the influence line for the
horizontal thrust at the abutments, and the influence line for the bending moment
at a point whose horizontal and vertical distances from the left-hand abutment are
x andy. Find the position and magnitude of the maximum bending moment as a
50 kN point load traverses the rib.
1.11 Determine the maximum bending moment and shear force that can occur at
a third point of a simply supported beam of 24m span when the following load
system traverses the beam.
load (kN) 5 10 15 10
spacing (m) 2 2 2
Also determine the maximum possible bending moment and shear force that can
occur in the beam.
2 STATICALLY DETERMINATE STRUCTURES
2.1 Simple plane trusses

A structure of this type consists of a number of members that are connected by
pins at the ends, all the members lying in a single plane. Of course most structures
that exist are three-dimensional, but in a large number of cases it is possible to
treat the problem from a two-dimensional point of view. An example of this would
be a bridge consisting of two parallel trusses connected by transverse members
carrying the roadway. If the loads that are transmitted by the transverse members
to the trusses are known, then each truss can be analysed as a two-dimensional
system.
As the discussion is being limited to pin-connecred members at this stage, no
bending moments will be developed in the members if the loads are applied
through the joints of the structure-they will only be subject to tensile and com-
pressive forces. If however a load is applied between the ends of a particular mem-
ber, that member will have bending moment and shear force developed in it, in
addition to an axial force.
N\N\ Warren Truss Warren Truss With Verticals
Pratt Truss Howe Truss
K Truss Roof Truss

Figure 2.1
45
Figure 2.1 shows five different types of bridge truss and one typical roof truss.
The top and bottom members, referred to as chords, need not necessarily be
parallel. The two inclined members at the ends are called end-posts and the rest
of the members are referred to as diagonals and verticals.
2.2 Stability and determinacy

In the previous chapter we discussed the question of determination of reactions
without any reference to the structure itself. It will now be necessary to decide
whether the complete structure is unstable, statically determinate or redundant.
F9r pin-jointed structures the unknown quantities will be the axial force in the
members arid the reactions at the supports. Thus if there are b members and r
reactions the total number of unknowns will be b + r.
For a plane pin-jointed framework it is possible to resolve for the forces in the
framework in two directions at every joint in the system. Thus for j joints a total
of 2j equations can be formed relating the unknowns. If
b + r = 2j the structure will be statically determinate (2.1)
b + r < 2j the structure will be unstable or a mechanism (2.2)
b +r > 2j the structure will be statically indeterminate or redundant (2.3)
It is necessary to qualify these statements to some extent, as it is quite possible
to conceive a structure that might be redundant in one part and unstable in
another, and yet a count of members, reactions and joints would satisfy equation
2.1. We can state here and now that it is no good blindly applying the above
relationships; it is always necessary to inspect a structure carefully at the same
time.
A B
Figure 2.2
The Warren truss (figure 2.2) has a total of 11 members and the number of
reactions is 3, 2 of these are at A and I at B. Thus b + r = 14. There are 7 joints
in the framework, so 2j = 14. In this case b + r = 2j and the truss is statically
determinate.
If the truss had been mounted on rollers at each end the value of r would be
2 and b + r = 13. Thus b + r < 2j and we have an unstable system.
It should be noted that the loading applied to the framework does not enter
into the question of stability or redundancy.
The result of applying equation 2.1 without examining a framework carefully
can be demonstrated by the very simple frameworks in figure 2.3. In all three
cases the equation b + r = 2j is satisfied. However for case a the framework itself
STATICALLY DETERMINATE STRUCTURES 47
Figure 2.3
is unstable (consider a small lateral displacement) and the reactions are redundant.
For b the framework is redundant and the reactions are unstable. Case c will be a
suitable statically determinate structure.
Let us consider case b in a little more detail. It has been stated that the frame-
work itself is redundant. The simplest framework that can be formed from a mini-
mum number of bars is that of a triangle (figure 2.4). This will be statically deter-
minate as long as the support conditions are correct. Suppose that it is required to
add one more joint D to the basic triangle ABC. If the problem is to remain stati-
cally determinate we can only connect D to the triangle by two bars.
/ .. 0
LSJ
A
Figure 2.4
C
This can be seen very simply if we consider a load applied in any direction at
D. It would be possible to resolve the load in two directions at D and hence find
the forces in the members meeting at D. If we add a third bar DA the frame
becomes statically indeterminate.
Thus if a framework is built up from a simple triangle and each joint added is
only connected by two members to the framework, the frame will be statically
determinate. This will lead to the following relationship for the members in a
statically determinate frame
b = 2j- 3 (2.4)
A number of texts on structures make use of this relationship but in certain
cases it leads to diftlculties and the relationship given by equation 2.1 is preferred.
There are a number of different methods for finding the forces in the members
of a framework and it is often a matter of experience and practice to decide which
one should be used. The sign convention used throughout will be that tensile
forces are considered positive.
2.3 Resolution at joints

As the name suggests this method consists of resolving in two different directions
at each joint in the framework. The directions need not be vertical and horizontal
and in many cases a quicker result may be obtained if other directions are con-
sidered. While the resolution is proceeding a check should be kept on the direction
of the forces, that is, whether a member is in tension or compression. The most
convenient method of doing this is to enter arrows on each member of a line
drawing of the framework. Thus ~ would indicate compression and -
would indicate tension. With practice, if the angles between members happen to
be simple ones, the forces can be written down straight away on the line drawing.
The frame should first be checked to see if it is statically determinate. It is
always worthwhile spending a few moments on an inspection of the frame. Some-
times certain members may not be loaded and if this can be seen straight away,
time may be saved later on. Often it will be necessary to determine the values of
the reactions before resolution can start.
Figure 2.5
For the framework in figure 2.5 we need not attempt to determine the
reactions at G and D as resolution can start straight away at joint A. However
after checking that the frame is statically determinate, we can see that the force
in EB is zero (vertical resolution at B). This will mean that the force in BC is
equal to that in AB. Again as the force in EB is zero, by resolving perpendicular
to EF at E, the force in EC is zero. In fact the forces in all members will be zero
except for those lying along AG or AD. (Consider joint C and then joint F.) If
this is so the forces in AE, EF and FG must all be equal as are those in AB, BC
and CD.
Resolution of forces at joint A
Vertical
F AE sin 30° = P therefore F AE = 2P tensile
Horizontal
F AB = F AE cos 30° therefore F AB = v3P compressive
Results
FAB =FBc =Fen =-y3P
F AE = F EF = F FG = 2P
We might question the necessity of having the vertical and diagonal members
in this frame as the forces in these members are zero. Basically the frame could be
constructed out of two members AG and AD. If these were continuous the frame
would be stable and the two member forces could be found by resolution at A. If
AG and AD are made up of several shorter members connected together, the
extra members will be required even though the forces in the members are zero.
The framework has to be stable and this stability is completely independent of
the actual loading system applied.
In our analysis we have ignored the dead weight of all the members themselves.
If this were taken into account there would be a force, for example, in member
FD.
0
1
22·5m
·~~G~H~J
15m - t - 1 5 m ---t--15m ----t-15m
50kN 70kN 40kN
Figure 2.6
A somewhat more complicated example is shown in figure 2.6. The frame is

statically determinate and by taking moments about A and vertical resolution for
the whole frame, the two reactions at A and B are
VA= 82·5 kN VB= 77·5 kN
By inspection FGn = 70 kN and FFG =FGH·
The lengths of members CD and FD will be required
CD = 16·8 m, FD = 27 m
In a case of this kind where the framework is symmetrical but the loading
system is not, it is probably best to start resolution at each end of the frame.
Joint At FAc = -82·5y2 = -116·5 kN
~ FAF = 82·5 kN
Joint B t FBE = -77·5y2 = -109·5 kN
~ FBH = 77·5 kN
Joint C-+ 15 -116·5 F

Fco x - - - - - CD = -92·3 kN
16·8 y2
116·5 ( 7·5)
t FcF= V2 - 92·3x 16 _8 =41·2kN
IS -109·5
JointE-+FE 0 x 16 _8 y 2 FED=-86·5kN
109·5 (
t F =- - - 86· 5 X - 7·5 ) = 38·8 kN
EH y2 16·8
Joint F t 22·5
FFDxTI =50-41·2 FFD =10·6kN
-+ FFG = 82·5 - ( 10·6 x ~~) = 76·6 kN

Joint H t 22·5
FoH x 27 = 40- 38·7 FoH = 1·6 kN
-+ FGH =77·5-(1·16x~~)=76·6kN
The analysis is now complete and a check has been obtained in that FFG and
FGH are the same. A further check could be obtained by resolution at joint D.
2.4 Graphical method

A series of force triangles or polygons are drawn for each joint in the structure.
They are connected together in one diagram to avoid the repetition that would
result if a separate polygon were drawn for each joint. In connection with this
method it is useful to employ Bow's notation such that the forces in the members
of the structure are easily identified.
It is first necessary to make an accurate scale drawing of the framework. The
spaces between all external and internal forces are lettered. The force diagram,
drawn to a suitable scale can then be started. It is essential to commence the
drawing from a joint at which there are only two unknowns.
The truss in figure 2.7a has all members the same length. By inspection
VA= Va = P and HA = Ha = 0. The spaces between the forces have been lettered;
note that space U extends right round the top outside edge of the truss and
separates VA and VB. The force diagram could be started from either A or B, as
at both these points there are one known and two unknown forces.
Starting from Band proceeding round this joint in a clockwise manner, we see
that from U to S there is a vertical upwards force of VB = P. This is drawn on the
force diagram and lettered such that sis above u, indicating that R 8 acts upwards.
The rest of the force triangle for joint B can now be drawn and is represented by
rsu.
s
c u 0
B
p
(a) (b )
Figure 2.7
We can now proceed to joint D, again working round this joint in a clockwise
manner urq will be the force triangle. Proceeding systematically round the truss
the force diagram can be completed (figure 2. 7b ).
The values of the forces can be scaled off the force diagram and their directions
found making use of Bow's notation. As an example the force in member CE is
represented by the distance qp. We proceed round joint C in a clockwise direc-
tion from Q toP-this is represented by the direction qp. An arrow is entered on
the line diagram, close to C, in this direction. If we repeat the process for joint E
we shall obtain an arrow in the opposite direction. Thus it can be seen that the
member CE is in tension.
The accuracy of the final result is entirely dependent on the accuracy of the
two scale drawings. In the majority of problems some idea of the accuracy can be
obtained when the last few lines are drawn on the force diagram. These should
meet at a point, but a small triangle or error may very well result. If the errors are
too large to be acceptable, it will be necessary to check both diagrams most care-
fully to find any major source of error.
2.5 Method of sections

In this method, ascribed to Ritter, an imaginary cut is made through the truss and
forces are applied to each part of the structure to keep it in equilibrium. These
applied forces will have the same values as the forces in the members that have
been sectioned. As there are only three equations of equilibrium it will not be
possible to find the values of the forces if more than three members are cut, in
separating the two parts of the truss, unless the forces in some of the members
are already known.
A great advantage with this method arises when forces are only required in
certain members of a truss; very often a section can be made and the forces
determined at once. In both the methods that have been discussed previously, it
would be necessary to start at one end of the truss and work steadily through the
joints until the required forces were found .
The forces in members AB, CD, AD and AC are required for the Pratt truss
(figure 2.8a).
(a) (b)
Figure 2.8
A section 1 is made cutting AB, AD and CD. The isolated left-hand part of the
frame is shown in figure 2.8b with the necessary forces for equilibrium.
'J:.MA
4FcD + 3P= 6 x 2~P
therefore
FeD= 3P tensile
'LMo
4FAB + 3P + 6P = 9 X 2!P
therefore
F AB =3iP compressive
The force in AD can be found very simply by considering the vertical equilib-
rium of the sectioned portion. The vertical component of the force in member AD
must balance the applied vertical loading and reactions on the left-hand portion of
the truss.
Thus the vertical component of FAD is
p
2!P- 2P=-
2
therefore
P 5 SP
FAD =- x- =- tensile
2 4 8
Figure 2.9
We can proceed in the same manner to find the force in AC, but a new section,
2, is required.
Vertical equilibrium for the left-hand portion of the frame gives
p
F AC = 2 compressive
A particular case arises with a K truss, where it is possible to section four

members. If the section is made in the manner shown in figure 2.9, the force in
CD can be found by taking moments about A, and that in AB by moments about
C. Also note that by horizontal resolution F BE = -FED.
2.6 Compound trusses

In the case of simple trusses it is possible to make a complete analysis of the struc-
ture by any of the methods previously discussed. For compound trusses however
it is very often necessary to make use of the method of sections in addition to
resolution at joints.
Several examples of compound frameworks are shown in figure 2.1 0. These are
often formed by connecting two or more simple trusses with sufficient members
to make the structures rigid. In figure 2.1 Ob, for example, there are two simple
triangles connected by three members. Figure 2.1 Oc shows two smaller roof trusses,
each consisting of half the span, again interconnected by three members.
A brief discussion follows on a possible method of solution for each problem
illustrated in figure 2.1 0.
Figure 2.1 Oa: This problem can be started by resolving at A then D. Difficulties
will arise however if we now attempt to resolve at C-five members meet at this
point, and the forces in only two of these have so far been found. If a section is
put through as shown, the forces in EG and FH and CH can be found. It is now
possible to return to C and resolve as only two unknowns remains. The rest of the
analysis can then proceed.
Figure 2.10b: A section is put through the three interconnecting members.
Moments can then be taken about the point of intersection X of two of the con-
necting members-this will give the force in the third member, and resolution can
then proceed normally.
Figure 2.1 Oc: This can proceed on similar lines to a.
Figure 2.1 Od: In this case there are three unknowns at every joint and it will
not be possible to make a start by resolving. Again it is not possible to put a sec-
tion through without cutting more than three members. If we examine the prob-
lem it consists of two basic triangles AEF and BCD connected by three members
AC, DE, and FB. The triangle AEF is shown isolated in figure 2.1 Oe. The values
of FAc, FoE and FFB can be found by resolving and taking moments. No further
difficulties will arise.
2. 7 Complex trusses
In this type of truss it will not be possible to solve directly using resolution at
joints and method of sections. Indeed it is often difficult to determine whether
CD
li...u.
/
UJ
u
or not a truss of this type is statically determinate. The usual test-by satisfying
equation 2.1-can be applied, and this will check that the number of members is
correct. It is much more difficult to decide whether these are in the correct
positions and it might well be that one part of the truss is statically indeterminate
and another part unstable.
F
(c)
Figure 2.11
Three examples of complex trusses are given in figure 2.11. In all these cases
there are at least three members meeting at every joint. It would of course be
possible to resolve at every joint and form a total of 2j equations in terms of all
the unknown bar forces-these equations could then be solved. This approach
would be somewhat tedious, but if used it would be best to employ the method
of tension coefficients which will be dealt with in the section on three-dimensional
statically determinate structures.
An alternative approach would be to proceed from one of the supports and
resolve as far as possible. Having reached a point where there are three unknowns,
let one of these be T. The resolution can now proceed, and the value ofT found
when resolving at the last joint in the structure. Once Tis known the rest of the
forces in the members will be known.
2.8 Virtual work

This is a useful method to assist the analysis of a complex truss. If the principle is
applied once to the structure we can determine the force in one of the members.
Having determined this force, the rest of the analysis can generally proceed by
resolution at joints. It may in certain cases be necessary to apply the principle of
virtual work a second time to find the force in a second member.
The method of application is very similar to that demonstrated in section 1. 7.
Basically a member of the truss is removed and replaced by an unknown force
applied at each end of the member. The frame will now have turned into a mech-
anism and a virtual displacement is applied in the direction of the removed mem-
ber. The resulting displacements of all the applied forces have to be calculated.
The total work done can then be summed to zero.
In certain simple cases it may be easy to calculate the displacement of the
applied forces in terms of the applied virtual displacement. If the problem is
complicated it is probably best to use a graphical approach-termed a displacement

diagram.
As an introduction let us consider the very simple example shown in figure 2.12.
The force in member AC is required, so this member is removed and a force equal
to FAc is applied-the frame will become a mechanism pivoting about B. We start
the displacement diagram from a pole o which also represents any fixed points, in
this case A and B. The point D can now describe a circular arc of radius BD, but
for a very small initial movement the motion will be perpendicular to D. This can
be represented by a line od of suitable length on the displacement diagram. The
motion of C relative to B is perpendicular to BC and the motion of C relative to D
is perpendicular to CD.
oab
(a) (b)
Figure 2.12
These two directions can be represented by lines on the displacement diagram

drawn from b and d respectively which must intersect at c. Note that in this case
we have a triangle bed which will be similar to BCD. Let od equal x. The point c
has moved horizontally and vertically through x/2. We require the displacement
in the direction of AC, that is, x/2.
The virtual-work equation can now be applied. Assuming that AC is in tension,
the horizontal displacement of Cis in the opposite sense to the force. Thus
X
-FAc -+Px = 0
2
or
As a more complicated case we shall consider the same truss that was discussed
in section 2.7. The force is required in member BE, this has been removed in
figure 2.13a. It is probably best to give E a known horizon tal displacement oe;
note Eison rollers and can only move horizontally. The point f can next be
fixed; d must coincide with o as the point D will not move for a small initial
movement of E; c can now be fixed and this can be followed by b and the dis-
placement diagram (figure 2.13b) is complete.
(a) (b)
Figure 2.13
The movement of E relative to B is given by be-we require the component in

the direction of BE, that is, bh. Neither B nor D move vertically and the vertical
movement of Cis given by cj. This implies that there is no virtual work term for
P 1 or P 3 , and that for P 2 is given by -P2 cj.
The complete virtual-work equation assuming BE in tension is
FBE bh -P2 cj =0
therefore
2.9 Statically determinate space structures

The type of structure to be considered in this section will consist of members that
are jointed at their ends such that complete rotational freedom exists and only an
axial force will act in the members. It will be necessary to determine whether a
three-dimensional structure is stable and statically determinate. We can proceed
on similar lines to the test for a plane structure.
The total number of unknown forces in the members will be equal to the num-
ber of members bin the structure; in addition to this there are the unknown
reactions r, making a total of b + r unknowns. If
b + r = 3j the structure will be statically determinate (2.5)
b + r < 3j the structure will be unstable
b + r > 3j the structural will be statically indeterminate
It is quite possible for equation 2.5 to be satisfied, but part of the structure
could be unstable and another part statically indeterminate. This is more difficult
to determine by inspection than in a plane structure, one reason for this being
that 'thinking' in three dimensions does not come automatically to the majority
of people. The simplest basic element for a space structure is a triangle-if a further
joint has to be fixed in space, it will be necessary to provide three more members
connected to the triangle. For each additional joint after the first three, it is
necessary to provide three members. We can often make use of this when checking
through a framework.
When solving a space structure we can use equations 1.1 and 1.2 to determine
the reactions, and then proceed to resolve at the joints in the structure. Unless
the problem is very simple it is probably best to proceed in a systematic manner.
2.10 Tension coefficients

Basically this method consists of resolving at the joints in a structure and forming
a number of simultaneous equations, which are then solved to find the unknown
forces.
Figure 2.14
Consider the vector force P (figure 2.14). This can be expressed as

P=Pxi+Pyj+Pzk
where Px, etc. are the resolved parts of Pin the directions of the three axes. These
can be found by multiplying P by its direction cosines.
Px =Pcosa Py =Pcos(3 Pz =P COS"'f
In a structure the values of the direction cosines for a particular member can be
expressed as the resolved length of the member in the x, y or z direction divided
by the length of the member. So that if a plan and elevation for the structure are
given it will be a very simple matter to determine the resolved length.
The term tension coefficient, T, for a member is defined as the force in the
member divided by the length of the member. So that if the tension coefficient
is multiplied by the resolved length of the member in a particular direction, the
resolved part of the force in that direction is obtained. (The term 'tension' is used
since at the start of an analysis all members are assumed to be in tension.)
The elevations and plan of a space structure are shown in figure 2.15. The
forces are required in all the members.
b=12 r=9 j=7
therefore
b + r= 3j
100kN I
I 't
6m
~~~~~~~~~~
l.j
Figure 2.15
and the frame is statically determinate. A further inspection of the structure

shows that members are correctly placed such that no part is a mechanism.
It is probably best to start the resolution at joint H. There are only three
unknowns at this joint and the tension coefficients for three members can be
found straight away.
Joint H
x 3THE + 3THF + 6THn =0
Y 3THE = 3THF
Z 4THE + 4THF + 4THn + 100 = 0
therefore
Joint E
X 3THE = 3TED
Y 3THE + 3TED + 6TEF =0
Z 4THE = 6TEB
therefore
50
TED= -25 TEF = 25 TEB =- -
3
Joint F
X 3THF = 3TFD
y 3THF + 3TFD + 6TFB + 6TEF = 0
z 4THF = 6TFc + 6TFB
therefore
50
TFD = -25 TFc = - -
3
Joint D
x 3T08 + 3TEn + 3T0 c + 3TFD + 6THn = 2TnA
y 3T08 + 3TED = 3Tnc + 3TFn
z 6TnA + 6Tnc + 6Tns = 4TnH
therefore
20
TnA = 10 Tns =Toe=-
6
The values of the tension coefficients can now be multiplied by the lengths of
the members to obtain the forces in kN.
Member HD HF HE ED DF FE FC EB AD BD CD BF
length 7·2 5·84 5·84 4·23 4·23 6 6 6 6·31 7·35 7-35 8·97
T.C. 25 -25 -25 -25 -25 25 -16·67 -16·67 10 3-33 3-33 0
force 180 -145 -145 -100 -100 150 -100 -100 63-2 24·5 24·5 0
2.11 Rigid jointed structures

The majority of structures of this type are highly redundant and are dealt with in
later chapters. In certain cases it is possible to make assumptions that will enable
an approximate solution to be obtained by statics. It is useful at this stage to have
some idea of the problems involved in solving rigid jointed structures. Three
examples are shown in figure 2.16.
The majority of engineering structures in everyday use have their members
rigidly connected at the ends. In steelwork joints are made by welding, rivetting or
bolting. There will probably be a small amount of slip at the joint in the latter
two cases, but usually the joint can be assumed to be rigid. In reinforced concrete,
where there is a joint between a column and a beam, part of the reinforcement in
the column is bent round into the beam so that a continuous joint is made.
n I11
(a) (b)
Figure 2.16
Since the joints are rigid the members will develop both bending moment and
shear force in addition to axial load. Thus for each member in the structure there
will be three unknowns, in addition to the support reactions being unknown. It is
possible to form three equations at each joint relating the unknown quantities by
resolving in two directions and taking moments. If
3b + r = 3j the structure is statically determinate (2.6)
3b + r < 3j the structure is unstable
3b + r > 3j the structure is statically indeterminate
Consider the structures in figure 2.16.
(a) b = 3 j = 4 r=6 3b + r = 15 3j = 12
therefore the structure is three times redundant.
(b)b=10 j=9 r=9 3b+r=39 3j=27
therefore the structure is 12 times redundant.
(c) b = 9 j = 6 r=3 3b + r = 30 3j = 18
therefore the structure is 12 times redundant.
nn
c._---41
(d)
(a) (b) (c)
Figure 2.17
If we have a structure with a cantilever portion (figure 2.17a) the end A of the
cantilever should be considered as a joint and the cantilever as a member
b=4 j =5 r =4 3b + r = 16 3j = 15
therefore the structure is singly redundant.
Alternatively the cantilever portion could be completely ignored.

A structure that has a mixture of pinned and rigid joints can probably best be
treated by first stiffening up all the pinned joints in the structure. The degree of
redundancy can then be found. We can then subtract the number of releases or
equations of condition that result from turning the rigid joints back into pinned
ones.
The portal-type frame (figure 2.17b) is the same as that in figure 2.16a except
for the pin joint at B. The original frame was three times redundant. By inserting
a pin at B we provide one equation of condition, that is, the moment at B is zero.
Thus the degree of redundancy is reduced to two.
For the frame in figure 2.17c the joint at C is first stiffened-there is no need
to alter the pinned support since the number of reactions can be calculated. The
stiffened frame is five times redundant. Re-introducing the pin at C will not pro-
vide three releases, that is, one for each member meeting at C. The effect of the
pin is shown in the subsidiary sketch (figure 2.17d). From this it is quite clear
that there are only two releases, thus the structure is three times redundant.
Figure 2.18
To produce a complete analysis of a statically determinate structure with some

rigid joints, we shall finally require the values of axial force, shear force and
bending moment for every member in the structure. The sign convention that was
originally adopted for bending moment, that is, sagging bending-moment positive,
was quite satisfactory for beam analysis, but this will no longer suffice when
dealing with frameworks consisting of beams and columns. The usual convention
to be adopted is that if part of a framework is viewed from the inside, a positive
bending moment will produce tension in the near edges of the member. This
requires some more explanation and will be discussed in conjunction with the two-
bay portal-type structure in figure 2.18.
An assumed bending-moment diagram has been drawn using the line drawing
of the frame as a base line. Consider ABED-if the bending-moment diagram lies
on the inside of this frame it will be positive and will produce tension on the inside
of the members.
For member DE the bending moment is positive over the centre portion and
negative towards the ends. Thus considering the lower edge it will be in tension
over the centre and in compression towards the ends. For member BE the bending
moment is negative at end E of the member but changes sign along the member and
is positive at B. So that considering ABED the inside edge of BE is in compression
at E and in tension at B.
Now let us consider the member BE from the point of view of BEFG. The
bending moment is positive at E and negative at B and is in tension at E and com-
pression at B. This means that the sign of the bending-moment diagram is depen-
dent upon the side of the member from which it is viewed, but the convention is
consistent as it produces tension, say, on one edge of the member and compression
on the other edge. We can give numerical values to points on a bending-moment
diagram but must be careful about the sign.
The convention we shall adopt is such that if we view part of a framework from
r1-
the inside, when the near edge of a member is in tension then the bending-moment
diagram will be drawn on the inside edge of the member, and can be considered as
positive.
20kN/m E
fY21~ P- ~ 21~
30 ~., I
- - \LkN
A B c
t 82LkN 30k
- sm --sm~
(a } (b l (C }
Figure 2.19
An analysis of the framework and a sketch of the bending-moment diagram

are required for the two-bay structure (figure 2.19a).
A check is made first to see if the framework is statically determinate. It is
necessary to make joints E and F rigid.
b =5 r =5 j =6 3b + r =20 3j = 18
The framework is twice redundant and two releases will be required. These are
provided by the pins atE and F.
L MF gives He = 0. L ME gives
5Vc = 60 x 2!
Vc = 30kN
Next it is worthwhile splitting the frame into two parts at joint E (figure 2 .19b ).
The forces required for equilibrium are entered on each part. Thus the left-hand
portion requires a vertical load of 30 kN and a horizontal load of 10 kN at E.
Resolve horizontally for the left-hand portion
Hs = (4 x 6)- 10 = 14 kN
therefore
VA= 47·6 kN
Resolve vertically for the left-hand portion

VB= (20 X 5) + 30- 47·6 = 82·4 kN
A check can be obtained by taking moments about A.

The axial and shear forces and the bending moments can now be found for each
member.
Member AD
Axial - 4 7·6 kN; shear force zero at A, 24 kN at D-linear
MoA = -72 kNm, MAD = 0-parabolic
Member DE
Axial-24 kN; shear force -47·6 kN at D, 52·4 kN atE-linear
20 X 5 2 4 X 6 2
MoE= 72 kNm, MED = (47·6 x 5)- - 2- - -2- = -84 kNm-parabolic
Member BE
Axial -82·4 kN; shear force 14 kN-constant
MBE = 0, MEB = -(14 x 6) = -84 kNm-linear (viewed from ADEB)
Member EF
Axial-10 kN; shear force 30 kN from E toG, -30 kN from G to F
MEF = MFE = 0; Mmax at centre= 75 kNm
Member CF
Axial -30 kN; shear force= bending moment= 0
It should be noted that a check can be made at all the joints by summing the
various components of forces and moments; the resultant should be zero in all
cases. This is shown for joint E in figure 2.20.
The bending-moment diagram (figure 2.19c) can now be drawn and salient
values attached. The maximum or minimum moment should be found for DE; this
occurs when the shear force is zero, that is at 2·38 m from A. The bending moment
at this point is
-72 + (47·6 2·38)- ( 20 2·38-2 ) = -16 kNm

X X -
2
An example of a structure that is statically indeterminate but that can be solved

by making an assumption is shown in figure 2.2la. The beams are very much more
Figure 2.20
rigid than the columns. Bending moments are required at the ends of all the
members.
If the beams are assumed to be infinitely stiff, they will not deform under the
action of bending moments, hence there will not be any joint rotation when the
load is applied. If in addition the effect of axial forces is neglected the frame will
deform in the manne shown in figure 2.21b. The columns in a particular storey
must deform in an identical fashion, hence they must be identically loaded. If we
look at b we can see that the point of contraflexure or zero bending moment must
occur at the mid-point of each column; we have in fact established an equation of
condition based on the approximations stated above. The column ACE, shown in
c, is broken into two parts at C with all the horizontal forces for equilibrium
entered on the diagram.
For column CE, taking moments about the mid-point H
MEc = 30 xI!= 45 kNm =McE
For column CA, taking moments about the mid-point G
MeA= 70 x 2 = 140 kNm =MAc
It is now possible to draw the complete bending-moment diagram. The moments
in the right-hand column are the same as the corresponding ones in the left-hand
column. As the sum of the bending moments at joint C is zero, Men = 185 kNm.
MEc
-
60kN 1- 6m ·j
t
M-i~~
-
E F
80kN
3m
"11. ~'1:.
--30
45
c D eo- .. h 40
t
4m
30--
G
18«;
A
~'
8
~ '
! '1c
(a) (b) (c I (d)
Figure 2.21
The same convention has been used to draw the bending-moment sketch
(figure 2.2ld). Bending moments are positive if they produce tension in the
inside edge of a member and positive bending moments have been drawn on the
inside of the frame.
2.12 Influence lines for statically determinate trusses

The virtual-work approach is the simplest method for sketching the influence line
for the axial force in a particular member of a truss. Before discussing this in
detail it is advisable to get some idea of the shape of these influence lines.
(b)
(c l
Figure 2.22
The influence lines for the forces in various members are required when a unit
load crosses from A to B (figure 2.22a). All the members of the pinjointed truss
have the same length /.
Chord member CD
The truss is sectioned as shown in the diagram. For a unit load to the left ofF and
distant x from A
LMF
..j3
F CD X 2 I= vB X 31
Now VB =x/51, therefore
2..j3 X
F CD = - - - compressive
5 l
For the unit load to the right ofF
LMF
y3
Fenx-l=VAx2l
2
Now VA= (51- x)/51, therefore
4 (51-x)
Fen= y3 Sl
compressive
Both of the expressions for Fen are linear and give a maximum value of
4y3/5 when the unit load is at F (figure 2.22b). The influence line for other
chord members can be found in a similar manner.
Inclined member CF
Making use of the same section, the vertical component of the force in CF will be
equal to the shear force in panel CDEF. The influence line for shear in a panel has
already been discussed in section 1.22. To obtain the influence line for the axial
force in CF, the influence line for shear force will have to be multiplied by 2/y3
(figure 2.22c). It will be noted that the member is in compression while the unit
load moves from A to E and reaches a maximum value of 2/Sy3 at E. From E to
F the force changes from compression to tension and has a maximum tensile value
of 6/5y3. At F. Thereafter the axial force decreases linearly, remaining in tension.
Ar~
e___ _
I
_t__ - -------
(b)
(c)
Figure 2.23
If the influence line for the axial force in CE were required, it would have
exactly the same shape and numerical values as that for CF except that all signs
would be reversed.
If a vertical member exists in a truss, a suitable section can usually be chosen.
The influence line would have the same shape and numerical value as the influence
line for shear force in an adjacent panel.
The virtual-work method will now be applied to the same problem.
The member CD is removed and equal and opposite forces FCD are applied at
C and D. These two points are given a virtual displacement L1 and approach one
another. Now both ACF and BDF are rigid bodies connected together at the pin
F. Thus the base line AEFB would deform as shown in figure 2.23b. The unit
load being displaced by 8.
Using virtual work
8
FCD X L1 + 1 X 8 =0 FCD =- J
The negative sign indicates that CD is in compression. The force is proportional
to the vertical displacement of the unit load, so that the displaced shape of AEFB
represents the shape of the influence line. It will only be necessary to find the
ratio 8 I L1 to fix the scale of the influence line.
Let 8 be the relative rotation of AF and BF. When the unit load is to the right
ofF
..:1= ..,f3 18 8=AFx8(51-x) ~=_i_(51-x)

2 51 L1 ..,f3 51
Note that the term (51- x)l5l is the value of VA. For the load to the left of
F, the value of 8 becomes FB x 8 (x/51) and 8ILJ becomes
2V3 X
5
For member CF we can proceed in a similar manner. With the member removed
ACE and DBF will remain rigid bodies connected together by two parallel links
CD and EF. When a virtual displacement is applied in the direction of CF the line
AEFB will displace to the shape shown in figure 2.23c. Once again the ratio of
8ILJ can be found in terms of x and the dimensions of the members.
It is suggested that the method of virtual displacements should be used to
sketch the influence lines. It is often easier to find a spot value for a particular
point on the influence line, rather-than trying to fix a general scale factor. For
example for member CD it would be convenient to place the load at F and hence
find the force in CD for this particular position. For member CF, place the load
atE, the value of FcF will be 2VBIV3 where VB has a value of 115. The maximum
positive value can be obtained immediately by remembering that the parts of the
influence line AE and BF are parallel.
As a fmal example consider a bridge formed of two Pratt trusses. The influence
lines for members CD, FG and EF are required (figure 2.24a).
C E
1 - . - - - - - - - - - 6 at 9m ---------~
(a)
(b)
(c)
(d)
~ 4 ___
-- -:_.,__
(e)
(f)
Figure 2.24
A four-wheeled vehicle of wheelbase 4·5 m crosses the bridge from A to B.
The front-axle load is 20 kN and the rear-axle load is 40 kN. Find the maximum
tension and compression in member EF.
The required influence lines can be sketched straight away and are shown in
figures 2.24b, c and d for members CD, FG and EF respectively. Numerical values
have been calculated and entered on the diagram at salient points.
Next consider the vehicle. As there are two Pratt trusses, we can consider one-
half of each axle load to be carried by each truss. Thus d will represent the force
in EF due to the front-axle load if the ordinates are multiplied by a factor of 10
kN. The rear-axle load can be represented to the same scale if an influence line is
drawn for a 'two-unit' load crossing the truss. The total force is required in EF
and this coutd be sketched as a function of the position of either the front or the
rear axle. Choosing the front axle, the influence line for the two-unit load is drawn
(figure 2.24e). This is similar to that of d, but has ordinates twice as large, and is
displaced to the right by 4·5 m. The graphs illustrated in figures 2.24d and e can
next be summed to give the total force in EF (figure 2.24t).
The maximum compressive force in EF due to the vehicle is 25/24 x 10 =
10·4 kN.
The maximum tensile force in EF due to the vehicle is 85/48 x 10 = 17·7 kN.
Problems
A 8
p 2.1
2.1 For the truss in P 2.1 all vertical and horizontal members are 5 m long. Verti-
cal loads of 5 kN act at C and D. Find the forces in all the members, and check
your results for members AC, AD and CD by the method of sections.
2.2 Find the forces in AB and DE and all the members that meet at C for the K-
type truss shown in P 2.2.
A 8
- - - - - 6 at 3m - - - - - - - 1
p 2.2
p 2.3
2.3 For a statically determinate framework, derive relations between the number
of bars, reactions and joints for the following cases.
(a) two-dimensional pin-jointed;
(b) two-dimensional rigid-jointed;
(c) three-dimensional pin-jointed.
For a and b sketch examples that are not completely statically determinate but
for which the relation holds.
The plane framework P 2.3 is pin-jointed and basically consists of an inner
square of side a/v'2 connected to an outer square of side 2a. The diagonal bar of
the inner square is horizontal. Find the force in AB when the horizon tal load P is
applied.
2.4 A plane pin-jointed structure consists of 9 nodes, 4 of which are pinned to a

rigid foundation. Sketch a suitable framework such that the complete structure is
statically determinate. Repeat for the case of a three-dimensional pin-jointed
structure.
.-------,.G
p
P2.4
The plan and elevation of a derrick-type crane are shown in P 2.4. All members
are pin-jointed. A, B, C, and D, E, F, lie at the corners of equilateral triangles of
10 m and 5 m side lengths respectively. DEF lies in a plane parallel to the ground
and AD = BE= FC. Find the forces in all the members when the vertical load Pis
applied at G.
2.5 The framework shown in P 2.5 has rigid joints.
It /p
--
-t A
p
C)
2
{ _.1..
! 0 F
~
2{
p 2.5
(a) Show that the framework is statically determinate.

(b) Find the reactions at D, E and F when pi = P.
(c) Sketch the bending-moment diagram stating the sign convention adopted.
2.6 In the structure shown in P 2.6 the members are vertical, horizontal or
inclined at 30° to the horizontal, and they cross without being connected. Show
that the frame is statically determinate and find the force in the member AD.
Hence find the forces in the other members.
t40kN
p 2.6
2.7 For the truss shown in P 2.1 draw the influence lines for the forces in mem-
bers AB, AC, AD, and CD. A 7 m long, uniformly distributed load of 5 kN/m
crosses the truss. Find the maximum force in member CD.
3 ELEMENTARY ELASTICITY, PLASTICITY
AND BENDING OF BEAMS
3.1 Sign conventions

It is necessary to adopt a sign convention when using a rectangular or cartesian set
of axes. A right-handed system will be adopted here. This means that if the axes
are taken in the sequence Ox, Oy, Oz, Ox etc. and the directions of any two of
the axes are chosen, then the direction of the third axis will be given by the right-
handed·screw rule.
y X
Y -ve +ve
( ()
X z X
(a) (b) (c)
Figure 3.1
Two examples of right-handed sets are shown in figures 3.1 a and b. Suppose
that the directions of x and z have been chosen in a. The sequence would be z, x,
y, so that Oy must point vertically upwards.
Next consider the section of a body in figure 3.1c. The exposed face will be
considered positive if the outward normal to the face acts in a positive direction
when referred to the coordinate system. This means that the right-hand face in
figure 3.1c is positive while the left-hand face is negative.
A force or component of a force or moment will be considered as positive if it
acts on a positive face in a positive direction, or on a negative face in a negative
direction.
This will lead to a tensile force as positive (figure 3.2a) and a compressive force
as negative (3.2b ). Positive shear force is shown at c and a positive bending moment
at d. These conventions are identical with those adopted in chapter 1, where for
example a sagging bending moment was considered as positive.
73
--
+
--
-t 1- + +I++
y
~I- +I_!
(a) (c)
-(1-
X
z +--1- +I~ +IJ+

(b) (d)
Figure 3.2
3.2 Stress
If a body that is under the action of external forces is imagined to be sectioned
and equilibrium maintained, then there must be internal forces acting on the cross-
section (figure 3.3).
dF
I Figure 3.3
On the small element of cross-section dA there is a force dF and a moment

dM. There is of course no reason why the force should act normal to the surface.
Total force acting F = ~ dF
The moment of the force about 0 could be written as
~dM+:L rxdF
If a set of axes is now introduced such that the x direction is normal to the
sectioned face, as dA tends to zero the value of dF/dA is termed the stress. This
stress is in fact a tensor. A vector has three properties, direction, magnitude and
sense. This would not be sufficient in the case of stress as it is also necessary to
know the direction of the normal relative to the element face.
It would be perfectly possible to resolve the small force into three components
dFx, dFy and dFz· Since dFx is in the direction of the normal to the surface it is
ELEMENTARY ELASTICITY, PLASTICITY AND BENDING OF BEAMS 75
called the normal or direct force. The other two forces act at right angles to the
normal and lie in the plane of the section, these are both referred to as shear
forces.
We now write
dFx = dFy dFz
a =- T
xy dA
T =-
dA (3.1)
xx dA xz
where a is a normal stress and T is a shear stress. The first subscript denotes the
direction of the normal to the face and the second subscript the direction in which
the stress is acting. It is usual to express a stress in terms ofN/mm 2 .
F 8
(a) (b)
Figure 3.4
A small parallelepiped of sides dx, dy and dz'is shown in figure 3.4a, its
centroid is at x, y, z, and it has been cut from the body such thqt the face ABCD
lies in the plane of the original section. The block is in equilibrium. All possible
stresses that can act have been shown on the diagram acting in a positive direction.
It will be seen that there are nine different stress components-three normal and
six shear. If we look directly at the front face EADH, we see the stresses indicated
in figure 3.4b.
shear forces on face ABFE and DCGH are Tyx dx dz

the moment about the centre of EADH is Tyx dx dz dy
shear forces on faces ABCD and EFGH are Txy dy dz
the moment about the centre of EADH is Txy dy dz dx
Now as the parallelepiped is in equilibrium there cannot be any rotation,

hence the sum of the moments must be zero. Thus
Tyx = Txy (3.2)

similarly Txz = Tzx and Tyz = Tzy
It can now be seen that the original nine different stresses have been reduced
to six. It should also be noted that it is impossible for Txy• say, to exist without
Tyx-a shear stress is always accompanied by a complementary shear stress.
3.3 Strain
c'
,'1
I
I
y I
I
I
I
I
I
I
c I
I
I
: I I
I
1:
I I
I I s'l
,'!1
,,-
dy 11 1 -'
L~---"~t~-:-=-: =~Ol
A
~-----dx------~1
18
X
Figure 3.5
A small rectangular lamina ABCD of sides dx and dy, the coordinates of A being
x,y, is shown in figure 3.5. The lamina is loaded in some manner such that it
displaces to A'B'C 'D', the displacement of A being u in the x direction and v in
they direction. We now want to try to write down expressions for the new co-
ordinates of A', B' and D'. There is certainly no difficulty about A' and it has
u,
coordinates x + y + v.
A convenient way of representing the horizontal displacement of B would be
to say that it was equal to the horizontal displacement of A plus the rate of
change of displacement in the x direction times the length, that is
ou
u +--dx
ox
The partial derivative is used as we only require the rate at which u has changed
withx.
Similarly the vertical displacement of B could be written
ov
v+--dx
ox
Thus the coordinates of B' are
ou
x + dx + u +-- dx, y + v +-- dx
ov
ox ox
In a similar manner it is easily shown that the coordinates of D' are
au av
x +u +-dy, y +dy +v +-dy
ay ay
The horizontal component of length of A'B' is
au
dx +-dx
ax
and the vertical component
av
-dx
ax
So that the length that was originally dx has become
2
ax dx ) + (av
{(dx + au ax dx
)2}1/2
= dx {
1+2
au
ax +
(au)2
ax + a-;
(av)2}1n
Now aujax is in any case a very small term so that it will be in order to neglect
terms such as (aujax) 2 and the length becomes dx(l + aujax). The increase in
length of AB is
dx ( 1 +au)
- - dx = -au dx
ax ax
If we define the normal or direct strain as
increase in length
e=
original length
e
XX
=(au_
ax
dx)/dx =au
ax
In a similar manner we can show that Eyy = avjay. The analysis could be extended
to a three-dimensional case and
au av aw
Exx =ax €
YY
=-
ay Ezz = az (3.3)
It should be noted that the displacements u, v and ware in the x, y and z

directions respectively.
The angle DAB, after the deformation has taken place, is no longer a right
angle. The shear strain 'Yxy is defined as the change in value of the right angle.
av av
-dx -
ax ax av
tan a= <=::::-
au 1 + Exx ax
dx + - dx
ax
similarly
au
tan~=
ay
The angles are very small hence tan a = a and tan ~ = ~.
a+~='Y
av au
=ax- +ay-
xy
In a three-dimensional case it can be shown that

av au aw au aw av
'Yxy = ax + ay 'Yxz = ax + az 'Yyz = ay + az (3.4)
The only assumptions that have been made in deriving equations 3.3 and 3.4 is
that all the displacements are small.
3.4 Relations between stress and strain

The fundamental experimental law on which linear elastic theory is based was dis-
covered by Robert Hooke in 1678. When subjected to uniaxial loading it was found
that for certain materials the extension was proportional to the applied load. If a
rod of linear elastic material with a uniform cross-section is loaded axially, the
stress is uniform over the cross-section and would be equal to the load divided by
the area of the cross-section. The axial strain would be the extension divided by
the original length, so that Hooke's law may be written
A constant of proportionality, E, which will vary from one elastic material to

another can be introduced-this is known as Young's modulus, or the modulus of
elasticity.
(3.5)
If careful measurements are made on the material under test it will be found
that the dimensions of the cross-section also change as the load is applied. The
strain in both they and z directions is proportional to the strain in the x direction
but opposite in sign. The constant of proportionality vis referred to as Poisson's
ratio.
Eyy = ezz = -VExx (3.6)
Typical values of E and v for three different materials are given below.
Mild steel Brass Duralumin
E(kN/mm 2) 210 100 70
v 0·3 0·35 0·33
If Pis the load applied to the rod and A the cross-sectional area axx = P/A.
Suppose that the rod is oflength 1 and extends by an amount e, then Exx = ejl.
Substituting into equation 3.5
PI
e=--
AE
It has been assumed that the direct stress is uniformly distributed over the
cross-section. If the load is applied to the rod by means of a testing machine the
stress distribution at the ends will certainly not be uniform. A section that was
plane in the unloaded state will no longer be plane when the load is applied. How-
ever at a small distance from the ends it will be found that the stress distribution
is sensibly uniform and it may be assumed that plane sections remain plane. This
is a particular application of what is known as St Vernant's principle.
E
QL-------------------------------
E
Figure 3.6
Not all materials exhibit this linear stress-strain relationship and those that do
will only behave in this manner for a limited value of strain. Mild steel for
example has a curve similar to that shown in figure 3.6. The portion OA
is linear; at A there is a yield point with a sudden drop in stress to a lower yield
point B. This is followed by what is termed aplastic range;BD, where there is a
large increase in strain for a small increase in stress. Thereafter the stress increases
more rapidly with work hardening until the ultimate stress is reached at E. It is of
interest to note that the strain at D is about twelve times that at A.
l
r1yy
Figure 3.7
A rectangular parallelepiped is shown in figure 3.7 with stresses axx• ayy• and
Gzz applied. The material is linearly elastic, homogeneous and isotropic, that is, it
has the same properties in all directions. Hence we can use the method of super-
position: apply each stress separately, determine its effect, and then add all the
effects together as follows
axx applied alone
exx =Gxx
E- fyy = fzz = -E
VGxx
ayy applied alone
€ = ayy = fzz =-
va
yy E fxx ;y
Gzz applied alone
ezz =Gzz
E- fxx =eyy = -E
VGzz
If all stresses are now applied simultaneously
fxx = ~ [axx - v(ayy + Gzz)] (3.7)
two similar expressions can be derived for Eyy and Ezz·

These expressions are the generalized form of Hooke's law. A very common
error is to assume that the uniaxial case of equation 3.5 holds on all occasions-in
other words the Poisson-ratio effect is ignored. A further relationship exists
between shear stress rand shear strain 'Y·
(3.8)
where G is another elastic constant called the modulus of rigidity or sometimes the
shear modulus. These linear relationships again only apply to certain materials
and will hold only for a limited value of shear strain.
3.5 ,Composite rods with axial tension or compression

The type of problem we shall consider is that of two rods of the same length, but
of different materials which have been rigidly fixed together at both ends. A load
Pis applied such that there is no bending in either rod. The question arises as to
what proportion of the load is carried by each rod.
Figure 3.8 shows the two rods of materials 1 and 2. As the rods are rigidly con-
nected at the ends, they must both extend by the same amount when the load is
applied or alternatively as they are of the same length the normal strains must be
the same.
,C A p
~r------:,:------~ ~------::,:---------11---r--::-+-'___,I-
~ 2 ~ ~ 2
~--------~--------~~8 p
1-~
Figure 3.8 Figure 3.9
Let P 1 and P 2 be the loads carried by each rod of areas A 1 and A 2 and Young's
moduli£ 1 and£2.
thus
also
From these
Note that
p
a 1 = _A_1_+_A_2(-E-2-/E-1-) (3.9)
It is worthwhile examining the expression for a 1 in a little more detail. The stress
is given by the total load applied divided by the area of material! together with a
modified area of material 2. This is the same as converting material 2 to an equiv-
alent amount of material 1. This process is referred to as transforming the section:
that is, converting it all to one material.
Now consider the effect of a temperature variation on the composite rod. Again
the restriction is made that axial effects only are considered. This could affect the
configuration of the material forming the rods as a symmetrical arrangement would
have to be used to avoid secondary bending effects. One possible arrangement
would be to have a circular rod of material 1 inside a tube of material 2. However
for the analysis we shall consider the two rods side by side-figure 3.9 with the
fixing removed at one end.
If a temperature rise Tis applied and the rods have coefficients of linear expan-
sion o: 1 and o: 2 where o: 1 > o: 2 , rod 1 expands to A while rod 2 only expands to B.
In fact the rods must fmish with the same length as they are really fixed at both
ends. The only way of achieving this is to introduce two forces each having a
magnitude P, to bring the internal force system to equilibrium. A compressive force
Pacts on material I and a tensile force P on material2 such that the fmal position
at both ends is at C. Both rods have an original length 1.
Due to the temperature change, T, the extension of rod I is kx 1 T and due to

the force P the extension is -PI/A 1E 1 . Note that the original length I is used for
this second extension and not the length /(I + a 1 T)-the reason for this being that
changes in length are very small compared with the original length.
PI
Total extension of I = la 1 T- - -
AIEl
PI
Total extension of 2 = la 2 T + __
A2E2
These extensions are equal-hence the strains are equal, and
therefore
(3.IO)
As an example a steel tie rod 20 mm in diameter is placed concentrically in a

brass tube of 3 mm wall thickness and 50 mm mean diameter. Nuts and washers
are fitted to the tie rod such that the ends of the tube are closed by the washers.
The nuts are first tightened to give a compressive stress of 30 N/mm 2 in the tube.
A tensile load of 20 kN is then applied to the tie rod. Find the resulting stresses
in both the tie rod and the tube (a) with no change in temperature, (b) when the
temperature rises by 60 o C. Es = 210 kN/mm 2 , Es = IOO kN/mm 2 , as=
12 X 10-6 tC, as= 20 X IQ- 6 tC.
Area of steel= 314 mm 2 , area of brass= 471 mm 2 • Initial load in brass=
initial load in steel therefore
30 X 471
stress in steel= = 45 N/mm 2
314
Now consider the effect of a 20 kN load applied to the unloaded system

assuming that the tube and rod are fixed together at their ends. They would
therefore extend by the same amount, the 20 kN being shared between them.
20 X 10 3
a = = 37·6 N/mm 2
s 314+471xWo
as= 37·6 x W8 = 18·8 N/mm 2

therefore
resultant stress in steel= 45 + 37·6 = 82·6 N/mm 2 -tensile
resultant stress in brass= 30 - 18·8 = 11·2 N/mm 2-compressive
Considering the temperature effect only, from equation 3 .I 0
314 X 471 X 210 X 100(20- 12) X 10-6 X 60

P= (210 X 314) + (100 X 471) = l3·I kN
The stress in the steel is 41·7 N/mm 2 tensile, and in the brass 27·8 N/mm 2
compressive. Therefore final stresses are
steel 82·6 + 41·7 = 124·3 N/mm 2 -tensile
brass 11·2 + 27·8 = 39·0 N/mm 2 -compressive
3.6 Principal stresses in two dimensions

The rectangular element in figure 3.10a is of unit thickness and is acted on by the
stress system shown. This is referred to as a plane stress system, since azz = 7 xz =
7 yz = 0. The triangular element in figure 3. lOb has been generated by sectioning
(a)
Figure 3.10
the rectangular element perpendicular to the xy plane. We now want to find the
stresses that must act on plane AC to maintain the triangular element in equilib-
rium. The obvious way of proceeding is to resolve forces along and normal to AC.
A fundamental point arises here in that it is essential to resolve forces and not
stresses.
Resolving forces normal to AC
a X AC = axx X BC cos e + a yy X AB sin e + 7 xy X BC sin e + 7 yx X AB cos e
In section 3.2 we showed that 7xy = 7yx therefore
a= axx COS 2 8 + ayy sin 2 8 + 7xy sin 8 COS 8 + 7xy sin 8 COS 8
or
a= !(axx + ayy) + !Caxx- ayy) cos 28 + Txy sin 28 (3.11)
If we resolve in the direction AC we shall find that
T = !(axx - ayy) sin 28- Txy cos 28 (3.12)
If r is zero then
2Txy
tan 28 = --"--- (3.13)
axx -ayy
therefore
2Txy
8 = ! tan -l ---=---
axx -ayy
or
8 =! tan-1
1t
2Txy +-
axx- ayy 2
There are therefore two planes inclined at 90° to each other on which the value
of the shear stress is zero. These are known as principal planes-the stress acting
is normal to these planes and is called a principal stress.
We can find the planes on which the direct stress is a maximum by differen-
tiating equation 3.11 and setting the result equal to zero.
-(axx - ayy) sin 28 + 2Txy cos 28 = 0

or
2Txy
tan 28 = --"-----
axx - ayy
This result was also obtained for the case when T = 0. So the principal stresses are
also the maximum and minimum direct stresses that can occur. To find the values
of these stresses it is only necessary to substitute the value of tan 28 into
equation 3.11. The arithmetic is somewhat tedious and an alternative method
exists.
In figure 3.10b assume that a is a principal stress, that is, r will be zero.
Resolve normal and along the direction of BC.
a X AC cos 8 = axxBC + TxyAB

or
a-axx
tan8 = - - -
Txy
a X AC sin 8 = ayyAB + TxyBC

or
a-a
COt 6= yy
Txy
Eliminating 6
(a- axx)(a- ayy)- T~y = 0
This is a quadratic in a which solves to give
(3.14)
these are the two principal stresses. Note that

(3.15)
We should also determine the maximum value of the shear stress and the plane
on which it acts. The maximum value will occur when dr/d6 = 0 in equation 3.12
or
tan 26 = _,ay'-'y'---_a..:.x::.:x (3.16)
2Txy
The plane on which the maximum shear stress acts is inclined at 45° to the
planes of principal stress (compare equation 3.16 with equation 3.13).
If the value of tan 26 from equation 3.16 is substituted into equation 3.12 the
maximum value of the shear stress can be found.
- "2:1 [( axx-
Tmax- ayy )2 + 4Txy ] 1/2 (3.17)
From the values of a 1 and a 2 given by equation 3.14 it can be seen at once
that
(3.18)
This is a most important relationship-it is very simple to remember that the

maximum shear stress is equal to one half of the difference between the principal
stresses.
3. 7 The Mohr circle diagram for stress

The theory that has been derived so far leads to fairly lengthy expressions for
principal stresses. An alternative method of presentation is by what is termed the
Mohr circle diagram; once the significance of this is appreciated it will be seen to
be a much easier method to use.
First consider the problem illustrated in figure 3.11a. Given a 1 and a 2 what
are the values of a and r? It can very quickly be shown by resolution of forces that
a=-!(a 1 +a2)+!(a 1 -a2)cos26 (3.19)
T = !(a 1 - a2 ) sin 26 (3.20)
If a set of orthogonal axes are taken with a horizontal and r vertical. Two
points (a., 0), (a 2 , 0) are entered on the diagram represented by OE and OF. A
circle with centreD is then drawn with FE as diameter. The length OD is given by
(a 1 + a 2 )/2 and the radius r of the circle by (a 1 - a 2 )/2. It is now possible to
find the coordinates a, r of point P where DP is inclined at 20 to the horizontal.
The horizontal coordinate is OD + r cos 20 and the vertical r sin 20. So that
and r = -!-(a 1 - a 2 ) sin 20. These expressions are identical to equations 3.19 and
3.20, thus indicating that the construction is correct.
values of a.,
The circie can also be constructed given values of axx• ayy and Txy· Hence
a 2 and 0 can be found.
(a)
(b)
Figure 3.11
The sign convention adopted for direct stresses will apply when constructing a
Mohr circle diagram, that is, tensile stress is positive. Our previous shear-stress
convention will not be satisfactory. However, let us now consider shear stresses
acting on a pair of opposite faces of the rectangular element in figure 3.1 Oa. If the
effect of the shear stresses is to form a positive couple along the z axis, then by
our sign convention the shear stresses will be positive. This would mean that axx
in the diagram is accompanied by a positive shear and ayy by a negative shear.
Effectively we are saying that when viewed from the front, an anti-clockwise
shearing effect is considered positive.
As an example, let us determine the principal stresses and their directions when
axx =50 N/mm 2, ayy = -30 N/mm 2, Txy = -30 N/mm 2 • The stresses are shown
on the rectangular element figure 3.12a. Using the theoretical result
al> a2 =!(50- 30) ± [(50+ 30) 2 + (4 X 900)] 112 = 10 ±50
therefore
30
~
30
t \
30+ \
I~
le
50~ I ~so F
\
I
I
I +30
-I
+ 30 I
30
(a)
(b)
Figure 3.12
The maximum shear stress is
60+40
- - - = 5 0 N/mm 2
2
2 X 30
tan 28 =( 50 + 30) = i
therefore
8 = 18!0
To construct the Mohr circle diagram, the two axes are first drawn (figure
3.12b). PointsP 1 andP2 with coordinates (50, -30) and (-30, 30) can now be
found. These are joined by a line which cuts the abscissa at D. The circle with
radius DP 1 or DP 2 can now be constructed. By measurement OE = 60, OF= -40
and 2(} = 37°. The maximum shear stress is of course equal to the radius of the
circle. Note that starting ftom P 1 on the circle and proceeding through an angle
of 37° anti-clockwise we come to a principal stress of 60 N/mm 2 • So that starting
from the plane on which the 50 N/mm 2 stress is acting figure 3.12a and proceeding
anti-clockwise through an angle of 18! o we shall come to a plane on which a
principal stress of 60 N/mm 2 is acting.
3.8 Principal strain in two dimensions

Suppose that a line MN was scribed on a piece of material that is strained and dis-
placed such that its new position is M'N' (figure 3.13a). The coordinates of M
beingx, y and those of N x + dx,y + dy. The displacements ofM' are u and v,
y y
d
ML~'
X X
(a) (b)
Figure 3.13
shown greatly exaggerated in the diagram. It is possible by making use of Taylor's
theorem to express the displacements of N' as
au
u +-dx +-dy
au av
v +-dy +-dx
av
ax ay ay ax
Now move the line MN so that it remains parallel to its original position, such
that M and M' are coincident (figure 3.13b). The horizontal and vertical displace-
ment ofN' relative toN are NP and PN' where
au au
NP=-dx+-dy
ax ay
, av av
PN =-dy+-dx
ay ax
The component of displacement along MN' is PN cos e + PN' sin e and if the
original length of MN was d/ the strain is
e=[(:~ dx+:; dy)cose+(:;dy+:: dx)sine]/d/

= Exx COS 2 8 + €yy sin 2 8 + 'Yxy sin 8 COS 8
or
_ 1( 1( 'Yxy .
€- "2" Exx + €yy) + "2" COS 28 +-Sin 28
Exx- €yy) (3.21)
2
The angle dO through which MN' has rotated is given by the normal component
of the displacement of N divided by the length.
PN I cos e - PN sin e
de=--------
dt
= [(avay dy + ~
ax dx) cos e - (au
ax dx + au
ay dy) sine] /dt
dv au
= - COS 2 8 - - sin 2 8 + (eyy- sin 8 COS 8
ax ay Exx)
Now we have already defined shear strain as being the change in value of a
right angle when deformation has taken place. So that it will be necessary to fmd
the equivalent rotation dO 1 of a line that was originally at right angles to AB. This
is easily determined by substituting 0 + rr/2 in place of 0 in the expression for dO
dO 1
av Sill
=ax . 2 0- au
ay
2 .
cos 0- (Eyy- Exx) srn 0 cos 0
The change in the angle is dO 1 - dO
'Y =(Exx - Eyy) sin 20

av + au)
+ (,ax ay (cos 2 0 - sin 2 0)
therefore
~ = !Cexx - Eyy) sin 20 + 'Y;y cos 20 (3.22)
If equations 3.21 and 3.22 are compared with equations 3.11 and 3.12 they will
be found to be identical in form if normal stress is replaced by normal
strain and shear stress by half the shear strain. This means that any method used
for the analysis of principal stress can also be used to analyse principal strains. So
that from equation 3.14 the principal strains can be written down as
(3.23)
3.9 The Mohr circle diagram for strain

The two equations 3.21 and 3.22 fore and r/2 may. be represented on a circle
diagram in a similar manner to the stress circle. e is plotted horizontally and r/2
vertically.
If a Mohr circle diagram has been obtained for stresses it is a straightforward
matter to draw the equivalent strain circle. Once the principal stresses have been
found, the principal strains can be determined by using Hooke's law in a two-
dimensional form. The strain circle can then be drawn.
The strain circle is however of more use when constructed from experimental
readings of strain. A number of different experimental devices exist for measuring
normal strain but there is no method for measuring the shear strain directly. This
means that the circle could not be constructed in the usual manner from two sets
of coordinates.
A very convenient method of measuring strain is by means of an electrical
resistance strain gauge. This relies on the fact that if a wire is extended its elec-
trical resistance increases. For particular materials the ratio of electrical strain to
mechanical strain is a constant, where the electrical strain is the change in resis-
tance divided by the original resistance. It is not the purpose of this book to
describe experimental methods and it will suffice to say that by using resistance
strain gauge rosettes it is possible to determine the strain in three different known
directions.
It is assumed that the three strain values Ea, Eb, Ec have been found and that
the strains are measured at relative angles a: and (3 (figure 3.14a). Equation 3.21
can be modified for principal strains by putting Exx = €1> Eyy = E2 and 'Yxy = 0,
so that
Ea = h€1 + €2) + !(EJ - €2) COS 28
Eb = !(EJ + €2) + !(EJ - €2) COS 2(8 +a:)
Ec =!(EJ +E2)+!(EJ -E2)cos2(8 +o:+(3)
These three equations can be solved for E1, E2 and 8. In rosette strain gauges
the angles a: and (3 between the gauges are equal and generally have values of 45°,
60° or 120°.
Y -+----Ea
'
''
' r-
-----"
1
I
I
I
I
(a) I
y
(b)
Figure 3.14
To construct the Mohr strain circle directly from the three strain readings it is
first necessary to arrange the gauge readings such that the one with the middle
value lies between the gauges with the highest and lowest values; the angles
between the gauges must be less than 90°. This rearrangement will be demon-
strated in a following example. A vertical line YY is drawn as the strain
origin; additional verticals are then drawn at scaled distances equal to Ea, Eb and
Ec (figure 3.14b). A point Pis chosen on the vertical line representing the median
value of strain. Lines inclined at a: and (3 are drawn to cut the verticals representing
Ea and Ec respectively at Q and R. A circle can now be drawn to pass through P, Q
and R. This is in fact the Mohr circle. The circle cuts the vertical representing Eb at
P and P'. We shall show that P', Q and R represent the correct points on the Mohr
circle diagram. These points most certainly have the true values of strain but it will
still be necessary to show that they are correctly orientated. The angle between Ea
and eb is a, so that on the Mohr circle diagram the two gauge points P' and Q
should sub tend an angle of 2a at the centre of the circle. This is so, as the chord
P'Q sub tends an angle at the centre of the circle equal to twice the angle at the
circumference. Similarly it can be shown that the angle subtended by P' and R at
the centre of the circle is 2{3. This means that the gauge points have the correct
orientation and that the constructed circle is indeed the required Mohr circle.
As an example of the construction of a Mohr circle diagram for strain, the
readings taken from a 120° rosette Strain gauge are 75 X lQ-S, 20 X lQ-S and
-50 x 1o-s. The gauge was mounted on a piece of sheet steel and the principal
stresses and their directions are required. The gauge readings (figure 3.15a) are
-50 0 20 75
75
~
z-~0
20
(a)
(b)
Figure 3.15
rearranged as in b. Note that the angles between the gauges are now less than 90°
and the median reading lies between the other two. A suitable scale is chosen and
the Mohr diagram constructed as previously described. From the diagram, figure
3.15c, the principal Strains are found to be 87·5 X lQ-S and -57·5 X lQ-5; the
angle 20 is 34°; so that the direction of the principal strain of 87·5 x 1o-s is at an
angle of 17° measured clockwise to the direction of the strain gauge reading of
75 x 10- 5 .
The general form of Hooke's law for the case of plane stress when principal
stresses are used is
1
€1 =-(at - va2)
E
1
e2 =-(a 2 -va 1)
E
If the two equations are solved for a 1 and a 2 we have expressions for principal
stresses in terms of principal strains
(3.24)
Assuming values of E = 210 kN/mm 2 and v = 0·3 we find in this case that
a 1 = 162 N/mm 2 -tensile a 2 = 72 N/mm 2 -compressive
3 .I 0 Relationships between elastic constants
v t
y
·D·
2
~ -- 02 El
E
to
(a) (b) (c)
Figure 3.16
A particular case of plane stress is shown in figure 3.16a where axx =a, ayy =-a
and Txy = 0. The Mohr circle for stress is drawn in figure 3.16b. Note that a 1 =a,
a2 = -a and T max = a, so that the maximum shear strain is a/G. The strain diagram
is shown in figure 3.16c and from this it can be seen that
'Ymax a
EJ =--=-
2 2G
From Hooke's law
Thus
a a
-=-(I+v)
2G E
or
E
G=--- (3.25)
2(1 + v)
So that if two of the three constants E, G or v are known, the third may be
determined theoretically.
A body that is subjected to hydrostatic pressure has the same compressive stress
or pressure acting on it from all directions. Due to this the volume of the body, V
will decrease by an amount d V. The volumetric strain is defined as d V/V. If the
compressive stress is a and the material behaves in an elastic manner the bulk
modulus K of the material is defined as
-a
K = (dV/V)
Let us consider a small cube of the material having unit edge length. Using
Hooke's law for a three-dimensional case, the strain along the directions of three
adjacent edges is identical and has a value
1
e =-(-a+ 2va)
E
The length of each side which was unity becomes ( 1 + e) and the new volume of
the cube is (1 + e)3. So that
dV (l+e) 3 -1 3a
- = ~ 3e = - - (1 - 2v)
V 1 E
Hence
a E
K =- - - - =- - - - (3.26)
3a 3(1- 2v)
E(l- 2v)
Both G and K are positive quantities so that from equations 3.25 and 3.26 the
following limits for v can be found
-1<v<!
We shall conclude this section with a discussion on the solution of a particular
problem. This may help to fix some of the ideas that have been developed so far
in this chapter.
Consider the case of two thin-walled circular tubes. One is of brass and the
other of steel. Both are the same length. The brass tube is inserted inside the steel
tube and is an exact fit at room temperature. The tubes are closed by rigid end
plates. The wall thicknesses of the tubes are tB and t 8 . These dimensions are small
compared with the mean radius r of the tubes. The vessel is filled with water. How
much more water has to be pumped in to raise the pressure by an amount p?
Two effects will have to be considered: the compression of the water due to
the increase in pressure; the changes in dimensions of the tubes due to the pressure
increase.
If the bulk modulus for water is K
V=1tr 2 l
-p
(dV/V) =K
Thus
This is a decrease in volume.

It is assumed that both materials remain elastic. When the pressure is applied
the internal tube will tend to expand and a radial stress aR will be developed
between the two tubes (figure 3.17a).
(a) (b)
Figure 3.17
Consider the inner brass tube, in particular the equilibrium about a diametral
cut figure 3.17b. For a unit length of tube where a 0 B is the hoop stress.
a08 x 2t8 = 2r(p - aR)
thus
(i)
for the steel tube

raR
aos = - - (ii)
ts
For equilibrium in the longitudinal direction with longitudinal stresses aLB

and aLs
or
aLBtB + aLsts =~ (iii)
So far there are a total of five unknown stresses aR, aoB, aos, aLB and aLs and
there are only three equations relating them. This means that two further equations
have to be found.
The length of each tube and the radius of each tube must extend by the same
amount. Otherwise they would not fit together after the pressure is applied. This
is referred to as compatibility of strain. Working in terms of strains, the longitudinal
strain is
(iv)
Hoop strain
1 1
Eo=- (aos- vsaLs) = - (aos- vsaLs) (v)
Es Es
There are now five equations numbered (i) to (v) which can be solved for the
unknown stresses.
If the length has increased by d/ and the radius by dr
d/ 2 dr
::::::-+--
1 r
Thus dV= (EL + 2E 0 )rtr2 l. Both ELand Eo are known.

The total change in volume can now be found and hence the volume of water
pumped in is this value plus the bulk modulus effect.
3 .11 Strain energy

If a spring is extended it is obvious that a certain amount of external work has
been done. Assume that the process is adiabatic, that is, there is no heat exchange
with the surroundings, and that the force is applied sufficiently slowly so that
there is no change in temperature of the spring and hence no change in internal
energy. If the change in kinetic energy is zero then the external work done must
be equal to the increase in potential energy of the spring. This particular form of
potential energy is usually referred to as strain energy and is denoted by the
symbol U.
In the case of a spring that remains linearly elastic the relationship between the
applied force P and the extension e can be written P = ke where k is the spring
constant. As the load increases from 0 to P0 the extension increases from 0 to e0
(figure 3.18).
dU=Pde = ke de
therefore
(3.27)
Figure 3.18
Exactly the same analysis will apply for the case of a rod of linearly elastic
material under the action of an axial tensile or compressive force.
The spring constant k will be replaced by AE/1 where A is the cross-sectional
area and I the length. Thus
(3.28)
Now consider the case of a rectangular parallelepiped of dimensions dx , dy , dz.

A uniform stress axx is applied, so that the dimension dx will change to dx(l + €xx).
The work done by the force axx dy dz during a small change in the length dx of
d(€xx) dx is
Uxx dy dz dx dexx
or the strain energy stored per unit volume.
Now
Thus
dV 2£ 2
By similar reasoning the strain energy stored per unit volume for a case of pure
shear will be given by
dU riy Txy'Yxy
-=-=---
dV 2G 2
This can be extended to a case of plane stress and indeed to the general case of
a three-dimensional stress system. For this general case
dU
dV = !(axxfxx + Oyyfyy + OzzEzz + Txy'Yxy + Txz'Yxz + Tyz'Yyz)
Now
Thus
dU 1 2 2 2
dV = 2E [axx + Oyy + Ozz- 2v(axx0yy + OxxOzz + OyyOzz)]
1 2 2 2
(3.29)
+ 2G [Txy +Txz +Tyz]
3.12 Pure bending of beams

A straight prismatic beam of homogeneous isotropic material is loaded in such a
way that the bending moment at any cross-section has a constant value of M This
immediately implies that the shear force is zero. Also there is no axial force
applied. As the bending moment is applied the beam will bend into a curved shape.
1I I (a)
I (b) (c)
M M
trrJ (d)
Figure 3.19
A small element of the beam is shown in figure 3.19a with two vertical lines
indicating plane sections. After the application of the bending moment it is
assumed that the straight lines could be deformed, as in b or c. However if a
rotation of 180° about they axis is considered it will be seen that neither defor-
mation is possible and the only form which satisfies this condition is that shown
in figure 3.19d. From this it may be concluded that plane sections remain plane
when pure bending is applied.
In figure 3.20a a small portion of a beam is shown bent, such that the radius
of AA in the xy plane is Ry- The two inclined lines were originally vertical and a
distance dx apart. It will be seen that the bending effect will compress some
fibres such as BB, and extend others. There must however be some position AA
where the fibres have not changed length. We shall refer to this particular position
as the neutral axis.
II
I Jl I \
I I
I I
I I
I 1
1 d0 1 R
j-'----'1 y
I I
I I
\
I
I
I
(a) (b)
Figure 3.20
The length of the line BB distant y from AA can be written as (Ry - y) dO. Its
original length must have been the same as AA, that is, Ry dO. So that
_ (Ry- y) dO-Ry dO _ y
Exx - Ry dO - - Ry (3.30)
Note that this equation has been derived solely from geometric considerations.
It has not been necessary to make any assumption about the way in which the
material behaves when loaded, apart from the fact that it must be homogeneous
and isotropic.
The cross-section of the beam is shown in figure 3.20b. A small area dais indi-
cated with coordinates y and z. The stress there is axx, and as it was originally
stated that the axial force is zero
J Gxx da = 0
Now make the assumption that the material of the beam is linearly elastic, so
that axx = Eexx· Thus
JEexx da= 0
or
_ _!_fyda=O
Ry
and
Jy da = 0
If the beam had been bent such that the radius of curvature was R z, by similar
reasoning it would be found that f z da = 0. So the position of the neutral axis is
defined-it must pass through the centroid of the cross-section.
The stress due to the radius of curvature Ry can be written axx =- (E/Ry)Y
and that due toRz will be Uxx = -(E/Rz)z. As the material is linearly elastic,
superposition can be used for the case when the beam is bent about two axes at
the same time, that is
axx =- (..§_y
Ry
+_§_ z)
Rz
(3.31)
It will be remembered that the moment of a force about a particular point can
be written as r x F. Taking moments about the centroid of the cross-section
dM =- (yj + zk) x (..!i... y da + ..!.._ z da)i

Ry Rz
or
(3.32)
and
My=- ..!...fyz da- £fz 2 da

Rz Rz
It should be noted from these equations that even if Rz is zero it is necessary in
this general case to apply moments about both the z andy axes to produce a
curvature in the xy plane. In fact Mz and My can be combined and a moment
applied about an inclined axis.
The reader may possibly recognise some of the integrals in equations 3.32.
Integrals of this type are used for finding moments of inertia. However we shall
refer to f y 2 da and f z 2 da as being second moments of area and denote them
by lzz and lyy respectively. fyz dais called a product second moment of area
and is denoted by lyz so that
Elzz Efyz
Mz=-- +-- (3.33)
Ry Rz
M = _ Elyz Elyy
y Ry Rz
If lyz is zero it will only be necessary to apply a momentMz to produce a
radius of curvature Ry. For this case equations 3.33 are much simplified and fzz
and I yy are then referred to as principal second moments of area andy and z are
called principal axes. The requirement is that f yz da should be zero. This is
certainly true if a cross-section has an axis of symmetry.
All the sections shown in figure 3.21 have at least one axis of symmetry, hence
the directions of the principal axes are known at once. For bending about the z
axis we may write
axx = Mz E
(3.34a)
Y Izz Ry
Similarly for bending about they axis

axx My __ E
(3.34b)
Z Iyy Rz
We shall confine our attention for the time being to bending about a principal
axis so that the expressions given by equations 3.34a and b may be used.
~
-r I
dy~
~
d
{
! 1--b-1
(o) (b) (c) (d)
Figure 3.21
It is necessary to be able to determine values of fzz and fyy· For the rectangle
in figure 3.2la
I
zz
= fy 2 da= f d /2
-d/2
b d3
by 2 dy=--
12
For the circular cross-section in figure 3.21 b it is easiest to determine Ixx.
= J0
R 7t R4
I 21tr 3 dr=--'-
xx 2
Now Iyy + Izz = Ixx and in this case Iyy = Izz• so that
nR 4
Iyy =Izz =4
For the Universal beam (figure 3.2lc) one suggested method of finding Izz is
to take the value for a rectangle equal to the maximum dimensions of the section
and then to subtract the values for two smaller rectangles each of the same size.
The derivation of the bending formula was based on a basic assumption that
the bending moment was constant. This would mean that the theory would have
a very limited application in practice. However the formula is still used when the
bending moment changes and the shear force is thus no longer zero. If the length
of the beam is long when compared with the dimensions of the cross-section the
errors introduced by the simple theory will not be great. The negative sign is
often omitted in equations 3.34, mathematically it should be present but in most
simple cases the sign of the stresses can be determined by inspection. For the
particular case when Mz, say, is constant, Ry will be constant and the beam will
bend into the arc of a circle.
As a fairly straightforward example of the application of the bending theory,
the maximum value of p, the uniformly distributed load, is required for the canti-
lever in figure 3.22. The cross-section is rectangular and the maximum stress is
limited to a 1 .
r-b+J
Figwe 3.22
01
From equation 3.33 it can be seen that for a given cross-section the stress will
be a maximum when both y and M have their greatest values. The maximum
value of y in this case is d/2 and the worst bending moment occurs at the support
and has a value of p/ 2 /2. The value of I is bd 3 /12, so that
and
In a large number of cases we are interested in the maximum value of y and

often the value of 1/Ymax is required for a particular section. This is referred to as
the section modulus and it is denoted by Z. It would of course be perfectly poss-
ible to have two values of Z for a particular section, for example, consider the Tee
section in figure 3.2ld, for bending about the z axis.
For a given maximum value of stress the bending moment will be a maximum
when Z is a maximum. For a symmetrical cross-section of given area the material
should be placed as far away from the neutral axis as possible. There are limi-
tations to this as other effects such as shear and stability have to be considered.
The Universal beam {figure 3.21c) makes good use of material and for a given
cross-sectional area would carry much greater bending moments than a rectangular
beam. For a beam of this type, the web {the vertical part) does not contribute
much to the bending resistance. A fairly close approximation to the moment
is obtained by only considering the flanges {the horizontal portion) and assuming
that the longitudinal stress is constant within them .
3.13 Bending of a composite beam

A cross-section of one type of composite beam is shown in figure 3.23a. It is
formed from a piece of material 2 sandwiched between two similar pieces of
material 1. Assume that there is perfect bond between the two materials. To
make the problem general, a bending moment M is applied about an axis inclined
at 8 to they axis. The maximum resulting tensile stress is required in material 1.
Et > £2.
rf--->r+--lt
~+-~~~~~d2
z t---t-'l""'"""""l-·
(b) (c) (d)

Figure 3.23
It will be necessary to work about the principal axes. These will be they and z
axes as shown. The bending moment can be resolved into two components
Mz =Msin8 My =Mcose
For bending about the z axis the strain and stress distributions will be of the
forms shown in figures 3.23b and c. The strain must be linear over the section
and has a maximum value of e 1 say. At the junction of the two materials the
strain must have the same value in each material and as E 1 > E 2 there will be a
sudden change in stress at the junction.
The maximum stress in material! is e 1E 1
So that the moment carried by material 1 is

2
M1 =nb(d~- d~)etE t "d;
The two moments must sum to Mz and the value of e 1 can be found
. bet
M sm e = -
3 3
{E 1d 1 - d 2 (£ 1 - E 2)}
6dt
The maximum tensile stress in material will occur at the bottom edge and has a
value E 1 e, therefore
(3.35)
For bending about they axis the strain distribution and the stress distribution
for each material will be linear. If the maximum strain is e2
therefore
The maximum tensile stress in material 1 will occur at the left-hand edge and
has a value E 1 e2
(3.36)
The maximum value of the total tensile stress will occur at the bottom left-hand
corner of the section and has a value given by the sum of the stresses given by
equations 3.35 and 3.36.
An alternative approach to this type of problem is to try and find an equivalent
section that consists entirely of material 1. This is referred to as transforming the
section (see section 3.5).
If the strain in material 2 at a particular distance y from the neutral axis is e
the stress is £ 2 e then the force carried by an element of width b and height dy is
E 2 eb dy. If this material is replaced by an equivalent amount of material 1 the
strain value would have to remain at e but the stress would beE 1 e. It is essential
that the force transmitted by an element of height dy should remain the same,
otherwise the bending moment would be altered. Hence the only possible modi-
fication is to the breadth, changing it to b 1 say
E1eb1 dy =Ezebz dy
Ez
b 1 = b - = bm
El
where m is referred to as the modular ratio.

For bending about the z axis the transformed section for the composite beam
we have been discussing is shown in figure 3.23d, where material 2 has been
turned into an equivalent amount of material 1.
The value of the equivalent I is

rtbd[- -b_(b- mb)d]
This will lead to the same expression for e 1 as the one already obtained.
The transformed section for bending about they axis is left as an exercise for
the reader.
3.14 Bending of unsymmetrical sections

The bending problems considered so far have been confined to sections which
have had an axis of symmetry and hence the directions of the principal axes have
been known. We shall now proceed to make use of equations 3.33 which were
derived from a general case of bending. These two equations can be solved
simultaneously to give values of E/Ry and E/Rz
E Mzlyy + Myfyz
Ry fzzfyy - Iiz
E -Myfzz- Mzfyz
Rz fzzfyy - Iffz
If these values are now substituted into equation 3.31 we shall obtain a
general expression for the longitudinal stress axx
_ -Mz(Ylyy - zlyz) + My(zlzz - ylyz)

0 xx - 2 (3.37)
fyyfzz- fyz
This rather cumbersome expression will give the value of the longitudinal stress
for any chosen pair of axes. It will be seen to simplify to the expressions 3.34a and
b for the case when fyz = 0.
The alternative approach as we have seen is to resolve the applied bending
moment into components acting in the directions of the principal axes. This is a
perfectly simple operation so long as the directions of the principal axes are
known; however in the case of an asymmetrical section we have no idea of the
position of the principal axes. It will be assumed that a particular set of axes y, z
have been chosen for an asymmetrical section (figure 3.24), also that it has been
possible to determine lyyJzz and fyz·
The principal axes Y and Z are assumed to be at an angle (J to they, z axes.
Consider a small area da with coordinates y, z or Y, Z.
Y = y cos (J - z sin (J
Z = z cos (J + y sin (J
J
lzz = Z 2 da= Jz 2 cos 2 8 da+ J2yz sin8 cos(] da + Jy 2 sin 2 8 da
Now
Figure 3.24
Thus
Izz =!(Iyy + Izz) + !(Iyy- Izz) cos 20 + Iyz sin 20 (3.38)
A similar expression to this can be obtained for I yy
Iyz=fYZda
= Jy 2 sin 0 cos 0 da- Jz 2 sin 0 cos 0 da + Jyz(cos 2 0- sin 2 0) da =0

since Y and Z have been defined as principal axes. Hence
Izz- Iyy
2 sin 28 + I yz cos 28 = 0
or
2Iyz
tan 20 = - - - ' - - (3.39)
Iyy- Izz
We should by now recognise the form of the two equations 3.38 and 3.39-they
are in fact similar to the expressions obtained for principal stress and strain. So
that once again the Mohr circle diagram can be used to good effect. Suitable axes
y and z are chosen for the section,/yy,/zz and Iyz are determined. Axes are now
set up so that second moments of area are plotted horizontally and product
second moments of area vertically. Points (Iyy ,/yz) and (Izz,- Iyz) are located.
These lie at the ends of a diameter of the circle, which can then be drawn. The
principal second moments of area are given by the values where the circle cuts
the horizontal axis.
To indicate the various calculations the principal second moments of area for
the unequal angle section in figure 3.25a will be found. The dimensions are in mm
and the thickness of the material is 6 mm.
lyz y
y
-It!-
z I z
I
I
(a) (b) (c)
Figure 3.25
It will first be necessary to locate the centroid and we shall consider the angle
made up of the two rectangles indicated.
The area of the section is 192 + 306 = 498 mm 2 • Taking moments about the
bottom edge
(192 X 22) + (306 X 3)
Yc = 498 = 10·3 mm
and moments about the left-hand edge
(192 X 3) + (306 X 25·5)
z = = 16·8 mm
c 498
Having located the centroid suitable axes as shown are chosen.
fzz = (Tt X 6 X 32 3 ) + (192 X 11·7 2 ) + (Tt X 51 X 6 3 ) + (306 X 7·3 2 ) =59 900 mm 4
lyy =(-b. x 32 x 6 3 ) + (192 x 13·8 2 ) + (-b_ x 6 x 51 3 ) + (306 x 8·7 2 ) = 126 600 mm4
It is also necessary to determine the value of lyz which is of the form f yz da.
Suppose that the product second moment of area is required for the rectangle
in figure 3.25c. An elemental area is shown in the figure as t dy
lyz = fyzt dy
Now z is constant for all similar small areas and is the horizontal distance from
the axis to the centroid of the rectangle, that is, z
lyz = Z I yt dy I yt dy = Ay
where y is the vertical distance from the axis to the centroid of the rectangle. So
that lyz = Axy. For the given case taking account of signs. Note it is perfectly
possible for lyz to be negative for a particular case.
lyz = (192 x 11·7 x 13·8) + 306(-7·3)(-8·7) =50 400 mm 4
The coordinates ( 126 600, 50 400) and (59 900, -50 400) are plotted and the
Mohr circle diagram drawn. From this the principal second moments of area are
found to be 154 000 and 33 000 mm 4 , and the angle 2() is 56°. This means that if
we measure clockwise from they axis through an angle of 28° we shall come to
an axis of the major principal second moment of area.
An alternative to drawing the Mohr circle diagram would be to substitute into
a similar expression to that of equation 3.14 for the principal values, and into
equation 3.39 for the directions.
3.15 Strain energy due to bending

For an elastic material a general expression for the strain energy stored per unit
volume is given by equation 3.29. In a case of pure bending the only stress that
will arise is axx and the equation will simplify to
dU a~x
-=-
dV 2E
Assuming that bending takes place about a principal axis and that a bending
moment M has been applied
yMz
a ==---
xx fzz
For a small area of cross-section dA and unit length of beam
Now
so that
u- I _z_
M2
2Efzz
The total strain energy in the beam of length I is then
U= f z __
M2 M21
z_ dx == __
0 2Elzz
z_
2Efzz
since Mz has been assumed constant.

If Mz is a variable, that is, we are no longer considering a case of pure bending,
a similar expression can be used for the bending strain energy
M} dx
u- - - f 1
o 2Efzz
(3.40)
3.16 Combined bending and axial force

The cross-section of a short column is shown in figure 3.26a. Oy and Oz are
principal axes. A vertical 'compressive' load Pis applied at the top end of the
column so that its position is defined bye, the eccentricity, and 8.
y
y
t
""'
./
J
b
z z "-.... v
d
(a) (b)
Figure 3.26
The load P can be replaced by an axial load P and a moment of value Pe. The
moment can be resolved into two components
My= -Pe sin() Mz =Pecos()
At any pointy, z the stress due to the axial load is axx =- P/A. From
equations 3.34a and b the stress due to My is -Pez sin 8/Iyy and that due to Mz
is- Peycos 8/lzz.
The total stress can be found by superposition.
P Pez sin () Pey cos ()
a =------ ----
xx A fyy fzz
Certain constructional materials are weak in tension; concrete and brickwork

are notable examples. If it is assumed in a particular design of a column made of
one of these materials that no tensile stress is allowed to develop, the position of
the applied load P will be limited and must lie within a certain area called the
core. We shall find the core for a rectangular section (figure 3.26b).
A = bd; lyy =bd 3 /12; lzz =db 3 /l2. The worst stresses arise when z = -d/2,
y = -b/2, so that
6e sin() 6e cos()
bd bd 2 bd2 = 0
bd
be sin () + de cos () =6
This represents the equations of straight lines that cut the z axis at ±d/6 and
they axis at ±b/6, and the core is as sketched in figure 3.26b.
3.17 Plastic bending of beams

The stress-strain curve for mild steel in both tension and compression will be
idealized as shown in figure 3.27. The upper yield point has been neglected and
the stress is assumed to remain constant at ay until work hardening takes place.
Ey E
Figure 3.27
We are not interested beyond this 'plastic' region as in this case large deflections
of the beam would take place.
The following assumptions will be made when a beam is bent
(1) the lower yield stress and the modulus of elasticity have the same value
in tension and compression
(2) the stress remains constant in the plastic range
(3) plane sections remain plane
( 4) the material is isotropic and homogeneous
ay
(a) (b) (c) (d) (e)
Figure 3.28
The section of a beam in figure 3.28a is symmetrical about one axis and a
steadily increasing bending moment is applied. If the stresses are below ay, the
yield stress, the neutral axis will pass through 0, the centroid of the section. In
figure 3.28b the top edge has just reached ay. The elastic section modulus Ze =
1/Ymax• and the moment at yield is given by
ayl
My =--=ayZe
Ymax
The effect of increasing bending moment is shown in figures 3.28c, d and e. In

c, part of the upper portion of the section has become plastic, but the whole of the
section below the neutral axis is still elastic. Yield has taken place over part of the
bottom portion of the section in d. Finally in e the beam has reached what is
termed the fully-plastic bending moment. This last diagram does not quite corre-
spond to the stress-strain curve (figure 3.27) since a small area of the section either
side of the neutral axis remains elastic. This however will have a negligible effect
on the value of the fully-plastic bending moment. It should be noted that there
has been a steady shift of the neutral axis towards the bottom edge of the beam.
This would not occur if the beam was of a completely symmetrical section. The
neutral axis will finally divide the section into two equal areas.
The value of the fully-plastic bending moment Mp will be required. If y 1 and
y 2 are the distances from the neutral axis to the centroids of the areas above and
below the neutral axis respectively, and A is the area of the cross-section, then
ayA
Mp =2 (Yt +Y2) (3.41)
The plastic section modulus is denoted by Zp where Mp = ayZp, then

A
Zp =2(Y1 +Y2) (3.42)
Since My= ayZe we can obtain the ratio

Mp Zp
-=-=a
My Ze
The ratio is only dependent on the dimensions of the section and is called the
'shape factor'. It is also sometimes referred to as the 'form factor'.
For the case of a rectangular beam of breadth band depth d
z e =bd6- z p =bd4-
2 2
therefore
a= 1·5
For universal beams, that is, !-sections, an average value of a about the major axis
isl·lS.
It can be seen that it will not be a difficult matter to calculate the value of the
shape factor for a particular section. As an example consider the tee section
(figure 3.29).
It is first necessary to find the position of the centroid. Taking moments of
areas about the lower edge of the section
- (10t 2 X St) + (10t 2 X 10·5t)
y = 20t2 = 7·75t
1= (n x 1or x r 3) + oot 2 x 2·75 2 t 2 ) + (-ht x 1ooot 3 ) + oor 2 x 2·75 2 t 2 )
= 235·4t 4
ELEMENTARY ELASTICITY, PLASTICITY AND BENDING OF BEAMS Ill
r----tot "lj_
t
f
J
TOt
-ld-
Figure 3.29
so that
235·4t
ze =-
4
- = 30·3t3
7·75t
For the fully plastic case the neutral axis will be at the junction of the flange
and web.y 1 = t/2, y 2 = 5t,A = 20t 2 , then
2
Z =20t
-- ( t) = 55t 3
5t+-
p 2 2
therefore
Zp 55
a = - = - = 1·82
Ze 30·3
If a particular stress-strain relationship is assumed it is perfectly possible to
calculate a theoretical moment-curvature relationship for a given cross-section.
We shall do this for the rectangular cross-section shown in figure 3.30a. The
T
~--1·-b--~·1
d
1 ----------- kd
1~1
5 10
1S.
(a) Xy (b)
Figure 3.30
material is ideal elasto-plastic, with a yield stress of ay. This means that the stress-
strain relation is similar to figure 3.27. Assuming that the yield stress has been
exceeded and that a total depth of beam kd remains elastic. The moment carried
by the elastic portion is
The moment carried by the plastic portion is
4bd
2
M= (1- k 2 )ay
Thus the total moment is
This can be expressed in a slightly different form if we make use of the fully
plastic bending moment Mp = aybd 2 /4
~ = (1- ~2 ) (3.43)
From the consideration that plane sections remain plane e = yjR, where y is
the distance from the neutral axis to any fibre, and R is the radius of curvature,
1/R = ejy = K, the curvature. Therefore K = 2ay/Ekd. If we put Ky = 2ay/Ed-the
maximum fully elastic curvature
K
-=-
1
Ky k
therefore
(3.44)
This equation will apply forK> Ky.

Figure 3.30b shows a plot of this equation. The value of M/Mp will be asymp-
totic to 1 when the ratio of the curvature is infinite. A similar shape of curve
would result for the case of a universal beam, the deviation from the straight line
would occur at a moment ratio of about 0·87, that is 1/ 1·15.
If a mild-steel specimen is tested in tension and the yield point is exceeded, it
will be found that the unloading curve is a straight line parallel to the original
linear part of the loading curve when the load is removed. From this it is seen
that the unloading process is linearly elastic with a modulus value of E. The speci-
men will not regain its original length and has what is termed a permanent set.
The same type of action takes place when a beam is bent and the yield point is
exceeded in parts of the cross-section. Every element of the beam will unload
elastically when the bending moment is removed. Resultant stresses will be locked
in the beam. The resultant force and bending moment due to the stresses must be
zero and the beam which was originally straight will remain permanently curved.
It is a straightforward matter to calculate the residual distribution and the final
value of the curvature. This will be demonstrated with the help of the example
below.
f--- b- j
T d
(0) (b) (c) (d)
Figure 3.31
A beam of rectangular section has a moment applied such that half the cross-
section becomes plastic with stress Oy (figure 3.3la). The moment is then removed.
It is first necessary to calculate the bending moment applied. This can either be
found from first principles or from equation 3.43.
Oybd2 . ll 2
M= - 4 - (1- u) = ~ aybd
_l_
Unloading will occur elastically when the moment is removed. One way of
considering the problem is to apply a moment of - M and determine what elastic
stresses this would produce. By superposition the sum of the two moments is zero.
The final stress distribution can be found by superimposing the two separate dis-
tributions due toM and -M. The unloading moment will produce a linear stress
distribution giving a maximum stress
M(d/2)
amax = bd2/12
where
therefore
The distribution is shown in figure 3.3lc. The maximum stress is greater than ay,
but this does not matter since this stress is of opposite sign to the loading stress
and does not in fact exist by itself. The final distribution is given in figure 3.31 d
where the two diagrams with a sign change for the unloading stresses have been
superimposed. The shaded portion gives the resultant distribution.
To find the final curvature it is necessary to calculate the initial curvature for
the loading condition and then subtract the change in curvature that takes place
when unloading.
For loading at y = ± d/4 the stress has just reached a value of ±ay, hence the
strain at this position is± ay/E therefore
Uy d
E 4R 1
or
4ay
-=-
Rl dE
The change of curvature when unloading is found from the fact that the value
of the stress is (11/8)ay aty = d/2
or
_l_=l_!_~
R2 4 dE
The final curvature is
Uy ll 5 Uy
dE(4 - 4 )=4 dE
3.18 Combined plastic bending and axial force
We shall consider the case of a rectangular section (figure 3.32a) which has an
axial load P applied. A bending moment M is then applied until the section
becomes fully plastic (figure 3.32b). It is obvious that when this stress distri-
bution is integrated over the section there must be a resultant force P.
J. P.
kd
T
(a) (b) (c) (d) (e)
Figure 3.32
This stress distribution can be thought of as being composed of two separate

distributions: c showing the fully-plastic stress distribution when P = 0 (that is, a
moment Mp applied); d showing a stress distribution equivalent to the load P
such that 2aybkd = P.
If a purely axial load is applied to the section, the load Ps producing a stress of
ay over the section is often referred to as the squash load. Thus
P= 2kPs
and
Pkd 2
M=Mp - - 2= Mp (1-4k)
Substituting fork = P/2Ps
~+ (!..)2 = 1 (3.45)
Mp Ps
This equation may be plotted in the form shown in figure 3.32e to give what is
called an interaction diagram. Any combination of M and Pis safe if i lies within
the boundary, and the cross-section is adequate. A point on the boundary indicates
that the section is fully plastic and outside the boundary is unsafe.
Problems
3.1 A rope of length I is required to carry a vertical load Pat one end. The
material is of density p and elastic modulus E. The cross-section A is circular but
varies such that the axial stress has a constant value a. Find an expression for A at
a distance x above P. What is the extension of the rope when the load is applied?
3.2 A circular brass tube of wall thickness 1 mm fits inside a steel tube of the
same wall thickness with a clearance of about 0·1 mm. The mean diameter of the
tubes may be taken as 100 mm for calculation purposes. Both tubes are of the
same length and are connected together at their ends by heavy flanges. Neglecting
any end effects, find the longitudinal stresses for the following cases
(a) a tensile load of 30 kN applied at the ends of the tubes
(b) a temperature rise of 80 °C
(c) a tensile load of 30 kN together with a temperature rise of 80 °C.
3.3 A vessel has a cylindrical centre portion oflength I and radius r, the ends are
hemispherical. The wall thickness tis small compared with the radius. For an
internal pressure P, find the stresses in the cylinder and the hemispheres.
A thin hollow steel sphere of radius rand wall thickness tis filled with water.
The temperature of the sphere and contents are raised by T°C and water is
pumped in until the pressure has risen by P bar. Derive an expression for the
volume of water that has to be pumped in. The vessel is then sealed and allowed
to cool to its original temperature. What is the final pressure of the water?
3.4 For a case of plane stress find the values of the principal stresses given axx =
2, 2, 2.
140 N/mm ayy =-50 N/mm Txy = 80 N/mm A rosette strain gauge (0°,
45°, 90°) is fixed to a steel sheet and the same stress system applied. The 90° arm
is inclined at 40° clockwise to the direction of the axx stress. Determine the strain
in the direction of the gauge arms.
3.5 From a theoretical analysis, the principal stresses at a point on a steel
member due to a certain load have been calculated as 65 N/mm 2 tensile and
138 N/mm 2 tensile. To confirm the analysis an electrical resistance strain gauge
rosette with arms at 120° was placed at the point. When the member was loaded
the following changes in the resistance of the arms were noted. dR 0 = + 0·122 n,
dR 120 = + 0·104 n, dR 240 = +0·04 n. The nominal resistance of each arm was
120 n and the ratio of electrical strain to mechanical strain or the gauge factor
was 2·1.
It is thought that one of the arms of the rosette may not be functioning
correctly. Show that the theoretical and experimental results are inconsistent and
suggest which arm is incorrect.
3.6 A rectangular beam is to be cut from a circular log of wood 180 mm in

diameter and used to span 2·4 m with simple supports at each end. Taking a
maximum longitudinal stress of 10 N/mm 2 , find the greatest uniformly distributed
load that the beam can carry.
3.7 Derive the following relationships used in the bending of beams and state any
assumptions on which they are based.
a M E
y I R
A rectangular beam of breadth band depth dis made of a material that has a
modulus E 1 in tension and E 2 in compression, E 1 > E 2 . Find the position of the
neutral axis and show that
M E*
-=-
[ R
where
3.8 Explain the terms (a) principal second moment of area; (b) product second
moment of area.
p 3.8
The section shown in P 3.8 is formed from sheet metal of thickness t where
t ~d. Find the values of the principal second moments of area. The section is
used as a cantilever of length l with the longer side vertical. Find the maximum
value of the longitudinal stress when a vertical point load Pis applied at the end
of the cantilever.
3.9 A short vertical concrete column has a regular hexagonal cross-section of side
a. A vertical load P can move along the z axis which lies along a line joining two
opposite corners of the hexagon. The maximum allowable compressive stress is
am and no tensile stress is permitted. Find the limits of the position of P and the
maximum value of Pin terms of its distance from the centre. If the load is not
restricted to the z axis define the area within which it must be confined if no
tension is to develop in the section.
3.10 A beam has a rectangular symmetrical hollow cross-section. The outer

dimensions are 2b by 2d and the inner dimensions are b by d. The z axis is parallel
to the b dimension. Find the ratio of the fully-plastic bending moment to the
maximum elastic moment for bending about the z axis. The material of the beam
is ideal elasto-plastic with a yield stress ay.
Find the moment required such that the stress has just reached ay at a distance
d/2 from the neutral axis. What is the residual stress at this position when the
moment is removed? Find also the maximum elastic moment of opposite sign that
may now be applied.
3.11 Show that the plastic moment of resistance of a mild steel beam of rectangu-
lar cross-section of breadth band depth d isMP = aybd 2 /4 where ay is the yield
stress and the material is ideal elasto-plastic.
A prestressing force is applied inwards at the bottom edges of such a beam on
the vertical centre-line of its section so that yielding just begins to occur. Show
that the effective plastic moment of resistance of the section is now increased
fromMP to HMP.
4 TORSION AND SHEAR EFFECTS
4.1 Torsion of a circular cross-section

A pure torque is applied to a circular rod of homogeneous isotropic material
(figure 4.1a). We shall consider possible deformations that can take place on the
section AA. In figure 4.1 b it is assumed that the left-hand portion, X, has a convex
surface so that the surface of the right-hand portion, Y, must be concave. This is
not compatible as may easily be seen by rotating the left-hand portion through
180° about A-A to become Y' (figure 4.1c). If we now compare band c, apparently
the same torque has produced a concave surface on Y and a convex surface on Y'.
Clearly this is impossible. A line that was radial is shown in c assumed deformed
into a curve when the torque is applied. A similar argument may be applied to
show that this is impossible. The details are left as an exercise for the reader.
I
A I (b) (c)
(d)
Figure 4.1
It may then be inferred from the above arguments that when a rod of circular
cross-section has a pure torque applied
(I) plane sections remain plane
(2) a radial line remains straight.
A thin sheet of material has pure shear stresses, T, applied, and would deform
in the manner shown in figure 4.2a. It is possible to bend the sheet round to form
118
TORSION AND SHEAR EFFECTS 119
the slit tube of radius r (figure 4.2b ). The edges could be fixed together in some
way. If the ends of the tube have rotated through an angle e with respect to each
other
re = l"f
Note that the shear stress 'flows' round the circular end of the tube. Taking
moments about the axis of the tube it can be seen that the same stress pattern
would be produced by applying an axial torque of value
T= 21tr 2 tr
where tis the thickness of the sheet of material.
(a) (b) (c)
Figure 4.2
A solid rod and indeed a hollow one could be thought of as a large number of
cylindrical shells are cemented together. A torque is applied and each shell will
rotate through the same angle e. For a shell at radius r
e =!:!... (4.1)
r
To derive equation 4.1 no mention has been made of the elastic properties of
the material. If the material is elastic
T rl
"(=- 8=-
G Gr
or
T
-=-
ae (4.2)
r
A section through a circular rod of radius R is shown in figure 4.2c, together

with a cylindrical shell of radius r and thickness dr. If the shear stress at radius r is
r then the tangential force on an element of length ds is r ds dr and the moment
about the centre is rr ds dr. The torque due to the shell is dT =r2 1r r2 dr and the
total torque on the rod can be written as
(4.3)
If the rod were hollow the limits of integration would extend between the two
radii.
r is a function of r as given by equation 4.2 for an elastic case, therefore
T= f
0
R GO 2rtGO R 4
-2rtr3 dr=-- - -
I I 4
Now the second polar moment of area for a circular rod is J where
rtR 4
J=-
2
thus
G8J
T=- (4.4)
1
which together with equation 4.2 gives
T T GO
-=-=- (4.5)
r J
This result will only apply to a circular section rod or tube of linear elastic
material. When a torque is applied to rods of other shapes the cross-section will
no longer remain plane but will warp or deform.
As a simple example of elastic-torsion theory, consider the problem of a rod of
circular section built in at each end. One-third of the rod is of radius rand the
other two-thirds of radius 2r. If the shear stress is limited tor, what is the maxi-
mum torque that may be applied at the change in section? Any stress concen-
trations are to be neglected.
The first point to note is that, at the point of application of the torque, both
parts of the rod must have twisted through the same angle 0. Using subscript 1 for
the shorter length and 2 for the longer length
r1 GO 3G0r
-=- rl=--
r 1/3 I
r2 GO 3GOr
-=-- r2=--
2r 21/3 l
therefore
therefore
The case of a rod of one material inside a tube of another material (the two being
rigidly connected at the common boundary) will not present any difficulty. The
relationship between shear strain and radius will be linear. As there will be a
change in the value of the modulus of rigidity at the common boundary, there
must be a sudden change in the value of shear stress at this point. The total torque
can be found by summing the torque for the rod and the torque for the tube.
4.2 Strain energy due to torsion

Referring back to equation 3.29, the strain energy per unit volume for a case of
pure shear would be given by
-=-
dV 2G
Consider a solid circular rod under the action of a torque T. For an annulus at
radius rand thickness dr the shear stress Tis given by TrjJ. Therefore the energy
stored in the ring of unit length is
T2r2
dU = x 2rtr dr·
J2 X 2G
Total energy per unit length equals
So that for a rod of length l
(4.6)
This could have been derived in a simple manner. If(} is the resulting angle of
twist when the torque Tis applied, the external work done is TO /2. This must of
course be equal to the strain energy stored. Now(}= Tl/GJ therefore
4.3 Combined torsion, bending and axial force

A rod of circular section is subjected to an axial force P, a bending moment M
about a horizontal diameter and an axial torque T. The question arises, what is
the maximum stress that will result?
Taking the x axis coincident with the axis of the tube and the z axis as hori-
zontal, the longitudinal stress produced by the bending moment can be found
from simple bending theory. If the rod has a radius r
4My
Oxx =-4-
1tY
This can be combined with the stress Oxx = Pj1tr 2 produced by the axial force,
assumed tensile; therefore
4My P
Oxx = - 4 - +-2
1tY 1tY
The maximum stress will arise when y has a maximum value, that is
= 4M
0
XX - - +P-
1tY3 7tY2
Oyy = Ozz = 0
The torque will produce a shear stress that will vary from zero at the centre of the
rod toT= 2Trf7tr4 at the outside.
To find the resulting maximum stress when all 'forces' are applied, it will be
necessary to perform a principal stress analysis. To simplify the arithmetic it will
be assumed that M = T =Pr. Thus
SP 2P
Oxx=-2 Ozz=O Txz=-2-
1tY Ti.Y
p
al 02 = - 2
1tY
r; ± !(25 + 16) 112 ]
5·7P 0·7P
al =-- a2 = - - -
7tr2 7tY2
3·2P
Tmax = 7tY2
4.4 Plastic torsion of circular rods

It will be assumed that the shear-stress-shear-strain curve for mild steel, when sub-
jected to pure shear, is similar in form to that assumed for tension and compression
in figure 3.27. For a case of pure torsion on a circular rod we shall further assume
(1) the stress remains constant in the plastic range
(2) plane sections remain plane
(3) the material is isotropic and homogeneous.
The torque required to make a section fully plastic can be found assuming that
the yield shear stress is Ty. Considering an annulus at radius r of thickness dr.
Moments about the axis give
dT= 21tr dr x rry
T=r y J.R21tr 2 dr= 2 1tr3 r

0 1 y
The fully elastic torque is nr3ry /2, therefore
(4.7)
It is not difficult to treat a section that is partially elastic and partially plastic.
A specific example will demonstrate the method.
(b)
Figure 4.3
A rod of circular cross-section, radius R, is made from an elasto-plastic material

with a yield stress T y. Find the torque required to produce a plastic stress for one
half the radius. The torque is then removed; the residual stresses and the permanent
angle of twist per unit length are required.
The stress distribution is shown in figure 4.3a. The torques carried by the two
parts of the shaft can be found separately and summed.
The central elastic portion can be treated as an elastic rod of radius R/2
1tR 3 ry
TE = -1-6-
T = Tp + TE =H1tR 3 ry
For the unloading process we shall make exactly the same assumption as for
the case of plastic bending, that is, the rod unloads elastically. Effectively this
means that we apply a torque of value Tin the opposite sense and calculate the
stress distribution. This has then to be combined with the initial stress distribution.
=(31/48)7tR 3 ryR =ijr
Tmax 1tR4/2 Y
This stress will vary linearly with r.

The two stress distributions have been superimposed in figure 4.3b, and the
residual stress distribution is shown shaded. The two triangular portions of this
diagram will have shear stresses of opposite sign such that the resultant torque is
zero.
The initial angle of twist e1 can be found for the radius at which the stress has
just reached r y, that is, R/2. The value of the shear strain at this point is ry/G, so
that edl = 2ry/GR. The change in shear stress at radius R during the elastic un-
loading process is (31/24)ry so that the change in angle of twist e2 will be given by
e2 31 ry
-=--
[ 24GR
The permanent angle of twist is given by
e,
---=--
e2 17 ry
[ I 24 GR
4.5 Shear stresses in beams
An element of length dx cut from a beam of uniform section is shown in figure
4.4a. The beam is subject to a varying bending moment applied about a principal
axis Oz. At the left-hand end A the bending moment isM and at the right-hand
end B the bending moment isM + dM
- - - dx
(a)--- B
Figure 4.4
(b)
At a height y above the neutral axis, the stress at section A will be axx = -My/I
and the stress at B will be axx + daxx = -(M + dM)y/1.
If we consider a small element of area dA at the height y, the force at section A
will be -My dA/I and the force at section B -(M + dM)y dA/I. There is then a
net difference in longitudinal force on the element of area dA given by
-dMy dA/1. So that there will be a difference in force of- J dMy dA/I on the
two areas shown shaded.
The elemen of the beam above the plane CDEF must be in equilibrium. The
only way of providing the necessary longitudinal force is by means of a shear stress
acting on the plane CDEF (figure 4.4b ). If it is assumed that the shear stress T yx is
uniformly distributed over the width of the beam b
Tyxbdx=- f dMydA
I
or
T =- dMfy dA
yx dx bi
From equation 1.10 dll-1/ dx =- Q, therefore
QfydA
Tyx = (4.8)
bi
The horizontal component of shear stress that has been found must be accom-
pahied by a complementary shear r xy, which of course acts in the vertical direction
on the face of the section. We now have a means of calculating the vertical shear
stress at any level on a beam.
An alternative method of writing equation 4.8 is
QA.Y
Tyx=bl
where A is the cross-sectional area of the beam above the level at which the shear
stress is required andy is the distance from the neutral axis to the centroid of the
area.
Note: The value of I is the second moment of area for the whole section. The
assumption has been made that the shear stress is uniformly distributed over the
width of the beam. Consequently it can be assumed that the vertical component
of shear stress at a particular level is constant.
The shear distribution for a rectangular section will first be found (figure 4.Sa).
At level CD
- - rd'2ybbidy-- IQ [d28 - h2]

Tyx - Txy - Q Jh 2 (4.9)
(a} (b)
Figure 4.5
This implies that the shear stress varies parabolically with a maximum value of
Qd2j8J at the neutral axis. Substituting for I gives
3 Q
T =- -
max 2 bd
Now Qjbd would be the average value of the shear stress, so that there is an
increase of 50 per cent above the average at the neutral axis.
Figure 4.6
The vertical component of shear stress for the circular cross-section (figure 4.6)
can be calculated, assuming the shear stress is constant over the breadth of the
beam. However, it is not possible for the shear stress to be vertical at CorD as
this would infer that the shear stress is crossing an unloaded boundary-an impossi-
bility. It is essential that the shear stress be tangential at C and D and the distri-
bution would appear as shown in the diagram. The maximum stress will still occur
at the neutral axis and has a value of 4/3 times the average shear stress.
In the derivation of the simple bending formula a= My/! it was assumed that
the bending moment was constant along the length of the beam, and it was per-
fectly in order to assume that plane sections remained plane. It was further
assumed that this latter statement was true even if the bending moment varied.
However, if a shear stress exists over the cross-section of the beam, this must
result from the effect of a shear strain. If we consider a cantilever with an end
Figure 4. 7
load P (figure 4. 7), the shear force is constant with a value of - P. Due to the
effects of shear strain, a section that was plane will no longer be plane but would
deform in the manner shown in the figure by the full lines. At the top and bottom
surfaces of the beam the shear stress is zero, therefore the line must be normal to
the top and bottom surfaces. The angle of inclination to the normal will be given
by r = r/G. Thus for a symmetrical cross-section, the worst angular displacement
will occur at the neutral axis. Here we have an example of a section warping, that
is, no longer remaining planar.
_[
(a)
1
l
d
D (b)
Figure 4.8
We shall now deal with two further cases. The first of these is the universal
*(~2
beam shown in figure 4 .8a. The shear stress in the web can be found as follows
J dA = ~ ( D ~ d
y
2 2
) + _ h2 )
To obtain the shear stress, the integral would be multiplied by a constant Q/bl. It
can be seen that the shear variation is again parabolic with a maximum value at
the neutral axis.
Tmin
QB
= 2b/
(D2-4 d2 )
T - -
Q [ -BD2
-
d2 B b ]
- -
max - 2bl 4 - 4 ( - )
Comparing these values it can be seen that there will only be a small difference
between them if b ~B.
If an approximation is made for I, omitting the web completely and also omit-
ting the second moment of area of the flanges about their own centroids
I~ 2B(D; d)( D: dr = ~ ( D2+ d)( D~ d 2 2

)
and the value ofT min becomes
b(D +d) bd
2Q Q
T. = ~- (4.10)
mm
This means that to a reasonable degree of accuracy we may assume that the shear
is carried by the web and the shear stress distribution is uniform over the web.
Figure 4.9
Finally the case of a thin walled tube will be derived (figure 4.9). For a solid
circular rod we have seen that the shear stress has to be tangential at the boundary.
We should then not be far wrong if we assume the shear stress in the case of the
tube to be tangential and uniform over the wall thickness t.
7 = Q f y dA = Q fe r cos cxrt da
bi -0 2ti
note b = 2t
Now I= 2nr 3 t, therefore

Q sin 8
r=
nrt
Twill be a maximum when 8 = 90°, that is, at the neutral axis T = Qjnrt. The
average shear stress is Qj2nrt so that in this case the maximum is twice the
average shear stress.
4.6 Shear-stress distribution in flanges

A similar piece of analysis to that introduced in the previous section can be used
to determine the shear-stress distribution in the flanges of the universal beam
{figure 4.1 Oa). A small length of beam dx is considered with a bending moment
changing from M toM + dM. It will be assumed that the flange depth t is small
.c=:-1 ~
1 -- ::~r: -- 1
II
II
II
It
1'
- -- ~ c....._
--~ _-_...,_1
(b)
Figure 4.10
compared with the total depth of the beam D, and so it will be a reasonable
approximation to assume that for bending about the z axis the stress Oxx in the
flanges is constant. It will also be sufficiently accurate to take y = D/2 in the
simple bending formula, that is, Uxx = -MD/21. The total force difference on the
two ends of the shaded area can now be written as
dMDat
21
If we isolate the portion of the flange beyond PQRS, it will be seen that to main-
tain equilibrium it is necessary to have a shear stress acting on the face PQRS.
Assuming this to be constant
dMDat
T tdx=----
zx 21
or
QaD
T =--
zx 21
This could be written as

QAy
T = --
zx tl
where A =at andy= D/2, a similar expression to that derived in section 4.5.
This shear stress must be accompanied by a complementary shear stress Txz
which will vary linearly from zero at the outside edge to a maximum value at the
junction with the web (figure 4.10b). The shear flow is shown by arrows on the
diagram. If the thickness of the flange is t and the thickness of the web is b, and T
represents the value of the horizontal shear stress at the web, then the value of the
vertical shear stress at the top of the web will be given by 2rt/b.
For other types of section it will be a straightforward matter to calculate the
shear-stress distribution so long as there is an axis of symmetry. However, in the
case of an asymmetric section certain difficulties will arise. These will be discussed
in conjunction with the section shown in figure 4.11 where tis small compared
with the other dimensions.
Figure 4.11
In the analysis used so far we have been fortunate in being able to work about
principal axes and so the simple bending formula was used. For the case shown we
shall assume that the only bending moment applied is about the z axis. Unfor-
tunately it is not known where the principal axis lies. It will therefore be necessary
to use the more complicated expression for the horizontal stress given by equation
3.37, which will simplify to
Mz(Yfyy - zfyz)
Oxx =- 2
fyyfzz - fyz
since My= 0. 0 is the centroid of the section and the axes are as shown. The total
longitudinal force on the shaded area is f a22 t da. Two integrals are required.
jytda=mta Jztda=(b-n-1)at
TORSION AND SHEAR EF FECTS 131
Thus the force difference on two elements dx apart is
- I /~ Iyz [Iyy(mta) - Iyz (b --n -

yy zz
2 %)at]
thus
Txz =Iyy I zzQ_ Iyz2 (Iyyma - Iyz [(b - n)a -- £]}

2
The shear-stress distribution in the flange is a combination of linear and para-
bolic relations. Note that the stress is zero when a = 0 and also when
a = 2(b - n - I YY
Iyz
m)
The distribution in the other flange can be found in a similar manner. It is left to
the reader to show that there will be a parabolic variation of shear stress in the
web .
4.7 Shear centre

If the shear-stress distribution is examined it will be seen at once that the vertical
components when integrated over the section must be equal to the vertical shear
force that is applied .
-l
r-----.___
- I
a
- a'
a
a'
d t t~
p
--
_[ - I
Q'
1-- b- 1 Q'
(a) (b) {c) (d )
Figure 4.12
For the channel iron (figure 4.12a) the horizontal shear stress in the flange will
vary in a linear manner from zero to a maximum value of Qbd/2I where t ~ b or
d. Integrating, we shall find that a horizontal force Q' = Qb 2 dt/4I is acting in the
flange.
Suppose that the channel iron is set up as a cantilever with a point load Pat the
end. The shear-flow diagram would be as shown in figure 4.12b and the resultant
forces as shown in figure 4.12c. Q is in equilibrium with the applied loading P,
and horizontal equilibrium is maintained by the two forces Q' acting in opposite
directions. Unfortunately they also provide an unbalanced couple of magnitude
Q'd which will cause the section to rotate if the applied load Pacts through the
centroid of the section. However it is possible to prevent this rotation if the
applied load is applied at a distance e from the centre line of the web (figure 4.12d).
Where
Pe = Q'd (4.11)
This particular point is known as the shear centre or the centre of shear for the
section.
4.8 Torsion of thin-walled tubes

When the torsion problem was discussed in section 4.1 the point was made that the
simple theory derived would only apply to circular sections. It is possible however
to derive a theory which will apply to a thin-walled tube of non-circular section.
(a) (b)
Figure 4.13
It will be assumed that the cross-section of the tube does not change with
length and that there are not any sharp corners or sudden changes in wall thick-
ness. This, however, does not preclude changes in wall thickness. Instead of refer-
ring to shear stress we shall use the term shear flow or shear force per unit length
of perimeter, that is, q = Tt.
A length of tube is shown in figure 4.13a. If the equilibrium of the small rec-
tangular element is considered (figure 4.13b) it will be seen at once that q 1 =q 2 =
q 3 = q 4 . From this we may infer that the shear flow is constant round the peri-
meter of the tube. A small element of the tube perimeter of length d s will have a
tangential force q ds. Taking moments about any point 0, dT =q dsa, where a is
the moment arm. Now a ds is twice the area shown shaded on the diagram, there-
fore
dT = 2q dA
or
T= 2qA (4.12)
where A is the area enclosed by the mean perimeter of the tube.

We have shown (equation 3.29) that the strain energy per unit volume due to
pure shear is r 2 /2G. For a unit length of tube the total energy stored will equal
the external work done when a torque Tis applied.
Now
T
rt=q=-
2A
therefore
1
1T8=- f -rds
T
2G 2A
1 -frds=-q-
8=- s~ (4.13)
2GA 2GA t
or
T
e = 4A 2 G
fds- t (4.14)
This theory is readily extended to take account of multi-cellular sections in

torsion. A particular example will indicate the procedure used.
n= -o
t -jt -
a
L
-.
j
*
I
2t
a-~ 2a
3!
(a) (b)
Figure 4.14
A torque Tis applied to a cellular structure of shear modulus G. The cross-

section is shown in figure 4.14a. The relation between G and(} is required and also
the values of the various shear flows. The effects of the sharp corners are to be
ignored. t ~a.
It can be seen at once from figure 4.14b that q 1 = q 2 + q 3 • For the left-hand
tube TA = 2a2q 1 and for the right-hand T 8 = 4a 2q 2 . Also for the left-hand tube
e =-l-[qla +2qla +(ql-q2)a]

A 2Ga 2 2t t t
while for the right-hand tube
e = _1_ [ q2a + 2q22a + (q2- ql)a]
B 4Ga 2 3t t t
It is essential that both tubes rotate by the same amount, that is, 8A = 8B.
From this relation
Now
Hence
Substituting into eA gives
4.9 Torsion of a thin rectangular section

y
t tt
y rr I
t
1
z II I
r b I
(a) (b)
Figure 4.15
We can obtain an approximation for the torsional rigidity of a thin rectangular

cross-section by using the theory derived in the last section. The solid section
where t ~ b (figure 4.15a) could be thought of as being built up of a number of
tubes, one of which is shown at a distance y from the neutral axis, of wall thick-
ness dy . If the wall thickness is constant for the tube Tis constant. The torque in
the tube is 2A T dy where A ::::; 2yb. Thus
dT
-=4byT
dy
From equation 4.14
or
dT
- = 8by 2 G8
dy
Equating the two values for dTjdy gives r = 2yG8. This indicates that the shear-
stress distribution is linear over the thickness of the section with a maximum value
aty = ±t/2.
therefore
T= f0
t/2
8bG8y 2 dy = !bt 3 G8
Compare this result with simple torsion theory which only holds for circular
sections T = JG8.
This implies that the equivalent value of the polar second moment of area for
the thin rectangle is bt 3 /3. It is interesting to compare this with
J =Iyy +Izz =_L(bt 3 + tb 3 ) = !?!__ (b 2 + t 2 )

TI 12
This theory can be used to derive approximate torsional constants for other thin-
walled sections. The channel iron (figure 4.15b) could be thought of as consisting
of three thin rectangles and so the value of J is~ bt 3 /3.
Problems
4.1 A hollow steel shaft with internal diameter one-half the external diameter
has to transmit 750 kW at 240 r.p.m. without the shear stress exceeding
110 N/mm 2 . Find the dimensions of the cross-section, and show that there is a
saving of material of about 22 per cent by making the shaft hollow instead of
solid. What will be the angle of twist per metre length of the hollow shaft?
4.2 A particular material has a relation between shear stress and strain
r= Oy -Ar 2
where G and A are constants. Find the torque required to produce a maximum
shear strain ofG/4A in a solid circular shaft, length l, radiusR. When the torque
is removed and shaft unloads elastically with a modulus G. Find the permanent
angle of twist in the shaft.
4.3 A solid shaft of radius R is twisted through an angle 8 by a torque T. Writing

the fully plastic torque as Tp, the angle of twist at first yield as By and the radius
at which the yield stress is reached as 'AR., plot graphs of T/Tp and X as functions of
8/By.
4.4 A circular steel shaft of length 2/ is rigidly fixed at both ends. One half of the
shaft has a diameter 2d and the other half a diameter of d. Assuming that the
material behaves in an ideal elasto-plastic manner, with a yield stress of T y, find
the value of the torque which, when applied at the mid-point, will cause half the
cross-sectional area of the shaft with the larger diameter to become plastic. Discuss
the residual stress distributions when the torque is removed.
4.5 A torsional assembly consists of a steel rod of radius r and a concentric steel
tube of internal radius 1.5r and wall thickness 0·1r. Both are of the same length
and the rod and tube are rigidly fixed together at one end. The free ends are fixed
together after a clockwise torque T has been applied to the rod. When the torque
is released, find the residual torque in the rod and the tube. An anti-clockwise
torque 4Tis now applied to the assembly. Both materials behave in an ideal elasto-
plastic manner with a yield stress T y. A separate test on the rod indicated that
yield first occurred at a torque of 2T. Assuming that the tube is fully plastic find
the depth of yielding in the rod.
4.6 Derive an expression for the shear stress at any point on the cross-section of
a beam subject to a shear force. Indicate the types of section to which the result
may be applied without serious error. The cross-section of a box beam is square,
the side being 120 mm and the thickness 5 mm. A shear force of 8 kN acts along
a line through the centre of the square and parallel to two of its sides. Determine
the maximum shear stress in the beam.
4.7 Sketch the shear distribution for both the webs and the flanges of the sections
shown in P 4.7, when subject to a vertical shear force.
D p 4.7
4.8 Find the position of the centre of shear for a circular tube of diameter D and
wall thickness t (where D ~ t), when the tube is slit along one edge parallel to the
axis of the tube.
4.9 A beam has an L-shaped cross-section, the sides of which are of equal length
a and constant thickness t (t ~a). A shear force Q acts normally across the beam
parallel to one of the section sides. Show that the maximum shear stress is
1·35 Qjat and sketch the shear distribution.
4.10 A thin-walled tube is to be formed from a particular strip of sheet metal and
welded longitudinally. Compare the torsional rigidity of a circular tube with that
of the best possible rectangular tube. Prove any formulae that are used.
4.11 A bar made from a material with shear modulus G has a square cross-section
of side a. Assuming that under torsion the stress trajectories form concentric
squares, find the torsional stiffness of the bar.
4.12 Determine the torsional rigidity of the tube with the section shown in
P4.12.
P4.12
5 DEFLECTION OF BEAMS
5.1 Introduction
In this chapter we shall consider the deflection of a beam that is subject to trans-
verse loading. The deflection can be thought of as being caused by two different
effects: the bending moment and the shear force. If the beam is relatively long
when compared with its depth, the shear force will only make a small contribution
to the deflection and can be neglected. For the time being we shall only consider
the deflection due to bending.
In section 3.12 it was shown that when bending takes place about a principal
axis coincident with the z axis
Mz E
-=- (5.1)
It is also shown in mathematical books that the equation for the radius of curvature
for a plane curve in the x, y plane is
(5.2)
where v is the displacement in they direction.

In the majority of cases of beams in structures the value of dv/dx is small and
(d2v/dx 2)2 <{I. Thus equation 5.2 can be written as
d2 v
(5.3)
Ry = dx 2
If equation 5.3 is now substituted into equation 5.1 the following relation is
obtained
(5.4)
138
DEFLECTION OF BEAMS 139
This equation gives a relation between the transverse displacement and the bending
moment. Note that a positive bending moment is associated with a positive curva-
ture. From now on we shall confine our attention to bending about the z axis and,
for ease of writing, will drop the subscripts in equation 5 .4.
It was shown in chapter 1 that dM/dx =- Q and that d 2M/dx 2 = p. Thus
2 v)
-d2 ( E dI - =p
dx 2 dx 2
If the beam is of constant cross-section, EI is constant and
d4 v
E!_=p (5.5)
dx 4
EI is called the flexural rigidity of the beam and is sometimes denoted by B.
5.2 Deflection by direct integration

The deflection curves for several straightforward cases can be found by direct
integration of equation 5.4. We shall ignore the dead weight of the beam.
Figure 5.1
A cantilever of uniform section with a point load Pat the end is shown in
figure 5.1. The bending moment atx from the encastre end isM= -P(l- x).
Thus
Integrating once
dv Px 2
EI-= -Pix+- +A
dx 2
Integrating again
-Plx 2 Px 3
E!v =- - + - - +Ax + B
2 6
At the fixed end of the cantilever both the deflection v and the slope dv /dx are
zero, enabling the two integration constants A and B to be found. x = 0, dv/dx = 0
therefore A = 0. x = 0, v = 0 therefore B = 0. Hence
Px2
E!v=--(31-x)
6
At the free end

p[3
v=--
3£!
and
dv P/ 2
dx 2£/
r·-1
The same results would apply for the deflection and slope under the load if it
was applied at some other point I' from the fixed end(/' would be substituted for
I in the two expressions).
p
Figure 5.2
For the case of a cantilever with a uniformly distributed load (figure 5.2), the
bending moment at x from the fixed end is
d 2v (/-x) 2
M=E!dx2=-p 2
dv pl 2x plx 2 px 3
£1-=---+----+A
dx 2 2 6
pl 2x 2 plx 3 px 4
Elv = - - - + - - - - - +Ax + B
4 6 24
When x = 0, dv/dx = 0 therefore A = 0; when x = 0, v = 0 therefore B = 0. Thus
px2
Elv = - - (6/ 2 - 4/x + x 2 )
24
For the particular case x =I
dv p/ 3
dx 6£!
and
p/4
v=---
8£!
Figure 5.3
The simply supported uniform beam with a concentrated load (figure 5.3) will
be somewhat more lengthy to solve. For x <a
d 2 v Pbx
EI dx2 =-~-
dv Pbx 2
EI-=--+A
dx 2/
Pbx 3
Elv =- - +Ax + B
6/
For x >a
d2v Pbx
EI dx 2 = - 1- - P(x- a)
dv Pbx 2 Px 2
EI- =----+Pax+ C
dx 2/ 2
Pbx 3 Px 3 Pax 2
Elv = - - - - - + - - +C:X +D
61 6 2
Four constants A, B, C, and D have to be found. Two end conditions are known-
when x = 0 and/, v = 0. Two other conditions are required-these can be found
from the fact that when x =a, the slope and the deflection must each be the same
whether the expressions for x < a or x > a are used.
It is then perfectly possible to obtain expressions for the deflection curve by
this means; however it could be extremely tedious. Consider the case of a simply
supported beam with three separate concentrated loads applied. This method of
approach will not be pursued any further as there is a very much neater method of
performing the integration.
5.3 Deflections using singularity functions or the Macaulay method

W. H. Macaulay first suggested that singularity functions could be of considerable
use in solving problems on the deflection of beams. Figure 5.4 shows five singu-
larity functions of the set.
(5.6)
The functions have a somewhat unusual behaviour. Dealing first with the two
negative indices, the functions are always zero except when x =a. Here they are
infinite but will integrate as follows
Jx [x- a]-
0
2 dx = [x- a]- 1
-IL
-2
[x-a[
1•-al ' 1•-al01 l_____c,
'----t---x
a a a
(a) (b) (c)
1·-{__L, 1·-{_L, a a
(d) (e)
Figure 5.4
The first of these is a unit doublet function or a unit concentrated moment, while
the second is a unit impulse function or a unit concentrated load.
For the other functions n;;;;. 0. When x <a the value of fn(x) is zero, but if
x >a the value of fn(x) is (x- n)n. The functions will integrate as follows
fo
x [x _ a]n +I
[x -a ]ndx= (n;;;;.O)
n+1
For n = 0 (figure 4.5c), the function is described as a unit step at x =a. For n = 1
(figure 5.4d) the function is a unit ramp starting at x =a.
y
p
tFf Ml
~ a-l b .I
fVs X
I: c
Figure 5.5
We shall now make use of singularity functions to write down an expression

for the loading applied to the simply supported beam (figure 5.5). It will be
remembered that a positive load is in the positive y direction. Concentrated loads
will be considered as figure 5.4b and uniformly distributed loads as step functions
(figure 5.4c).
p (x) = VA [x - 0] - I - P J[x - a] -I +M J[x - b] - 2 - p [x - c] 0 + VB [x - !] -I

This may be integrated to fmd the shear force since dQ/dx =-p
-Q(x) = VA[x- 0] 0 - Pdx- a] 0 + Mdx- b] - 1 - p[x- cP + Va [x -1] 0
Integrating again, remembering that dM/dx =- Q
1 1 o P [x - c 12 1
M=VAx-0]
[ -Pdx-a] +Mdx-b]- +Va[x-1]
2
This expression holds for any positive value of x. To find VA and VB take x just
greater than I, substitute into the expressions for Q and M which will be zero. It
will be seen that the resulting equations are in fact the same as resolving vertically
and taking moments about B. It would, of course, be easier to determine VA by
statics.
If we now substitute into equation 5 .4, two more integrations can be performed
to give both the slope and deflection at any point. Omitting Va since the beam
terminates at x = I.
Eldy = VAx 2 _ Pt[x-aF +Mt[x-b] _p[x-c] 3 +A

dx 2 2 6
, VAx 3 Pdx-a] 3 Mdx-b] 2 p[x-c] 4 A

Ely =-6- - 6 + 2 - 24 + X +B
A and B can be found by the use of suitable end conditions.

The use of singularity functions has been described here as a method for
initially writing down the load applied to a beam. For finding the deflection curve
it is usually much simpler to write down an expression for the bending moment
straight away using singularity functions. Thus for the beam in figure 5.3
d2v Pbx
EI-=--P[x- a] 1
dx 2 I
EI~ = Pbx 2 _ P[x- a] 2 +A

dx 21 2
Pbx 3 P[x- a] 3
E/v = - - - +Ax +B
61 6
When x = 0; v = 0 therefore B = 0. Note [x- a] = 0. When x =I, v = 0 therefore
P(l - a) 3 Pbl 2
A= 61 6/
The complete deflection curve is given by
Pbx 3 P[x- a] 3 P(l- a) 3 x Pbl 2 x
E/v=--- + __ _
61 6 61 61
If the deflection is required under the load, put x =a

Pa 2 b 2
v=----
3EII
When the load is at the centre of the beam the maximum deflection is obtained,
that is a= b = I/2 and
PI3
v=---
48£1
It can be seen that the singularity-function method is simple to apply. An
origin is chosen at one end of the beam and an expression is written down for the
bending moment such that every load on the beam is included. Functions such as
[x - a] are integrated according to the rules set out at the start of this section.
Do not forget that [x- a] cannot be negative . A slight difficulty can arise when a
uniformly distributed load extends over part of a beam, this however will be
illustrated in the following example.
P•¥ 8 (
.f!~
A
v.tf - - t .I. tva

,I, l_j
Figure 5.6
The deflection curve is required for the simply supported beam in figure 5.6.
The reactions can be found by taking moments: VA= 4P/3; V 8 = SP/3.
It should now be possible to write down an expression for the bending moment
when x is measured from the left-hand end. If the uniformly distributed load is
considered, the bending moment is- p [x - 2/] 2 /2. This expression however will
not suffice once x > 3I, as the distributed load terminates when x = 3I. A simple
way of dealing with this is to continue the load to the right-hand end of the beam
and add an equal distributed load acting upwards from B to the free end. The
expression for the bending moment is now written
- p[x-2/] p[x-3I] 2
2
+.:_;--~
2 2
Thus
d 2 v 4Px ,[ I]o 2P[x-2I] 2 2P[x-3I] 2 SP[ ]1
EI-2 =--4Pt1 X- -- +- +-X- 3I
dx 3 I 2 I 2 3
There is no reason to write down the bending moment due to the point load P
since the beam terminates when x = 4I.
EI ?..!!_ = 4Px2 - 4PI[x -I] 1 + 2P [x- 2I] 3 + 2P [x- 3I] 3 + SP [x- 3I] 2 +A
dx 6 I 6 I 6 6
4Px 3 4Pl[x-I] 2 2P[x-2I] 4 2P[x-31] 4 SP
Elv = Ji:- 2 +l 24 +-I 24
3
+ 18 [x - 3I] +Ax+ B
When x = 0; v = 0 therefore B = 0. When x = 31; v = 0 therefore
p[3
0 = 6Pl 3 - 8Pl 3 - -
12 + 3AI ·, A = "36
25 Pl 2
The complete deflection curve has now been determined.

All the deflection problems discussed so far have been statically determinate.
However there is no reason why the same methods should not be used for statically
r
1-----a-----+--b--l
~(lA ~)MB
~~------------~------~8~
~~X +~
Figure 5.7
indeterminate beam problems. A fairly straightforward case is shown in figure 5. 7,

where we shall apply beam-deflection theory to determine the fixing moments at
the ends of the beam.
dv VAx 2 P[x- a] 2
EI- = --- --MAx - +A
dx 2 2
VAx 3 MAx 2 P[x- a] 3

E!v = - - - - - - +Ax +B
6 2 6
When x = 0, dv/dx = 0 and v = 0 therefore A =B = 0. When x =I, dv/dx = 0 there-

fore
When x =I, v = 0 therefore

VAI 3 MAI 2 P(l- a) 3
----- =0
6 2 6
Solving the two equations gives

Pab 2
MA=y
The moment at B can be found by taking momentsM8 =Pa 2 b/1 2 .

Using these integration methods the complete deflection curve for the beam is
obtained. This can be a disadvantage if a point deflection is required, since a sub-
stitution will have to be made into the final equation for a particular value of x.
The method then becomes somewhat tedious and other means exist which will
give a more rapid solution.
5.4 Moment-area methods
VB
I A
.... .::
8
(a ) (b )
Figure 5.8
There are two theorems associated with moment-area methods and it is essential
to have a full understanding of them before attempting an application. Figure 5.8a
shows part of the elastic curve of a deflected beam. We shall be concentrating on
two points, A and B. Now
or
JB
A
M
-dx=
EI
fd2v
dx 2
- ]B =8s - 8A
-dx= [dv
dx A
(5.7)
This is the first theorem which states that the difference in slope between two
points on a beam is equal to the area of the M/EI diagram between the two points.
The M/EI diagram is shown in figure 5.8b and the shaded area is required.
d 2v - - JB -dx
JB-Mx-EIdx- --fB --xdx=
A dx
A 2
[ xdv]B
dx
dv
dx A A
(5.8)
If the origin is now shifted until it is below A
JA
B Mx dx
---=xs8B-
E/
z;B +vA (5.9)
where x = 0 at A . xs8 B is represented by CD in figure 5 .8a, and the complete

expression is equal to the distance AD.
The second moment-area theorem may now be stated as follows. The moment
about A of the M/EI diagram between points A and B will give the deflection of
point A relative to the tangent at point B.
~A
s(
(a)
.I
(b)
Cr-,
I ',
I '-..
(c)
Figure 5.9
We shall first apply the moment-area theorems to find the slope and deflection
under the load for the cantilever {figure 5.9a). The M/EI diagram is drawn in
figure 5.9b. The difference in slope between A and B is given by the area of the
complete M/EI diagram, that is, Pl 2 /2EI. Now the slope at A is zero, and so the
slope at B is Pl 2 /2EI
Taking moments about B for the complete area of the M/EI diagram will give
the deflection at B, as the tangent at A is horizontal.
p[2 p[3
v = 2 /x--=--
B 1 2£/ 3EI
If moments had been taken about A, this would give the deflection at A relative
to the tangent at B, that is, distance AC {figure 5.9c). The distance AD is in fact
required. Now AD= CD- AC, where CD= 18s
pf2 I pf2 pf3

vB = 2EI X I - 3 2EI =3 EI
r.~~-------- 1 I
(a )
~ 8£1
(b)
Figure 5.10
The central deflection is required for the beam with a uniformly distributed
load (figure 5.1 Oa).
As the beam is symmetrically loaded, the slope at the centre of the beam is
zero. Taking moments about A for one half of the M/EI diagram will give the
required deflection. A certain knowledge of the areas of parabolas and the position
of their centroids is an advantage.
2 p/ 2 l 5 l 5 p/ 4
v =--x-x-x-=---
c 3 8£/ 2 8 2 384£/
The applications so far have given rapid resuls. The case of a point load on a
simply supported beam (figure 5.1Ia) will however require more thought. We
shall assume that the deflection under the load is required.
(a)
~ £/I
(b)
Figure 5. 11
The M/EI diagram is drawn in figure 5.11 b and a sketch of the deflected form
of the beam in c. The first moment-area theorem is not much use here as the slope
is not known at any point on the beam.
If moments about Bare taken for the complete M/El diagram the distance v 1
will be obtained. The value of eA can then be obtained by dividing this distance
by I.
Pab 2
v =--xjb+-- Pa 2b ( b+- a) =--(2b+a)
Pab
I 2£/1 2£/f 3 6£/
therefore
Pab
(}A= - ( 2 b +a)
6£!1
Now
Pa 2 b
v2=a(}A=--(2b+a)
6£/l
v 3 can be obtained by taking moments for the portion AC of the M/EI diagram
about C.
Pa 2b a
v =--x-
3 2£!1 3
Pa 2b Pa 2 b 2
vc = v 2 - v 3 = --(2b +a- a)=--
6£!1 3£!1
Had the beam been built in at both ends (figure 5.12a), the moment-area methods
would enable the fixing moments to be determined quite easily. The bending-
moment diagram has been sketched (figure 5.12b ). If the beam is uniform this
can also represent the M/EI diagram.
Applying the first moment-area theorem to the whole of the beam results in
the fact that the area of the M/EI diagram is zero, then
Pab l (MA + Ms)l
-x-- =0
I 2 2
1---a-rb---1
~A C B~
(a)
(b)
Figure 5.12
If the second theorem is applied about end B for the whole of the beam, the
first moment is zero.
Pab
- X-a - +Pab
( b +a) -X-b (2b)
- - M l Xl- - (M - MA)-lX -l = 0
l 2 3 l 23 A 2 B 23
If these two equations are solved it will be found that
Pab 2 Pa 2 b
MA=T MB=p
The deflection at any point can easily be found by applying the second theorem
about that point and taking account of the M/EI diagram between the point and
one end of the beam.
The moment-area theorems lend themselves equally well to deflection problems
where the beam has a variable section, so that the value of I changes. In cases of
this kind it is probably best to sketch both the bending-moment and the M/EI
diagrams.
£1, E/1
(b)
11..
EI (c)
Figure 5.13
The beam in figure 5.13a has a larger cross-section over the central portion of
the beam. Sketches of the bending-moment diagram and the M/EI diagram are
shown in figures 5.13b and c. No difficulties should arise in the application of
the theorems, but the evaluation of the various integrals may be somewhat tedious.
In certain cases either graphical or numerical integration could be used to advantage.
An extension of the moment-area theorems is known as the conjugate-beam
method. With a little thought it can be seen that the slope at any point on a beam
is equivalent to finding the 'shear force' at that point when the beam is loaded
with the M/EI diagram. The deflection at a point is equivalent to finding the
'bending moment' for the same loading. The equivalent beam is called the 'con-
jugate' beam and the M/EI diagram is referred to as the 'elastic load'. Slight com-
plications arise over the support conditions of the conjugate beam. At an internal
support the deflection is zero and hence the 'bending moment' must be zero. This
can be achieved by the insertion of a pin instead of the support. At a built-in end
the deflection is zero. This is achieved by making the corresponding point on the
conjugate beam into a free end. In a similar manner a free end on the beam
becomes a built-in end on the conjugate beam.
5.5 Use of standard cases

Table 5.1
p
__JA
t
~
l -o + c t
~
tp
- • ~3
1 c::-----1•
~~-----.+. f.l2EI
1-f
Table 5.1 gives a list of deflections and some slopes for certain loading on beams.
The results can sometimes be used in conjunction with the principle of super-
position, symmetry and anti-symmetry and Maxwell's reciprocal theorem. We shall
have to anticipate Maxwell's theorem which is proved in the next chapter and is
now quoted in a somewhat restricted form : For a beam of linear elastic material
the vertical deflection at B caused by a unit vertical load at A is equal to the verti-
cal deflection at A caused by a unit vertical load at B.
To illustrate the application of this approach the three problems, figures 5.14
to 5.16, will be discussed.
In figure 5.14a the deflection at the end of the cantilever is required when the
load is applied at B.
Since the portion BC of the cantilever is unloaded it will remain straight and
the deflected form will be as shown in figure 5.14b. The deflection at Cis equal
to the deflection of B plus the deflection of C relative to B.
vc =vB +(I- a)(JB
Pa 3 Pa 2 Pa 2
vc = 3EI + 2EI (! - a) = 6EI ( 3 ! - a)
(a)
(b)
Figure 5.14
In figure 5.15 the deflection is required at B for the point load at C. In this
case we may use the reciprocal theorem and instead of applying the load Pat C
and finding the deflection at B, we apply the load P at B and find the deflection
at C. In fact this has already been done in the previous example. Hence
Pa 2
Vs =- ( 3 / - a)
6EI
In the final example (figure 5.16a) the deflection at the centre of the simply
supported beam is required. The effect of this load can be considered as the sum
of problems in b-symmetric loading, and inc-anti-symmetric loading. The
A B
~
__
I
~a; -1
Figure 5.15
deflection at the centre due to the anti-symmetric loading will be zero. This leaves
us with the problem in b to solve where the slope at the centre will be zero. This
is equivalent to finding the deflection at the end of a cantilever of length 1/2 loaded
as shown in d.
This expression will of course simplify.

It would have been possible to approach this last problem in a slightly different
manner making use of a reciprocal theorem. The problem is equivalent to applying
a central load P to the beam and finding the deflection at a distance a from the
support.
+p c
At +B
~a-i I
+;
(a)
+~
+ (b)
t
+1 t
+ (c)
t;
tf ~
+; (d)
Figure 5.16
5.6 Deflections due to shear

As was mentioned in the introduction to this chapter, there are deflections due to
shear forces as well as deflections due to bending moments. Figure 5.17a shows
~
tLJ
[:)
(a) (b) (c)
Figure 5.17
part of a cantilever of rectangular section that has a point load applied at the free
end, so that the shear force is constant. The shear stress, as was shown in chapter 4,
will vary from zero at the top and bottom edges of the beam to a maximum value
at the neutral axis. This will mean that a rectangular element at the neutral axis
will deform as shown in figure 5.17b. The slope of the top and bottom edges will
be
dv Txy
-=r=-
dx G
However another rectangular element at the top surface will not deform as the
shear stress there is zero. There is only one way of accommodating the deformities
that take place at various distances from the neutral axis, and that is for the section
to warp as shown by the lines that are no longer straight in figure 5.1 7a. So the
plane sections do not remain plane unless the bending moment is constant and
hence the shear force is zero.
Difficulties arise near a concentrated load where there is a sudden change in
shear force (figure 5.17 c). According to the above discussion sections would have
to warp as shown-clearly an impossibility in the vicinity of the load. In fact a
point force is also an impossibility since it would produce an infinite stress. The
strain system we have derived will not be correct near a concentrated load but will
be reasonably accurate for other parts of the beam. Uniformly distributed loads
produce no difficulties since the shear force is gradually varying.
With these reservations, the slope of the elastic curve may be written as
dv Tmax KQ
- = - - =-- {5.10)
dx G AG
where A is the area of the cross-section and K is a constant to convert the average
shear stress to the maximum shear stress for a particular section. For a rectangle
the value of K would be 3/2 while for a universal beam K would be 1 and A would
be the area of the web.
To find the deflection due to shear it is only necessary to integrate equation
5.10. This should not present any difficulties and if necessary singularity functions
can be used.
We shall now find the ratio of shear deflection to bending deflection at the end
of the cantilever with uniformly distributed load shown in figure 5.2. A rectangular
cross-section will be assumed of breadth band depth d.
The bending deflection was found to be v = -p/ 4 /8EI (section 5.2). For shear
Q=-p(l-x)
Thus
dv Kp
-=--(1-x)
dx AG
v =- Kp
AG
(zx - x2
2
) +c
The constant Cis zero as v = 0 when x =0. Thus the shear deflection at the end of
the cantilever is- Kpl 2 /2AG. The ratio of shear to bending deflection is
4KEI
AG/ 2
Now for the rectangular section K= 3/2,1= bd 3 /l2, A= bd alsoE/G = 2{1 + v).
Thus the ratio becomes {1 + v)d2 /1 2.
It can now be seen that the statement at the start of the chapter was correct-
shear deflection could be ignored when compared with bending deflection unless
the span to depth ratio is small.
Problems
5.1 A uniform cantilever of length lhasa uniformly distributed load of p per unit
length and a load Pat the mid·point. Calculate the deflection at the free end. The
free end is to be raised, so that it is at the same height as the built-in end, by a pure
couple applied at the tip. Find the value of M to do this. What will be the slope at
the free end after the couple has been applied?
5.2 A light uniform beam of flexural rigidity EI is simply supported at A and D.

A uniform load of p per unit length is applied to AB and a clockwise couple of
value pl 2 /8 is applied at C. AD= !, AB = l/2, AC = 3!/4.
Find the deflection curve for the beam and the value of the deflection and the
slope at B. A third support is now introduced at B such that it is on the same level
as those at A and D. Find the reactions at the supports when the above loading is
applied.
5.3 A horizontal beam of length!, freely supported at the ends, carries equal
loads Pat distances a and (l- a) from one end. Find the deflection under the
loads and show that when the loads divide the beam into three equal parts, the
deflection under the loads is
s PI 3
----
162 EI
Find the central deflection when loads of P and 3P are applied at third-points of
the beam.
5.4 A fixed-ended beam of length l carries a uniformly distributed load p. Its

cross-section is optimised such that its bending stiffness is proportional to the
magnitude of the bending moment it has to carry. Find the points at wnich the
bending stiffness is zero, assuming that these are not at the ends. Also find the
deflection curve assuming that the end bending stiffness of the beam is EI.
5.5 Find the end deflection of a cantilever oflength l, constant width b, and
height h which increases linearly from zero at the free end to h0 at the built-in end,
when it is subject to a uniformly distributed load of p per unit length.
5.6 A cantilever has an adjustable prop at one end and carries a uniformly distri-
buted load, the maximum allowable bending moment being M Calculate and
compare the load-carrying capacity for the two cases
(a) the prop at the same level as the built-in end
(b) the prop adjusted for maximum carrying capacity.
5. 7 A Universal beam has a web thickness of 10·1 mm and an effective web depth
of 358 mm (Ixx = 26 900 cm4 ). It is simply supported and carries a uniformly dis-
tributed load. Investigate the ratio of shear deflection to total deflection for span
to depth ratios of 10, 15 and 20.
6 VIRTUAL WORK AND ENERGY METHODS
6.1 Introduction
Much of the work covered in the last three chapters has been devoted to the
behaviour of an individual member when subjected to various types of loading.
We shall now investigate framed structures made up of a number of individual
members that may either be pinned together at their ends or rigidly jointed. A
number of ideas and methods will be presented that will enable the deflections of
points on such structures to be found; or the forces may be found for the mem-
bers in statically indeterminate cases. The chapter also provides an introduction
to a later chapter on stiffness and flexibility methods, which are fundamental to
a modern approach in structural analysis and can in fact be applied to all types of
structures, including framed structures, plates, and shells.
6.2 Strain energy and complementary energy

We have already discussed the strain energy that is stored in a stressed member in
section 3.11. Let us again examine the case of a member made from a linear elastic
material and subject to a steadily increasing tensile force F. The load-extension
curve is shown in figure 6.1a.
~ ~~
(a) (b)
Figure 6.1
As the force increases from F to F + dF let the extension increase from e to

e +de. The external work done is (F + dF/2) de~ F de. This will result in strain
energy being stored in the member, represented by dU, and equal to the shaded
157
area on the diagram. U is the total strain energy stored in the member and it will
be represented by the area under the load-extension curve. That is
rei
U= Jo Fde (6.1)
For a member of constant cross-section A, length I and modulus of elasticity

E, e = Fl/AE, and on substituting into equation 6.1 the strain energy stored can be
expressed as
Now examining the shaded area dCin figure 6.la

rF, (6.2)
dC R:: e dF C = Jo e dF
Cis called the complementary energy and is sometimes denoted by U*.

For a linear elastic material C = F 1 e 1 /2, represented by the area above the
load-extension curve. It will be seen that the values of the strain and comple-
mentary energies are equal.
The load-extension curve for a non-elastic member is shown in figure 6.1b.
The same initial arguments will apply such that
fe, fF,
U= Jo Fde C= Jo edF
It will not be possible to integrate these expressions unless the relationship

between the load and extension of the member is known. The strain and comple-
mentary energies will not be equal unless the relation is linear.
In figure 6.Ic the same load-extension curve is shown for the non-elastic mem-
ber, with an initial extension applied, due for example to a temperature increase
before the load is applied. In this case the complementary energy is represented by
the shaded area; the strain energy however is the same as for the member without
the initial extension. We shall find that complementary energy, even though it has
no particular physical meaning, will play an important part in certain energy
theorems that will follow later in this chapter.
Now consider a pin-jointed structure that consists of members all made from
the same linear elastic material. An external force P 1 is applied to a particular
point 1; this will induce forces in the members and strain energy will be stored in
them. The total strain energy stored must be equal to the external work done. As
the members are linearly elastic the deflection of the loaded point and indeed all
other points that are not restrained will increase linearly with the applied load, so
that finally the external work done will be P 1d 1 /2 where d 1 is the deflection of
point 1 in the direction of P1 • Neither is there any need to restrict the structures
to pin-jointed statically determinate cases. The structure can be highly redundant
and have rigid joints. The external work done will still be given by P 1 d 1 /2 and this
must still be equal to the total strain energy stored in the members, which may be
due to axial, bending, or torsional effects.
VIRTUAL WORK AND ENERGY METHODS 159
6.3 Flexibility coefficients
Figure 6.2
The pin-jointed structure in figure 6.2 is constructed from members that are
linearly elastic. This will mean that the deflection in any particular direction at a
particular point will be proportional to the load applied at any other point.
We shall define flexibility coefficients as follows:
! 11 is the displacement at point 1 due to a unit load applied at point 1.
! 21 is the displacement at point 2 due to a unit load applied at point 1.
fn, !12, etc. follow from these definitions.
The first subscript gives the point and direction at which the displacement is to
be measured, the second subscript gives the point of application and the direction
of the applied unit load. It can be seen that it is perfectly possible to have two or
more different subscripts for the same point if the displacement is measured in a
different direction to that in which the load is applied. Or indeed we may require
the deflection at a point for a moment applied at that point.
Let us refer again to the framework. If P 1 was applied by itself the vertical
deflection of point 1 could be written f 11 P!> the vertical deflection of point 2,
f 21 Pb and the deflection of point 3 in the direction of P 3 , fJ 1P 1 . If the three
loads are all applied to the structure, by making use of superposition, the follow-
ing equations for the deflections of points 1, 2, and 3 can be written down.
dt =fuPt + f12P2 + f13P3

d2 = f21P1 + f22P2 +123 P3 (6.3)
d3 =IJtPt + IJ2P2 +[33P3
So if we can find the values of the flexibility coefficients, the displacement
due to the applied loading can be found for any point.
The three equations 6.3 can be written in the following matrix form
(6.4)
or in a more general form
d=FP (6.5)
It is of interest to examine the relation between P 1 and d 1 from equation 6.3
dt p2 p3
-=In + /12 - + ft3 -
Pt Pt Pt
Even though the material of the structure is linearly elastic, the relation
between P 1 and d 1 will not be linear unless the ratios P 2 /P 1 and P 3 /P 1 remain
constant, that is, we have what is called proportional loading, all the loads are
steadily applied in the same proportions such that they reach their maximum
values together. For this type ofloading the relation between the loads and their
corresponding displacements will all be linear and it can be seen at once that the
external work done is given by
(6.6)
The case where the loads are applied in a random manner must be examined in
P,C P,LL P,LL

more detail. To simplify the discussion it will be assumed that P 1 is first applied,
followed by P 2 and then P 3 •
d, d2 d3
Figure 6.3
The graphs of applied load against deflection for the three loads are shown in
figure 6.3. For the case of P 1 there will be a linear portion while P 1 rises to its
maximum value, followed by two horizontal portions while P2 and then P 3 are
applied, where the load P 1 does not change but the corresponding deflection
increases. The work done will be equal to the area under the curve and will clearly
not equal P 1d 1 /2.
If we now concentrate our attention on a particular member of the structure,
as it is made from linearly elastic material the graph of the force in the member
against the extension must be linear whatever the manner of application of the
external forces; so the energy stored in the member will always be the same once
the loading in the structure is complete. This will apply to all the members of the
structure, so that the total strain energy stored must be independent of the loading
sequence. This in turns means that the external work done must also be constant
and independent of the loading sequence. The final deflection of each point on
the structure is also independent of the manner in which the loads are applied.
We have obtained an expression in equation 6.6 for the external work done for
proportional loading, and this must represent the work done for any other type of
loading.
6.4 Maxwell's reciprocal theorem

In order to determine the deflection of the three points in figure 6.2 it would be
necessary to know the values of nine flexibility coefficients, equation 6.3. This
however is not quite true as it can be shown that [ 12 = f2 1 and that the matrix is
in fact symmetric.
Consider the same structure (figure 6.2) with the case of P3 = 0. Let us obtain
an expression for the work done when P 1 and P 2 are applied. It has already been
shown that the work done is independent of the way in which the loads are applied.
P 1 is applied first: the deflection at point 1 can be written as [ 11 P 1 so that the
work done is [ 11 Pf /2. Now apply P2, the deflection of point 2 will be [ 22 P2, and
the work done f2 2 P 2 /2. When P 2 is applied point 1 will deflect through a further
distance of [ 12 P2 and as the load P 1 is displaced through that distance the work
done will be f1 2P 2P 1 • Consideration of the graphs in figure 6.3 will help to make
this point clear.
The total work done or energy stored is
!fuPf + !i22Pi + f12P2P1 (6.7)
The structure is now unloaded and the two loads applied in the reverse order. A
similar argument will give the work done as
lf22Pi + !fuPf + f21P1P2 (6.8)
As the work done is independent of the loading sequence the two expressions
6. 7 and 6.8 must be equal. If the common terms are removed we are left with
f12 = [21
This can be written in a more general form
(6.9)
Expression 6.9 is a form of the Maxwell reciprocal relationship. This states that
for a linear elastic system, the displacement at point i due to a unit load applied at
point j is equal to the displacement at point j due to a unit load at point i. The
loads and the displacements must of course correspond, so that when the displace-
ment is measured at i it is in the same direction as the unit load applied at i. We
made advance use of this theorem in chapter 5.
This proof can also be carried out making use of matrix algebra as follows.
Instead of having two loads on the structure, assume that there are n, so that
the work done or energy stored is given by
W=-!P1d1 +!P2d2 ... +!Pndn (6.10)
Now both P and d are column matrices and as the work done is a vector dot
product we can express equation 6.10 as
W= !PTd
Now d = FP, (6.5), therefore
(6.11)
If we take the tranpose of both sides of equation 6.5

dT = pTFT (6.12)
Now as the work done is a scalar the expression for W could also be written
W= !dTP
Substituting from equation 6.12 for dT
W= !PTFTP (6.13)
Examining the two expressions for W, equations 6.11 and 6.13, it can be seen at
once that
(6.14)
This implies that the flexibility matrix is symmetrical and that
fij = fji
6.5 The flexibility approach

The elastic beam shown in figure 6.4a is statically indeterminate to the first degree.
We require the reactions due to the applied loading. The problem could be solved
by making use of singularity functions. If the reaction at B is assumed to be V 8 ,
the other reactions could be found in terms of V 8 and the applied loading. A
v~ t~
ic
A' '8
(al
+Ff +~
(b)
t +
t t (c l
fv8
Figure 6.4
bending-moment equation can be written down that will apply for any point on
the beam, and this may then be integrated twice to give the deflection equation.
There are three end conditions-the deflection is zero at A, B, and C, so that the
constants of integration may be determined and the value of VB can be found.
An alternative approach would be as follows. The problem is first simplified by
making it statically determinate. There are several different ways of doing this,
any of which would be suitable. However one particular choice can lead to a more
rapid solution than another. The choice is really a matter of experience. In this par-
ticular problem any one of the supports could be removed, or a pin could be
inserted at any point on the beam such that the bending moment at that point
becomes zero. We shall make VB zero by removing the reaction at B (figure 6.4b ).
The vertical deflection at B due to applied loads P 1 and P 2 must be calculated,
and this is a fairly straightforward matter. The deflection could be found for the
two loads applied at the same time or for each load applied separately, and the
individual deflections found at B and then summed. Let the total deflection be d 1.
(b)
(c)
(d)
(e)
Figure 6.5
The applied loading is next removed and a vertical force equal to the value of
the unknown reaction V 8 is applied at B (figure 6.4c). Due to this load point B
will move vertically upwards and this deflection can be calculated in terms of VB.
Let the deflection be d 2 .
In the original problem it is known that the true deflection at B is zero, and
this occurs when Pb P 2 , and VB are all applied together. Superimposing the
deflections due to figures 6.4b and c the resulting deflection is zero.
(6.15)
d 1 is known, and d 2 is known in terms of Vs. Thus the value of Vs can be found
from equation 6.15. This equation relating the deflections is known as an equation
of compatibility or an equation of geometry.
This type of solution is the basis of the flexibility approach, but in order to
gain further insight we shall now discuss the problem in figure 6.5.
It is seen that the structure is three times redundant. Instead of choosing the
three reactions at B, C, and D as the redundants we shall take MA the fixing
moment at A, together with the vertical reactions VB and Vc. If these redundants
are all released the problem is turned into the simply supported beam of figure 6.5b.
The displacements that correspond to the releases have to be calculated with
the externally applied loading in position. Let these displacements be 8AL• dBL•
and deL·
The applied loading is next removed and unit forces are applied in turn to the
points of release (figures 6.5c, d and e). Inc a unit moment is applied at A, and
we can find the corresponding rotation at A together with the vertical deflections
at B and C. As the moment applied is unity, the displacements will be the flexi-
bility coefficients fAA, fsA and fcA· A unit load is applied at Bind and at Cine.
There will be three compatibility equations. In the original problem the rota-
tion at A and the deflections at both B and C are zero.
8A= 8 AL+fAAMA+fAB VB +fAc VC =0

ds =dsL + fsAMA +Iss Vs +fsc Vc = 0 (6.16)
de =deL+ fcAMA + fcsVs + fccVc = 0
or writing this in matrix form
8Al
ds = r8ALl
dsL + rfAA
fsA fAs
Iss fAcl
fsc rMAl
Vs =0 (6.17)
r
de deL leA fcs fcc Vc
The column matrix on the left-hand side represents the displacements that
occur in the original structure at the points of release. In the particular case dis-
cussed this is a null matrix. This might not always be so, and in order to keep the
approach general we shall denote this matrix by dR. The first column on the right-
hand side represents the displacements that occur at the points of release with the
original loading applied to the released structure, and these will be denoted by
dRL· The 3 x 3 matrix is the flexibility matrix F and the last column matrix con-
tains the redundant reactions and is R.
We may now write
(6.18)
Equation 6.18 is the fundamental equation of the flexibility approach when

solving redundant structures.
The quantities that are really required are contained in R, and these can be
found by solving equation 6.18 in the form
R= F- 1 (dR- dRL) (6.19)
where F- 1 is the inverse of the flexibility matrix F.
6.6 Kinematic indeterminacy

When the stiffness approach is used, the unknowns are chosen to be the joint dis-
placements, as opposed to the flexibility method where the unknowns were taken
as the redundant reactions. This will mean that finally we shall need to know the
displacement of every joint in the structure. The number of unknowns is equal to
what is called the kinematic indeterminacy.
B .----------, c
~A B
(a)
i
A D
0
(b)
Figure 6.6
Two examples illustrating how the degree of kinematic indeterminacy is

obtained are now discussed. The first is the beam in figure 6.6a.
End A is built in so that no deflection or rotation can take place there. At B
vertical deflection is prevented, but it is possible for rotation and horizontal dis-
placement to occur. This particular beam would then have a kinematic indeter-
minacy of 2. Very often axial effects can be neglected when compared with
bending, in which case the horizontal movement of B can be omitted, reducing
the kinematic indeterminacy to one.
The second example is the portal frame in figure 6.6b. As the feet of the portal
are both pinned, rotation can take place at both A and D. At B there are three
unknowns, two deflections and a rotation, the same will apply at C. In all, the
unknown displacements total 8. However if axial effects are omitted we can reduce
this by 3 as B and C will not deflect vertically and the horizontal deflection of B
will be the same as that of C. This means that the kinematic indeterminacy is
reduced to 5.
6. 7 The stiffness approach

As has already been mentioned, when the stiffness approach is used the displace-
ments of the joints are taken as the unknowns. For the beam in figure 6. 7a, if
axial effects are neglected the kinematic indeterminacy has been shown to be one.
The unknown is the rotation at B.
The first step in the analysis is to fix end B such that no rotation can take
place. If the external loading is now applied a moment will be developed at B. It
is a fairly straightforward matter to determine the value of this moment; at this
stage we shall denote it by M 8 (figure 6.7b).
The external loading is next removed from the beam and a couple equal and
opposite to M 8 is applied and the joint allowed to rotate. It is possible to calculate
the resulting angle of rotation 8 8 (figure 6.7c).
(a) i
Me
-!' ~
~~-----------'~----------~~
~ ~
(b)
,....
~~----------------~!
Me
(c) T
Figure 6.7
These two results can be superimposed and the resulting moment at B will be
zero, leaving B pinned as it was in the original problem. 8 8 will be the value of the
unknown rotation at B. Once this has been found the values of the reactions can
be determined.
We shall next attempt to obtain a more general approach using the stiffness
method and we shall base the discussion on the beam problem shown in figure 6.8a.
This particular problem has already been discussed using a flexibility approach in
section 6.5.
The beam is seen to be three times kinematically indeterminate if axial effects
are neglected. The displacements 8 8 , 8c, and 8 0 corresponding to the joint rota-
tions at B, C, and D will be taken as the unknowns. Proceeding in a similar manner
as in the previous example, the beam is built-in at B, C, and D. It is necessary to
calculate the values of these restraints when the external loading is applied. Any
applied forces corresponding to the unknown displacements are omitted at this
stage. In this example the moment M applied at D will be left out, leaving the
loads PI> P 2 , and P-3 in position (figure 6.8b).
The moments at B, C, and D for the applied loading have to be found. Thus
M8 y will be caused by P 1 acting on span AB together withP2 on span BC. Three
moments are required, May, Mcy, and MoF·
When the individual terms are calculated for a flexibility matrix, a single unit
load corresponding to one of the releases is applied and the resulting displacements
that correspond to all the releases have to be calculated. These would then form
one column of the matrix. The equivalent to this in the case of a stiffness matrix
is to apply a unit displacement at one of the restraints that have been introduced,
..~ 8
t~ c
(3 ~M
~
~A
i (a) i
tP, ::-.,MaF
t~ ~MCF
t~ ;;;.(OF
~ m ~ ~
(b)
(c)
(e)
Figure 6.8
and determine the 'forces' that are developed at the positions of all the intro-
duced restraints. One column of the stiffness matrix will result. So that in our
problem if a unit rotation is applied at B, with C and D restrained, the moment
required to produce this rotation will be the stiffness coefficient kBB and the
restraining moments will be kcB and knB, where the latter is of course zero. Unit
rotations are next applied to joints C and D (figures 6.8d and e), and the various
stiffness coefficients found.
By the use of superposition we can state that the original forces applied to the
structure corresponding to the unknown displacements, will be equal to the forces
developed in the restrained structure together with the corresponding forces due
to the unit displacements multiplied by the actual displacements. The original
force applied at both Band Cis zero in each case; at D however there is an exter-
nally applied clockwise moment M The following set of equations will result
0 =MBF + kBBeB + kBc8c

O=McF +keBOB +kcc8c +kcn8n (6.20)
M=MnF +knc8c +knn8n
These can be written in matrix form as
[ ~]
M
= r:::)
MoF
+ r~:: ~:~ k~o)
0 koc koo
[::)
Bo
(6.21)
The column matrix on the left-hand side contains the forces applied to the
original structure corresponding to the chosen unknown displacements. We shall
denote this matrix as PR. The first column matrix on the right contains the forces
developed in the restrained structure when the external loading is applied. This
will be PF. The large matrix is the stiffness matrix K. It is to be noted that several
zeros appear in this matrix, whereas in the flexibility matrix this was not so. It
can be seen that it may be easier to determine the stiffness matrix as fewer terms
have to be calculated. The final column matrix is formed from the unknown
displacements.
Equation 6.21 can be written in the more general form
~=~+~ ~~
and solved in the form
(6.23)
where K- 1 is the inverse of the stiffness matrix, which if equation 6.23 is com-
pared with equation 6.18 is seen to be a flexibility matrix.
Having introduced the ideas of the stiffness and flexibility approaches we shall
now discuss the application of work and energy methods to structures.
6.8 The method of real work

This approach has already been discussed in section 6.2 where it was stated that if
a single load is applied to a structure the work done is P 1 d 1 /2 and this is equal to
the energy stored in the structure. So that if the energy stored can be found there
will be no difficulty in the determination of d 1 .
Figure 6.9
The vertical deflection of C is required when the load P is applied to the pin-
jointed truss in figure 6.9. The vertical and horizontal members are oflength I and
the cross-section of all members is A.
To find the energy stored in the members of the truss it will first be necessary
to find the forces in the individual members. It is probably best to set out all
results in tabular form.
Table 6.1
member length area force u

AB A p k
BD l A -P k
BC y2l A y2P 2y2k
ED l A -2P 4k
DC l A -P k
AD y2l A y2P 2y2k
~ (7 + 4y2)k
The strain energy for a particular member is found from F 2 l/2AE and in this
case k = P2 l/2AE. The total strain energy is found by summing the last column
in the table and has a value of
(7 + 4y2) P 2 1
2 AE
This must be equal to the external work done, which is Pd/2 where d is the vertical
deflection of C.
-!Pd = (7 + 4y2) p2f

2 AE
PI
d=(7+4y2)-
AE
The beam problem in figure 6.10 has already been solved in chapter 5 where
the deflection under the load was required. The strain energy stored in this case
Figure 6.10
will be due to bending action (shear effects are neglected), and was shown in
section 3.15 to be given by
f M 2 dx
2£1
If x is measured from support A, M = Pbx/1 where x < a. The energy stored

frpm A to Cis
fo
a p2b2x2 dx = p2b2a3
2Eil 2 6EI1 2
When x > a the expression for the bending moment is more complicated and it
would be much easier to measure x from B. M =Pax/! where x <b. It can be seen
at once that the energy stored from B to Cis p2a2b3 /6Eil 2 therefore
This approach would work equally well if the beam was of variable section; it
would only be necessary to include a suitable expression for I in the integration.
The method of real work is seen to be very simple to apply, but unfortunately
it has severe limitations in the type of problem that it will solve. If a number of
loads are applied to a structure the forces in the members can be found and the
strain energy summed, this will be equal to the external work done, which will
be of the form ~ Pd/2 but it is impossible to find the deflection of an individual
point. Often the deflection is required for a point on a structure at which the load
is not applied or the deflection at a particular point in a different direction to the
load is wanted. Examples would be the vertical deflection of D or the horizontal
deflection of C for the frame in figure 6.9. The method of real work again cannot
help in such cases. It is also limited to cases that are statically determinate.
It is therefore necessary to look for an approach that is far more general in its
character and the next section discusses such a method.
6.9 Method of virtual work

Virtual work has already been used to determine the forces in the members of a
statically determinate framework (section 2.8). It was assumed that there were no
large changes in geometry when the loads were applied to the structure.
We shall now consider the virtual-work method in a slightly different form,
often referred to as Mohr's equation of virtual work, with particular reference to
a pin-jointed framework.
We have already seen in the method of real work that the external work done
is equal to the energy stored in the members; we could also equate the external
work to the work done on the members. If we now consider a member that is in
tension with an axial force F m and the length of the member is changed by a
virtual displacement dem considered as an extension, then the work done on the
member is Fm dem. (Note that this will not be Fm dem/2). The same would
apply if the member was in compression and the member shortened. If however
the member was in tension and the virtual displacement shortened the member,
work would be done by the member and this would be considered as negative in
a work balance equation.
Let us now consider a complete pin-jointed structure under the action of an
external set of loads P; the structure is in equilibrium and the force in a particular
member is Fm. For the present we shall assume that all members of the structure
are inextensible apart from member m, and that a virtual displacement dem in
the form of an extension is applied to this member. As a result of this a number
of the joints in the structure will move to new positions and work will be done
by the applied loading system. Let dLl j be the corresponding displacement of the
load Pj ( that is, the displacement in the direction of Pj)· Equating the external
work done to the work done on the member m
~ Pj dLlj = Fm dem
If all the members of the structure were allowed to extend
~Pj dLlj = ~ Fm dem (6.24)
So that we may state that the virtual work done by the external loads is equal
to the virtual work done on the members.
It will be essential to find a compatible system of deformations (d Llj, dem) so
that the particular information required in a problem can be found easily. Before
embarking on this it should be pointed out that the actual loading system applied
to the structure has been used together with a compatible system of virtual dis-
placements. The principle would apply equally well if actual displacements and an
artificial or virtual set of loads and forces were used; the forces would of course
have to satisfy the equilibrium of the system. If actual loads and forces are used
together with actual displacements we are of course again using the method of
real work.
The derivation of the virtual-work method has been discussed in relation to a
pin-jointed structure, but there is no reason why the general principle should not
be applied and extended to take account of bending moment, shear force, and
torque if these exist in a particular problem.
6.10 Virtual work applied to statically determinate systems

In section 6.9 it was stated that when the equation of virtual work was applied,
either virtual displacements and the actual loads on the structure could be used, or
the actual displacements and an imaginary or virtual set of loads. The latter will
be of more use when dealing with structural problems. The applications that follow
are often referred to as unit load methods.
Figure 6.11 shows a Warren truss with external loads P 1 and P2 applied. The
deflection of joint E in the direction of the arrow is required.
So long as there is no large distortion, the equation of virtual work 5.1 can be
written
~PLl =~Fe (6.25)
The extensions e of all the members due to the externally applied loads P 1 and
P2 can be found. It is first necessary to find the forces in all the members and from
Figure 6.11
a knowledge of the load-extension curve for each member, the individual exten-
sions are found. We shall make use of these extensions together with the actual
displacements on the framework in the equation of virtual work as they are known
to be a compatible set.
It would be most convenient if we could have only the displacement we are
seeking on the left-hand side of equation 6.25. It is quite easy to introduce a set
of hypothetical forces that will achieve this; in fact all that is required is a unit
load applied in the direction of the required displacement at E. This load will pro-
duce forces in the members of the framework, which can be found by resolution
and will be denoted by the set F '. The set F' and the unit load are of course in
equilibrium.
The left-hand side of equation 6.25 now becomes 1 x LIE. The complete
virtual-work equation can now be written
LIE= F~ceAc + F~necn + · · ·
where the right-hand side is summed for all the members. The value of e for a
member is the extension due to the applied loads P 1 and P 2 •
If all the members are uniform and linearly elastic the values of e can be found
from Fl/AE where F is the force due to the applied loads. Thus
L1
E
=""Lt F'Fl
AE
(6.26)
For the more general case of a curved member of variable section, considering
axial effects only, it is necessary to find the extension e of a small element ds
where the axial force is F and the axial force due to the unit load is F'. In this
case
Ll ="" F'F ds (6.27)
L AE
For bending effects, the bending moment at a particular point is determined
with the applied loading in position; the corresponding value of e is the change in
angle of the small element ds and is given by M ds/EI. With the unit load in
position the bending moment M' is determined at the same point. Applying the
virtual-work principle
(6.28)
where the integral extends over a particular member and the summation is for all
the members.
It can be seen that it is possible to extend the above principles to take account
of shear force and torsion.
Shear L1 Lf KQ';~s
= (6.29)
Torsion L1 Lf T'~/s
= (6.30)
The constant K is the one introduced in section 5 .6 and J is the equivalent polar
second moment of area as mentioned in section 4.9.
If a member is subjected to several 'forces' at the same time the virtual-work
equation will be formed from the summation of the various terms that appear on
the right-hand sides of equations 6.27 to 6.30.
c D E
Figure 6.12
As a numerical example of this approach we shall find the vertical deflection of

D for the truss in figure 6.12. All members have the same length of 5 m. The cross-
sectional areas of the top and bottom chord members are all 2500 mm 2 and the
rest of the members 2000 mm 2 . E = 210 kN/mm 2 .
The analysis is best set down in tabular form (table 6.2). The first columns of
the table give details of the members. Column 4 gives the force in the members
due to the applied loading. Column 5 gives the extension of the members. In
column 6 is shown the force in the members due to an applied unit vertical load
at D. Column 7 is the product of columns 5 and 6, and the summation of this
column will give the vertical deflection of D. This has a value of 2·87 mm.
The working in the table could have been somewhat shorter as both the struc-
ture and the applied load are symmetrical and also the deflection at D is required.
Thus the contribution from members in similar positions, for example, AF and GB
will be the same. This means that the results from one half of the framework
could have been doubled and added to those from FG.
Deflections due to temperature change can be treated in a similar manner. The
extension e of each member would be the change in length of the member due to
a temperature rise, and would be of the form !exT where ex is the linear coefficient
Table 6.2
2 3 4 5 6 7
Member Length m Area F e p' F'e

mm 2 kN mm
AC 5 2000 -57·8 -0·69 -0·578 0·399

AF 5 2500 28·9 0·276 0·289 0·08
CF 5 2000 57·8 0·69 0·578 0·399
CD 5 2500 -57·8 -0·552 -0·578 0·319
FD 5 2000 0 0 -0·578 0
FG 5 2500 57·8 0·552 0·867 0·478
DE 5 2500 -57·8 -0·552 -0·578 0·319
DG 5 2000 0 0 -0·578 0
EG 5 2000 57·8 0·69 0·578 0·399
EB 5 2000 -57·8 -0·69 -0·578 0·399
GB 5 2500 28·9 0·276 0·289 0·08
r-a---t·t-· p
--b--J•i
~+~A------~c------------~st~
T lal T
(b)
l?t
I (c l
(d)
+'·---)-+--1- - - - - - - ,
++
(e)
I
(f)
Figure 6.13
of expansion and T the temperature change. Once the extensions have been
obtained the working will follow the same steps.
Virtual-work principles will next be applied to fmd the deflection and slope
underneath the load P for the simply supported beam in figure 6.13a. We have
already solved this problem by alternative means. The bending-moment diagram
M for the applied load system is first drawn. A unit load is next applied at C
(figure 6.13c) and the bending-moment diagramM' drawn. The integral
f (M'M ds/EI) has to be found. In this case we may put x = s.
Measuring x from the left-hand support where 0 < x <a, M' = bxfl;
M=Pbx/1 and
Measuring x from the right-hand support where 0 < x < b, M' = ax/1,
M=Pax/1 and
So that
Since the value of de is positive it will be in the same direction as the applied unit
load namely downwards.
To find the slope underneath the load Pit will be necessary to apply a unit
moment at C (figure 6.13e). The resulting bending-moment diagram is shown at f,
when 0 < x < a,M' = -x/1, M=Pbxfl and
f aPbx 2 Pba 3
o - E/1 = - 3E/1 2
2
when 0 < x < b, M' = xfl, M =Pax/! and
so that
Pab Pab
1 x Oc =- -2 (b 2 - a2 ) =- - (b -a)
3EI1 3Ell
The unit moment was applied in a clockwise direction so that if the expression
for Oc is positive it means that the elastic centre-line of the beam has rotated in a
clockwise direction. It can be seen that if b > l/2, ec is positive and if b < 1/2, Oc
is negative.
As an example of a problem that is subject to both bending and torsion,
consider the case of a steel rod of circular cross-section and radius r bent into a
quadrant of radius R, where R :l!> r. The rod is built in at one end and a load P
applied at the free end normal to the plane of the rod (figure 6.14). The deflection
of the free end is required in the direction of the load.
/
/
/
/
/
.-"
c ..:-::!_\ --------- 8
Figure 6.14
A point D on the rod is subject to both bending and torque such that
M =PR sin 0 and T =PR (I - cos 0).
If a unit load is applied at the free end in the direction of the required
deflection, M' = R sin 0, T' = R (1 -cos 0), so that
d
B
f
= MM' ds
El
f
+ - -ds
El
-
TT'
where ds = R dO
_ PR
-
3
- I[ +2E
Er 4
- --2
1tG 4
(3" )]
So far all the examples discussed have been for structures made from linearly
elastic materials. However this restriction was not applied when the virtual-work
applications were discussed; it was only necessary to be able to determine the
value of e with the applied loading in position.
The simple framework in figure 6.15 has a vertical load of 200 kN applied and
the horizontal deflection that results is required. The relation between the load
and extension for member AC is F = 200e - 20e 2 while that for BC is F = 400e.
For both expressions the load is measured in kN and the extension in mm. For AC
(
e =5- 25- 20
p)112
VIRTUAL WORK AND ENERGY METHODS I77
c
~ 200kN
Figure 6.15
For BC
F
e=-
400
Due to the 200 kN applied load the forces in AC and BC are each 200/y3.
If a unit horizontal load is applied to the right at C, FA.c = I, Fsc =-I,
therefore
It should be noted that it was necessary to solve a quadratic equation to deter-

mine the value of e in terms of the load for member AC. Difficulties of using this
approach would be met if the relation had been a cubic equation or higher.
6.II Energy theorems and their application to statically determinate structures

We shall first derive two energy theorems from basic principles, the first of these
is generally known as Castigliano's theorem Part I .
Figure 6.16
Consider a structure that is in equilibrium under a set of applied loads P. No

restriction is to be placed on the material of the structure. If the applied loads are
increased by dP the points of application of the loads will deflect through d.J say.
The external work done at joint j will be Pi dLij, so that the total external work
done is ~pi dLii, and this must equal the increase in stored energy, that is, strain
energy in the structure, so that
dU= L Pj dLij
The load-displacement curve for joint j is shown in figure 6.16. The increase in
strain energy due to an increase dLii could be written as dU= 3U/3Lij) dLii and the
total change as
dU= 2 au -dLI·
3LI·J 1
Equating the two expressions
(6.31)
This relation is known as Castigliano's theorem Part 1.
B c
I I
I I
pk I I
I I
I I
I I
I I
ol El
A
I:J.k
Figure 6.17
So that if the total strain energy is found for a particular structure and differen-
tiated with respect to the deflection at a particular point, the load is obtained at
that point. If the point is not subject to a load then equation 6.31 can be written
as 3U/3Lii = 0. This relation is sometimes referred to as the First Theorem of
Minimum Strain Energy, and depends on the fact that the strain energy will have
a stationary value if an unloaded joint of a structure is given a small displacement
about its equilibrium position.
The second theorem concerns complementary energy which has already been
discussed for a member subject to axial load in section 6.2.
Consider the same system as for our previous theorem, that is, a structure in
equilibrium with external loads P. If the load at joint j alone increases by dPi the
increase in complementary energy can be written
ac (6.32)
dC= ap. dPi
J
As a result of this increase the deflection of most of the joints in the structure
will increase. Figure 6.17 shows the load-deflection curve for joint k. Note again
that no restriction has been placed on the material of the structure. The portion
AB shows the effect as the initial1oads are applied. When dPj is added the value
of the load Pk remains constant but the deflection increases by dLik giving rise to
the horizontal part of the graph. Thus the value of the complementary energy does
not change. External work will be done however, and this is equal to the area
BCDE.
As a result of the increase dPj there will not be any increase of complementary
energy at any of the joints apart from joint j, where the increase can be written as
Llj dPj.
The two expressions for the complementary energy can now be equated
ac
-ap. dP·J = Ll·J dP·J
J
or
ac (6.33)
= Llj
apj
This is known as the First Theorem of Complementary Energy.
It can be seen that if the complementary energy for a particular system can be
found and differentiated partially with respect to the load at a particular joint,
then the deflection at that joint will be found.
If the structure undergoes a temperature change in addition to the application
of an external loading system, the complementary energy for each of the members
could be similar to that shown in figure 6.1c. Thus applying the complementary
energy theorem to a particular joint in the structure will give the total deflection
at that point due to both the temperature change and the external loads.
Basically there is no reason why either of the theorems should not be applied
to a redundant structure apart from the fact that it is impossible to calculate
either the strain or complementary energy unless the forces are known in the
members.
If we restrict the material of the structure so that it is linearly elastic the strain
energy will be equal to the complementary energy, that is, C = U. It is possible to
write equation 6.33 as
au
-=LI· (6.34)
ap. J
J
These two derived theorems will next be applied to determine deflections of

certain points in different structures.
Let us suppose that the vertical deflection of point F is required in the structure
shown in figure 6.11. It would first be necessary to determine the complementary
energy for each of the members, and the total complementary energy C would
then be known. If the complementary energy theorem is used in the form
acjaP1 = J F the vertical deflection of joint F is obtained.
A slight complication arises if say the deflection of E is required in the direction
of the arrow as there is no applied load at E. This difficulty can be overcome _by
the introduction of a load PE in the required direction. The complementary energy
is found for all the members in terms of the applied loading system and the extra
load at E. This can now be differentiated with respect toPE· The value of PE can
then be set to zero in the resulting expression.
The complementary energy for an axially loaded member is given by Jf• e dF
where F 1 is the force in the member. For a linearly elastic member the energy will
become F 1 etf2 or F[l/2AE. The total complementary energy will be "2:, F 2lj2AE.
F will be a function of PI> P 2 and PE. It is perfectly possible to square the
expression for F before differentiating, but this sometimes leads to a rather cum-
bersome expression. It is often preferable to differentiate in the following manner
a ""' F 2 Z aF z
apE L 2AE = L F apE AE (6 .35 )
The problem already solved by virtual work will be solved using complemen-
tary energy. For convenience the figure has been redrawn in figure 6.18. The ver-
tical deflection of D is required. The calculations are set out in table 6.3. The first
three columns are identical with table 6.2. It will be necessary to introduce a fur-
ther external force Pat D. Column 4 gives the total force in each member due to
Figure 6.18
the 50 kN applied loads and the vertical load P. In the next column the force is
differentiated with respect toP. Column 6 is formed by multiplying together the
values in columns 4 and 5, at the same time Pis set to zero and the result
multiplied by 1/AE.
The sum of column 6 gives the required deflection that is, 2·87 mm.
It can be seen that there is very little difference in the calculations that are
required when virtual work is compared with complementary energy. In one case
a unit load is applied, and in the other a load P which is fmally set to zero.
The correspondence between the two methods can be demonstrated in a more
theoretical manner as follows.
The force in a member due to a unit load was denoted by F', so that if a load
P was applied instead of unity the force in a member would be F'P.
The total force in a member can be expressed as F" + F'P where F" is the force
in the member due to the external loading system, which is of course independent
ofP.
If Cis the complementary energy in the member
ac ac
-=--
aF
ap aF ap
Now F =F" + F'P therefore apjaP =F' and acjaF = e, the extension of the
member, therefore
ac
-=eF
,
ap
Summing for all the members in the framework where C is now the total com-
plementary energy
ac
- = Ll= 2 eF
aP
This expression is identical with the virtual-work method of solution.
Table 6.3
2 3 4 5 6
Member Length Area F aF aF l

m mm 2 kN F-·-
ap ap AE
AC 5 2000 -57·8 -0·578P -0·578 0·399

AF 5 2500 28·9 + 0·289P 0·289 0·08
CF 5 2000 57·8 + 0·578P 0·578 0·399
CD 5 2500 -57·8 -0·578P -0·578 0·319
FD 5 2000 0 -0·578P -0·578 0
FG 5 2500 57·8 + 0·867 p 0·867 0·478
DE 5 2500 -57·8 -0·578P -0·578 0·319
DG 5 2000 0 -0·578P -0·578 0
EG 5 2000 57·8 + 0·578P 0·578 0·399
EB 5 2000 -57·8 -0·578P -0·578 0·399
GB 5 2500 28·9 + 0·289P 0·289 0·08
~ 2·872
Suppose that the vertical deflection of point F had been required. Using com-
plementary energy the fundamental approach would be to replace the 50 kN load
at F with a load P. The calculation could then proceed in the usual manner, and
after the differentiation process the value of P could be set to 50 kN.
The truss is symmetrical, as is the loading. It is possible to replace both the
50 kN loads by P and only treat one half of the frame, doubling up the comple-
mentary energy for this half and adding in that for member FG. A little care has
however to be exercised.
Apply a load Pat F and a load P 1 at G where P = P 1. Consider two symmetri-
cally placed members, say, AF and GB. For a unit vertical load at F let the force
in AF be a and that in GB be b. It follows that for a unit vertical load at G, the
force in AF would be b and that in GB would be a.
Thus for the applied loading

aF
F AF = Pa + P 1b - =a
aP
FGB=P1a+Pb aF =b
ap
Putting P 1 = P
2. FaF
- = P(a + b) 2 (6.36)
aP
If a load Pis placed at F and G
aF
F AF =P(a + b) -ap =a + b
aF
FGB=P(a+b) -=a+b
ap
therefore
2. FaF
- = 2P(a + b) 2
ap
(6.37)
It can be seen, comparing equations 6.36 and 6.37 that if both loads are
replaced by P, then the resulting deflection will be twice the true value.
£[ t 2£/ £1
t-1- 2
I
(a)
I +
2 -----t
:x
tf t tiff
~+
·~ 2
(b)
2
Figure 6.19
Both the complementary energy theorem and the virtual-work method can be
applied to find the deflection of points in statically determinate beams. When
complementary energy is used it is of course necessary to be able to calculate the
complementary energy in the beam. For the case of a linearly elastic material this
will be equal to the strain energy stored, that is, JM 2 ds/2£/. As an application
we shall find the deflection at a quarter-point for a simply supported beam of
changing section with a central load (see figure 6.19a).
It is perfectly possible to introduce a load X, say, at the quarter-point and
proceed in the usual manner. Unfortunately to find the complementary energy it
would be necessary to perform four separate integrations, one for each quarter of
the beam. A certain amount of working could be saved. The value of ac;a X is
finally required, and it is possible to differentiate under the integral sign, that is
ac = f M(3M/3X) dx
ax , EI
this however would still require four slightly simpler integrals. If a little thought
is given to the problem much labour is avoided.
We could first apply the reciprocal theorem, change the problem and find the
deflection at the centre for a load P applied at the quarter-point. If the load at
the quarter-point is now replaced by the sum of two loadings, one symmetrical
the other anti-symmetrical, the anti-symmetrical case will result in no deflection
at the centre, and we are left with the problem in figure 6.19b to solve. This is
symmetrical and it will only be necessary to find the complementary energy for
one half of the beam.
Measuring x from the left-hand side
0< x< ~ M=( p; X )x ~~ = ~

l
-<x <-l M=-+-
Pl Xx 3M x
4 2 8 2 ax
-=-
2
J M3M/3x ax 2 (f114 (P+X)x 2 dx +fz;z(Pl+4Xx)x dx}

4£/ 16£/
EI 0 l/4
There is no point in integrating the terms that have X present since the
expression obtained will give the required deflection when X is set equal to zero,
so that deflection
d = 2 [fl/4 Pxz dx +fl/2 Plx dx] = __!_!__ pz3

o 4£/ 1/4 16£/ 768 EI
The problem of the two pin-jointed members with one member of non-linear
material shown in figure 6.15 can be solved by complementary energy. It is of
course necessary to calculate the complementary energy in the form C = f e dF.
An alternative approach using strain energy and Castigliano Part 1 would be as
follows.
It is assumed that when the 200 kN load is applied C displaces to C' with ver-
tical and horizontal components of displacement dv and dH (figure 6.20). So
that the extension of AC would be y3dv /2 + dH/2 = e 1 say and that of BC
y3dv/2- dH/2 =e 2 say.
Now U= f Fde, then for AC
F= 200e- 20e 2 U= s:I (200e- 20e 2) de
Figure 6.20
for BC
F=400e U= J 0e
2
400ede
The total strain energy is now known in terms of e 1 and e 2 and the substitution
for e 1 and e 2 is made, in terms of dv and dH.
Arithmetic can be saved by the fact that
au au ae
-=---
aL1ae aLl
for AC
for BC
Two statements can be made using Castigliano's theorem

au
-=200kN
au
-=0
adv adH
[ 200 ( ~3 dy + ~H) - 20 ( r]
~3 dy + ~H ~3 + [ 400 ( ~3 dy _d;)] ~3
= 200
or
450dy- 50y3dH- 10y3 ( ~3 dy + d; r
=200
and
or
These equations solve to give dv = 0·53 mm and dH = 0·33 mm. This solution
turns out to be more complicated as far as the arithmetic is concerned when com-
pared to the virtual-work solution.
Before proceeding further we ought to consider the type of approach we have
been using when applying the various methods that have been discussed. When
using virtual work or complementary energy we are in fact using a flexibility
approach. The structure is first assumed to be in equilibrium under the action of
loads P 1-Pn and the required deflection d 1 at a particular point 1, say, is deter-
mined. This could of course be written as
d1 = fuP1 + f12P2 + · · · + f1nPn
The various flexibility coefficients could be determined by virtual work or
complementary energy.
With Castigliano's theorem, the displacements are assumed to be unknowns,
and using compatibility (that is, the members must fit together both before and
after the external loads are applied), the extensions of the various members are
found in terms of the displacements. From a knowledge of the properties of the
members the individual member forces can be found. Finally equilibrium equations
are formed, which can then be solved for the unknown displacements. This is a
stiffness approach.
We shall next discuss a graphical approach for determining deflections of pin-
jointed statically determinate structures.
6.12 Williot-Mohr displacement diagram

All the methods that we have considered so far give the displacement in a particu-
lar direction of one point in the framework for each application of the theorem
used. A graphical approach can be used and this will give the displacement of
every joint in the framework, the results being displayed on one diagram.
Figure 6.21a shows a basic triangular pin-jointed framework with a vertical
load P applied at A. The resulting deflection at A is required.
The forces in all the members can be found, and from a knowledge of the
properties of the members the extension of each member can be determined.
Let these be esc, eAS and eAC; the latter two values will of course be negative
as the members are in compression.
It is theoretically possible to draw the frame in its displaced position when the
external load is applied. As C can only move horizontally its new position can be
fixed; it will be at C' a horizontal distance esc from C. The new position of A
will be at the intersection of the arcs BA and CA; the first will have a length
BA- eAs and centre Band the second a length CA- eAc and centre C. The new
position of the framework is shown in figure 6.2la. In fact the extensions of the
members are small compared with their lengths and it would be completely
impossible to attain any accuracy by this method.
ob c
a•
(a) (b)
Figure 6.21
Williot decided on a graphical method of construction that overcame these

difficulties. The displacement of one end of a member relative to the other end is
considered to consist of a displacement in the direction of the member equal toe,
the extension due to the force in the member, together with a displacement due
to the rotation of the member; since the value of e is small compared to the
length of the member it is sufficiently accurate to take the component due to the
rotation as being perpendicular to the original direction of the member. The dis-
placement diagram consists only of extension and rotation elements.
A fixed point o is chosen which will also represent any fixed points in the
framework; in the given example B is a fixed point (see figure 6. 21 b). Relative to
B, C moves horizontally through a distance esc; this is drawn as be on the Williot
diagram to a suitable scale, possibly 10 or 20 times full size. Relative to B, A
shortens by eAB represented by ba'. The direction of the rotation component is
known, perpendicular to AB, but its length is unknown; this direction is shown
by a dotted line. The process is repeated considering the displacement of A rela-
tive to C, this consists of a length ca" = eAC and a rotation term. The intersection
of the two dotted lines is at the point a. oa will give the displacement of A. The
vertical and horizontal components can be measured from the diagram.
A certain amount of care has to be exercised when drawing a Williot diagram.
Usually the drawing 'grows' rather rapidly and it is best to make a rough sketch
first, enabling a suitable scale to be chosen such that the diagram does not become
too large for the size of paper used.
The pin-jointed truss in figure 6.22a is pinned to the wall at A and Band has a
vertical load applied at E. The forces in the members would normally have to be
found and then the extension of each member calculated. In this example the
extensions are given, and are entered on the line drawing.
There are two fixed points A and B and we can start the Williot diagram from
these (figure 6.22b ). The sequence after fixing the pole, which will also represent
A and B, is c, d then e. The only point to note is that since member DC has zero
force, and hence no extension, there will only be a rotation component.
It is not always possible to start drawing a Williot diagram unless an assumption
is made-the assumption may very well not be true and it will be necessary to
correct the diagram when it is completed. A typical assumption in figure 6.23a
might be that the displacement ofF relative to A is horizontal and has no rotation .
With this assumption a horizontal line representing eAF can be drawn, and the
rest of the Williot diagram will then follow. When the diagram is complete it will
be found that B has both horizontal and vertical components relative to A. Aver-
tical component is of course impossible as the support at B only allows horizontal
movement. The frame in its distorted position is shown on figure 6.23b, and it
can be seen that it is necessary to rotate the distorted frame about A until B' lies
on the horizontal through A.
We shall first consider the effect of rotating the original framework about o
through a small angle dO. The displacement of A will be Ao x dO in a direction
perpendicular to Ao. This is represented by oa on the displacement diagram
figure 6.23c. The displacements of all the other points in the framework can be
r
oab
I
I
I
I
1', I
I '
I ',
' ', I
I ' I
~-------~'-\
/
/
/
-1 D -1 /
,,
/
,,
/
(a)
,/
, /
/
,/
~e
(b)
Figure 6.22
drawn. The resulting figure abcdef is similar to the original framework but rotated
through 90°. This diagram is known as the Mohr rotation diagram.
Returning to figure 6.23b the correct displacements could be found by rotating
the displaced diagram about A through an angle dO which would be given by the
vertical displacement of B divided by the length AB. An alternative would be to
rotate the original frame through the same angle dO in the opposite direction. The
true displacements would then be given by the displacements of corresponding
points measured relative to the distorted framework. This latter method is the one
that is used in conjunction with the Williot-Mohr diagram.
The framework in figure 6.23a has the extensions of the members shown on the
diagram. The Williot diagram figure 6.23d has been drawn on the assumption that
member AF only extends and does not rotate.
It can be seen that point B has a large vertical displacement. To correct this the
framework will have to be rotated through an angle bf/ AB about A. The Mohr
(c)
(d)
Figure 6.23
diagram is next superimposed on the Williot diagram. This involves drawing a

diagram similar to the original framework on the base ab'. The point b' is easily
located since horizontal movement is only possible at B, so that b' must lie on the
horizontal through b and it also has to lie on the perpendicular through a. The
Mohr diagram is ab'c'd'e'f'.
The displacements of all the points can now be found. The true displacement
of point D for example is given by dd'.
6.13 Virtual work applied to a statically indeterminate truss

The problem in figure 6.24a will first be discussed. The cantilever AB was originally
supposed to rest on the spring of flexibility [, the top of which is at the same level
as the support at A; unfortunately the cantilever is slightly bent and as a result
there is a clearance of X. A uniformly distributed load p is now applied, end B
VIRTUAL WORK AND ENERGY METHODS IB9
comes into contact with the spring and compresses it. The force in the spring is
required.
The required force will be assumed to be R. The problem can be broken down
into a series of steps. If the spring is removed the deflection at B due to the load p
will be p/ 4 /BEl downwards. The forceR in the spring acts on the end of the canti-
lever (figure 6.24b) and by itself would cause a deflection upwards. The forceR in
the spring will cause a compression of [R. We should now be able to make use of
compatibility and determine a relation between the various displacements.
~r ~1------------,+R
+R
1
(a) (b)
Figure 6.24
The downwards deflection of B on the cantilever is

p/4 R/3
BEl 3El
which must be equal to the initial clearance plus the compression of the spring. So
that
This solution makes use of the flexibility approach. However it was not neces-
sary to use virtual work since all the deflections involved could be written down .
The problem of finding the forces in the truss in figure 6.25 will however not be
quite so simple.
All the vertical and horizontal members are of length 2m; all members have a
cross-section of I 000 mm 2 ; E = 210 kN/mm 2 .
Figure 6.25
It will be seen at once that the truss is statically indeterminate to the first
degree. Any member can be chosen as the redundant apart from BC or DC, which
are excluded since the forces in them are known by resolution. BE is chosen as
the redundant and it is assumed that the force is tensile and has a value of X.
If we imagine BE to be cut at each end, in order to maintain equilibrium,
forces equal to X must be applied at joints Band E tending to pull them together;
while member BE is also subject to tensile forces X. The first step in the analysis
will be to find the forces Fin all the members due to the applied load and the
forces X acting at B and E. This can be done in two separate parts. First the forces
are found due to the 50 kN load. Next unit forces are applied at Band E in the
same direction as the forces X. A further resolution will give the forces F' in the
members due to the unit load. The total force FT can now be found in any mem-
ber by superposition. The result from the first resolution is added to X times the
result from the second resolution.
The virtual-work method used in section 6.10 to calculate the deflection of a
point on a truss can next be applied. The relative displacement of B to E can be
found from the expression
Ll =""
L....,
F'FT!
AE
where FT = F + F'X. The summation includes all the members apart from BE.
The extension of BE is given by Xl/AE.
Now the movement of B relative to E must have the same value but be of
opposite sign to the extension of BE, hence
"" F'FT! = _ XI (6.38)

L...., AE AE
The same result may be obtained by summing the virtual work terms for all the
members including BE and equating the result to zero. This is the way in which we
shall proceed.
The calculation is best set down in tabular form and this is shown in table 6.4.
The column headings are all given and further explanation is not required apart
from the fact that k = AE/1 where l = 2 m.
From equation 6.38 it can be seen that the sum of the elements in column 8
will be zero. Therefore
lOOk(I + ) 2 ) +XkG +2y2)=0
X= -39-4 kN
Once the force in BE is known, the forces in all the members can be found.
The approach here has been flexibility. The structure was first made statically
determinate, and displacements were then found due to both the external loads
and the redundant force X. Finally compatibility is applied, in that the relative
approach of B and E is equal to minus the extension of member BE.
<
l....,"
-
Table 6.4 c::
>
t""'
2 3 4 s 6 7 8
~
l"
Member Length Area 50kN Unit Total force Extension (S) X (7) :;.:
CB -./21 A S0-./2 0 50-./2 lOOk 0

z>
0
CD l A -50 0 -SO -50k 0 t!j
BD l A -50 -11v'2 -(SO+ X/-./2) -(SO+ X/V2)k (50/-./2 + X/2)k zt!j

BA l A 50 -1/-\/2 SO -Xfy2 (SO- X/V2)k (- SO/V2 + X/2)k l"
Cl
DE l A -100 -1/-\/2 -(100 + X/-./2) -(100 + X/V2)k (I OO/y2 + X/2)k -<
DA -./.21 A S0-./2 I S0-./2 +X (100 +-./2X)k (100 +-./2X)k :s::
t!j
BE y2l A 0 I X -./2Xk -./2Xk
Ed
0
0
rn
\0
-
If the external loading had been accompanied by a temperature rise of T, the
forces in the members would be altered. It would not however be difficult to take
account of the temperature rise. In fact all that would be altered in the analysis
would be the value of e, which means that the values entered in column 7 of table
6.4 would have to be changed and they would now have the form
FTI
-+exiT
AE
At the same time it is easy to take account of any lack of fit. We shall suppose
that the truss is being constructed, and that member BE is the last to be placed in
position and that it is unfortunately too short by an amount A. It is however
forced into position by straining. The force X in the member due to both the
lack of fit and the external loads can be found from the final equation of com-
patibility, which would state that the relative movement of B and E together with
the extension of member BE must be equal to the initial lack of fit.
~ F'FTI =A ( 6 .3 9)
~ AE
where the summation of the left-hand side of equation 6.39 includes member BE.
We shall solve a numerical example on the same frame a little later on, but
making use of complementary energy.
It should be noted that we can use the virtual-work method of approach for
members made from non-linear materials, it is only necessary to have a knowledge
of the load-extension characteristic for each member.
6.14 Engesser's theorem of compatibility

It is first necessary to establish this theorem and we shall make use of the redun-
dant pin-jointed truss in figure 6.26a. When the truss was built there was an initial
lack of fit of member AC, it was too short by an amount A. The member was
forced into position and the total force F AC in member AC is required.
Figure 6.26
The complementary energy C 1 can be found for all the members in the truss
omitting member AC for the time being. This will arise from two causes, the
externalloadingP 1 and P 2 and the force FAc applied at C (figure 6.26b).
Making use of the first theorem of complementary energy
ac1 _ Ll
aFAc - c
where Lie is the displacement of joint C in the direction ofFAC.

Next consider member AC. The force in the member is FAC and the comple-
mentary energy CAc for the member can be found.
Applying the first theorem of complementary energy to this member alone
acAc
--=eAc
aFAc
where eAC is the extension of member AC due to the force F AC.
Now the displacement of joint C and the extension of AC must together be
equal to the original lack of fit of AC
Llc+eAc=;\
or
ac. + acAc =A
aFAc aFAc
Instead of finding the values of C 1 and CAC separately the total complementary
energy could be found
and
ac
--=;\
aFAc
or in a more general form
ac (6.40)
ap
-=;\
This is known as Engesser's theorem of compatibility.

If member AC had been a perfect fit in the first place before the loadingP1 ,
P2 was applied, the value of;\ would be zero; Engesser's theorem could then be
written
ac
-=0 (6.41)
ap
No restriction has been placed on the properties of the load-extension curves
of the members. If however the material is linearly elastic the complementary and
strain energies will be equal. The statements aujaP =;\,and aujaP = 0 for no lack
of fit, are sometimes referred to as Castigliano's theorem of compatibility.
Both virtual work and Engesser's theorem will give an identical final equation
to solve, that is, the compatibility equation is the same. In most cases the analysis
is also very similar.
If we consider the same problem again (figure 6.25), the complementary energy
in the members can be found from L FjljAE. However as the expression is to be
differentiated with respect to X, this is best carried out under the summation sign,
in the form
If this is done for one member, say BD

aFT __ 1
FT =-(so+~) ax v2
In other words the values so obtained are of course identical to column 8 in

table 6.4.
There is really no point in repeating the analysis so we shall consider the same
framework with an initial lack of fit.
It will be assumed that when the truss was constructed, member BE was the
last member to be put in position and that it was too short, in this case 1·2 mm.
Let us find the value of the load that must be applied at C in a vertical direction
so that member BE may be fitted. Also, when the member has been fitted the
load at C is removed. What is the value of the residual force in BE?
Let the force required at C be P and the force in the member BE be X. The
first part of the problem can be solved by making use of the first theorem of
complementary energy, and in this case C = U The value of Pis required such
that there will be a movement of B relative to E of 1.2 mm. We shall introduce a
load X in this direction and finally set it to zero.
The forces in the members are found in terms of P and X in table 6.5 column 4.
The complementary energy can be found in the form
F]l
C= 22AE
If this summation is carried out for all the members apart from BE and the
result differentiated with respect to X, we can obtain the movement of B relative
to E with P applied at C, and X applied at B in the direction BE. However for the
first part of the problem X is zero; so that once the differentiation has been
carried out X is set to zero and the result equated to the initial lack of fit. It is
very much easier to carry out the work in the form
This has been set out in the table, columns 5 and 6.1/AE has been put equal
to k. Note that member BE has been included in the table as it will be required in
the second part of the analysis.
Table 6.5
2 3 4 5 6
aFT aFT l
Member Length Area --
ax
FT F
Tax AE
CB y2/ A y2P 0 0
CD l A -P 0 0
BD l A -(P + Xjy2) -1/y2 (P/y2 + X/2)k
BA A P- X/y2 -1/y2 (-P/y2 + X/2)k
DE l A -(2P+ X/y2) -1/y2 (2P/y2 + X/2)k
DA y2l A y2P+X 1 (2P+ y2X)k
BE y2l A X 1 y2Xk
Column 6 can be summed and X put equal to zero

2P(l + 1/y2)k = 1·2 X 10- 3
k = 2/(1000 x 210) m/kN
therefore
P= 36·8 kN
For the second part of the problem Engesser's theorem of compatibility can
be applied with an initial lack of fit. In this case the frame is redundant and the
force in the redundant member BE is X while the value of Pis zero. There is no
need to carry out any further resolution, etc., as the existing column 6 in the
table can be used
aFT I
~ FT--=X
P=o ax AE
where X= 1·2 mm.
It is of course necessary this time to take account of the complementary energy
stored in member BE. The summation of column 6 with P set to zero will give the
required result.
X(i + 2y2)k = 1·2 X 10- 3
X= 29·2kN
It is interesting to observe the relatively high load that arises in the member BE
for a small initial lack of fit.
6.15 Trusses with several redundants

The methods that have been discussed so far will be equally applicable to a truss
that has more than one redundant member. The truss in figure 6.27 has three
redundant members; choosing these as BH, CJ, and DK, the forces in these mem-
B c 0 E
Figure 6.27
bers can be denoted by X 1, X 2 and X 3 • The complementary energy can be found

in terms of the external loading and these unknown loads. If there is no initial
lack of fit Engesser's theorem can be applied in turn to each redundant member.
Thus
(6.42)
Three simultaneous equations will result and these can be solved for X b X 2
and X 3 . It can be seen that the method of approach will become rather tedious as
the number of redundants increases.
So far we have only considered the cases of trusses with redundant members.
Redundant reactions will however not create any special difficulty.
B c 0
Figure 6.28
Figure 6.28 shows a truss with two redundant members and one redundant
reaction. Choosing BG and CH as the redundant members with forces X 1 and X 2
and letting the vertical reaction at G be X 3 , the forces in all the members can be
found. Engesser's theorem can be applied in the usual manner to BG and CH, to
give two simultaneous equations in terms of the three unknowns. The truss rests
on the support at G, and assuming that this support does not sink or move when
the loads are applied, the first theorem of complementary energy can be applied
setting the value of L1 to zero, or
aFT I
2FT ax3 AE = 0
and hence a third equation will result.
The virtual-work method can of course be applied to the same type of problem.
For figure 6.27 the redundant members are imagined to be cut, to make the
truss statically determinate. The forces in all the members can now be found in
terms of the external loads. Unit loads will next be applied in the direction of
each cut member and the forces found in the rest of the members of the truss.
The total force in a member will be of the form
F+F~X1 +F~X2 +F~X3 =FT
where F~. F~ and F~ represent the force in the member due to unit loads being
applied in the directions of BH, CJ and DK respectively. The extensions of the
members can next be calculated and a similar analysis to that used in section 6.4
can be applied.
Thus for BH say
(6.43)
Two similar equations will result for CJ and DK. It should be emphasised that the
equations will be identical to those obtained by applying Engesser's method.
The truss in figure 6.28 can be solved in a similar manner. Two equations
result from the redundant members and the third from the statement that the
deflection of the support at G is zero. We should perhaps examine equation 6.43
in a little more detail. We shall consider the terms individually.
Once the frame has been made statically determinate the first term ~ FF~ 1/AE
gives the deflection of B relative to H. The second term ~ (F~ ) 2 X 11/AE gives the
deflection of B relative to H with the external loads removed and X 1 only applied.
The third term gives the same deflection with only X 2 applied, and the fourth
term with only X 3 applied. The complete equation states that the sum of all
these terms is zero. We have met this kind of equation before in section 6_5
when discussing flexibility methods. If we compare equation 6.43 with one of
those in equation 6.16 the form is identical. We shall rewrite the equation as
"" FF~ l +X ""(F~ ) 2 1+X "" F~F~1 +X ""F~F~l = O (6.44)

LAE 1 LAE 2 L AE 3 L AE
The summation terms with the Xs outside are flexibility coefficients and the three
equations could be written down as a flexibility matrix.
[~ FF~ (1/AE)]
~FF;(Z/AE)
~FF~(l/AE)
~F~F~(l/AE)
~ (F~) 2 (l/AE) (6.45)
~F~F~(l/AE)
6.16 The trussed beam

It is not proposed to discuss in detail the application of virtual work and energy to
redundant beams, as other methods already described often give a rapid solution in
straightforward cases. Suffice it to say that if complementary energy is used,
forces X, Y, etc. are chosen as the redundants, which could be moments or
reactions. The moment is written down in terms of the applied loading and the
redundant forces. For an elastic beam the complementary energy will be given by
f M 2 dx/2EI. If the supports have not deflected we may make use of the first
theorem of complementary energy and state that acjaX= acjaY= 0. The result-
ing equations are solved to find the redundants.
It is interesting however to consider the case of a trussed beam which is really
a composite structure. It consists of a simple beam that is strengthened by the
addition of bracing members.
Figure 6.29
A typical example of a trussed beam is shown in figure 6.29. The beam will be
subject to both bending and axial force . The stiffening structure is pin-jointed and
so the members are only loaded axially.
The beam is formed from two 254 mm x 76 mm steel channels 7 m long. For
each channel fzz = 3367 cm 4 ; A = 3603 mm 2 . The bracing members are made of
steel, and pinned at A, C, and B between the two channels that are placed back
to back. The cross-section of the tension members is 760 mm 2 and that of the
compression member 1600 mm 2 . The force in the member DC is required when
the beam carries a uniformly distributed load of 20 kN/m.
There are really only two basic approaches to this problem. It can be treated as
one structure or it can be split up into a beam and a truss. We shall start by treat-
ing it as one structure and will use an energy approach.
The problem is singly redundant and CD will be regarded as the redundant
member. Let the compressive force in CD be X.
FAD = FBo = 1·27 X -tensile
The axial force in the beam is 1·17X- compressive.
The vertical reactions at A and B will each be 70 kN. When the bending
moment is calculated for any point on the beam allowance must be made for the
vertical component of the force in AD or BD.
Measuring x from the left-hand side
M =(7o-X)x - 20x2
X 2 2
O< x < 3·5 m.
The complementary energy is of the form
I M 2 dx
2EI
+""'L... F 2l
2AE
where the first term refers to the bending moment in the beam and the summation
in the second term extends to all the members including the beam.
Using Engesser's theorem acjaX = 0
ac
ax
= JM(aM;ax) dx + ~ F aF _z_ = 0
EI ax AE
2 [fJo
3 5
" (20x 3
4
+ Xx 2 _ 70x 2 )dxx 10 8 ]
4 2 6734£
1·17 2 xXx7xl0 6 [1·27 2 xXx3-8x10 6 ] Xx1·5x10 6

+ +2 + =0
7206£ 760£ 1600£
X= 74·3 kN
The alternative process is to split the problem up as shown in figure 6.30 into
a simply supported beam and a truss. There is an unknown reaction X between
them at C. This will also be the force in the member CD. The connecting fact or
Figure 6.30
compatibility in this case is that at C the deflection of the truss and the deflection
of the beam must be the same. The first theorem of complementary energy could
be used to find the deflection of the truss at C. The deflection of the beam can
easily be found from the list of standard cases in table 5.1. The calculation will be
quite straightforward and will include a number of the terms from the previous
solution and it will not be repeated here.
6.1 7 Virtual work and energy methods applied to frames

In this section we shall confine our attention to cases that only have a small num-
ber of redundants. Effects due to axial loads will be omitted as in general they
will be small when compared with bending effects.
The problem of the portal with pinned feet in figure 6.31 will be solved using
energy and virtual work. It is also perfectly possible to make use of moment area
theorems, but this is left as an exercise for the reader.
El Constant
H y
:A. /(
A B
Figure 6.31
In all frames the junction between two or more members will be regarded as
rigid, so that the joint as a whole can rotate but there is no relative rotation of
the members with respect to each other. This particular problem is only statically
indeterminate to one degree, as an equation of condition gives rise to the fact that
VA= Vs =plj2andHA =Hs =H
Making use of energy
It will be necessary to obtain an expression for the bending moment at any point
on the portal. Choosing an origin at A and adopting the convention that a positive
bending moment produces tension on the inside edge of the frame
aM
MA-C =Ms-D = -Hy -=-y
aH
pix px 2 aM
Me o = - - - - H h -=-h
- 2 2 aH
U=c=JM2dx
2E1
Since the supports are pinned acjaH = 0 ; or
f M(aMjaH)
EI
dx = 0
The integration is of course carried out along the length of a member.
2 f
o
Hy 2 dy
h
EI o
1
2
f
+ ( - -plhx phx 2
2
) dx
- + - - + H h 2 -=0
EI
2Hh 3 pl 3 h pl 3 h
-- - - +-- +Hlh 2 =0
3 4 6
p/2
H=-....!.._-- {6.46)
4h(2h + 3/)
Once the value of His known the portal has become statically determinate.
Making use ofvirtuai work
The method developed in section 6.10 for finding the deflection of a point on a
beam can be applied
Ll=JM'Mdx
EI
Let us modify the problem by making the support at A into a roller. The hori-
zontal force is now zero, and CD will be the only member with a bending moment
present (figure 6.32a).
M _pix px 2
C-D- 2 --2-
( a)
B -
I
(b)
Figure 6.32
A unit load is next applied in the direction of the horizontal reaction with all
the applied loading removed (figure 6.32b ).
M;,__c =M~-D = -y M~-D = -h
If the applied load is replaced and a force H instead of the unit load is applied
pix px 2
MA-C =Ms-D = -Hy Mc-D =2 - l - Hh
In the original problem the horizontal deflection at A is zero.
M'Mdx =O
J
EI
i
or
2 f
o
Hy 2 dy + 1 ( - -plhx
h
£/ o 2
phx-
- +-
2
2
+Rhz) -dx =0
EI
This equation is of course identical with that obtained by using an energy

approach.
As a further example of the application of the unit-load method to a problem
where there is more than one redundancy, we shall consider the same portal frame
but with both feet built-in. This will introduce a further unknown, namely the
moment at the feet (figure 6.33a).
---
I
Ia) Ib) lc l
Figure 6.33
From conditions of symmetry HA =Hs =H; VA= Vs =plj2;MA =Ms.
Choosing the two redundants asHA and MA, the bending moment at any point
on the frame could be written in the form
M=Ms +M:H+M~MA
where Ms is the bending moment at the point for the statically determinate frame ,
that is, both HA and MA are equal to zero. Mi is the bending moment with no
external load and with M A = 0, H = 1. M~ is the bending moment with no external
load, withHA = O,MA = 1.
The loading systems for determining Mi and M~ are shown in figure 6.33b and
c. The bending moments are set out in table 6.6.
Table 6.6
Member Ms u; M~
AC, BD 0 - y
pix px2
CD - h
2 2
The horizontal deflection at A is zero in the original problem
(6.47)
The rotation at A is zero. There is of coutse no reason why the unit-load tech-
nique should not apply to rotations as well as deflections.
1
El
f (Ms + M 1I H + M2MA)M2
I I
dx = 0 (6.48)
Equations 6.4 7 and 6.48 can of course be written as a flexibility matrix as

follows
f MsM{El dx f M;M~ dx
EI
H
f f
+ =0
MsM~ dx (M~) 2 dx
EI EI
Inserting the values from table 6.6
Hj=O
JlM
r;-hdx+2Jt-ydyll
J0 12 dx + 2 J0 12 dx A
[3
.!!._- (hl + h 2 )H + (l + 2h)MA =0
12
Whence
p/3 p/3
M - H=--'---
A-12(2/+h) 4h(2/+h)
It should be emphasised that this is not necessarily the best method of approach
for solving this problem. In actual fact we shall find later that there is a very rapid
solution if the method of moment distribution is used.
If point loads are applied to the frame there is no basic change in the approach,
using energy or the unit-load methods. All that it will mean is that the expressions
for the bending moments in the released structure could be somewhat more
complicated.
6.18 Ring and arch problems

Consider a ring loaded by two equal and opposite forces acting at the ends of a
diameter. The dimensions of the cross-section are small compared with the radius
R of the ring. Axial effects will be neglected in comparison with bending. The
maximum bending moment in the ring is required (figure 6.34a).
If the ring is cut through as shown at b, forces must be applied to keep the
two halves of the ring in equilibrium. The applied force P must be reacted by P/2
at each cut, and there will be bending momentsM 1 . There also is the possibility
that shear forces H can exist. However if both halves of the ring are examined it
will be seen that the system of forces shown is not compatible. Each half is identi-
cally loaded apart from H which acts in opposite directions on each half of the
ring. This is of course impossible and so the value of H must be zero.
(a) (b)
Figure 6.34
Consider the upper half of the ring (figure 6.34c). The bending moment at an
angle e is
Me =M 1 -
PR
2(1- cos e); O<e <2
TC
C= U=f M 2 ds = fM 2 Rd8
2EI 2EI
There will be no need to calculate the value of the energy for more than one
quarter of the ring.
There must be no rotation of the ring at the cut. Applying the first theorem
of complementary energy
ac
--=0
aMl
or
JM aMt
aM R d8 =O
El
aM
-=1
aMl
therefore
_!_J"''
EI 0
2
[M 1 - P
2
(l- cos e)] R de= o
M=PR(+-;)
1
So that
(1 1)
PR
M=PR - - - --(l-cos8)
2 7t 2
8=0
8 = 7t/2
PR
M=--
1t
The bending moment changes sign and there will be some value of 8 where the
moment is zero. The maximum bending moment is -PR/7t.
The problem will be taken a stage further to find the change in diameter at
right angles to the line of action of P (see figure 6.35).
Figure 6.35
A force H will have to be introduced in the direction of the required displace-

ment; H will finally be set to zero.
PR
M=M1- 2(1- cos 8) -HR sin 8
A value of M 1 has already been found for the case where H was zero. The
question arises whether or not this value of M 1 changes by introducing H. Most
certainly M 1 is a function of H, as we should find if we used above expression for
M and determined M 1 from acjaM 1 = 0. But asH is set equal to zero at a later
stage of the calculation there is no reason why H should not be made zero when
finding the value of M 1 • This means that the value of M 1 obtained from the first
part of the calculation can be used.
M1 =PR(1-;)
Now
~=ac =f1t' 2MaM Rd8. aM=-Rsin 8

2 aH 0 aH EI ' aH
therefore
E;L1 = tn/ 2
[PR(1-; )- ~R (1- cos B) -HR sin B] (-R sin B)R dB
At this stage H may be put equal to zero
EI ~2 =Jrr/ 2
PR(cos B- __!_)( -R sin B)R dB
2
=PR (_!_- __!_)
3
0 1t: 4
1t:
PR 3
0·138-L1 =
EI
The basic approach to a two-pin arch is very similar to that of a pin-ended
portal frame (section 6.17). We shall use the method of virtual work.
If the arch is released at the right-hand abutment so that horizontal movement
can take place, that is, H = 0, the bending moment Ms can be written down for
Figure 6.36
any point on the arch due to the external loads. The bending moment M for a unit
horizontal load applied at B with the external loading removed can also be found.
The horizontal movement at B due to the external loading is
If a horizontal load His applied at B to the unloaded arch the horizontal

movement of B is
Hf(M')2 ds
EI
In the original problem no movement is allowed at B. By superposition
f ----"---
M M ds - H J(M')2 ds = 0
5
EI
1
EI
If the origin of the coordinate system is set up at A (figure 6.36), M' = y and
f MsY ds _ Hfy 2 ds = 0
EI El
therefore
H = _f_M_s_Y_d-'s/_E_1
fy 2 ds/EI
The evaluation of these integrals is not always easy. An arch consisting of a

segment of a circle with constant EI does not present difficulties.
Another straightforward case is that of an arch of parabolic form but with a
secant variation of I. If Ic is the value of the second moment of area at the crown
and(} is the slope of the arch at the point considered I= Ic sec e, ds = dx sec e,
therefore
ds dx
-=-
EI Eic
It might seem rather arbitrary to choose a secant variation. Arches in practice

do not in general conform to this variation in I, but the value of H obtained by
this much easier calculation is often good enough for a first analysis.
H~~~ H
'1~ 1 - + - - - - - - - --------+-1 Fl ¥
Figure 6.37
As an example we shall fmd the value of the horizontal thrust for a pinned
arch of parabolic form with span I and rise h when a point load is applied at a
distance a from the left-hand abutment (figure 6.37). I has a secant variation.
(/- a)x
M s =P1 I O<x <a
a
Ms = P 1 1 (/ - X) a<x <l
J
ds
My--
s EI-
fa P1 -
(/-a)
1- xy dx
E'lc
+
a
f'
P 1 1 (/- x)y dx
0 a Eic
For the parabolic arch

4hx
y=y(l-x)
Substituting for y and integrating
f MsY ds
EI
=~ (al4- 2a3z2 +a4l)
3l 3 Elc
8 h2t
f y 2 ds=--
15 Elc
H = 3.__ p 1 (a - 2a3 + a4 )
8 h 12 t3
In this problem the position of P 1 is a variable and the expression that we have
derived will give the influence line for H if P 1 is set equal to unity.
The case that we derived in detail presented no difficulties in integration.
Cases can arise where the integration cannot be carried out directly and a numeri-
cal approach must be used.
While discussing the two-pin arch we ought to consider two other effects: one
is the effect of axial forces, often referred to as rib shortening, and the other is
the effect of a temperature change.
Consider first the effect of an axial force of value F where the slope of the
arch is e.
An element of the arch of length ds will shorten by an amount F ds/AE.
The change in length dx is
F ds cos e H ds
----=--
AE AE
if shear forces are neglected.
Therefore the change in length of the arch in a horizontal direction is
f Hds/AE.
The basic equation for a two-pin arch if rib shortening is taken into account
will become
J M 8 yds -H fy2
ds =Hf~
EI EI AE
The effect of rib shortening will only affect the value of H for the case of a flat
arch, and can usually be omitted if the rise to span ratio is greater than 10.
The temperature change in an arch can be treated in a similar manner to a
lack-of-fit problem.
If the arch is released at one end the horizontal movement due to a temperature
change T would be
J~ aTdx=ail
where a is the coefficient of linear expansion.
A force H would now have to be applied to the end of the arch to reduce this
movement to zero, that is
If the arch has external loading and a temperature change takes place the total
value of H can be found by superposition.
A number of arches are built-in at the abutments, and can be treated in a
similar manner to the two-pin arch. The concept is no more difficult, but the
degree of redundancy is higher and the arithmetic becomes tedious. For the two-
pin arch we can use the fact that the horizontal displacement is zero; in addition
for the built-in arch the rotation at an abutment is zero. There are special
methods such as column analogy and elastic centre that help to reduce the
arithmetic, but they are beyond the scope of this present work.
6.19 Redundant trusses using the stiffness approach
The redundant structures that have been discussed so far have been solved using a
flexibility approach together with virtual work or energy. We shall conclude this
chapter by discussing a problem that can be solved more easily if a stiffness
approach is used.
~
A F
Figure 6.38
The truss in figure 6.38 is redundant to the fourth degree, but it has only two
unknown displacements, the vertical and horizontal displacement of G, so that it
is kinematically indeterminate to the second degree. This means that on a count
of unknowns alone it would probably be better to use stiffness rather than
flexibility.
The inclination to the horizontal of a typical member of length lis denoted by
a. Let the vertical and horizontal displacements of G bed 1 and d 2 respectively.
Using geometric relations, that is, applying compatibility, the extensions of all
the members can be calculated in terms of d 1 and d 2 . Since we shall proceed using
a virtual-work method, it is probably best to calculate the extensions first of all,
in terms of a unit vertical displacement of G and then in terms of a unit displace-
ment to the right of G.
For member AG
extension due to unit vertical displacement = 1 sin a
extension due to unit horizontal displacement = 1 cos a
If now instead of unit displacements d 1 and d 2 are applied together
total extension of AG = d 1 sin a+ d 2 cos ex
The force in AG can next be found from the load-extension curve F =f(e). Thus
F=f(d 1 sina+d 2 cosa)
and in the case of a linear elastic system

AE
F= (d 1 sinex+d 2 cosex)-1-
The forces in all the members of the truss will be of a similar form
If the horizontal equilibrium of G is considered by resolving the forces in all
the members horizontally and vertically, two simultaneous equations will result in
terms of d 1 and d 2 • These can be solved and if the forces are required in the mem-
bers, the values of d 1 and d 2 can be substituted into the expression for F. The
virtual-work approach differs slightly from this and will be adopted since it is
more systematic and will apply to more complicated problems.
Virtual-work equations can be written down in terms of the actual forces in
the truss and a virtual compatible set of displacements. Thus the correspondence
to the unit-load method will be to use a unit displacement.
Consider first a unit vertical displacement, the extensions due to this have
already been found. Applying virtual work
P1 x 1= 2: Fsinex
For a linear elastic system

. 2 . AE
P1 = 2: (d 1 sm ex+ d 2 sm ex cos ex) - 1
For a unit horizontal displacement

AE
P2 =2: (d 1 sin ex cos ex + d 2 cos 2 ex) - 1-
These equations are solved ford 1 and d 2 • The forces in all the members can be
found.
The various summation terms on the right-hand side of the equilibrium
equations will be stiffness coefficients if d 1 and d 2 are omitted. The equilibrium
equations written in the form of a stiffness matrix are as follows
p1 2: sin 2 ex -AE
1- 2: sin ex cos ex -AE
1- d1
=
p2 2: sin ex cos ex -AE 2
AE
1- 2: cos ex- d2
I
An alternative approach which would give an identical stiffness matrix, would

be to make use of Castigliano's theorem part 1, which was derived in section 6.11.
The theorem can be written in the form oU/o!:i =P.
If we consider the truss that was shown in figure 6.38; under the action of the
applied loads the vertical and horizontal displacements of G are d 1 and d 2 respect-
ively. The extensions of all the members meeting at G can be found in terms of d 1
andd 2
From a knowledge of the load-extension characteristics of the members the
strain energy of all the members can be found.
If the members are linearly elastic
U= 2 1 A£2
--- e
2 l
21
= -AE
- (d 1 sino: +d 2
2 l
cos o:) 2
oU/od 1 =P1 , oU/od2 =P2 , therefore
""AE
~-1 - (d 1 sino: coso:+ d 2 cos 2 o:) = P 2
These equations are identical to the ones obtained previously by the virtual-work
approach.
Problems
6.1 For the pin-jointed frame shown at P 6.1 the cross-section of all tension
members is A and that of the compression members 2A. The length of AB is l.
Using the equation of real work show that the vertical deflection of B is approxi-
mately 1·8Pl/AE. Find also the horizontal deflection of D.
c
p 6.1
6.2 All the members of the mild steel pin-jointed framework P 6.2 have a length
of 5 m and a cross-section of 5000 mm2. Find the deflection of the joint E when a
40 kN load is applied as shown.
p 6.2
6.3 The pin-jointed truss at P 6.3 has all members equal in length and cross-
section. Determine the reaction at C.
p 6.3
6.4 A thin uniform rod forms a circular arc subtending a right angle at its centre
0 (P 6.4). The rod is rigidly fixed at A, and is loaded at B by a force P acting in
the plane AOB and making an angle <P with BO. Find the value of <P which will
make the deflection of B in the direction of P a maximum.
B
P6.4
6.5 A ring of radius R is made from a uniform rod whose diameter is small com-
pared with R. The ring is subjected to three equally spaced forces P whose line of
action lie in the plane of the ring and pass through its centre. Show that if only
bending effects are considered that the greatest bending moment in the ring is
p~(;- J3)
6.6 The pin-jointed framework P 6.6 has member BD vertical. AD and CD are
the same length and inclined to BD at 30°. The relation between the extension
and the load in members AD and CD is e =aF and for BD e =bF 2 , where a and b
are constants and ajb = 3/2. Find the force in BD when a vertical load Pis applied
at D.
p 6.6
6.7 A circular hoop, made of material whose weight per unit length is w, is sus-
pended from the end of a vertical wire. The dimensions of the cross-section are
small compared with the diameter d of the hoop. Find the bending moment at
the point of suspension.
6.8 At P 6.8 is shown a structure in which two bars AD and EG are pinned to
rigid supports at A, D, E and G. A wire EBFCG is fixed atE and passes over
frictionless pulleys at B, F and C to G where it is again anchored. The wire is
10 mm too short before being fitted. Find the tension in the wire when a vertical
load of 100 kN acts at F, also find the deflection of F. The second moment of
area of the bars is 4000 cm 4 , the cross-sectional area of the wire is 60 mm 2, the
length I is 1 m. E = 210 kN/mm2.
6.9 The plane frame shown at P 6.9 consists of eight identical members pinned to
each other. Calculate the forces in the members AB and BC due to the load P by
(a) considering compatibility of displacements between C and B, and Band A, and
(b) by the use of an energy theorem.
~G F
p 6.9
E D
6.10 An open coiled spring with pitch angle a has N complete turns of mean
radius R. The wire is of diameter d, where d ~ R, Young's modulus E, modulus
of rigidity G. Derive an expression for the twist induced by a torque T if the
spring is prevented from compressing or extending.
6.11 A portal frame has a beam of length I and vertical legs of length 1/2 pinned
at the feet. A point load Pis applied at a point distant 1/3 from one end of the
beam. If EI is constant find the reactions at the feet.
6.12 A semi-circular slender arch of radius R and uniform flexural rigidity EI has
fixed supports and carries a vertical load Pat its highest point. Calculate the
reactions at the supports. Determine the deflection corresponding toP.
7 MOMENT DISTRIBUTION AND
SLOPE DEFLECTION
Both moment-distribution and slope-deflection methods can be applied to similar

types of problems, namely continuous beams and rigid jointed frameworks.
Moment distribution has an arithmetical solution in which a number of successive
corrections are applied to an initial set of assumed moments. In the slope-deflection
method, equations are written down for the equilibrium of all the joints in the
structure; when these equations are solved simultaneously, the moments acting at
all the joints can be found.
7.1 Moment distribution

The moment-distribution method was first introduced by Hardy Cross in 1930;
it is essentially a displacement or stiffness approach. Deformations due to axial
loads and shear forces will neglected in comparison with those due to bending.
The general concept of the method is as follows.
Imagine a framework with all external loads removed and clamps available at
all joints so that any joint can be clamped to prevent rotation or left free to
rotate. With the clamps fixed and no loading, the bending moments at the joints
will all be zero if the dead load of the structure is not considered. If the external
loading is now applied to the framework, moments will be developed at each end
of the members that are loaded. It is a fairly straighforward job to calculate the
values of these moments. The clamp at one joint is now released. If the sum of
the moments acting at the joint is not zero, there will be an out of balance
moment. This will cause the joint to rotate until equilibrium has been attained.
The rotation will cause moments to be developed at the remote ends of all the
members meeting at the joint. This joint is then clamped, and another joint
released. The process is repeated for all the joints in the structure and for several
cycles until every joint is very nearly in equilibrium and the clamps may be left in
the released positions.
When a joint is out of balance we shall need to know what proportion of the
out of balance moment is carried by the individual members meeting at the joint.
If the joint rotates through an angle 8 to reach the equilibrium position, the
moments will be distributed among the members according to their stiffnesses.
The moment that is developed at the far end of a member due to a rotation of
the near end, will be some proportion of the moment at the near end. The ratio
of the two moments is called the carry-over factor.
214
MOMENT DISTRIBUTION AND SLOPE DEFLECTION 215
7.2 Sign convention

It is best to adopt a new sign convention when using moment distribution.
Terminal moments will be considered as the moments which act on the ends of
the members and not those acting on the joint. All quantities will be considered
positive if they are clockwise, whether they be moments, rotations, or the dis-
placement of one end of a member relative to the other.
(a)
(b)
Figure 7.1
Figure 7.1a shows a beam that is built in at both ends. MBA is positive while
MAB is negative. In 7.1 b rotations have taken place at ends A and B. eA is positive
and eB is negative. A positive displacement is shown at c when the right-hand
support has moved downwards relative to the left-hand support.
7.3 Stiffness and carry-over factor

The rotational stiffness or the moment per unit rotation of a member that is
pinned at end A and fixed at end B will be written as s AB· Thus sAB is the
moment required at end A to produce a unit rotation there. If the member has a
uniform cross-section or a cross-section that has a symmetrical variation, then
SAB = SBA·
When a moment is applied at the pinned end A, a moment will be developed at
the fixed end B. cAB is defined as the carry-over factor, that is, the proportion of
the moment at end A that is developed at end B. Thus for unit rotation at end A
the moment required there will be sAB and the moment developed at end B will
be CABSAB·
Figure 7 .2a shows a straight uniform member of length I pinned at end A and
fixed at end B. A moment sAB is applied at end A, which will cause the member
to deflect (b) and a moment cABs AB is developed at end B. The bending-moment
diagram is shown in c, the sign convention adopted for this diagram being that
B~
(a)
(b)
·~~
L---------------~
:I ~
(c)
-ve CAB SAB
Figure 7.2
sagging bending moment is positive. As the beam is uniform the value of EI will
be constant, so that c can also represent the M/EI diagram.
We can now make use of the moment-area theorems. The tangent at B is hori-
zontal and the rotation at A is unity, so that the area of the diagram between A
and B is unity.
(7.1)
The deflection of end A relative to the tangent at B is zero.
1 [I2(SAB + CABSAB) I
£/ 3- CABSABI X 2I] = 0 (7.2)
Hence (1 + cAs)/3::: cAB or cAB= 1/2. Substituting for cAB in equation 7.1.
or
4EI
SAB = - - (7.3)
1
So that the stiffness of a uniform member is 4£1/1 and the carry-over factor is
1/2. For non-uniform members it is quite possible to use a similar approach to
find the values of stiffness and carry-over factor.
7.4 Distibution factor

Figure 7.3 shows four members that meet at joint A, the far ends of all the mem-
bers being ftxed. A positive moment is applied at joint A and we need to know
what proportion of Miscarried by each member. If the members are rigidly
jointed together at A there will be no relative rotation between the members at
the joint. Under the action of M the joint A will rotate through an angle 8 say.
Figure 7.3
The moment developed by member AB will be sA 8 8

the moment developed by member AC will be sAc8, etc.
so that
and
MAB SAB
--=--=
4EAahaflAB
M ~s ~ 4EI/l
This quantity is defined as the distribution factor for member AB.

In general E will be constant for all the members meeting at a joint, therefore
DF _ haflAB (7.4)
. ·AB- ~ Jj/
7.5 Fixed-end moments

These are the moments that will be initially developed at the ends of the members
when the external loading is applied and all joints are in the clamped position.
An arbitrary loading system is applied to member AB with both ends flxed
(ftgure 7.4a). The bending-moment diagram is shown at b; this is ftrst drawn
for the beam in the simply supported condition and the negative portion is added
for the ftxed-end moments.
Let the area of the bending-moment diagram for the simply supported beam
be A 1. x A is the distance from end A to the centroid G of the diagram and Xs the
distance from end B.
(a)
(b)
Figure 7.4
The moment of the M/EI diagram about A is zero, and assuming EI is constant
/2 /2
Ax A =MBA 2 + (MAB- MBA) 6
or
6AxA
- 1-2 - = 2MBA +MAB (7.5)
Similarly the moment of the M/EI diagram about B is zero, hence

6AxB
- 1-2- = 2MAB +MBA (7.6)
Solving equations 7.5 and 7.6

2A 2A
MAB = f (2xB - XA) MBA= f (2xA - XB)
These two expressions will give the value of the fixed-end moments for any loading
on a straight uniform beam. There are however several standard cases that are
often required and these will be derived below. It is important to note that if
both of the values of MAB and MBA are positive when derived from the above
expressions, the implication is that MAB is anti-clockwise and MBA is clockwise.
This would mean that M AB would have a negative value when the usual moment-
distribution sign convention is used.
Consider a built-in member with a uniformly distributed load p (figure 7 .Sa).
The bending-moment diagram for a simply supported beam would be parabolic
with a maximum height at the centre ofp/ 2/8.
2 p/ 3 I
A=-- xA =xB =-
3 8 2
therefore
Note for moment-distribution sign convention
p/2 p/2
MAB=-- MBA=- (7.7)
12 12
For the same member with a concentrated load (figure 7.Sb ), the bending-moment
diagram would be triangular with a maximum height of Pab/l under the load.
tr~ (a)
~a --·~1 ~----- b ----~~~

» A
I 1 --------~1
Bl
(b)
Figure 7.5
A =Pab/2, xA =(2a + b)/3;x 8 =(a+ 2b)/3, thus

2xA - x 8 =a 2x 8 -xA=b
Hence
(7.8)
adopting the new sign convention.
j -C d-----1
~A ~
ldxl- ~I
I
X
Figure 7.6
The expressions for the fixed-end moments due to a concentrated load are
extremely useful for deriving further fixed-end moments in cases where the
loading system can be represented as a function of x, the distance along the beam
from a support. As an example consider the case of a short uniformly distributed
load (figure 7.6). For an element of the load oflength dx
(/- x)2
MAB = -p dx x t2
So that for the total load
MAB = - Jc d+c
px
(l-x) 2
l
2 dx
and
MaA = f d+c
c
px 2 - -
(1- x)
z2 d x (7.9)
These expressions can be evaluated quite easily for a specific case.
7.6 Examples
As a simple example, the bending moments at all the supports are required for
the beam in figure 7.7a. Both AC and CB have the same values of EI.
cr 4m
-tSOkN
~~Sm .t. (a)

10m ~~
D.F. ~O
.!fl. 0~
3 3
(b)
F.E.M. 0 0-72 48
DIST.
~ 48 2 4 - - - - - - -
C.O. 24 ~12
SUM u ~~ ~
Figure 7.7
It is necessary to calculate the values of the ftxed-end moments.

AC is unloaded, therefore MAC =MeA = 0
-50 X 4 X 62
Mea= 102 -72kNm
50 X 4 2 6
=48 kNm
X
Mac= 102
Next the distribution factors at joint C have to be found. Stiffness of CA =

4EI/5 and of CB = 4EI/IO, therefore
1/5 2
D.F. for CA = 115 + 1110 =3
and D.F. for CB = 1/3. The values of the distribution factors have been entered on
a line drawing of the beam in figure 7. 7b. Next the values of the ftxed-end
moments are entered at their appropriate points underneath the beam. It is now
possible to start the distribution process. As the joints A and B are rigid they will
never be unlocked. Examining joint C it can be seen that if it was unlocked it
would rotate, since the sum of the moments meeting there is not zero. On unlock-
ing this joint it will rotate until the sum of its moments is zero, i.e. a moment of
72 kNm will have to be distributed at the joint, in the ratio 2/3 to CA and 1/3 to
CB, that is, 48 kNm to CA and 24 kNm to CB. These values are entered on the
diagram. The distributed moments will cause further moments to be developed at
the far ends of the members, the carry-over factor in each case being 1/2. Thus
the carry-over moment at A is 24 kNm and at B is 12 kNm. These values are
The distribution is now complete for this simple case, since the moments
balance at the intermediate support. The final moments are found by summing
the values in the diagram, thus
MAc= 24kNm MeA =48kNm
McB = -48 kNm MBc = 60 kNm
Next consider the same example as before but with a pin joint at B (figure 7.8).
Assuming for the time being that joint B is fixed, the fixed-end moments will be
the same as before. The distribution factors are the same except for joint B.
The new factor is 1, whereas the old one was 0. The reason for having a D.F. of 1
is that each time joint B is unlocked a balancing moment equal and opposite to
the out of balance moment has to be applied.
Starting the distribution at joint B, obviously the moment of 48 kNm is incor-
rect, since the true support there should be pinned, and the resultant moment
should be zero. This is achieved by adding -48 kNm at B and will cause a carry-
c
l SOkN B
~~ J2 +J 1 1t
F.E.M. 0 0 -72 48
OIST. 0 0 0 -48
c.o. 0 0 -24 0
DIST. 0 64 32 0
c.o. 32 0 0 16
DIST. 0 0 0 -16
c.o. 0 0 -8 0
DIST. 0 5·3 2·7 0
c.o. 2·6 0 0 1·4
DIST. 0 0 0 -1·4
c.o. 0 0 -0·7 0
DIST. 0 0·4 0·2 0
c.o. 0·2 0 0 0·1
DIST. 0 0 0 -0·1
SUM 34·8 69·7 -69·8 0
Figure 7.8
over moment of- 24 kNm at C. The joint B is now clamped and the one at C
undamped. The total out of balance moment is 96 kNm and this is distributed
as 64 kNm to CA and 32 kNm to CB. These moments will in turn cause carry-
over moments of 32 and 16 kNm at A and B. Thus joint B is again out of balance,
and the distribution is continued until the residuals are small enough to be neg-
lected. The final moments are MAc= 34·8 kNm, MeA= 69·7 kNm; McB =
-69·8 kNm; and MBc = 0.
It can be seen that the convergence in this solution has been very slow. If a
pinned end has to be treated in this manner the working becomes tedious. Fortu-
nately it is possible to modify the stiffness factor and obtain a rapid convergence.
However before considering this, an alternative method of setting out the calcu-
lation is shown in figure 7.9.
32 46
2-s -46
0·2 16
34·6 ::..!§_
1-4
~
0·1
.;;.Q;l
0
Figure 7.9
The distribution factors are entered in boxes on a line drawing of the beam.
The ltxed-end moments are written down in their respective places and distri-
bution is commenced. So that there shall not be any confusion over which figures
to use, a line is drawn under the balancing moments as soon as they have been
7. 7 Modified stiffness
Pinned end
The member AB in figure 7.10 is pinned at end B. Consider the following sequence
of operations. A unit rotation is applied to A with B clamped. A is then clamped
and B released. The moments resulting from this operation are shown on the
diagram. Note that in the fmal state end A has a unit rotation applied and end B
has no moment.
A 8
SAS
Figure 7.10
Thus if the far end of a member is pinned, the stiffness becomes
SAB(l - CBACAB).
For a uniform member sAB = sBA = 4EI/l, cAB= cBA = 1/2. Hence the modi-
fied stiffness is 3EI/l.
An alternative method of stating this is to say that the stiffness is reduced to a
factor of 3/4 of the normal stiffness.
Symmetry
The member AB (figure 7.1la), is symmetrically loaded. This implies that if the
member is uniform then the fmal moments at each end of the member must be
the same but of opposite sign. Thus 8A= -8B·
A B
~
(a)
~, ~
~cABs:~
5AB
(b)
,,~
-CBA5 BA -SBA
(c l
Figure 7.11
Consider the following sequence of operations. Fix end B and apply a unit
rotation at end A, and the moments shown at b will result. Fix end A and apply
a unit negative rotation at end B, and the moments shown at c will result. Sum
the moments that are developed. Note that in the final state end A has a unit
positive rotation and end B has a unit negative rotation, thus the total moment
must be equal to the stiffness of the member.
MAB = SAB- CBASBA
MBA= cABsAB- sBA
Again for a uniform member sAB = sBA = 4E/fl, cAB= c 8 A = 1/2. Hence
2EI
MAB =-1-
and
2EI
MBA = - -1-
Thus the modified stiffness is 2EI/l, and is reduced to a factor of half the normal
stiffness.
Antisymmetry
For this type ofloadingMAB =MBA and for the uniform beam eA = eB·
Apply a unit positive rotation at end A with end B fixed (figure 7 .12b ). Next
apply a unit positive rotation to end B with end A fixed (c). The moments for each
(a)
~. ~
-~ CAB 5 AB
SAB (b)
(c)
Figure 7.12
case are shown in the diagram and these values can be summed at each joint. Note
that in the final state that both A and B have a unit positive rotation, thus the
final moments must be equal to the stiffness.
MAB = SAB + CBASBA
MBA= CABSAB + SBA
Substituting the usual values for s and c will give
6EI
MAB =MBA=-~-
Thus the modified stiffness is 6EI/l, and the stiffness has increased to a factor of
3/2 of the normal stiffness.
Let us reconsider the problem shown in figure 7.8 and this time make use of a
modified stiffness to take account of the pin end at B. The values of the fixed-end
moments will remain unaltered. There will however be new distribution factors at
C. Stiffness of CA = 4EI/5 and of CB = 3EI/l0, therefore
4/5 8
D.F.forCA=4/5+3/10 11
and the D.F. for CB is 3/11.

These values, and those of the fixed-end moments are entered on the line
drawing (figure 7.13). The distribution should be started at joint B, the pinned
end. To reduce the moment there to zero a moment of -48 kNm is required,
which results in a carry-over moment of- 24 kNm to CB. Once the moment at
B
~----'c~[5l
34·9 69·8 -72 48
m m -24 ~
..l2.1. 0
-69·8
Figure 7.13
the pinned end has been reduced to zero no further moments will be developed
at this end if a modified stiffness has been used. In other words there is no carry-
over moment from CB to BC when joint C is balanced. To complete the analysis
of this problem only one further distribution is required, namely at joint C. This
will result in a carry-over moment to A but not to C. The distribution is shown
completed in figure 7.13. This agrees very well with the previous result. A com-
parison of the arithmetic required in the two distributions (figures 7.9 and 7 .13),
shows clearly the advantage of using a modified stiffness.
~3mf 40 kN 40kN f3m -~
li!!fJ/#M, ~
I I + I + 2! t I +I I
I- 2m 6m -----J 12m 6m f-2m~
A B C (o) 0 E F
.iQ. -90 90 - 240

40 ~ 25 50
- 40 li -190
190
(b )
Figure 7.14
The beam in figure 7.14a carries a uniformly distributed load of 20 kN/m in

addition to the two point loads. It can be seen at once that span CD is symmetri-
cally loaded, and thus only half the problem need be treated if a modified value
of stiffness is used for CD. In addition to this the beam is pinned at Band E (even
though there is an overhang), so a reduced stiffness of 3EI/l can be used at C for
CB.
The fixed-end moments are first calculated.
20 X 22
MaA = 2 =40kNm
20 X 6 2 40
""62
X 3 )3
Mac=- ( - 1-2 - + = -90 kNm
Mea =90kNm
20 X 12 2
McD =- 12 = -240 kNm
Distribution factors
CB = 31/6 = i CD= ~
3!/6+2x2!/12 5 5
The distribution factor for BA must be zero since the end moment BA is 40 kNm
and must remain at this value during the distribution.
The distribution is shown in figure 7.14b and is started at joint B. The final
momentMsA is 40 kNm, thus the final value of Msc is -40 kNm. To achieve
this balance a moment of 50 kNm has to be added to BC, and this causes a carry-
over moment of 25 kNm at CB. The joint at B is completely balanced and it is
only necessary to perform one further distribution at C. The reactions can be .
found by taking moments about suitable points.
'LMc
20 X 82 )
6 VB = ( 2 + 120 - 190
= 95 kN
20 X 20 2 )
12Vc= ( 2 +(40x15)-(95xl8)-190
= 225 kN
An alternative procedure would be to split the continuous beam up into simply
supported beams, and to find the reactions (figure 7.15a). The reactions due to
the bending moments acting alone (b) are added to these to form the total
reactions (c).
~
t tf&:!
W.@ff..@!W'
t 80t t t
~
40
+
80 120 120
(ol
(40 190) (190 190)
t
25
t
25
t0 +
0
(bl
+
95
+
225
(C)
Figure 7. 15
As a further example, the bending moments are required at the joints and
supports for the portal frame in figure 7.16.
This will turn out to be a very simple calculation if a moment-distribution
method is used.
F.E.M.
MBc = ---u-
- pb2
McB
pb2
=U
The member BC is symmetrically loaded.
D.F.
4Elja 2b
BA= =
4Elja + 2EI/b a + 2b
a
BC=--
a + 2b
Distributing at joint B
The moments at C and D can be written down straight away from these
results.
7.8 Deflection of supports and sidesway

The beam problems that have been treated so far have had rigid supports. The
question naturally arises as to how the moment-distribution method could be
modified to treat problems in which one or more of the supports sinks.
Figure 7.17a shows a member AB that is rigidly built in at each end, and the
end B has moved vertically through a distance Ll with respect to A. This is a posi-
tive displacement according to the sign convention in section 7.2.
~A
~--=-B~,J (a) ~
Figure 7.17
Imagine the beam remains straight with the same displacement A applied (b).
The chord rotates through an angle if>= A/1, and each support rotates clockwise
through the same angle such that the bending moment at any point on the beam
is zero. To get back to the original configuration as shown at a it is necessary to
rotate each end through an angle -1/>.
Figures 7.17c and d show the moments developed at each end of the beam
when ends B and A are each rotated through an angle of -1/>.
Summing the moments will give the total effect due to the rotation of both
ends. Thus
MAB = -(SAB + CBASBA)if>
MBA= -(sBA +cABSAB)if>
With the usual substitution for s and c in the case of a uniform beam,
6EII/>
1-
MAB=MBA=-- or -6EIA
p (7.10)
For the problem in figure 7 .18a the deflection at support B is required such
that the bending moment at B is zero. E = 210 kN/mm 2 ,J= 41 000 cm4.
It will first of all be necessary to find the bending moment at B without any
deflection of the support.
tBOkN t60kN 20kN/m
~3m-- --~sm-
A -Is - 8 -m==--Fsm-+:1
6m C 0 E
(a)
------14/7 3t7 t-------t5tn

-60 60 -22·5 67·5 -41 ·7 41·7 ~
-9·1 -18·2 - 5·7 ..=1li... -0·8 .;;ll -40
~ ~ .:..1.il.. - 6·8 -13·6 40
-69·5 40·9 1·6 _2L __ll.
-0-7 -0·4 0·2
::'ToT ...Q;L -ST'i
52·2
(b)
4/7 3t7 5/11 6'11

-100 -100 56·3 56·3 0 0 0
___l__U 32·3 -12·8 -25·6 -30·7
~ 1·6 24·2 12·1 -6·6
-83·0 0·1 -2·8 - 5·5 - 0·3
:s&o' --1:2.. 0·6 ":37-7
-0·2 -0 ·3
...!Ll. "37-6
66-0
(cl
Figure 7.18
Fixed-end moments
80 X 3 3
MAB =- ~ = -60 kNm MBA= 60kNm
60 X 6 X 22 60 X 62 X 2
Mse =- 82 = -22·5 kNm Me 8 = 82 = 67·5 kNm
20 X 52
Men =- =- 41·7 kNm Moe= 41·7 kNm MoE=- 40 kNm
12
Distribution factors
Joint B
BA = 116 = i BC = ~
1/6 + 1/8 7 7
Joint C (D is a pin end)
CB = 1/ 8 = 2_ CD= i
1/8 + 3/4 X 1/5 11 11
The distribution has been carried out in figure 7.18b.
The bending moment at B is 40·9 kNm and this has to be reduced to zero by
lowering the support at B.
Consider all the external loading removed and let B move vertically through a
distance .:1 with all joints in the clamped position. Moments of -6EI.:1/l 2 will be
induced at the ends of the members AB and BC. Let us assume (incorrectly) that
6EI.:1fl 2 for member AB is equal to 100 units. Thus the end moments at A and B
are -100. Since El is constant for the whole beam the end moments at B and C
will be 100 x 6 2/8 2 = 56·3 units; these will be positive, since there is an anti-
clockwise rotation of the chord.
These moments are distributed as shown in figure 7.18c. The distribution fac-
tors are of course the same as those used in the first distribution.
The final moment MBc = 66 units.
Now for AB
6£/..:1 6 X 210 X 41 000 x 10- 2 .:1
-----z2 = 36 = 14 350..:1 kNm
where .:1 is measured in metres. This was assumed to be equal to 100 units. Thus
the true fmal moment MBA after distribution is
66
-14350..:1 X-= -9450..:1
100
Now with no deflection at B the first distribution gives MBA= 40·9 kNm. To
reduce this moment to zero will require a deflection of 40·9/9450 metres or
4·32mm.
A rigid framework that is symmetrical and that has symmetrical loading
applied, can be analysed in the same manner as a beam with only joint rotation
and no displacement taking place; this was illustrated by the example on the por-
tal frame in section 7. 7. If however the system is asymmetrical due either to the
layout of the members or to the loading, then the joint rotation will be accom-
panied by translation, or sidesway. The joint translations are relatively small
although they may have a considerable effect on the values of the bending
moments. In the majority of cases there will not be sufficient change in the
geometry of the structure for this to be taken into account in the analysis.
The portal frame in figure 7.19a has a lateral load applied. The deformation
due to joint translation alone is shown greatly exaggerated in b. This will of course
be accompanied by joint rotation. The two column members have deformed in a
similar manner to the beam in figure 7 .17a, so that an identical analysis can be
used to fmd the resulting moments.
~~------------~
(a) (b)
Figure 7.19
cl
75
---
100kN 31
~
B
15m ~
75
I I 10m
A
~
(a)
Dl ~
--~87·5
231 231 -100
12·5
-231 -231
':87.5 (c l
(b)
-269 -269
(d)
Figure 7.20
Consider first the numerical example in figure 7.20a. If sidesway is allowed to

take place with the joints clamped against rotation
6£/.t:J
MAs =MsA =Men =Moe= -102
Let this be equal to -100 kNm, say . .tJ is the horizontal displacement of BC.
When the joints are unclamped to allow rotation to take place at B and C, the
member BC will deform antisymmetrically, thus a modified distribution factor
can be used in the moment distribution.
D.F.
41/10 1 3
BA = - - - - - ' - - - - - CB=-
41/10 + 6 X 3//15 4 4
The distribution has been carried out in figure 7.20b using the arbitrary sidesway
values of -100 kNm.
Column AB is shown isolated in figure 7.20c with moments of -75 and
-87·5 kNm applied at Band A respectively. To keep the column in equilibrium,
shear forces of value (75 + 87·5)/10 = 16·25 kN are required. Thus to produce
the final moments shown in b a lateral force to the right of (2 x 16·25) = 32·5 kN
(for two columns) would have to be applied at B.
In the original problem 100 kN is applied at B, thus the true resulting moments
are obtained by multiplying the final moments in the sidesway distribution by a
factor of 100/32·5. The final values are shown at din kNm units.
The same portal is shown in figure 7.21a but with a vertical load applied to
the member BC.
In dealing with a problem of this type it will be necessary to perform two
separate distributions. First it is assumed that sidesway is prevented, and a normal
distribution is carried out. From the results, the force required to prevent side-
I 200 kN
8 r 3/
-....::1.
-194
1 139
sm-l---1om -3
5
10
14 =28
83 42
-124 -247
296 148
-444 222
(a)
2/3f-----l
-123
T4
~~195
..?:1139 ..;l.
291 20·9 -139
291 209:0- -62

.li. (b)
---~96 ~ 70
..lQ.. --=L
(c)
2 ..;l
96 -70
-175 158
1'9 1'9
-194 139
194 -139
~ ...;.12
175 -158
96 -70
,;li -22
74 (d) =92
Figure 7.21
sway is found. An arbitrary sidesway is next applied with the original loading
removed. After distribution the lateral force required to produce this sidesway
can be found. From this result the correct sidesway can be determined to reduce
the lateral load from the no-sidesway solution to zero. The true final moments
will be the sum of the first distribution and a certain known fraction of the
second distribution.
No-sidesway distribution
F.E.M.
MBc = -444 kNm McB = 222 kNm
D.F.
BC = CB = j BA = CD = !
The distribution has been carried out in figure 7.21b. The two columns are
shown isolated at c. The total shear is
195+96 139+70
-- 1-0 - - 10 = 29·1- 20·9 = 8·2 kN
Thus a force of 8·2 kN to the left is required at B to prevent sidesway.

For sidesway effects, arbitrary sway moments of -100 are chosen for the
columns and the distribution will be identical to that shown in figure 7.20b. This
results in a lateral force of 32·4 kN applied to the right at B. To produce a force
of 8·2 kN the final moments in figure 7.20b must be multiplied by a factor of
8·2/32·5.
If the moments resulting from the no-sway distribution and the reduced
moments from the sway distribution are summed as shown in figure 7.21d, this
will result in zero force applied at B, as specified in the original problem.
The final moments are as follows
MAB = 74 kNm MBA= 175 kNm
Mco =-158 kNm Moe =-92 kNm
7.9 Frames with inclined members
For structures of this type the analysis can proceed on somewhat similar lines to
the previous analysis. It is however best to determine the shear equation, which
will be a relation between the final moments and the external loads. The method
of approach will be discussed with reference to the problem in figure 7.22a, where
the relative Elf/ values of the members are shown in circles.
The simplest method of forming the shear equation is by means of a displace-
ment diagram and virtual work. The method of instantaneous centres can be used
instead of a displacement diagram. All joints are considered as pinned and a small
clockwise rotation q, is applied to AB; the resulting displacement diagram is shown
in figure 7.22b.
CD
ob =AB x q, ox = 9t/> oc = ox x -
12
=it~>CD
Thus the rotation of CD is !t~> clockwise.
bx = 3tf> ex =!t~> x 4 = 3t/> be =6t/>
Hence the rotation of BC is q, anticlockwise. The horizontal movement of Cis 91/>.
The vertical movement of E is 41/> - 31/> = 1/> upwards.
Making use of the virtual work equation.
31/> 31/>
MAB!J>+MBA!J>-MBc!J>-McBI/> +Mco 4+Moc 4+20x 91/>- 90x 1/>= 0
Now MBA+ MBc =0 andMcB + Mco =0, and the shear equation will simplify to
4MAB + 8MBA + 7Mco + 3M0 c = -360
A no-sidesway distribution is next carried ·out (figure 7 .22c).
l
20kN
L
9m b
12m (b)
35·0
(al -25·0 ~
----a-s 0·3
0 -0·8 -1-6
l3m --6m L.m-j t;:g H
-7-4 -11.·8
131·2 "·4 22·2
146 2 (i':2 -26·6 -53·3
-40
=:a! -0·3 80
__Q:1 1·7 2t3
-5·2 -2·6
7·8 15·6
-46 7 -23·4
-50 -100
240 240
12·5
0"2
...1.:£
11·1 (c)
-133·1
~
-1·3
m
-11·6 -9·5 48·9
-120 13-9
-25 35 -48·9
':'TJ.'9
-35
-1·6
:;rr
12·5 (e)
-28·1
::'iQ.7
-17·4
Figure 7.22
F.E.M.
M 8 c =-40 kNm Mcs = 80 kNm
D.F.
BA = CD = ! BC = CB = j
The moments from this distribution are substituted into the left-hand side of
the shear equation. This gives a value of -47·2 kNm, indicating that sidesway is
present. For no sidesway the value would be -360 kNm.
Next a sidesway is applied such that the frame deforms to the right. In the
previous analysis the rotations of members were as follows
AB=<P BC=-<P CD=i<t>
For these rotations the ratio of the side sway moments (- 6EI<P/l) can be
found
3<1>
MAB :M8 c :Men= -6 X 2</>:6 X 4</>: -6 X 2X 4
that is
-12:24:-9
These values multiplied by 10 have been taken as the arbitrary sway moments,
and are distributed in figure 7 .22d. The final moments from the distribution will
be in the correct ratio for the required final sidesway moments, and can be sub-
stituted into the left-hand side of the shear equation.
-(133·1 X 4)- (146·2 X 8)- (131·2 X 7)- (110·7 X 3) =- 2952
Ifk is the multiplying factor for the sidesway solution -47·2- 2952k = -360,
whence k = 0·1 06.
The final moments will be given by the sum of the no sway moments and
0·106 times the sidesway moments. These are shown in figure 7.22e.
7.10 Rectangular multi-storey frames

A solution can be obtained for this type of frame using the procedure employed
for single-storey structures; unfortunately the process is rather lengthy since a
total of n + 1 separate distributions are required, where n is the number of storeys.
Basically the procedure is as follows
(1) Carry out distribution with sidesway prevented. Lateral forces will be
required at each storey and these can be found from the resulting moments.
(2) Remove all external loading and apply an arbitrary sidesway to the top
storey only. Distribute and find the loads at each storey that would cause this
sidesway.
(3) Repeat process 2 for each storey in turn. It is then only necessary to
combine the solutions in the correct proportion.
A ,.---t------,
Figure 7.23
For the two-storey frame in figure 7 .23. Let the lateral force at A be PAt and
B be P 81 from the no-sway distribution. Let the lateral force at A be PA 2 and B
be P 82 from the top-storey sidesway. Let the lateral force at A be PAJ and B be
P 83 from the bottom-storey sidesway.
Since there are no lateral forces actually applied
PAl +ktPA2 +k2PA3=0
and
P 81 +k1P82 +k 2 P 83 =0
The two equations can be solved fork 1 and k 2 .
The final moments will be given by the sum of the moments from the no-sway
solution and k 1 times the moments from the top-storey sidesway and k 2 times
the moments for bottom-storey sidesway.
7.11 The slope-deflection equation

This in an alternative method to that of moment distribution, and will solve simi-
lar types of problem. Basically a number of simultaneous equations are formed
with the unknowns taken as the angular rotations and displacements of each joint.
Once the equations are solved the moments at all joints may be determined.
A member AB of length l has a loading system denoted by P 1 , P2 and is
allowed to deform as shown in figure 7.24a. End A rotates through 8A• and end B
through 8 8 , while B displaces vertically through ..1 with respect to A.
The various rotations and the deflection can be taken individually as shown in
figure 7.24b toe, and the end moments caused by each can be written down.
The total moments are found by summation
A ....:------::---.
~
(d)
-(sBA+ CAB 5AB)~
~ F. E.M. F. E.M. ~
(e)
Figure 7.24
For the case of a uniform straight member

4EI
SAB = SBA = - CAB= CBA =!
l
MAB = -2EI [ 28A + 8a- -3Ll]

1- 1- + F.E.M.
MaA = -2EI [ 3Ll ]

1- 28a + 8A- - 1 + F.E.M. (7 .11)
Note that in the analysis no sign has been allocated to the ftxed-end moments.
'This will of course be the same as that developed in section 7.5. These equations
are known as the slope-deflection equations.
As a simple example we will treat the ftrst problem that was solved by moment
distribution (figure 7. 25).
It is best to list all quantities that are zero.
_jt
SOkN
~A E/ + El B ~
~Sm ------<•:_1: --=-4-m-lOm__J
Figure 7.25
In this case OA =0 8 =0 and all Lfs are zero.

F.E.M.
McB = -72 kNm M 8 c = 48 kNm
Next the slope-deflection equations are applied to each end of the members.
2EI 2EI
MAc= - 5- (0 + 0 8 -0) + 0 MeA= - 5- (20c + 0- 0) + 0
2EI 2EI
McB = 10 (20c + 0- 0)- 72 MBc = lo (0 + Oc - 0) + 48
Since there is no external moment applied at C the sum of the moments there is
zero.
~Mc=O
4k0c + 2k0c- 72 = 0
where k = 2EI/10, therefore kOc = 12.
Substituting this value into the original equations for the moments gives
MAc = 24 kNm, MeA= 48 kNm, M 8 c = 60 kNm.
A B C D
I--sm
~ ~
./1
Figure 7.26
A further example is shown in figure 7.26 where the beam is continuous and is
supported by pins at B and C, the ends A and D being free. The movement at Dis
required when A is raised by d.
L1Bc = 0; .dAB= d, Oc = y/3 since CD remains straight.
MAB =Men =McB = 0
2EI
McB =6 [20c + 0 8 ] =0
therefore 0 8 = 2y/3
Mse = 62El [20 8 + Oe]
Ely
= - 3-
5- 20A + Os- S = 0
= -2El[
MAB 3d]
therefore OA = 3d/l0- y/3
MBA
2El [ 20s +OA -S
=S 3d]
~ M8 =0 therefore
Ely+ 2El [4y +3d - L- 3d]= 0
3 5 3 10 3 5
9d
y=-
55
As an example of a framework with sidesway present, the frame in figure 7.27
will be solved using slope deflection. The Ll terms for AB and CD will be the same.
OA =On =0, Lise =0

F.E.M.
P1 P1
Mse =-4 Mes =4
MAB = ll
2El [ 3 Ll ]
Os - 2.1 MBA=
2El
2/ [
20s - 2/
3 Ll ]
Moe =-2El [
1- Oe - -3LI
1-
] Meo = -2El [ 3LI ]
1- 20e - -1-
2El [
Mse =2z 20s +Oe
]
-4
P1
2El [ ] P1
Me 8 =2l 20e+Os +4
~M8 =0
-2El [ 40 8 +0e-- P1
3LI] --=0
21 21 4
~Me=O
-2El [ 60e+0 8 --
P1
6LI] +-=0
21 I 4
r· r' EI ct
£1
1
I
1
2/ £/
D
~
A
Figure 7.27
There are at present two equations relating the three unknowns OB, Oe and Ll.
A third equation can be formed by considering the shear in the columns.
MBA + MAB Meo +Moe
----'=-----=.:= + =0
21 l
3LI 24LI
30B - -1- + 120e- - 1
- =0
therefore
9LI
OB=--40e
l
The three equations can now be solved to give

22 PZ 2 2s PZ 2
(JB = 304 E/ Oe =- 304 EI
3LI 26 PZ 2
-=---
l 304 EI
35 57
MAB = 304 Pl MBA =-Pl
304
48 2
Meo = - 304 Pl Moe = 304 Pl
We shall make use of some of the individual terms in the slope-deflection

equations as stiffness coefficients in the later chapter on matrix methods.
Problems
7.1 Determine the bending-moment diagram and the reactions for the continuous
steel beam in P 7.1. hB = 8000 cm 4 , /Be= 20 200 cm4 ; leo= 11400 cm 4 .
fF:ts::::t:lo
/p=10kN/m t lOkN
~7m 10m 2m
p 7.1
7.2 For the beam in P 7.1 how far would the support at B have to sink to make
the bending moment there zero?
7.3 The beam and the columns in P 7.3 have the same cross-section. Find the
bending moments at the joints.
70kNr2ml- t.m -----t2mt 7DkN
A 8 C l 0
FJ
t.m
p 7.3
7.4 Find the force exerted on each of a long line of circular pegs by a long mild-
steel strip laced through them. The pegs are 10 mm diameter and spaced at inter-
vals of 240 mm with their centres lying on the same straight line. The strip is
20 mm wide and 1·25 mm thick.
7.5 The framework shown at P 7.5 has stiff joints but is pinned at D and H. In
addition to the point loads which act at mid-span, the member BF carries a uni-
formly distributed load of p per unit length, and member CG, 2p per unit length,
where pi= P. Find the moments at the ends of the various members of the
framework.
A El E
t
El tp El I
8 u
2EI
fP ~
i
2El 2El I
c
3El
3El 3El 21
0
'11
Hl '11
3/
p 7.5
242 STRUCTURAL TIIEORY AND ANALYSIS
7.6 Find the bending moments for the rigid jointed frame P 7 .6. All the members
have the same cross-section.
tlOOkN C
p 7.6
7.7 A single-storey two-bay frame consists of a horizontal beam ABC rigidly

connected to vertical stanchions AD, BE, CF, of equal length, which are rigidly
fixed at D, E, F. The relative values of EI/1 for the members are given by AB =
AD== CF = 2, BC == 1, BE = 3. Determine the shear force carried by each stan-
chion when the frame is subjected to a horizontal force P applied at A in the
direction ABC. Discuss how you would solve the problem of a two-storey single-
bay portal frame.
7.8 Sketch the framework shown at P 7.8 when it is deformed by the momentM
Make use of the slope-deflection equation to determine the angle of rotation of
the centre joint. All members have the same length and the same stiffness K.
t~-----T------~~
('*M
___.________.___ ~+
p 7.8
7.9 Determine the bending moments for the rigid jointed frame P 7.9. All mem-
bers have the same cross-section.
c
4m
l Bm
0 _j
p 7.9
7.10 The mild-steel beam ABC in P 7.10 is encastre at C, supported on a simple

support at B, and constrained at A. Vertical movement is prevented at A and the
rotation at A is resisted by an elastic bending moment M(kNm) = 2000 8, where 8
is the rotation in radians. Determine the moments at A, B and C when the point
load is applied at the centre of AB. I= 2530 cm 4 .
~
tSOkN 8 c~
~~45m + 45m_j
p 7.10
8 STIFFNESS AND FLEXIBILITY METHODS
8.1 Introduction
The basic ideas of the stiffness and flexibility approach to the solution of struc-
tural problems were introduced in chapter 6. We shall now proceed to develop
these methods in more detail and apply them to specific problems. The solutions
to the various problems will be set out in a manner that would be suitable for a
computer application so that all the fundamental processes can be demonstrated.
As the working will be done by hand it will be necessary to confine our attention
to fairly simple structures otherwise the arithmetic will become impossibly tedious.
Even so it will be found that the working in most cases will be lengthy and the
reader will be able to suggest alternative methods of solution which would give a
much more rapid solution. It will appear obvious that once a problem becomes
more complex, then a great deal of time and trouble will be saved by making use
of a computer together with a stiffness or flexibility approach.
8.2 Outline of the stiffness method

In this section the various steps will be outlined for the complete solution of a
problem using the stiffness approach. In fact when it comes to the solution of a
particular problem using a computer, slightly different procedures are used. A
particular example of this is that it is much more efficient of computer time to
solve a number of simultaneous equations rather than inverting a matrix.
{1) It is necessary to replace any external loading that does not act at a
node by equivalent loads at the ends of a member and acting at these nodes. This
means that the structure is first analysed with the external loading in position
and all the nodes fully restrained. As a result of this a set of fixed-end forces PF
will be obtained. A pin-jointed structure is a special case and the ends of the
member are not restrained against rotation.
(2) The stiffness matrices for all the individual members are next found. In
doing this the orientation of a particular member is ignored and a system oflocal
coordinates is used for each member. One axis is coincident with the principal
axes of the cross-section.
244
STIFFNESS AND FLEXIBILITY METHODS 245
(3) It is next necessary to determine the stiffness matrix for the complete
structure. For this it will be necessary to have only one set of coordinates: the
system coordinates. We shall require some method such that we can transfer a
member stiffness from local to system coordinates. A transformation matrix will
be used for this purpose. The question will of course arise as to how the complete
stiffness matrix is obtained from the individual matrices.
(4) It will now be possible to write down the stiffness relations for all the
nodes in the structure. In general matrix notation
P= PF + Kd (8.1)
where P represents a set of forces applied at the nodes and d is the corresponding
set of displacements at the nodes. The displacements can be divided into groups,
the ones that are unknown and the ones that are known. In most cases the latter
will have zero values. A typical case might be that of a cantilever where it is known
at the fixed end that a number of displacements are zero. Corresponding to the
zero or known displacements dR, will in general be a set of unknown support
constraints PR. The unknown displacements dL will in general occur at nodes
where the loads PL are known and it may well be that a number of these loads
are zero-valued.
The force or load matrix P and the displacement matrix d may each be par-
titioned in the following manner
The forces that arose from the initial restrained condition can be partitioned
in a similar manner, that is
The matrix for K may also be partitioned such that equation 8.1 becomes
(8.2)
(5) If the matrix is expanded we may write

PL = PFL + KLLdL + KLRdR
thus
If we consider the case where all the terms dR are zero, which is true if all the
constraints at the supports are rigid
dL = Ki:L(PL - PFL) (8.3)
As KLL• PL, and PFL are all known, the displacements dL may be evaluated.
Another relationship can be obtained from equation 8.2.
PR = PFR + KRLdL + KRR dR
thus
PR = PFR + KRLKi:L(PL - PFL) + (KRR - KRLKi:[KLR) dR
or for the case when dR is zero
PR = PFR + KRLKLl(PL- PFL) (8.4)
So that the displacements of all the nodes and the values of the support con-
straints are now known.
(6) It will be remembered that all the forces and displacments have been
written down using system coordinates. The forces at the ends of the members
will be required to complete the solution. These will consist of two parts, firstly
those arising from the fully constrained condition and secondly those that arise
when the structure is allowed to displace and reach its equilibrium under the
applied loading. In terms of member coordinates we may write
P=PF +Kd
It may very well be easier to determine the end reactions for a member in
system coordinates and then rotate these into the member coordinates. This will
become apparent in the examples later in the chapter.
y
z
Figure 8.1
The sign convention shown in figure 8.1 will be adopted. A right-handed triad
is used for the x, y, and z axes. It will be seen that an axial force or displacement
is positive if it acts in the positive sense of the axis concerned. The right-handed
screw rule is used to give the positive direction for moments or rotations. This
accords with the convention used in chapter 3.
8.3 Stiffness matrix for a space structure member

We shall first examine the case of a general member in space and determine the
stiffness matrix of the member. Once this general case has been obtained the
matrix can be simplified and used to treat plane frames and other simpler types
of structure.
Probably the easiest way of determining the individual terms of the matrix is
to apply a unit displacement to one end of the member, which can be either a
deflection or a rotation, while all other displacements are kept at zero. The various
restraints or end forces that are required will of course be equal to the various
elements in the stiffness matrix.
It will be assumed that the member is uniform and that the x axis is coincident
with the longitudinal axis of the member, while they and z axes lie along principal
axes. We shall only deal with displacements at node 1 of the member since the
effect of displacements at node 2 can be written down by inspection, taking care
about signs.
The slope-deflection equation will be of considerable help in finding the values
of the various elements in the matrix. For the case of zero fixed-end moments
2EI( 2e
M 12 == - 1- 1 +e2 3d)
- - 1-
The member is shown in figure 8.2. Each 'force' which acts on the end of a
member and displacement of the end of the member will have two subscripts, the
first denoting the direction of the 'force' or displacement, and the second, either
1 or 2, denoting the end of the member being considered.
z
Figure 8.2
For unit displacement at end 1 in direction x
EA EA
Px 1 ==-
l Px2 == - -1-
For unit displacement at end 1 in direction y
12Elz 12Elz
Pyt = ~ Py2 =- ---z3
For unit displacement at end I in direction z

12Ely
Pz1 =p-
6Ely
Myl =- 13
For unit rotation at end I in direction x
Mx l -
- GJx M x 2 = - GJx
1 1
For unit rotation at end 1 in direction y

4Ely 2Ely
Myt =- 1 - My2 =- 1 -
For unit rotation at end 1 in direction z

M _ 4Elz 2Elz
zl- 1 1-
Mz2 = -
6Elz 6Elz
Pyt =p Py2 =-p
The complete stiffness matrix can now be written down (equation 8.5). The
terms so far found can be entered; the rest follow from the fact that the matrix
must be symmetrical.
This could be expressed as
Ku
P= [
K21
The stiffness matrix that has been derived is 12 x 12, and a matrix of this size
would exist for each member of a rigid jointed space frame. It can be seen that it
will be well nigh impossible to solve problems of this type by hand, hence access
to a computer is essential. It is however possible to reduce the matrix to deal with
certain classes of problems. The next sections will be devoted to this end.
EA EA
Px! 0 0 0 0 0 0 0 0 0 0 dx!
I l
12£/z 6Elz 12£/z 6Efz
Py! 0 -~-3- 0 0 0 0 -~-3- 0 0 0 dy!
72 [2
12Ely 6Ely 12Ely 6Ely Cll
Pz! 0 0 0 0 0 0 -13 0 p 0 dzl ...,
13 12 ::;;
'Tj
Glx 0 - Glx
Mx! 0 0 0 0 0 0 0 {)XI
0
ztTl
I 0 ! I Cll
Cll
6Ely 4Ely ' 6Ely 2Ely >

My! 0 0 0 0 i 0 0 0 0 Oyi z
12 I T I t:l
''
'Tj
6Efz 4Efz i 6Efz O OO 2Efz 1:""
Mz! 0 -- 0 0 0 -: 0 - - {)ZI
P I : 12 I (8.5) ~
... --- .. ---------------------- ...... -------------------- ... ---------- ........ ---- .. -~---- --- ...... ---- ... ------------ ... ---------------------------------.-.------- ;;
EA ' F
Px2 0 0 0 0 ojEA 0 0 0 0 0 dx2
~
12£/z 6Elz
! 12Elz 6Elz a:
Py2 0 -~3- 0 0 0 - [2 i 0 -~-3- 0 0 0 dy2
''
72 ~
12Ely 6Ely ' 6Ely 0
12Ely t:l
Pz2 0 0 0 0 0 0 0 0 dz2 Cll
z3 z2 p z2
GJ, Glx
Mx2 0 0 0 0 0 0 0 0 0 0 8x2
I I
6Ely 2Ely 6Ely 4Ely
My2 0 0 0 0 0 0 0 0 8y2
z2 l z2 I
6Efz 2Elz 0 _ 6Elz 4Elz
Mz2 0 0 0 0 0 0 0 ()z2
7 l 7 I N
~
\0
8.4 Pin-jointed plane trusses

The only force acting in a member for this class of problem is axial. This implies
that all the 'forces' apart fromPx 1 andPx 2 in equation 8.5 are zero and we are
-1]
left with
[ Pxt] = EA [ I [dx1] (8.6)

Px2 1 -1 1 dx2
It has already been mentioned that it will finally be necessary to work in one
set of coordinates that apply to the whole system, whereas at present we are
using a separate coordinate system for each member, with the x axis coincident
with the member axis. If the system x axis does not coincide with the x axis of a
particular member, a displacement dx 1 in the member system would give rise to
displacements and hence forces in both the x andy directions of the system. It is
therefore better at this stage to include Py 1 and Py 2 when writing down the stiff-
ness matrix even though the stiffness terms in the matrix are zero. So that
equation 8.6 becomes
Pxl 0 -1 0 dxl
Pyt EA 0 0 0 0 dyl
=- (8.7)
1 -1 0 0
Px2 dx2
Py2 0 0 0 0 dy2
X
;(
, ,,
p;2
jf
I
I
I
I I
PY1
(a) (b)
Figure 8.3
In figure 8.3 a member is shown with end forces, using member coordinates.
The same member is shown at b using system coordinates, which are denoted by a
prime as are the end forces. It is essential for both sets of forces to be equivalent.
By resolution it can be shown that at node 1
Pxt =P~ 1 cosO +P; 1 sinO
Py 1 = -P~ 1 sin 0 + P; 1 cos 0
A similar relation exists at node 2. So that in matrix form we may write
Pxl cos e sine 0 0 p~l

Pyl -sine cos e 0 0 p;l
Px2 0 0 cos e sine p~2
Py2 0 0 -sine cos e p;2
This could be written in the shorthand form

P=TP' (8.8)
where T is a transformation matrix.
This particular matrix may very well be familiar to the reader. In matrix theory
for a three-dimensional system, if a positive rotation 8 is applied about the z axis,
the rotation matrix is given by
r cos e sine
R= l-s~ e cos e
0 ~l
This degenerates to the matrix Tin a two-dimensional system.
(8.9)
It can easily be shown that the same relation exists between displacements as
that between the forces, thus
d=Td' (8.10)
If we now consider the matrix
cos e sin 8]
[
-sin 8 cos 8
it can be seen straight away that the inverse will be given by
[
cos 8 -sin 8]
sin 8 cos e
and this is equal to the transpose of the original matrix; we have what is termed an
orthogonal transformation matrix. Thus
T-1 = TT
therefore
(8.11)
In local coordinates P = Kd and in system coordinates P' = K'd'

TP' = Kd = KTd'
therefore
also
P'=K'd'
thus
K'=TTKT (8.12)
If this transformation is carried out we shall fmd that
cos 2 8 sin 8 cos 8 i -cos2 8 -sin 8 cos 8

,
K =-
AE sin 8 cos 8 sin 8
2 i -sin 8 cos 8 -sin 8 2
---------------------------------------------------------------------
1 -cos 2 8 -sin 8 cos 8 : cos 2 8 sin 8 cos 8
:
-sin 8 cos 8 -sin 2 8 : sin 8 cos 8 sin 2 8
(8.13)
If the matrix is expressed in the form
we can see that K~1 =K;2, K~2 = K;1 = -Ku.

The stiffness matrix for each member of a truss can now be found in system
coordinates. The stiffness matrix for the complete structure is then assembled.
We shall first have a brief discussion on the formation of this matrix.
We shall suppose that we have a number of members all pinned together, at
end 1 of each member; end 2 in each case is pinned to a different rigid support.
If a displacement d' is given to the common end, the force acting on the end of
each member would be of the form K'd' so that the total force is given by~ K'd'
or(~ K') d'. It can be seen that the complete stiffness matrix for a structure will
be formed by summing the individual stiffness matrices for the members meeting
at a joint.
Once the stiffness matrix has been formed, equation 8.3 may be used to deter-
mine the displacements, and these in turn enable the forces in the members to be
found.
Once the displacements are known the reactions can be determined from
equation 8.4.
The complete solution will probably require the evaluation of the member
end reactions. These can now be found quite easily. In terms of member coordi-
nates the following equation would apply.
P=PF +Kd (8.14)
However it is probably easier to work in system coordinates and then apply a
rotation matrix to transfer to member coordinates.
8.5 Example of a pin-jointed redundant truss

The vertical and horizontal members of the truss shown in figure 8.4a are all of
the same length /. All members have the same cross-section and are made from
the same material.
~1 ~3
(a) (b)
Figure 8.4
It will be seen that the nodes have all been lettered, although it is more usual
when using a computer to use numbers. The letters have been used to avoid any
possibility of confusion with the displacements, which will be numbered. The
members will be referred to in terms of the nodes at their ends. For a particular
member the end that has the letter nearer the start of the alphabet will be end 1.
For transferring from a member to a system matrix the angle 8 is measured from
the x axis such that the rotation is positive with respect to the z axis {right-hand
corkscrew).
The problem we have chosen is a very simple one-even so, since there are two
unknown displacements at each end of the nodes C and D it will be necessary to
solve four simultaneous equations or invert a 4 x 4 matrix.
Details of the various members are set out in table 8.1.
If a computer is used, the coordinates of each node are fed into it, from which
it can work out the various rotation matrices.
Table 8.1
Member Length Area 8 cos 8 sin 8 cos 2 8 sin 2 8 sin 8 cos 8
AC y2l A
1
y2 y2
1 1
2
1
2 !
AD A 0 0 0 0
BC A 0 0 0 0
1 1
BD y2l A
y2 y2
1
2 ! 1
-2
CD A 0 -1 0 0
In figure 8.4b all possible displacements of the nodes have been numbered; the
directions are in system coordinates and in the positive x andy directions. The
numbering sequence is not the best but has been chosen to illustrate a point further
in the development. It would in fact have been better to use displacements 1 to 4
for nodes C and D and 5 to 8 for nodes A and B.
Using equation 8. 7 the stiffness matrix for member AC is given by
0 -1 0
E4 0 0 0 0
K --
AC- ..J2Z
-1 0 0
0 0 0 0
Or in system coordinates
1 2 7 8
! 1 -1 -1 1
I AE 1 1 -1 -1 2
KAc = ...j2Z
-1 -1 1 ; 7
-1 -! ! 1 8
The numbering introduced at the top and side of the matrix corresponds to
the numbered displacements at the nodes of the member and will be of assistance
when the matrix for the whole structure is assembled. Element 27 (row 2 column
7) would represent the force produced in direction 2 by a unit displacement in
direction 7.
The stiffness matrices for all the members can now be written down in system
coordinates as follows.
1 2 7 8
-1 -1 1
0·354EA 1 1 -1 -1 2
K~c=
1 -1 -1 1 1 7
-1 -1 1 8
1 2 3 4
1 0 -1 0 1
I EA 0 0 0 0 2
KAo=-
1 -1 0 1 0 3
0 0 0 0 4
STIFFNESS AND FELXIBILITY METHODS 255
5 6 3 4
-1 -1 1 5
0·354EA -1 -1 6
K~D =
1 -1 1 1 -1 3
1 -1 -1 4
5 6 7 8
1 0 -1 0 5
, EA 0 0 0 0 6
KBc=-
1 -1 0 0 7
0 0 0 0 8
7 8 3 4
0 0 0 0 7
, EA 0 1 0 -1 8
KCD =-1- 0 0 0 0 3
0 -1 0 4
The stiffness matrix for the truss can now be assembled and this will be an
8 x 8 matrix since there are four nodes, each with a possible displacement in two
directions. A particular element in the matrix is formed by the addition of any
corresponding terms from the individual stiffness matrices. For example element
77 is the sum of 0·354 from KA.c; 1 from Ksc; 0 from KeD; that is, a total value
of 1·354.
There is a connection between nodes A and B formed by the wall, which will
have a very high stiffness value. So far as our calculations are concerned it would
not matter what value is used for the stiffness as it will be assumed that the dis-
placements at both the nodes are zero. Hence the stiffness can be taken as a null
matrix.
1 2 3 4 5 6 7 8
1·354 0·354 -1·0 0 0 0 -0·354 -0·354

0·354 0·354 0 0 0 0 -0·354 -0·354 2
-1·0 0 1·354 -0·354 -0·354 0·354 0 0 3
0 0 -0·354 1·354 0·354 -0·354 0 -1·0 4
K'= EA
1 0 0 -0·354 0·354 1·354 -0·354 -1·0 0 5
0 0 0·354 -0·354 -0·354 1·354 0 0 6

-0·354 -0·354 0 0 -1·0 0 1·354 0·354 7
-0·354 -0·354 0 -1·0 0 0 0·354 1·354 8
We really require the matrix when partitioned to be in the form shown in

equation 8.2. We know the loads that are applied at C and D and so the stiffness
elements corresponding to directions 7, 8 and 3, 4 will form the sub-matrix KLL.
It was mentioned when the displacements were first numbered that the best
sequence had not been chosen. It would have been possible to choose a different
numbering sequence such that the submatrix KLL appeared in its correct place in
the top left-hand corner. We can however easily modify the stiffness matrix to
conform to our requirements. All the elements must be kept together in a par-
ticular row or column but the sequence of the elements can be altered. This means
that the sequence of the numbering of the rows and columns will be changed, but
any particular element will remain unaltered.
7 8 3 4 1 2 5 6
1·354 0·354 0 0 i -0·354 -0·354 -1·0 0 7

'
0·354 1·354 0 -1·0 ! -0·354 -0·354 0 0 8
0 0 1·354 -0·354! -1·0 0 -0·354 0·354 3
EA 0 -1 -0·354 1·354l 0 0 0·354 -0·354

-
K = - ........ - .. --- ................ .. - ............ - .. ~--- .... -- ....... -- ...... --- ........... -----
I 4
1 -0·354 -0·354 -1·0 0 i 1·354 0·354 0 0
-0·354 -0·354 0 0 ! 0·354 0·354 0 0 2
-1·0 0 -0·354 0·354! 0 0 1 0 5
0 0 0·354 -0·354l 0 0 0 0 6
At this stage we shall deviate somewhat from the theory suggested at the start
of the chapter where it would be necessary to invert the matrix KLL· From
equation 8.2 with system coordinates and dR zero
(8.15)
For the problem we are solving, equation 8.15 will consist of four simultaneous
equations which may be solved to give the displacement components at C and D.
We should next form the load matrix. In the original problem there is only one
load directly applied at a node, Pat D. Since this acts in the negative y direction
the value of the load is - P. The load matrix would be the column matrix
783 41256
{0 0 0 -P 0 0 0 0}
Other forces arise at the nodes when the truss is in the fully restrained position;
these will be due to the uniformly distributed load acting on member BC. The
forces that arise are in an upward or positive direction at nodes B and C and have
values of P/2, they correspond to displacements 6 and 8. The matrix PF will be
given by
3 4 2 5
0 0 0 0 0
We can now form the right-hand side of equation 8.15 and indeed write the
equations out in full as follows
1·354 0·354 0 0 d7 0
p
0·354 1·354 0 -1·0 ds -
EA 2
0 0 1·354 -0·354 d3 0
0 -1 -0·354 1·354 d4 -P
which when solved gives

PI PI
d7 =0·723- d 8 = -2·77-
EA EA
PI PI
d3 =-0·78- d4 =-2·99-
EA EA
The deflections of all the nodes in the structure are now known.
From equation 8.4
PR = PFR + K~LdL
so that in this case
7 8 3 4
0 1 -0·354 -0·354 -1·0 0 0·723 1·5 1
0 2 -0·354 -0·354 0 0 -2·77 0·72 2
PR = +P =P
0 5 -1·0 0 - 0·354 0·354 - 0·78 -1·5 5
p
- 6 0 0 0·354 -0·354 -2·99 1·28 6
2
So thatP 1 = 1·50P, P 2 = 0·72P, P 5 = -1·5P, P 6 = 1·28P.

The complete solution to the problem will require the values of the member
end actions or forces. Using the stiffness matrix in terms of the system coordinates,
the end actions of member BC become
p~ 0 0 -1 0 0
p
P~y 0 0 0 0 0
= 2 +P
p~ 0 -1 0 0 0·723
p
P~y 0 0 0 0 -2·77
2
where P~ is the end force at node B in the x direction. Thus P~x = - 0·723P,
P~y = 0·5P, P~x = 0·723P, P~y = 0·5P. The system and member coordinates are
in the same direction for this member. These end forces indicate that the member
is in tension.
We shall next look at member AC.
P:.U 0 1 -1 -1 0
P~y 0 -1 -1 0
= + 0·354P -1 0·723
P~x 0 -1 1 1
P~y 0 -1 -1 -2·77
p;.._x =p;.._y =0·725P

P~x =P~y = -0·725P
It is now necessary to rotate these forces through an angle of 45° clockwise.

The forces will then be member orientated.
PAx 1 0 0 0·725
PAy p -1 1 0 0 0·725
=-
Pcx y2 0 0 1 -0·725
Pcy 0 0 -1 -0·725
Hence PAx = 1·02P, PAy = 0, Pcx = -1·02P, Pcy = 0.

The result could be obtained more rapidly as follows.
P=Py +Kd
=TP):, +TK'd'
If the matrix TK' is evaluated it will be found to be
cos 8 sin 8 -cos 8 -sin 8

EA 0 0 0 0
(8.16)
l -cos 8 -sin 8 cos 8 sin 8
0 0 0 0
The final member end forces are
Member Near end Far end

X y X y
AC 1·02P 0 -1·02P 0
AD 0·78P 0 -0·78P 0
BC -0·72P O·SP -0·72P O·SP
BD -I· lOP 0 I·IOP 0
CD -0·22P 0 0·22P 0
8.6 The space truss

The relation between the end load and the displacement for a pin-jointed space
frame can be obtained from equation 8.5 and has the following form in member
axes
Pxl 1 0 0 -1 0 0 dxl
Pyt 0 0 0 0 0 0 dyl
Pzl 0 0 0 0 0 0 dzl (8.17)
Px2 -1 0 0 I 0 0 dx2
Py2 0 0 0 0 0 0 dy2
Pz2 0 0 0 0 0 0 dz2
Again it will be necessary to be able to transfer from a member stiffness K to a

system orientated stiffness K'. For this we shall make use of direction cosines
(figure 8.5).
x'
Figure 8.5
11 is the cosine of the angle between the x and x' axis

m 1 is the cosine of the angle between thex andy' axis
n 1 is the cosine of the angle between the x and z' axis
12 is the cosine of the angle between they and x' axis
and so on. Then
(8.18)
Pxt It mt nt 0 0 0 P~t
Pyt 12 m2 n2 0 0 0 p;l
Pzl 13 m3 n3 0 0 0 P~t
= (8.19)
Px2 0 0 0 It mt nt p;2
Py2 0 0 0 12 m2 n2 p;2
Pz2 0 0 0 /3 m3 n3 p~2
The reader is probably familiar with the rotation matrix.
(8.20)
so that equation 8.18 may be rewritten
(8.21)
or
P=TP'
K can be transformed into K' in exactly the same way as for a plane truss, that
is
If the manipulations are carried out the following expression will be obtained
if ltmt I tnt -I~ -ltmt -ltnt

ltmt mf mlnl -ltmt -m~ -mlnl
K'=AE I tnt mlnt n~ -l1n1 -mlnt -n~ (8.22)
I -I~ -ltml -!tnt ~~ ltmt ltnl
-ltml -m~ -mlnl ltmt m~ mtnl
-/1n1 -mtnl -n~ I tnt mtnt n~
It will be noticed that only the direction cosines of the x axis relative to the x'
axis are required.
The values of 11 , m 1o and n 1 can be found quite easily if the coordinates of all
the nodes are known.
The solution of a problem will follow along similar lines to that of a plane
truss, and it is not proposed to solve an actual example here.
8. 7 Continuous beams
In this type of problem it is assumed that the bending action is confined to one
plane and that there are no axial or torsional effects. This means that the general
stiffness matrix will reduce to equation 8.23. It will not be necessary to consider
member and system coordinates in this case since the two axes can be made
coincident.
12£/z 6Efz 12Elz 6£/z

Pyt
p [2 f3 f2 dyt
6£/z 4£/z 6Elz 2£/z

Mzl (]zl
[2 I [2 I
(8.23)
12Elz 6£/z l2Elz 6£/z
Py2 -I3- [2 -~3-
f2 dy2
6£/z 2£/z 6£/z 4£/z

Mz2
[2 l f2 I
(]z2
The solution of a continuous-beam type of problem can be illustrated by the

fairly straightforward example shown in figure 8.6a. This has been chosen with
fixed ends so that the number of unknown displacements is small and hence the
p
p::T
t2P tp
~
B
.L _J_.L ~~
2 2
l
£1 0
,I
(a)
+1 t3 fs p
~ { ~ /a
(b)
Figure 8.6
size of matrix to be inverted is also small. It would appear at first sight that there
is only one unknown, the rotation at node B. There is however a complication to
the problem since there is a sudden change of section at C. One method of dealing
with this is to introduce another node at C, making a total of three beams. Pro-
ceeding in this way we shall also be able to find the displacements at all supports
and also that at C.
The beam is restrained and all possible displacements are numbered (figure 8.6b).
It will be noted that a sequence of numbering has been adopted such that the
nodes have been taken in order along the beam, and two displacement numbers
allocated to each node. The K matrix for each member can be written down, and
each of these will have the form of equation 8.23.
1 2 3 4
24 12/ -24 12/ 1

EI 121 8/2 -12/ 4/2 2
KAs=p
-24 -12/ 24 -12/ 3
12/ 41 2 -12/ 8/2 4
3 4 5 6
24 12/ -24 121 3

EI 12/ sz2 -12/ 4/2 4
Ksc=y
-24 -12/ 24 -12/ 5
12/ 4/2 -12/ 8/2 6
5 6 7 8
12 61 -12 61 5
EI 61 41 2 -61 2z2 6
Ken=-
z3 -12 -61 12 -61 7
61 2/2 -61 4/2 8
It will be noted that the three unknown displacements are 4, 5, and 6. Time
will be saved if the system stiffness matrix is formed at once with the partitioned
term KLL at the top left-hand corner. The terms will then not need regrouping.
5 6 4 3 1 2 7 8
36 -61 -12/ : -24 0 0 -12 61 5
-61 12/2 4/2! 12 0 0 -61 2/2 6
-12 4/2 16/ 2 ! 0 12/ 41 2 0 0 4
-------------------- . ------ .... -- .. -- ... -.. -... -.. --------- .. -------- .. -.. -- --- .. ---- ..
I
I
I
EI
K=- -24 12/ 0 48 -24 -12/ 0 0 3
z3
0 0 12/ -24 24 12/ 0 0 1
0 0 4/2 -12/ -121 8/2 0 0 2
-12/ -61 0 0 0 0 12 -61 7
61 2/2 0 0 0 0 -61 4/2 8
(~ ~) (~ ~)
At:-p-----p""""':t 8 t:-p-----p.....:.. C -------0
Figure 8.7
The matrix of loads applied at the nodes is known and that due to the imposed
restraints can be found (figure 8. 7). Using the same order of terms as in the overall
stiffness matrix
-P p -2P 5
PI PI
0 6
4 4
PI PI
0 4
12 12
P-Py= 0 2P = -2P 3
0 p -P
PI PI
0 - 2
6 6
0 0 0 7
0 0 0 8
Now KLLdL = PL - PFL
36 -61 -121 ds 2P
EI PI
[3 -61 1212 41 2 d6
4
PI
-12 41 2 1612 d4
12
Solving these equations gives

350 Pl 3
d- ----
5 --4608 EI
18PI 2
d6 -
-4608£/
291 Pl 2
d- --
4 --4608 EI
So that all the displacements are known.

Now PR =PFR + KRLdL

2P -24 121 0
p 0 0 121 CSOIJ
18
PI p
PR = -
6
+--
46081
0 0 41 2 -291
0 -12 -61 0
0 61 21 2 0
Hence P 3 = 3·87P, P 1 = 0·245P, P 2 = -0·086PI, P 7 = 0·885P, P8 = 0·445PI.
A check is obtained since P 3 + P 1 + P 7 = SP.
A calculation of the end forces for member AB is set out below: the subscript
to each force P again refers to a force in the direction of a particular displacement.
pl p 24 121 -24 121 0

PI
p2 - EI 121 812 -121 41 2 Pl2 0
6 +-
13 4608£/
p3 p -24 -121 24 -121 0
PI
p4 - 121 41 2 -121 812 -291
6
therefore P 1 =0·245P, P 2 = -0·086P1, P 3 = 1·755P, P4 = -0·672PI.
8.8 The plane frame
The bending action in this case is confmed to one plane, there are no torsional
effects, and axial forces are taken into account. This means that the matrix will
be similar to that shown in equation 8.23, but allowance must be made for axial
effects. Thus the matrix becomes
Pxl -EA 0 0
__
,'''EA
0 0 dxl
I I
12Efz 6Elz 12Elz 6Efz
Pyl 0 -13- [2 0 12 dyl
13
6Efz 4Efz 6Elz 2Elz
0 0
Mzl [3 T
= .................................................................................................................................................................
1 I
ezl
EA EA
Px2 - 0 0 0 0 dx2
I
12£/z 6Elz 12Efz 6Elz
0 0 -13-
Py2 13
-
12 [2 dy2
6Elz 2Efz 0 6Efz 4Efz

Mz2 0 ez2
[2 I z2 I
(8.24)
Once again we must transfer member coordinates to system coordinates.
Equation 8.9 gives the rotation matrix R for a rotation (J about the z axis. The
required matrix can be expressed as
(8.25)
As in equation 8.12
K'=TTKT
Making the substitution i\ = cos (J and p. = sin 8 the rotation matrix becomes
R= [ -~ ~ ~] (8.26)
Writing I =Iz
2 12/ 2
AX +/2JJ.
12/
AXJJ.-[2XJJ.
6/!J.
I ii -A"'2 -[2JJ.
12/ 2 12/ 6/JJ.
-AXJJ.+p"AJJ. - f
6/A i 12/ 12/ 6/A

- : -AXJJ.+- XJJ. -AJJ. 2 - - X 2 -
6/JJ. 6/A
I \ 6! 12 ~~X I
- ~ i TJJ. I ll
I I
(8.27)
6/JJ. 6/JJ.
I I
6/A A JJ. 2 +12/A2 6/A
- - - -
I 12 I
6/JJ. 6/X 6/JJ. 6JA
ll 41
I I I I
It should be noted that if the matrix is partitioned as shown by the dashed

lines and expressed as
[ K~l K~z]
K21 K22
then K~ 1 = K~ 2 apart from the terms 6I/l where the signs are changed. Also
K~2 = K7t.
As an example we shall find the displacements and the forces at the nodes for
the problem in figure 8.8a. The two members are identical and have the following
properties: A= 104 mm2; I= 500 x 1Q6 mm 4; E = 210 kN/mm 2. The uniformly
distributed load has a total value of 3P.
The following quantities will be useful when setting up the stiffness matrix:
I/l = 10s mm 3 ; IjZ2 = 20 mm2.
;-.
4
---- 4m -
(a l (b)
Figure 8.8
For member AB: A= 0; j.( = I.

2 3 4 5 6
240 0 -6 X 10 5 -240 0 -6 X 10 5
0 104 0 0 -104 0 2
I E -6 X 10 5 0 2 X 109 6 X 10 5 0 109 3
KAB =-
I -240 0 6 X 10 5 240 0 6 X 10 5 4
0 -10 4 0 0 104 0 5
-6 X 10 5 0 109 6 X 10 5 0 2X 109 6
For member BC: A= 4/5; JJ = 3/5
4 5 6 7 8 9
4685 - 36 x 10 4 - 6486 -4685 -36 X 104 4
'
K8 c=/
E
[ M~
4685
- 36x 10
4
3754
48 X 104
48 X 104
2 X 10 9
- 4685
36 x 104
-3754
- 48 x 104
48 X 104
109
5
- 6486 - 4685 36 X 104 6486 4685 36 X 10 4 7
- 4685 - 3754 - 48 X to" 4685 3754 - 48 x to• 8
- 36 x 104 48 X 104 w• 36 X 104 - 48 x 104 2 X 109 9
It would now be possible to assemble the complete stiffness matrix, but as we

have seen in the previous examples we shall only require K{_L and KRL to com-
plete the calculations. Time and effort is saved if these sub-matrices are written
down.
From figure 8.8b it is seen that the unknown displacements are composed of
the horizontal and vertical displacements and the rotation at B, and also the
rotation at C where there is a pin joint. The matrix Ki.L will then be a 4 x 4
matrix with terms involving the displacements at 4, 5, 6, and 9: so that we need
to abstract terms for each member stiffness matrix and add corresponding terms.
In fact the matrices KLL and KA.L are formed from columns 4, 5, 6, and 9. We
can then separate the matrices by shifting the lines corresponding to 4, 5, 6, and 9
to the top of the matrix.
4 5 6 9
6726 4685 24 X 104 -36 X 104 4
I E 4685 13 754 48 X 104 48 X 104 5

KLL =-
l 24 X 10 4 48 X 104 4 X 10 9 10 9 6
-36 X 104 48 X 104 10 9 2 X 109 9
4 5 6 9
-240 0 -6 X 105 0
0 -10 4 0 0 2
I E 10 9 0
KRL=- 6 X 10 5 0 3
l
-4685 -4685 36 X 104 36 X 104 7
-4685 -3754 -48 X 104 -48 X 104 8
The load matrix PL is a null matrix since none of the applied loading acts at
the nodes. We shall next have to find the column matrix PFL· The fixed-end
forces for each member are given below.
1 2 3 4 5 6
AB { -1·5P 0 1250P -1·5P 0 -1250P}
4 5 6 7 8 9
BC {O 5P 5000P 0 5P -5000P}
therefore
1 2 3 4 5 6 7 8 9
PL- PFL ={1·5P 0 -1250P 1·5P -5P -3750P 0 -5P 5000P}
Applying equation 8.3
6726 4685 24 X 104 -36 X 104 d4 1·5P

E 4685 13 754 48 104 48 X 104 ds -5P
-
X
=
l 24 X 104 48 X 104 4 X 109 10 9 d6 -3750P
-36 X 10 4 48 X 104 10 9 2 X 109 d9 5000P
Solving these equations gives

PI
d4 = 104 X 10- 5 -
E
6 _5 PI
d 5 =-78· x 10 E
_8 PI
d6 = -186 X 10 E
_8 PI
d 9 = 380x 10 E
Substituting I= 5000 mm; E = 210 kN/mm 2 gives the following values for
displacements, where Pis in kN
d 4 = 0·0248P mm d 5 = -0·0187Pmm
d 6 = -4·421 x 10-5P rad d 7 = 9·058 x 10-5P rad
The values of the reactions at the supports can be found from

PR = PFR + K~LdL
The matrix K~L has already been formed and dL is now known, and we shall
find that
P 1 =-1·5P+0·864P=-0·636P
P2 = 0 + 7·858P= 7·858P
P3 = 1250P-1233P = 17P
P7 = 0 -2·363P= -2·363P
P8 = SP - 2·858P = 2·142P
To fmd the individual member forces

P' =P~ +K'd
is applied for each member, where K' is the member stiffness matrix in system
coordinates.
For member AB
P~x = -1·5P+ 0·864P= -0·636P p'Bx = -1·5P- 0·864P = -2·364P
p;,._y = 0 + 7·858P = 7·858P P~y = 0- 7·858P= -7·858P
p;,._z = 1250P- 1233P= 17P P~z = -1250P- 3089P= -4339P
For member BC
P~x = 0 + 2·364P = 2·364P Pbx =0- 2·364P= -2·364P
P~y = 5P + 2·858P = 7·858P Pby = 5P- 2·858P= 2·142P
P~z = 5000P- 660P = 4340P Pbz = -5000P + 5000P= 0
These results can be transferred to member oriented axes by making use of
the rotation matrix (equation 8.27).
8.9 Grillages
In this type of structure the members all lie in a single plane but the loading is
applied in a direction that is normal to this plane (figure 8.9a). A typical example
of this would be a system of beams that forms the floor of a building. In general
not a great deal of accuracy will be sacrificed if axial forces and My are neglected.
This will mean that for a member whose longitudinal axis is in the x direction, we
shall only concern ourselves with the moments and forces in figure 8.9b and their
corresponding displacements. Note that torsional effects are introduced.
(a) (b)
Figure 8.9
The stiffness matrix can be found by omitting terms from equation 8.5, and
becomes equation 8.28
12Elz 12Efz
[3 0 -~3- 0
Glx Glx
0 0 0 0
1 l
6Elz
0 0
f2
= 12Efz
0 -~3- 0
Glx Glx
0 0 0 0
l l
0 0
(8.28)
The rotation matrix R is the same as that given in equation 8.26. The combi-
nation of sine and cosine terms will be similar to that for a plane frame when trans-
ferring from member to system coordinates, equation 8.27. It is not proposed to
write out the K' matrix; this can be obtained from other more detailed works on
stiffness and flexibility matrix methods. It is also not necessary to work an
example since the procedure is the same as that used for plane frames.
8.1 0 Special cases

A pinned support in a rigid jointed structure presents no special difficulty-this
was seen in the example that was solved. In effect all that results is a further state-
ment that the value of the moment at the support is zero, and the rotation at the
support becomes an unknown displacement.
If symmetry is present the order of the stiffness matrix can be considerably
reduced, it will only be necessary to treat one half of the structure. It will be poss-
ible to set to zero certain displacements in the plane of symmetry.
The effects due to temperature change can be included very readily. If a mem-
ber is free to expand when the temperature rises the extension due to a tempera-
ture rise ofT would be laT. If the ends of the members are fixed, a compression
force of value El aT would be developed in the member. This effect can be included
in the initial state when we consider the structure to be built in at the nodes and
determine the fixed-end effects. It is only necessary to modify equation 8.1
slightly.
P= Py +Pr +Kd (8.29)
where Pr is a columh matrix representing the forces induced by the temperature

change. The matrix Pr can of course be partitioned into Pn and PrR. and the
various equations introduced at the start of the chapter may be suitably modified.
8.11 Outline of the flexibility method

A brief description of the flexibility approach was given in chapter 6. Basically a
redundant structure was first made statically determinate by removing a suitable
number of restraints or redundancies. There can of course be a number of differ-
ent forces that could be considered as the redundants in a particular problem.
With the structure now statically determinate we have what is sometimes referred
to as the primary structure. Displacements corresponding to the redundants are
obtained for the primary structure.
When the external loading is removed we have what is referred to as the second-
ary structure; unit loads are applied in the directions of the redundants and the
displacements in these directions are determined. We have shown that
(8.30)
where L referred to an applied load and R to a redundant.

It is possible to make the solution of a particular problem more general than
this, but equation 8.30 will have to be written in a more specific way. A matrix
FRL can be introduced which will give the displacements corresponding to the
redundants in the primary structure when unit loads are applied in the same direc-
tion as the applied loading L. The flexibility matrix F in equation 8.30 should be
more specific and would be better written FRR as it refers to the displacements
corresponding to the redundants when unit loads are applied in the direction of
the redundants. Hence
(8.31)
We stated that we were trying to make the application more general. This
might mean that we require the displacements of various points on the structure.
It will be seen that the displacements corresponding to applied loads L can be
written
(8.32)
Equations 8.31 and 8.32 can be combined and written as a partitioned matrix
as follows
(8.33)
It will be noted that the displacement matrix d has a subscriptS. The reason
for this is to differentiate between the systemS and a member M. We shall use
member and system flexibilities in the same way as we did in dealing with the
stiffness approach. It is easier here however to make use of the letters S and M
The partitioned matrix in 8.33 should be F8 .
We shall need to transfer from a member flexibility matrix to a system flexi-
bility matrix. This can be demonstrated by using an energy approach.
The external work done when a set of loads P are applied to a structure can be
written as
(8.34)
In this case P will consist of external loads L and the redundants R. It is obvious
that the system flexibility matrix must be used for this purpose.
If the flexibility matrix were determined for an individual member, the energy
stored in the member when the external loads are applied could be written down
in terms of the force in the member PM and the flexibility matrix. Summing the
energies for the individual members will give the total energy stored. This can be
written as
!PLFMPM (8.35)
Now the external work done must equal the total energy stored.
pTpsP = PLFMPM (8.36)
It is quite possible to make use of a transformation matrix TMs to find the
values of the forces in the members in terms of P
(8.37)
Hence from equation 8.36
or
(8.38)
Now the transformation matrix can be partitioned in terms of two sub-matrices,
one containing the terms involving L and the other with terms involving R.
Thus
It follows that
FLL = TLLFMTML
FLR = TLLFMTMR = FkL
(8.39)
FRL = TLRFMTML =FIR
FRR =TLRFMTMR
Very often the displacements of various joints in the structure will be required.
We shall denote these by the column matrix dJ.
If the end forces in the members were known it would be quite easy to use a
transformation matrix to transform the member end actions PM to system
orientated PJ.
(8.40)
TMJ would consist of member end forces due to unit forces applied in the
direction of system oriented loads or displacements.
According to equation 8.11 and using a slightly different notation since we
have several different transformation matrices
dJ=TLJdM
Now
Thus
We may also write
therefore
(8.41)
Use can be made of other transformation matrices, such that when unit loads
are applied in the directions of the loads L or redundants R the values of the
reactions or constraints at the supports C are
PeL= TeLL PeR= TeRR
The support reactions or constraints can be found from
Pe =PeL+ PeR= TeLL+ TeRR (8.42)
We are now in a position to set down the various steps that have to be followed
when using the flexibility method together with a member approach. It will be
assumed that a complete solution is required, and the following may have to be
found: forces at the ends of members, support reactions, displacements of the
various nodes in the structure, some of which may be points of application of
loads, and finally the deformation of the various members.
(1) It is first necessary to determine the degree of redundancy of the struc-
ture, and then to select the various quantities that will be known as the redundants
R.
(2) If the loading on the structure is not all applied through the joints it
will be necessary to replace any of these loads by equivalent loads acting at the
joints. These will of course be in the opposite sense to the member end reactions
and fixed-end moments. The total applied loads L are known.
(3) The flexibility matrices for the individual members are next found and
hence the flexibility matrix FM is assembled for all the members in the structure.
(4) We next consider the structure with the redundants removed. Unit
loads are applied in the directions of L and R and the following transformation
matrices determined. TML• TMR• TeL• and TeR·
The further transformation matrix TMJ is also determined. Unit loads are
applied in the direction of joint or node displacements that are required and the
resulting member end forces are determined.
(5) The flexibility matrices FLL, FLR• FRL and FRR are determined from
equation 8.39.
(6) The matrix dR is next obtained; this may very well be a null matrix.
dR = FRLL + FRRR
Thus
(8.43)
If however dR is zero
R=-F.RkFRLL (8.44)
the redundants will now be known.
(7) The support reactions can now be found
Pe =TeLL +TeRR (8.45)
(8) The member end forces can be found from
PM= TMJL + TMRR (8.46)
Some of the end forces will have to be corrected if the loads L were not all
applied through the joints. The method of correction will become apparent in an
example that follows.
(9) The displacements of joints can be found from
dJ=TLJFMPM (8.47)
or in the case of a joint to which a load Lis applied
dL = FLLL + FLRR (8.48)
8.12 Member flexibility matrix

It is a fairly straightforward matter to derive the flexibility matrix for a member
shown in figure 8.1 0. Once again the x axis has been made to coincide with the
longitudinal axis of the member; they and z axes are both principal ones. We
require the displacements that result from applying a unit force in a particular
direction at node 2 when node 1 is fixed. It is obvious for an axial force of unity
in the x direction, 1, that the only displacement is in the x direction and has a
value of l/EA. Similarly a unit moment about the x axis, 4, will produce a
rotation of l/GJx.
y
5
z
Figure 8.10
The rest of the flexibility coefficients can be determined from cantilever

theory. For example a unit force in direction 2 will result in a displacement in
that direction of l 3 /3Elz, and a rotation or slope at the end (corresponding to 6)
of f/2Elz.
The rest of the terms are left for the reader to derive.
The complete matrix can be assembled for the member and will be 6 x 6.
Note that d 1 to d 3 are deflections, d 4 to d 6 rotations,P1 toP 3 forces, andP4 to
P 6 moments. This numbering system has been chosen in preference to that used
in the case of a stiffness matrix since it will be easier to describe member end
displacements and actions.
EA 0 0 0 0 0
0 0
0 0
(8.49)
0 0 0 0 0
ds 0 0 0 0 Ps
Ely
0 0 0 0
This matrix is of course written for a general member in space and may be
reduced for a particular class of problem. We shall in fact treat two problems, the
first being the pin-jointed truss that was solved using a stiffness approach in
section 8.5. For a case of this type, the flexibility matrix will degenerate to one
term only, namely 1/EA.
If a little thought is given to the matter it will be seen that the complete flexi-
bility matrix for the members will be formed from the individual flexibility
matrices without changing any of the terms, the reason for this being that end 1
of the member has been regarded as being built in, and we are only interested in
the displacements that will occur at end 2. It can be seen in general we shall have
a band matrix. For the case of a pin jointed frame this will become a matrix with
the leading diagonal having terms of the type 1/EA and all other terms will be zero.
It should be noted that this differs from the stiffness matrix, where a particular
term in the system matrix was formed from a number of terms from the member
matrices transformed into system coordinates. The reason for this was that we
were interested in the force required to produce a unit displacement in a particular
direction in the structure.
8.13 Example of a pin-jointed redundant truss

The problem to be solved by a flexibility approach has already been solved by
stiffness in section 8.5. The members all have the same cross-section and are made
from the same material (figure 8.11 a). The vertical and horizontal members have
the same length /. It will be found convenient to letter each of the members.
Every node in the frame is numbered such that a horizontal component of force
or displacement can be recognised (b).
Following the various steps that we have set down in section 8.11 it is seen at
once that the structure is singly redundant. It is suggested that member d is con-
sidered as the redundant member.
The uniformly distributed load is replaced by loads each of P/2 acting at nodes
2 and 3 (figure 8.11c).
The flexibility matrix for the structure can now be assembled. Considering the
members in alphabetical order
1 0 0 0 0 a
0 1 0 0 0 b
I
FM=- 0 0 1 0 0 c
EA
0 0 0 y2 0 d
0 0 0 0 y2 e
Unit loads are applied at nodes D, B, and C, these being the nodes at which
external loads are applied. The unit loads will be along the same line as the
external loads but will be applied in the positive sense according to the directions
of the node displacements or forces 4, 6, and 8. The forces are found in the
(a) p
-:t[SJO -i2 0
~[S]o
-1t
'1 -V,
r
Pa: 1!
~t 1
-1 l -1 1
(d) T'4:1 (f)
1
0 -7'!
~[SJ1~
;.~=1
R=1
1 ...1.
V2
0 0
--fit
J_
at
:..+(g)
~ ~
0 0
0
(j)
Figure 8.11
individual members as are the reactions at the various supports (figures 8.11 d, e
and f) .
We can now write down the transformation matrices TML and TeL
p4 p6 Ps
0 0 0 a
0 0 b
TML = 1 0 1 c
.....j2 0 -../2 d
0 0 0 e
p4 p6 Ps
-1 0 -1 1
TeL= 0 0 0 2
1 0 5
-1 -1 -1 6
A unit load is next applied in the direction of the redundant member (figure
8.11g). This will enable the transformation matrices TMR and TcR to be written
down.
1 0
a
..J2
1
1 2
b ..J2
TMR = ..J2 TcR
1 0 5
c
..J2 1
6
1 d ..j2
e
If a complete solution to the problem is required it will be necessary to apply

unit loads in directions 3 and 7 (figures 8.11h and j). The transformation matrix
TMJ is then obtained.
p3 p1
0 1 a
0 0 b
TMJ = 1 0 c
0 0 d
0 0 e
The only other matrix that will be required is that of the load matrix L.
L=P [-1]
-!
-!
4
6
8
We are now in a position to determine the matrices FLL• FLR• FRL• and FRR·
The method used is set down in equations 8.39.
0 1 -../2 1 0 0 0 0 0 0 0
1
FLL = -
EA
l~
0
0 0 0
-../2 ~] 0
0 0
0 0
0
0
0
0
1
0
0 1
0 0 0 ../2 0 -../2 0 -../2
0 0 0 0 ../2 0 0 0
1 l3·83 0 3·83]
=- 0 0 0
EA
3·83 0 4·83
I l-2·707]
FLR = EA 0
-3·414
I
FRL = EA [-2·707 0 -3·414]
I
FRR = EA [4·332]
The force in the redundant member can now be found. The value of dR is zero,
hence
-3414]Hl
R = -F.RkFRLL
R = -P[4·332] - 1 [-2·707
0
R 5 =-1·02P
The support reactions are found from
Pe =TeLL+ TeRR
Pe1 -1 0 -1 0 [ -1·02 J
Hl+P
0 0 0 1
Pe2
=P
Pes 0 1 ../2
Pe6 -1 -1 -1 0
1
../2
Pe 1 = 1·5P;Pe 2 = 0·72P;Pe 3 = -1·5P;Pe 4 = 1·28P.
The member end forces are found from
PM= TMLL + TMRR
Pa 0 1
0 0 -1
y2 [ -1·02]
pb 0 0 -! 1
y2
= p +P
PC 0 -! 1
y2
Pct -y2 0 -v2
Pe 0 0 0 1
Pa = 0·72P; Pb = 0·22P; Pc = -0·78P; Pct = 1·10P;Pe = -1·02P.

The displacements of the positions at which external loads applied are given by
dL = FLLL + FLRR
d4l = - r3·830
d6
PI 0
0
3·83ll-1l
0
P/l-2·707l[-1·02]
_, + - 0
l EA EA
d8 3·83 0 4·83 -! -3·414
d
d 4 = -2·99(Pl/EA); d 6 = 0; 8 = 2·77(Pl/EA).
The displacements of other joints can be found from
dJ =T_LJFMPM
[d3]
d 7
=~[0
EA 1
0
0 0
0
0 ~] 0
0
1 0
0 0
0
0
0
0·72
0·22
0 0 0 0 -0·78
0 0 0 y2 0 1·10
0 0 0 0 y2 -1·02
d 3 = 0·78(Pl/EA); d 7 = 0·72(PljEA).
The problem has now been completely solved giving the forces in all the mem-
bers, the reactions, and the displacements of the nodes in the framework.
8.14 Example of a plane frame

As a further example of the application of the flexibility approach we shall solve
the problem of a plane frame taking axial loads into account. The frame is shown
in figure 8.12a and is the same frame that was solved in section 8.8 using a stiff-
ness approach. It will be possible to reduce the member flexibility matrix to a
3 x 3 matrix.
dl -I 0 0 pl
A
1 12
d2 =- 0 p2 (8.50)
EI 3 2
d3 0 l p6
2
The frame is redundant to the second degree and it is proposed to make the
frame statically determinate by the removal of the horizontal and vertical
reaction at C. Directions of system displacements and forces are shown numbered
I to 9 in figure 8.12b. The frame consists of two members a and band in figure
8.12c we have each member coordinate system together with member end load or
displacement directions, which are labelled a2 b6 etc.
The fixed-end moments and loads are next calculated, and when these are
reversed in direction they will become the loading system applied at the nodes (d).
Loads are applied in system directions 1, 3, 4, 5, 6, 8, and 9.
The two members a and b are identical and so the flexibility matrices will be
the same. For the case where I= 5 x 10 8 mm 4 ; A= 104 mm 2 ; 1 = 5000 mm
0 0
25
0 -x 106 2500
3
0 2500
The complete member flexibility matrix can now be assembled and will again
be a banded matrix.
Unit loads are now applied at the nodes in the positive sense of directions I, 3,
4, 5, 6, 8, and 9, that is, the points at which external loads are applied. The
reactions are found at the supports and the forces at ends 2 of the members. The
effects are shown in figure 8.12e to 1 where two diagrams are shown for each unit
load, the first showing the load applied and the reactions that result, and the
second the member end actions. k for example shows the effect of a unit load
being applied in direction 8. This will result in a reaction of -1 at 2 and moment
of -4000 at 3. For member a there is a force of I at a1 and a moment of 4000 at
a6; for member b there are two forces, 3/5 at b1 and 4/5 at b2.
281
3m
I
5
T>r
+....... 4
6 a5
I ./---· -4m -
a
fal {b) !cl
SP
lp .......
2
t 3750P
fdl fel It!

The matrices TML and TeL can now be written down.
Pt P3 P4 Ps P6 Ps p9
0 0 0 1 0 1 0 al
0 0 -1 0 0 0 0 a2
TML =
0 0 0 0 4000 a6
0 0 0 0 0 i 0 bl
0 0 0 0 0 ~ 0 b2
0 0 0 0 0 0 b6
p3 p4 Ps Pr. Ps p9
TeL= l-~· 0
0 -1
0 0 -1
0 0
0
0
-1 or
0 2
0 -1 5000 0 -1 -4000-1 3
Unit loads are now applied in the directions of the two chosen redundants,
that is, 7 and 8 (figures 8.12m and n). This enables the matrices TMR and TeR
to be found. The load matrix Lis also written down.
i 1
l
R7 Rs
0 1 al -1250 3
R7 Rs
l-1
-1 0 a2 ~ 4
-3000 4000 a6 -10 L=P -5 5
TMR = TeR = 0 2
~ i bl 8000 -4000 3 -3750 6
-i ~ b2 -5 8
0 0 b6 5000 9
The following matrices can now be found from equations 8.39.
0 0 0 0 0 0 0
0 0 0 0 0 0 0
25
0 0 - X 10 6 0 -2500 -1 X 107 -2500
l 3
FLL=- 0 0 0 5 X 104 0 5 X 104 0
EI
0 0 -2500 0 4000 1
0 0 -1 X 107 5 X 104 4000 214 X 10 5 6000
0 0 -2500 0 1 6000 2
0 0
0 0
1583 -1000
1 X 104
FLR=-- 0 5
EI
-0·55 0·4
-2598 2140
-0·7 0·6
- 1x 104 [ 3537 -2598]

FRR- El -2598 2140
EI [2606 3163]
F.Rk = 1 X 1010 3163 4307
The values of the two redundants are found from

R= -F.RkFRLL
where FRL = FLR·
Performing the matrix operation gives
R 7 =- 2·364P R 8 =2·142P
The support reactions can be found from
Pc=TcLL+TcRR
r-1~
-~l '
0 -1 0 0 0
Pc=P 0 0 -1 0 -1 --1250
-1 5000 0 -1 -4000 ~
-5
-3750
-5
5000
J
+P
~
8000
-1
0
0
-1
-4000
r-2·364J
2·142
P1 = -0·636P; P2 = 7·858P; P3 = 17P.

The member end forces are given by

PM= TMLL + TMRR
It will however be necessary to correct the member end forces since the
external loads are not applied through the nodes. The true values of the member
end forces will be obtained by the addition of the ftxed-end actions. These are
the same as the load matrix L but with the sign changed. For end 2 of member b
it is of course necessary to resolve the loads in the directions b1 and b2.
TMLL TMRR
Pal -10 2·142 0 -7·858
Pa2 -1·5 2·364 1·5 2·364
Pa6 =P -18 750 +P 15 661 +P -1250 =P -4339
pbl -3 -0·606 3P -0·606
pb2 -4 3·132 4P 3·132
pb6 5000 0 -5000 0
Finally the displacements of the ends of the members can be found from
dJ=T_LJFMPM
Or in the system coordinates where there are any loaded nodes

dL = FLLL + FLRR
These latter values are set out below
dl 0 0 0
d3 0 0 0
d4 5·938 X 107 -5·886 X 10 7 0·052 X 107
PI PI PI
ds -5 X lOS +- 1·071 X lOS = - -3·929 X lOS
EI EI EI
d6 -22 500 21 571 -929
d8 -1·073 X 10 8 1·073 X 10 8 0
d9 -27 500 29 402 1902
Substituting I= 5000 mm; E = 210 kN/mm 2 ; and/= 5 x 10 8 mm 4 : d 1 = 0;

d 3 = 0; d 4 = 0·0248P mm; ds = 0·0187P mm; d 6 = -4-421 x w-s P rad; d 8 = 0;
d9 = 9·058 X IQ-Sp rad.
8.15 Choice of stiffness or flexibility approach

It could perhaps be argued that the choice is simple to make and depends on com-
paring the degree of kinematic indeterminacy with the degree of redundancy in a
particular structure. If there are more unknown displacements than redundants
the flexibility method should give the shorter solution. This may well be so when
working by hand.
However if a computer is used the stiffness approach is generally to be pre-
ferred. It is possible to write a general programme which can quite easily be
modified to take account of specific cases. In the case of flexibility a general
programme is much more difficult to write. Again a stiffness approach requires
no particular choice of redundants whereas with flexibility it is necessary to
decide at the outset what the redundants should be. A particular set of redundants
may very well give a more rapid solution than another set. This means that the
speed of the result can depend on the skill of the operator.
One topic has deliberately been omitted from the subject material covered.
This is the connection matrix, which can be used to advantage when using a com-
puter. This and other more advanced methods will be found in books that are
devoted to the use of matrix methods in structures.
Problems
8.1 The members of the pin-jointed steel structure P 8.1 all have the same cross-
section. Using a stiffness approach, determine the displacements of D and the
forces in the members.
A
8.2 Find the stiffness matrix for the beam shown in P 8.2.
p 8.2
8.3 Find the stiffness matrix for the frame shown in P 8.3. All members have a
cross-section of A and a second moment of area of I.
p 8.3
8.4 If axial displacements can be neglected in P 8.3 determine the rotation of the
joint at B.
8.5 The load Pis applied at the centre of the beam of the portal frame in P 8.5.
Determine the displacement of Busing a stiffness approach. It may be assumed
that B and C do not deflect horizontally.
p
8 t 2! c
-~
I
1
p 8.5
8.6 Solve problem 8.1 using a flexibility approach.
8.7 Solve problem 8.5 using a flexibility approach.

9 INFLUENCE LINES FOR STATICALLY
INDETERMINATE BEAMS
9.1 Beams with two spans

In chapters 1 and 2 the virtual-work or virtual-displacement method was found to
be the most convenient approach when sketching the influence lines for various
functions in statically determinate structures. A similar method can be employed
when the system is statically indeterminate. The application of virtual work for
determining the influence lines both for statically determinate and indeterminate
cases is known as the Miiller-Breslau principle. Basically this states that the
influence line for a function will be represented by the deflection of the load line
of the structure when a displacement is applied corresponding to the function.
For the case of a statically determinate structure the influence line will consist of
a series of straight lines, whereas that for the indeterminate structure will be
curved.
A two-span continuous beam is shown in figure 9.1a. We shall first discuss how
the influence line for the reaction at A might be found.
If end A is given a unit vertical displacement and the beam is restrained to stay
on the supports at B and C, it will take up a shape as shown in b. The actual dis-
placement curve can be found by beam deflection theory.
Making use of virtual work
VA X 1-Pd=O
If Pis now set equal to unity then VA =d. This means that the deflection
curve represents the influence line for the vertical reaction at A.
It is possible to obtain the influence line for shear force or bending moment in
a similar manner. For the shear force atE a parallel mechanism is inserted as in
section 1.20. Equal and opposite forces Q are then applied. The resulting displace-
ment curve which will represent the influence line for shear force is shown in c.
Note that at E the tangents to the two parts of the curve must have the same slope.
It would be very easy to sketch the incorrect diagram for the left-hand portion of
the curve. If Q is imagined to act alone this would appear to be straight. There is
however a moment in addition to the shear force and the curvature shown in the
diagram will result.
For bending moment a pin is inserted and equal and opposite moments applied
to either side of the pin, which will cause the beam to deflect as shown in d. The
287
t--x----lp
r-a-j t
E D
(a)
(b)
""'JJ;~----:.......;.;;"'
~ (d)
(e)
Figure 9.1
values of the applied moments are increased until the relative displacement of the
two portions of the beam is equal to one radian.
By virtual work
MO- Pd= 0
Put () = 1 and P = 1; then M =d.
Instead of obtaining the influence line for the shear force and bending moment
atE from first principles, it is possible to modify the influence line for VA·
Considering the shear force first, with a unit load applied-when the load is to
the left of E the shear force atE has a value of (VA - 1), and when the load is to
INFLUENCE LINES FOR STATICALLY INDETERMINATE BEAMS 289
the right of E the shear is equal to VA. Thus it is only necessary to modify the
first part of the influence line for VA• to give the influence line for Q shown in
figure 9.lc.
The bending moment at E, with the load to the left of E is given by
ME = VA a- l(a- x)
With the load to the right of E
Thus if a similar diagram to figure 9.lb is drawn, but with a height a instead of
unity at A, this will represent the influence line for bending moment when the load
is to the right of E. The other part has to be modified by subtracting a - x. This
is shown in e.
The influence line for the reaction at A can be also found by a slightly different
approach using flexibility coefficients. If the support at A is removed, when the
applied load Pis in position A would deflect downwards a distance fA 0 P. With
the external load equal to VA applied at A the point A would deflect upwards an
amountfAAVA·
Thus- fAnP +fAA VA= 0 since the true deflection at A is zero. Therefore
VA= fAn P
fAA
Now fAn= InA and puttingP= 1
VA= InA
fAA
so that the influence line for VA is given by the ratio of the deflection at D to that
at A when a unit load is applied at A.
The influence lines for both shear force and bending moment can be found by
a similar approach.
9.2 Betti's reciprocal theorem

This is an extension of the Maxwell reciprocal relationship that was derived in
chapter 6.
Two different loading systems are applied to the same linear elastic structure
(figure 9.2). With the loading system in 9.2a the deflectionsd 33 and d 43 are
measured, corresponding to the directions of the loads that are applied in system
b. Similarly for this system where the loads P 3 and P4 are applied, the deflec-
tions d 1b and d 2 b are measured in the directions of the loads that are applied to
system a. Now
d3a = htPt + [32P2
d4a = f41P1 + f42P2
(a) (b)
Figure 9.2
and
dtb = fnP3 + ft~4
d2b = f23p3 + f24p4
Multiply the loads on system a by the corresponding displacements on "System b
(9.1)
Next multiply the loads on system b by the corresponding displacements on

system a
(9.2)
Now we know that from the Maxwell reciprocal relationship that / 13 = ht.
etc. If the two expressions are multiplied out and compared term by term they
will be found to be identical.
Betti's theorem states, for a linear elastic system with two different sets of
loads applied, that if the first set of loads is multiplied by the corresponding
displacements produced by the second set of loads, the result equals that
obtained by multiplying the second set of loads by the corresponding displace-
ments produced by the first set of loads.
This theorem can also be demonstrated by the use of simple matrix algebra
da = FPa
db= FPb
Pidb = PiFPb = (PiFPbf
since F = FT
9.3 Applications of Betti's theorem to influence lines

We shall consider the problem that has already been discussed, that of finding the
influence line for the bending moment at E (figure 9.3a).
f
tA E fs t
(a)
~~(
t
(b)
~
___,_'_________a__'_,_,_,__
SEA eEB
ooc:::::::::::: ~
(c)
Figure 9.3
The same beam with a pin inserted at E and unit couples applied at either side
of the pin is shown at b and in the displaced position at c. eEA is the slope of
member AE atE and eEB is the slope of BE at E. dis the deflection corresponding
to the unit load. eE is the actual slope atE in the original problem and ME is the
bending moment corresponding to this.
Applying Betti's theorem
MEeEA + MEeEs - 1 x d = 1 x eE - 1 x eE
therefore
(8EA + eEB) is the rotation of the two parts of the beam at E, and if this is set
equal to unity the same expression results as that obtained by the application of
virtual work.
The influence line for bending moment at B is required for the beam in figure
9.4a. Note that there is a different cross-section for each span. The deflected form
of the beam with a pin inserted at B and unit moments applied is shown at b, and
the resulting bending-moment diagram at c.
~~--x ---.;·r l [2
t.A-----t,---------<•~t-.B---/2
. __jC
(a)
(b)
(c)
(d)
(e)
(f)
(g )
Figure 9.4
Applying the second moment area theorem about A for span AB, and about
C for span BC, the slopes at B can be found . Then the first theorem can be applied
to fmd the slopes at A and C.
(J - ___}j__ (J BC = __}]:,_
BA - 3Eft 3£/2
(J - _11_ (J - __}]:,_
AB- 6£/1 CB- 6£/2
Applying the second moment area theorem again the deflection under the unit
load at a distance x from the left-hand support can be found. The displacement
in the y-direction is
xl 1 x3
v=-----
6£11 6EI 1 lt
therefore
v = _x_ (1 2 - x 2) 0<x < 11

6£1111 1
We have already shown by the application of Betti's theorem that
Ms = v/(OsA + Osc), therefore
Ms = x{lf - x 2 )/_l_(!J_ + l2)

6EI 1 lt 3E /1 /2
x/2 ( If- x 2 )
= 211 11h + 1211
If xis measured from the right-hand support at C
xi (l~-x )
Ms=-
1 2
O<x <I2
212 1112 + 1212
The moment at B is negative and is sketched at d.
The reaction VA could now be found by statics. An alternative approach how-
ever would be as follows: the hinge inserted at B is left in position and the end A
of the beam is given a unit vertical displacement (e).
Applying Betti's theorem
VA x 1 - 1 x d +M 8 x (/> =0
therefore
VA =d-MB(/>
The value of dis {1 1 - x)/11 for x < 11 and(/>= 1/11 • For x > 11> d = (/> = 0.
l1-x x/2( If-x 2 )
VA =-~-1-- 2lf l1l2 +l2l1 ;
0 <x < 11> x measured from A.
-x/ 1 ( 1~ -x 2
v --- )
A- 21112 f1/2 +12/l
0 < x < 12 , x measured from C.
A sketch of the influence line is shown at f.
The influence line for bending moment at any other point can easily be deter-
mined. Since a pin has been inserted at B the statically determinate bending-
moment diagram can be drawn; this has then to be corrected for continuity at B.
Consider the bending moment at the mid-span of AB. A pin is inserted at the
point and moments are applied until the two parts of the beam have a unit
rotation (figure 9.4g).
Mxi-Ixd+M8 x!=O
Ms
M=d--
2
The influence line can now be drawn.
In the last two cases the system has been made statically determinate, and the
influence line for the required quantity drawn and then corrected. This method
of approach can be used for beams with a higher degree of redundancy.
9 .4 Multi-span beams
The influence line is required for the reaction at A (figure 9.5). If the supports at
B and C are removed (b) the influence line can easily be drawn. This will need to
be corrected since the values of VA should be zero when the unit load is at B and C.
(a)
~ (b)
___
~~ ....._
(c)
~-[7 (d)
(e)
Figure 9.5
The influence line for the reaction at B is next determined when the beam is
simply supported at A, B, and D. This will be the deflection curve for the beam
simply supported at A and D with a load applied at B such that the deflection at
B is unity (c). In a similar way the influence line for the reaction at C when the
beam is simply supported at A, C and D can be drawn (d).
The original influence line can now be corrected. The apparent reaction da
can be reduced to zero by adding the correct proportion of the reactions given by
c and d. Thus
and
dc 1 x K 1 +I x K 2 +de= 0
From these two equations we can determine the value of the two constants K 1
and K 2 • The correct influence line for VA can now be drawn. It will consist of
figure b plus K 1 times figure c plus K 2 times figure d. A sketch of this is shown
at e.
The influence line for bending moment at a point on a beam can be found by
a similar approach. As a numerical example we shall find the influence line for
bending moment at B for the beam with three equal spans (figure 9.6).
--o.-~·t-c- -- ---to
(a)
(b)
(c)
Figure 9.6
If the supports at B and C were removed the required influence line would be
as shown at b with a maximum value of 21/3 at B and a value of 1/3 at C.
The reaction at B with the beam supported at A, B, and D is now found. We
require the deflection curve for a simply supported beam with a load at B.
d 2 v 2Px
E I - = - -P[x -I]
dx 2 3
x measured from A.
dv 2Px 2 [x -1] 2 A
EI-=---P +
dx 6 2
2Px 3 [x-1] 3
Elv = - - - P +Ax+ B
18 6
At x =0, v =0 therefore B =0. At x =31, v =0 therefore A= -5P1 2 /9.

We shall confme out attention to a few salient points. These are set out in
table 9.1 together with the corresponding deflection if the value at B is taken as
unity. k = P13jEI.
Table 9.1
X v Relative
value
l 19
--k 0·594
2 72
32
--k 1·0
72
31 69
--k 1·078
2 144
28
21 --k 0·875
72
51 31
--k 0·484
2 144
We have considered the beam supported at A, B, and D and have the influence
line for the reaction at B. Due to the reaction at B alone the reactions at A and C
will be some proportion of RB and the bending moment at any point will be
proportional to a length times the value of RB·
Since the problem is symmetrical the values in the table may be used to find
the influence line for the reaction at C when the beam is supported at A, C and D.
The influence line for the bending moment at B will be a constant times the value
of this influence line.
It should now be possible to combine the three influence lines to find the value
of the constants, as in the original problem the bending moment is zero when the
load is at B or C. At B
21
- + K 1 + 0·875K2 = 0
3
AtC
l
-+0·875K 1 +K2 =0
3
whence K 1 = -1·63/ and K 2 = 1·09/.
The influence line for bending moment at B can now be found; the values are
set out below at l/2, 31/2 and 5//2.
l 31 51
2 2 2
B.M. stat. det. 0·3331 0·51 0·167/
Correction for VB - 0·968/ -1·7551 -0·791
Correction for V c 0·5281 1·172/ 0·6461
~ - 0·1071 -0·0831 0·0231
A sketch of the final bending moment diagram is shown in figure 9.6c.
Problems
9.1 A uniform beam is of length I and uniform stiffness£/. The left-hand end is
simply supported and the right-hand end is fixed. Calculate the deflection curve
when a rotation (} is induced at the left-hand end by applying a moment to it.
Hence find the influence line for the moment induced at the left-hand end by a
unit vertical load acting on the beam when both ends are fixed.
9.2 Starting from first principles prove the Miiller-Breslau theorem. Obtain the
deflection curve for a simply supported uniform beam of flexural rigidity EI and
length l when a moment is applied at one of the supports. Use this result to deter-
mine the influence line for the support moment at the centre B of a two-span
uniform beam of length 2/, given that the supports A, B, and C are level and un-
yielding. Deduce the influence line for the moment at a point D at the mid-span
of AB.
9.3 A cantilever of length lhasa prop placed at the free end. Find the influence
line for the force in the prop.
9.4 A simply supported beam of span 3/ has an intermediate support at I from

one end. The flexural rigidity of the beam is EI. Determine the influence line for
the reaction at the intermediate support.
9.5 A symmetrical parabolic arch of span land rise h is pinned at the abutments.
The slope of the arch at any point is !f> and the flexural rigidity is given by
EI cos 1/>, where EI is the value at the crown. Determine the influence line for the
horizontal thrust.
10 STABILITY OF COLUMNS
10.1 Introduction
Let us consider the case of a perfectly straight long slender rod of uniform cross-
section and made from a linear elastic material. The rod is first tested in tension.
An axial load is steadily applied and increased until the yield stress is approached..
If a small lateral disturbing force were applied during the test the rod would bend
slightly, but would return to the original straight state as soon as the lateral load
is removed.
If the same rod is next tested in axial compression it will remain in a state of
equilibrium until a certain critical load is reached. Thus it initially behaves in a
similar manner to the rod in tension. At this critical load the rod will remain bent
and will not return to its original straight state after the application and removal
of a small lateral load. The average compressive stress would be found to be very
much less than the yield stress of the material in compression.
There is then a very big difference in the behaviour of such a member when it
is loaded in tension or compression, and we shall seek to examine the compression
behaviour in more detail.
In practice it is not possible to obtain a perfectly straight uniform member,
and the material will not be perfectly homogeneous. Any deviation from perfec-
tion is equivalent to having a slight eccentricity of loading, producing bending in
addition to the axial thrust. The bending action is equivalent to the small lateral
load that was applied when the axial load reached a certain value at which
instability occurred. We shall refer 'to the load at which instability occurs as the
Euler critical load.
To obtain further insight into the question of instability we shall examine the
problem in figure 10.1 a in some detail.
Member AB is straight and may be considered as rigid. It is pinned at B, and
restrained at A by the two springs each of stiffness k. An axial force P is applied,
and also a lateral forceR. As a result, end A displaces through a horizontal dis-
tance x. The forces acting on end A are shown at b. While x is small, taking
moments about B
2kxl= Px + Rl (10.1)
298
STABILITY OF COLUMNS 299
X
x;;
p
Pc
(a) (b) (c)
Figure 10.1
If P = 0, x = R/2k. If we put this value equal to x 0 , equation 10.1 can be

expressed as
Xo
x=--_,_- (10.2)
(1-P/2k[)
Note that x becomes infmite when (1 - P/2k[) = 0, or P = 2kl. We shall call

this the critical load and denote it by Pc, so that 10.2 can be re-written
X
-=---
xo 1 -P/Pc
A graph of this is sketched in figure 10.1 c. It can be seen that asP increases
the rate of increase of x increases, and when P =Pc the displacement xis infinite.
This portion of the curve however does represent a stable situation. Consider any
point on the curve and displace A such that the value of x is increased slightly,
keeping P constant. For equilibrium in this new position a larger value of Pis
required; since this does not exist, when A is released it will return to its original
position. Now consider the negative part of the curve, which implies that the force
R is in the same direction as the restoring force in the springs. Again for any point
on the curve we have equilibrium. Displace A such that x becomes further nega-
tive, keeping P constant. For equilibrium a smaller value of P is required, but as
this does not exist a condition of inbalance exists and x will become still more
negative and we have an unstable situation.
Thus member AB is stable up to the value of the critical load but unstable
above this value.
We shall now proceed to derive values for the Euler critical load for struts
(that is, compression members) with various end conditions.
10.2 Euler critical loads

The classical derivation for the critical load of a strut pinned at each end
follows.
It is assumed that the strut is initially straight and that the load applied is
axial, and also that the material is linear elastic.
The strut in figure 10.2 is shown in its deflected form under the action of an
axial load P.
Figure 10.2
So long as the lateral displacements are not large we can write for the bending
moment
or
where a2 =P/EI.
The solution to this differential equation is
v = A sin ax + B cos ax
Certain end conditions apply
when x = 0, v = 0 thus B = 0
when x = 1, v = 0 thus A sin al = 0
This means that either A is zero or sin al = 0. The first would imply that the
value of v is zero everywhere and the member is perfectly straight.
If sinal= 0, al = 0 or n1t (n = 1, 2, 3, etc.). Discarding the solution al = 0
n 2 1t2 EI
P = a 2 EI = --=---
12
The smallest value of P occurs when n = 1 and the first critical load for a pin-
ended strut is
1t2EI
PE=~ (10.3)
The subscript E will be reserved for this particular critical load. For all others the
subscript C will be used.
We should note that the displacement corresponding to the load Pis the
shortening of the strut due to the bending effect. The deflection curve is of a
sinusoidal form but the value of A is indeterminate and could in fact have any
value. We can see that we have a situation where the value of the applied load is no
longer proportional to the displacement even though the material is still linearly
elastic.
The higher critical loads correspond to different buckling modes. For example
the case when n = 2 could be obtained if the centre point of the strut was con-
strained such that it could not move laterally.
The particular case that we explored was for a strut with pinned ends. Suppose
that the ends had been built in; this would mean that in addition to the axial load
P there would be a fixing moment M 1 . The differential equation would be modi-
fied to
d2v
EI dx2 = -Pv +Ml (10.4)
The solution of the complementary function would be identical to the pinned-

end case, but there would be a particular integral-a constant-to consider as well.
It can be seen that Euler critical loads can be found for other end conditions
by a similar approach. It is perhaps easier to make use of the solution that we
have already obtained for the pin-ended strut. For a number of other cases the
solution of the differential equation will always be part of the sine curve v =
A sin ax. So long as suitable axes are chosen the load Pc which gives the correct
displacement curve will be the Euler critical load for the particular set of end
conditions.
~ ~ D F
E
Figure 10.3
Figure 10.3 shows a continuous sine curve, and it can be seen that for the
case of a pin-ended strut the first Euler critical load rt 2 EI/1 2 would correspond to
the length AB. AC would correspond to the second critical load. Denoting the
distance between two points of contraflexure as L, that is, AB = L
rt 2EI
-L2
p E- -
If we express L as a fraction of the actual length of the strut we can obtain the
values of the critical loads corresponding to the higher modes. For AC, 2L = l
therefore Pc = 4rt 2 EI/1 2 , the second Euler critical load.
We shall next examine other possible end conditions.
CD represents a strut with end D built in and end C free from any restraint.
When/= CD,L = 21.
DE represents a strut with both ends built in. When l =DE, L = l/2 and
41t2 EI
Pc = -~2- = 4PE
It is also possible to derive the case for a strut that is built in at one end while
the other end is constrained such that it can only move along the original axis of
the strut. This is represented by CF. Cis a point of contraflexure and CF is a
tangent to the sine curve. The approximate length of the equivalent strut will be
given by the horizontal component of CF.
1tX
v=Asin-
L
dv A1t 1tX
-=-cos-
dx L L
Now for line CF, v = mx where m is the slope. At F
m= (~) = A1t cos 1tXy

dx F L L
and
Vy A . 1tXy
m=-=-sm--
xy xy L
therefore
A . 1tXy A1t 1tXy
- s m - - =-cos--
xy L L L
The solution of this equation gives an approximate value of xyjL as 1·43; or
l = I ·43L. Thus
I·43 21t 2EI
Pc ~
/2
= 2·05PE
It is suggested that the reader should attempt to obtain this result by the con-
ventional method of writing down the differential equation. A very common error
in setting this equation down is to omit the restraining force that is required at
the pinned end normal to the applied load P.
Figure 10.4
There is one further case that we shall consider with particular end conditions,
and that is the problem of a strut with equal and opposite end moments applied
in addition to the axial load (figure 10.4). The ends are pinned.
The differential equation is
or
d2v M
- +a 2 v=--a 2
dx 2 P
whence
.
v =A sm ax + B cos ax - pM
when x = 0, v = 0, therefore B = M/P. When x = 1/2, dv/dx = 0 therefore
M a!
A =-tan-
p 2
thus
M( 2 sin ax + cos ax - 1)
v = p tan a!
Note that a relation has been obtained between x and v in this case.
The maximum deflection occurs when x = 1/2 and
v = ; (sec ~I - 1) (10.5)
It is of interest to find the stiffness of the strut, that is, the moment for unit
rotation.
The slope at the ends will be required, and this has a value of
Ma al
-tan-
p 2
Thus
M p
s=-=
e a tan al/2
Ths stiffness will become zero if tan al/2 = oo, that is
a! rc
-=-
2 2
or
The Euler critical load is reached at zero stiffness.

10.3 Strut with initial deformation

It has already been mentioned that it is impossible to obtain a perfectly straight
structural member. The initial deviation from perfect straightness could be written
down in the form v 0 = f(x).
Figure 10.5 shows an imperfect column of uniform section with pinned ends.
When the load Pis applied there will be a resulting change of curvature that can
be represented by
Hence
or
(10.6)
It is worth noting at this stage that the term E/(d 2 v0 /dx 2 ) could arise from a
different problem-that of a strut with a lateral load-where it would represent
the bending moment at a point on the strut when the lateral load is considered
alone.
We shall make the assumption that the initial deformation can be represented
as a Fourier series.
n=oo
2:
nTtX
v0 = an sin--
n =1 1
thus
For pinned ends v =0 whenx =0 or I. Thus A= B =0.
v=
~ 1-
n~ an sin(mrx/1)
(P/n 2PE)
If we take the ratio of v/v 0 for the nth term, that is,
(10.7)
1/1-~
n2PE
we have what is termed an 'amplification factor' -the factor by which the initial
component of deflection would be multiplied when an axial load Pis applied to
the strut.
It was mentioned that equation 10.6 also applies for the case of a laterally
loaded strut. If it is possible to represent the bending moment due to the lateral
loading in the form of a Fourier series, the resulting deflection equation will be
identical with equation 10.7.
10.4 Struts made from ideal elasto-plastic materials

So far it has been assumed that the material of the strut behaves in an elastic
manner and that the yield point of the material has not been exceeded. The Euler
critical load was found to be
r: 2 EI
PE=--
12
The second moment of area I can be represented in the form Ak 2 where A is
the cross-sectional area and k the radius of gyration. Thus
r:2Ek2
aE =--12- (10.8)
This is the stress that would arise at the Euler critical load. The ratio of 1/k is
referred to as the 'slenderness ratio' of the strut.
Figure 10.6 shows a plot of aE against 1/k for a steel strut. This graph will of
course be incorrect once the value of the yield stress has been reached. Assuming
a mild steel with a yield stress of 240 N/mm 2 , the corresponding value of 1/k is
about 93. This means that the theory that we have derived will only apply for the
case of struts with a slenderness ratio that is greater than this value.
If an ideal elasto-plastic stress-strain relation is assumed, when the slenderness
ratio is less than 93 the column will not fail by elastic buckling. The value of the
stress would reach ay and the column could be said to squash. Hence the load
applied is often referred to as the 'squash load' Ps.
The above discussion has been based on the assumption that the strut was
initially straight and remained straight. We shall now describe what may take
place with imperfections in the strut.
Consider first the case of an ideal pin-ended strut of linear elastic material. The
Euler critical load would be reached before any deflection occurred. A graph of
400
300
PLASTIC
11e
N/mm2
200
ELASTIC
100 STABLE
100 200
Figure 10.6
load P against central deflection would be a horizontal line A of value PE (figure

10. 7). If the strut was imperfect and there was initial eccentricity present the
central displacement would be given by equation 10.7, and a curve similar to B
would result assuming the material remains elastic.
Again consider a perfect strut of elasto-plastic material. The Euler load will be
reached before any central deflection occurs and the stress is uniformly distributed
over the section. For a small displacement, bending stresses will occur and these
Figure 10.7
will reduce the stresses on the convex side and increase those on the concave side.
At some stage the stress on the concave side will reach ay. Further displacement
may mean that the stress on the convex side will become tensile and the load
carrying capacity is reduced; there will be a deviation from the PE line and a curve
similar to C will result. If initial imperfections are present the curve would initially
follow B until yield is reached and then lie below the C curve at D. For the case
when the squash load Ps is reached before the Euler load, if a small displacement
is imposed, the load would have to be reduced to maintain equilibrium and a
curve such as E would result. If initial eccentricity is present the curve would be
similar to F.
Most practical struts would have a load-deflection characteristic similar to that
shown at F.
10.5 Double-modulus theory

This is also referred to as the reduced-modulus theory. The stress-strain curve for
the material of the strut is assumed to be similar to that shown in figure 10.8a.
The elastic limit is denoted by ay and the load applied to the column is such that
(P/A) > ay. When the load is applied a small lateral displacement is allowed to
take place near the centre of the strut; this will tend to reduce the stress on the
convex side and increase the stress on the concave side.
(a) (b) (c) (d)
Figure 10.8
It will be assumed that the strut has an axis of symmetry and that the displace-
ment is applied in the direction of this axis; and that plane sections remain plane.
The cross-section of the strut is shown in figure 10.8b. Before the lateral defor-
mation took place the stress was ap and the corresponding strain €p. With the dis-
placement applied, the upper surface of the strut becomes concave and the value
of the strain increases by a small amount €t. to a value of €p + e 1 , while that on
the lower surface decreases by e2 to €p - e2 (figure 10.8c). If the tangent modulus
on the stress-strain curve at A in figure 10.8a is E T, the increase in stress corre-
sponding to the strain e 1 , is approximately e 1 ET =at. At the bottom surface the
strain has been reduced and the unloading process will occur elastically and the
stress change will be e2 E = ac. The final stress distribution is shown in figure 10.8d.
The total load applied to the cross-section has remained unchanged. Thus
f dl
ac b dy -
f.d2 atb dy = 0
0 0
Now
If the radius of curvature isR, e 1 =dtfR, e2 =d2jR.

E
_!
fd' by dy - -E fd• by dy = 0
R o R o
therefore
(10.9)
where At.Y 1 and A 2 ji 2 represent the first moments of area about the horizontal
axis.
Equation 10.9 together with the relationship d = (d 1 + d 2 ) enables the position
of the neutral axis to be found.
Taking moments about the neutral axis
J:• achY dy + J atbY dy =M

0d
2
By making the substitutions for a and e

1
R (ETit +Eh) =M (10.10)
where I 1 and I 2 represent second moments of area about the neutral axis for each
part.
This means that the flexural rigidity has been modified to ETI 1 + EI 2 •
If I is the second moment of area of the cross-section about the centroid, we
could write
or
(10.11)
and this represents a reduced modulus of elasticity or double modulus.

For the case of a rectangular beam of depth d and breadth b, equation 10.9
becomes
bd~ Ebdi
ET----=0
2 2
therefore
d 1 + d 2 = d therefore
Now I= bd3 I 12, therefore
(10.12)
The critical load P0 will of course be given by

- rt2Eoi
Po--12 - (10.13)
10.6 Tangent-modulus theory

The double-modulus theory was based on the assumption that the axial load
remained constant while a small lateral displacement was applied. If however the
load continued to increase the stress and strain distributions over the cross-section
would differ from those shown in figure 10.8a. It is quite possible that the strain
could increase at both of the edges of the section, the rate of increase being
greater at the top than at the bottom. If it is assumed that E T remains constant
for fairly small changes of strain, the stress increase at the top could be written
e1ET, and that at the bottom e2ET, with a linear distribution over the cross-
section.
For this particular case equation 10.10 would become
ETI =M (10.14)
R
where I is the second moment of area about the centroid of the cross-section.
The critical load PT of the strut using the tangent modulus can be written
down straight away as
rt2ETI
PT=-
12 - (10.15)
The value of the tangent-modulus critical load will be less than the value
obtained using double-modulus theory, and this in turn will be smaller than the
Euler critical load.
A number of tests have been carried out on practical struts and it is usually
found that the actual critical load lies somewhere between the values given by
the tangent-modulus and double-modulus methods; the tangent modulus generally
gives the closer result.
10.7 Practical strut formulae

In section 10.3 we derived an expression for the lateral displacement of a strut
that initially was not straight, and for which the deviations could be written
down in the form of a Fourier series.
If equation 10.7 is examined it will be seen that the first term will predominate
when the applied load approaches the value of the Euler critical load. If all the
other terms are omitted
a sin (rtx/1)
v=
1- (P/PE)
The maximum deflection occurs when x = 1/2, that is
aPE
Vmax =PE-p
thus
aPEP
Mmax =Pvmax =PE _ p
The maximum stress in the strut will occur when the value of the bending
moment is a maximum, and this will be at the surface of the strut on the concave
side.
When the maximum stress reaches the value ay
P aPEP dt
a=-+-~-
Y A PE-p Ak 2
where d 1 is the distance from the neutral axis to the outer fibre and Ak 2 is the
value of the second moment of area about the neutral axis.
Put aE = PE/A, a= P/A and 1'/ =ad tfk 2
(10.16)
or
a2 - [ay + aE(l + TJ)] a+ ayaE = 0
Solving this quadratic equation gives
a=!{[ay +aE(l +11)]- {[ay +aE(l +TJ)] 2 -4ayaE} 112 } (10.17)

Hence the value of Py (figure 10.7), is known, the value of the load at which
yield first occurs in the strut. This particular expression was first obtained by
Perry. It is not possible to solve equation 10.17 unless the value of 11 is known for
a particular strut.
The value of Py will be less than the actual failure load PF of a strut. Robertson
conducted a large number of tests on struts and equated the failure load to Py; he
then solved equation 10.16 for values of TJ. The values of 11 would be less than the
true values for 11 for particular struts asPy < PF. An average value of 11 obtained
by this method was 0·003!/k. If this value is used, equation 10.17 may next be
solved to obtain a value for the collapse load. This method is referred to as the
Perry-Robertson formula.
In practice the actual failure load of a strut is lower than either the critical
load PE or the squash load P5 .
We shall consider the ratios PFfPE and PFfP5 . For struts with a high value of
the slenderness ratio (PFfPE) ~ 1 and for a low value of the slenderness ratio
PFfPs ~ 1.
If a linear relationship is assumed between PF/PE and PF/Ps such that
PF/PE = 1 when PF/Ps = 0, and PF/Ps = 1 when PF/PE = 0, the equation of such
a line may be written
or
1 1 1
-=-+- (10.18)
PF PE Ps
PF = Aay /1 ::E (ir

+ (10.19)
This is a very similar expression to that suggested by Rankine apart from the
fact that ay(1t 2E was replaced by an arbitrary constant.
10.8 The Southwell method

This is an experimental method that can be used to obtain the critical load of a
strut. The strut is loaded in an axial manner and it is possible from certain obser-
vations to predict the value of the critical load without the necessity of applying
a force of sufficient magnitude to cause failure of the strut.
We shall use the expression from equation 10.7 once again. If the loading P
applied to the strut is somewhere near the value of P£, the first term in the series
will predominate and the deflection can be written as
a 1 sin (1tX/l)
v=
1- (P/PE)
The fundamental component of v 0 was given by a 1 sin (11:xjl) so that the dis-
placement from the initial state is
. 1tX (
a 1 sm- p )
l PE-P
The maximum deflection will occur at x = l/2, and has a value
A= atP (10.20)
PE-P
Equation 10.7 was derived for the case of a strut with both ends pinned, but it
is possible to apply the above analysis to struts with other end conditions. Instead
of assuming a Fourier series it may be assumed that the critical modes vet> ve 2 ,
etc. are known for the strut and that it is possible to express the initial deflection
of the strut in terms of these modes. Hence
It can also be shown that when a load Pis applied to the strut
00
v= 7 1 - anven
""
(P/Pcn)
(10.21)
This equation will apply for any end conditions. Thus
v- v 0 = ~ anVenf( P~n -1)

Once again the first term will predominate when the applied load is somewhere
near Pet, therefore
(10.22)
where A is the deflection measured at a suitable point on the strut. This expression
is identical to that given in equation 10.20 except thatPE is replaced by Pc. For
the following discussion Pe will be used since it represents a more general case.
Equation 10.22 can be written in the following form
~=~+~ (10.23)
P Pc Pe
This gives a straight line relation between A/P and A. The slope of the line is
1/Pe and the intercept on the A-axis gives the value of a 1 (figure 10.9a). In a
practical test however there will be considerable deviation from the straight line
at lower values of P; the reason for this is that only the fundamental term was con-
sidered, but other terms can have a considerable effect when Pis small when
compared withPc.
Equation 10.23 can be written in a different form
P p Pc
-=--+- (10.24)
A a1 a1
This equation gives a straight-line relationship between P/ .:::1 and P. The inter-
cept on the P-axis gives the value of the critical load, and the reciprocal of the
slope gives the value of a 1 (figure 10.9b). Once again in a practical test there will
be a deviation from the straight line when Pis small compared with Pe.
(b)
Figure 10.9
The results from an actual experiment on a strut are incorporated in an

example at the end of the chapter.
It is possible to extend the use of a Southwell plot to a complete structure and
hence obtain an estimate of the critical loading that would cause collapse.
10.9 Energy methods

In this method of approach the energy that is stored in the strut is equated to
the external work done.
The bending moment at any point in the strut can be written in the form
d2 v
M=EI dx2
I
The energy stored in a member due to bending action alone is
JM2- -dx
=
2EI
fl
0
E(d2v)2
I - dx 2
dx 2 (10.25)
where dx:::::: ds.

The energy stored and the work done due to axial compression of the strut is
omitted, and the assumption is made that when the critical load is reached, the
strut suddenly deforms. The external work done will be equal to the critical load
times its displacement.
Now
ds = dx [ 1 + ( dx 2
dv )2]112 :::::: dx [ 1 + 1 (dv
dx )2]
1 (dv ) 2 dx
ds-dx=--
2 dx
If both sides of this equation are integrated it can be seen that the left-hand
side represents the shortening of the strut due to the action of bending, or the
displacement of the critical load.
The external work done can then be written
Pc f 1(dv)2
0
l
--
2 dx
dx (10.26)
Note that it is assumed that Pc remains constant while the deformation is

taking place.
Equating the two expressions in equations 10.25 and 10.26
fo EI (d2v)2
l
-
dx
dx 2
(10.27)
This expression is the Rayleigh formula for the critical load of a strut. If the
deflected form of a particular strut is known or assumed the expression can be
evaluated. If the true deflected form is known the result obtained will be etJ_ual to
the true critical load of the strut. This however is very unlikely to be known and
in general a guess will have to be made for the deflection curve. The resulting value
of PR will always be higher than the true critical load. This point will be discussed
after the Timoshenko method has been derived.
For the case of struts with both ends simply supported or with one end built
in and the other end free, the equilibrium equation can be written in the form
d2 v
EI dx 2 = -Pv (10.28)
We have already shown that the energy stored in the strut due to bending
action is
Making use of equation 10.28 this may be expressed as
JP 2EIdx 2v 2
Equating this to the external work done (equation 10.26)
PT = s: (~Y dx/J: v~~x (10.29)
Equation 10.29 is known as the Timoshenko method for the critical load of a
strut. It is often more difficult to evaluate than the Rayleigh expression since the
term involving I appears in the denominator, and for a variable value of the
STABILITY OF COLUMNS 3I5
second moment of area it may not be possible to perform an exact integral. This
often means that a numerical method of integration must be used.
It can be shown mathematically that the Timoshenko method gives a result
that is nearer to the true critical load, and that both methods will give values that
are higher or at best equal to the true critical load. However an exact proof of this
is rather difficult and we shall confine ourselves to the following discussion. If the
true deflection curve be chosen then either method will give a result that is equal
to the true critical load. For any other chosen deflection curve however, the strut
has to be forced into this mode by the addition of external restraints such as
moments and forces. The effect of these restraints is to make the strut more rigid
and hence the critical load is raised above the true value.
Example
A tapered column of length lis built in at the bottom end. The second moment of
Ir
area at any cross-section is given by
lx =!(: +
I is the second moment of area at the top of the column, and x is measured from
the top of the column. Find the load at which the column will buckle when
a= (1 + V2l).
Assume v = v 0 sin TCX/21
dv 1t cos TCX
---v - -
dx - 21 ° 2/
PR = ~; ; 122 f (x 2 +lax +a 2)(I- cos 7tt) dx j{(I- cos 7t;) dx

EI 2
a2 412 3
[1
= -1t- - +a/ 2 +a 2 l +2z3
3
- +--
7t2 7t2
4al ]/I 2
Now a= 2-411 and a 2 = 5·8/ 2 , therefore

4·I2EI
PR = -1-2-
p _ 1t 2 EI/f 1 I - cos(1txjl)
T - 4la 2 0 x 2 + 4·82/x + 5·8/ 2
The denominator has to be evaluated by a numerical method. Dividing the

column into ten equal parts and making use of Simpson's rule, gives a value for
the denominator of 0·1 04/l and PT = 4·07 EI/1 2.
It is in fact possible to obtain a theoretical result for this particular column.
The value of the Euler critical load being Pc = 4·046£!/1 2 •
We can see that both Rayleigh and Timoshenko give results that are too high,
but Timoshenko lies closer to the true value.
In this chapter, the strut has been discussed in isolation. It is of course per-
fectly possible for members of a structure to collapse through instability. A pin-
jointed structure presents no difficulty as the force in the members is axial. A
rigid jointed structure on the other hand is a much more difficult problem since
the members will be subject to end moments in addition to axial loads.
Other forms of instability can occur in structural members. In addition to
collapse by strut action certain types of cross-section even though subject to an
axial load may fail due to a rotation about a longitudinal axis; this is referred to as
torsional instability. A further form of buckling can occur in a loaded beam; here
it is possible for a sideways deflection to occur together with a rotation-termed
lateral buckling. It is felt that these topics are beyond the scope of this present
work and it is suggested that the reader should consult works of a more specialised
nature.
Problems
10.1 A strut of length lis clamped at one end and pin-jointed at the other. For
length l/2 from the clamped end the flexural rigidity is£/, and for the remainder
of the length the strut can be treated as having infmite flexural rigidity. Find an
equation from which the Euler critical load can be obtained.
10.2 A slender uniform column is vertical and built in at the foot. It supports an
axial load at the top, which is free to move. In order to raise the value of the
critical load it is proposed to stiffen the upper half of the column. Show that the
critical load cannot be increased by more than about 20 per cent by this means.
The equation e =cote is satisfied bye = 0·86 approximately.
10.3 A strut initially straight, oflength land flexural rigidity El, is built in at one
end. An axial load Panda moment Mare applied at the other end, but lateral
moverr.ent there is prevented. Show that the slope of the strut at this end is given
by the following expression, where a 2 = P/EI
e = aM(2- 2 cos a!- a! sin a/)
P sin a/- a/ cos a!
10.4 Show that an initially straight, uniform pin-ended strut, which is not allowed
to extend at its ends and which is uniformly heated to a temperature T above the
surroundings, will buckle when T= Tc where Tc =7t 2 k 2 /aL 2 , k is the least radius
of gyration, L the length, and a the coefficient of linear expansion.
A uniform boiler tube of length 2L is built in to heavy diaphragms and passes
through a supporting diaphragm, which constrains it in position, but not in
direction, at its mid-point. Show that it will buckle when the temperature is uni-
formly raised by approximately 2·04rc 2 k 2 /a.z2 provided the ends are immovable.
10.5 A strut oflength lis encastre at one end; the other end is supported in such
a way that if a transverse displacement d occurs there, it is resisted by a transverse
restraining force kd. Show that the buckling load Pis given by the solution of the
equation
tan a/ P
--=1--
Ci. k
where a 2 =P/EI.
10.6 A tie of length land bending stiffness EI is imperfect, being slightly curved
such that v = v 0 cos (rex//) where vis measured from a straight line between the
ends and x is measured from the mid-point of this line. If the ends of the tie are
pinned, find the tensile force P required to reduce the initial eccentricity v 0 at
the centre of the tie by half. A lateral force F is now applied to the centre of the
tie in addition to the tension. Find the magnitude ofF required to reduce the
central eccentricity to zero.
10.7 A pin-ended strut of length l has an initial bow given by v 0 =A sin (rex//)
where A is a length small compared with/, and the axes x andy are at one end of
the strut. Show that the lateral deflection d at the centre of the strut due to an
end load Pis given by d =A/(PE/P- 1), where PE is the Euler critical load.
10.8 A pin-ended strut has length land flexural rigidity EI. It is laterally sup-
ported by an elastic medium. The medium resists the lateral movement d of each
point on the strut by applying a force kd per unit length at that point. Taking the
buckling mode to be a half sine-wave, find the critical value of the axial force P
either by solving the differential equation or by an energy method.
10.9 In a test on a pin-ended strut the applied axial compressive force P, and the
resulting lateral central deflection v, were measured as follows
v(mm) 0·23 0"38 0·55 0·75 0·96 1·27 1·63 2·04
P(kN) 6·85 8·90 9·80 10•54 11·20 11·75 12·10 12·50
Using a Southwell plot, determine the critical load.
PLASTIC ANALYSIS OF
11 BEAMS AND FRAMES
11.1 Introduction
When elastic theory is used to design a statically indeterminate structure in mild
steel, the stresses that arise when the complete external loading is applied must
not exceed a certain value, termed the working stress. The ratio of the yield stress
of the material to the working stress is called the safety factor. In a structure made
from a linearly elastic material the safety factor is also the ratio of the load
required to produce this yield stress to the working load.
At the design stage it is not possible to predict all the applied loads exactly. It
is quite possible for an overload to be applied sometime in the life of a structure.
Also there may be defects in the materials used or there may have been poor
workmanship during the course of construction. Extra stresses can arise due to
differential settlement of supports and from a variety of other causes. It can be
seen that if a structure were to be designed to a stress very close to the elastic
limit, it is quite possible that the yield stress might be exceeded sometime in the
life of the structure. This would not necessarily mean that the structure would
collapse since the yield stress is well below the ultimate stress. However if a suitable
safety factor is introduced the design stress will be well below the yield point. As
an example a typical value for mild steel would be about 1·5.
In some cases a design based on elastic theory can be extremely conservative
and wasteful of material. As an illustration we shall consider the case of a beam
with a constant cross-section with both ends built in, and carrying a uniformly
distributed load. From elastic theory we can easily show that the maximum bend-
ing moment occurs at the ends of the beam and that the value there is twice the
value at the mid-point of the beam. The ends of the beam are therefore the critical
sections and the maximum load that is allowed to be carried by the beam is
governed by the working stress not being exceeded at the ends.
Let us suppose that the beam is loaded until the yield stress is reached at the
supports. Further load can still be applied without the beam collapsing. As the
load is increased plastic hinges will form at the ends of the beam. Even when these
are fully developed the bending moment at the centre of the beam will not be
sufficiently great for the yield stress to be reached there. Loading could in fact
be continued until a plastic hinge occurs at the centre of the beam. At this stage
collapse is said to have taken place. This example has been discussed to show that
318
PLASTIC ANALYSIS OF BEAMS AND FRAMES 319
there is still a considerable reserve of strength in a redundant structure even
though the yield stress may have been reached at some point.
The plastic design method is based on calculating the load required to produce
sufficient plastic hinges in the structure to turn it, or at least part of it, into a
mechanism. This load would then be divided by the load factor as opposed to the
safety factor and the value of the working load determined. In practice of course
the problem would be presented the other way round. The working loads would
be known approximately and for a particular load factor the sections of the
various members could be determined.
One word of warning should be introduced at this stage. If there is the possi-
bility of instability occurring or if the deflections have to be kept to a minimum
it may not always be possible to use a plastic method of design.
11.2 Collapse of redundant beams

In chapter 5 the moment-curvature relationship was derived for a rectangular
cross-section made from an ideal elasto-plastic material. It was also discussed for
a universal beam. In this chapter we shall idealise this relationship as shown in
figure 11.1. This will mean that once Mp is reached at a particular point on a
beam, large changes of curvature can occur for no increase in bending moment.
It is as though a hinge had been inserted in the beam, hence the term plastic
hinge.
Xy X
Figure 11.1
The beam in figure 11.2a is built in at both ends. It carries a uniformly dis-
tributed load that increases in value until collapse of the beam occurs. The
bending-moment diagram when the beam is completely elastic is shown at b. This
can be considered as two separate diagrams, the free bending-moment diagram, a
parabola of maximum height pl 2 /8, and the reactant diagram, a rectangle of
height pl2j12. It can be seen that the maximum bending moment occurs at the
ends of the beam and the value at the centre is only half this value.
If a moment-curvature relationship similar to that shown in figure 11.1 is
assumed, the value of the load when the bending moment has reached the fully
plastic value of Mp at the supports is given by
12Mp
Py = -
12 -
(o)
fl v %
(b)
2
'SJ
(c)
Figure 11 .2
The bending moment cannot increase beyond this value but an increase in
curvature can take place and plastic hinges form at each end of the beam. The
beam from now on will behave as though it is simply supported at the ends. The
value of the bending moment at the centre of the beam at this stage is Mp/2 or
PyP /24.
As the uniformly distributed load is further increased the central bending
moment can increase until its value isMp. At this stage a further hinge will occur
at the centre and the beam has become a mechanism. Any further increase of
load would cause the beam to collapse.
The incremental load Pi required to produce the further increase in bending
moment of Mp/2 at the centre is given by
Pi/2 Mp
--:::-
8 2
therefore
4Mp
Pi== 72
We can now find the total value of the load that causes collapse
::: (1 2 + 4 ) Mp::: 16Mp

Pc z2 j2
Note that the collapse load is only dependent on the value of Mp and does not
depend in any way on the shape of the moment-curvature relationship. It has
however been useful in our analysis to use a particular relation since we have been
able to predict the load that caused the first plastic hinges to form at the ends of
the beam.
We shall next endeavour to find out something about the central deflection of
the beam.
It has been assumed that the beam behaves in an elastic manner until the load
Py is reached, so that the first part of the deflection curve will be linear and the
value of the deflection is given by
Pyl4 M 12
d=--=~
384EI 32EI
Once the two end-hinges have formed the beam behaves as though it is simply
supported, and the incremental load-deflection relationship will again be linear
with a maximum incremental deflection
5 Piz4 5Mpl 2
di = 384 EI = 96EI
The total deflection at the centre of the beam when all hinges have formed and
collapse is imminent is
M 12 M 12
d=-p- (~ +f6)=-p-
E/ l2EI
The load-deflection curve consists of distinct linear phases as shown in the
sketch (figure 11.3). Point X corresponds to the formation of hinges at the ends
of the beam and point Y to the formation of the central hinge. It must be empha-
sised that the shape of the curve is dependent upon the moment-curvature
relationship that is assumed. In actual practice the sharp kinks in the diagram
would be rounded as the formation of a hinge would not take place instantaneously
d
Figure 11.3
One interesting point is that the value of the collapse load is .independent of any
displacements that may occur at the supports as the beam is loaded, or indeed of
any initial imperfections. It is required only that the supports should be capable
of developing the fully plastic bending moment.
To demonstrate this fact we shall consider the extreme case of a rotation at
the ends of the beam, such that before the load is applied there is an initial
bending moment at each end of the beam of -Mp/2, that is, hogging. The
supports at the ends of the beam are now clamped in this position and the load
applied. Plastic hinges will form at the ends when the bending moment there is
-Mp, an increase in value of -Mp/2. The load required to produce this state of
affairs will be precisely half that of the original case.
The value of the bending moment at the centre with no load applied was
-Mp /2 and the change due to the application of Py is Mp/4. Thus the value at
the centre when hinges form at the ends is
Mp Mp Mp
--+-=--
2 4 4
To form a plastic hinge at the centre requires a final value of bending moment
Mp, an increase of SMp/4 therefore
Pip 5Mp
-=--
8 4
lOMp
Pi=-1z-
so that the total load needed to produce a complete collapse is 16Mp/l 2 , the
same value as that obtained with no initial rotation at the ends. What in fact we
are saying is that the final bending moment (figure 11.2c) is independent of the
deflections and rotations of the supports.
Next consider a continuous beam of two spans, each of length l. The a!lplied
loading is shown in figure ll.4a. The value of Mp is required such that collapse
just occurs under the loading system.
(ol
(b)
(c)
(d)
Figure 11.4
Using the same basic method of approach the statically determinate bending-
moment diagram can be drawn. It is possible to make the problem statically
determinate in several different ways, but probably the simplest is to put a cut
through the beam at B. This will give rise to triangular bending-moment diagrams
of heights Pl/2 and Pl/4 in spans AB and BC respectively (figure 11.4b ). It is next
necessary to add the reactant bending-moment diagram. A bending-moment
diagram has to be introduced at B to restore continuity. The reactant bending-
moment diagram will be triangular with a maximum height at B equal to the
value of the final bending moment at B. The combined bending-moment diagram
is shown at c. There are three possibilities for a maximum bending moment, at B,
D, or E.
We shall first assume that there is a maximum bending moment at B and D,
and that plastic hinges have formed at these points. This means that span ADB
has collapsed and
or Mp =Pl/3. So a beam must be selected capable of carrying a fully plastic

moment of Pl/3.
So far we have satisfied two requirements. Part of the beam has become a
mechanism, and the system of applied forces and the moments are in equilibrium.
It is quite possible that the solution obtained might give a bending moment greater
than Mp at some point in span BC. If this is so the solution is an unsafe one. It is
an easy matter to check. The maximum bending moment in span BC will occur at
E and the value there is
a value well below the derived value of Mp. This implies that the solution we have
obtained is the correct one and that M..;; Mp at all points. The latter inequality
satisfies the yield criterion. So that three conditions have in fact been satisfied,
equilibrium, mechanism, and yield.
Let us suppose by some mischance that originally the incorrect failure mech-
anism was chosen and it was assumed that hinges would form at B and E (figure
11.4d). Thus
therefore
Pl
Mp=6
Equilibrium and mechanism have both been satisfied, but a check on the value of
the bending moment at D shows that the value there is 5Pl/l2, very much in
excess of the suggested value of Mp = Pl/6. In other words the yield criterion is
not satisfied. We do know that from this piece of analysis the beam chosen must
have a fully plastic bending moment greater than Pl/6 and we have found a lower
bound to the problem. It is easy to obtain an upper bound from the same analysis.
If a beam was chosen with Mp = 5Pl/l2 failure would not occur. So that we may
state
Pl 5Pl
-..;;;Mp..;;;-
6 12
The correct value of Mp obtained from the first solution must of course lie
between these bounds.
11.3 Load factor

The term load factor was mentioned in the introduction to this chapter. If a par-
ticular statically determinate structure has a set of working loads Pw applied, the
bending moment at any point will be some function of Pw. If the loads are all
increased by the same factor A, since the structure is statically determinate, the
bending moments will all increase by A. When Mp is reached at any point on the
structure, a hinge forms and the structure becomes a mechanism and collapse
occurs. The value of Ac to cause collapse is called the collapse load-factor.
If aw is the working stress in the material, the working bending moment Mw
would be given by awZe where Ze is the elastic section modulus. Now AcMw must
be equal to the fully plastic bending moment Mp. In chapter 3 it was stated that
Mp = ayZp where Zp was the plastic section modulus.
AcMw = ayZp
Mw = awZe
therefore
(11.1)
A reasonable value of ay/aw for mild steel is 1·5, so that

Zp
A=1·5-=1·5a
Ze
For a rectangular section a= 1·5 Ac = 2·25

For a universal beam section a = 1·15 Ac = 1· 725
Using these load factors would mean that the working load for a statically
determinate case would be the same whether elastic or plastic design is used. In a
redundant problem using elastic design the bending moment at any point is a
function of the load applied, but using plastic design the bending moment would
be a different function. In general the allowable working load is greater using
plastic design.
11.4 Basic theorems

Certain conditions apply when a structure is on the point of collapse.
Equilibrium
The system of bending moments must be in equilibrium with the applied loads.
Yield
The bending moment at any point on the structure must not exceedMP.
Mechanism
The bending moment must be equal to Mp at a sufficient number of points such
that the structure, or at least part of it, becomes a mechanism.
There are three basic theorems, which we shall not prove here, concerned with
the three conditions set out above. A set of working loads is postulated and a load
factor A is determined with certain conditions applicable.
Uniqueness theorem
If all the conditions are satisfied at the same time then the value of the load
factor is unique.
Lower-bound or safe theorem
If a load factor is found such that equilibrium and yield are satisfied then a safe
solution has been obtained. This means that Mp has not been exceeded anywhere,
but it also is possible that the structure has not yet collapsed, since there may not
be sufficient plastic hinges formed to produce a mechanism. In fact further load
would have to be applied, so A.;;;; Ac·
Upper-bound. or unsafe theorem
If a load factor is found so that a mechanism is formed then possibly an unsafe
solution has been obtained. It may well be that for the chosen mechanism the
bending moment at some other point in the structure is greater than Mp. The
yield condition has been forgotten. If this is so, then an incorrect mechanism
has been chosen and a different mechanism would have formed at a lower load,
A;;,Ac·
It can be seen that it should be possible to work from a lower-bound solution
towards the unique solution or from an upper-bound to a unique solution.
11.5 Graphical analysis

This method has already been applied to simple beam problems. Basically it con-
sists of drawing a statically determinate bending-moment diagram, and then super-
imposing the reactant bending-moment diagram. A decision has then to be made
as to where the plastic hinges are likely to occur. As an illustration of this method
we shall first discuss the solution of a portal frame with pinned feet and then pro-
ceed to the more complicated case of a portal with a pitched roof and built-in feet.
The portal frame in figure ll.Sa has pinned feet and has a load P 1 applied at
the centre of BC, and a side load P 2 at B. The members all have the same cross-
section.
To make the frame statically determinate the horizontal force at D is assumed
to be zero (b), thus the vertical reaction at D is (P 1 /2) + (P 2 h/l). The bending-
moment diagram for this loading system is shown at d. The relative values of the
bending moments at Band E are dependent on the values of P 1 and P 2 and also
on the dimensions of the frame , so that it would be perfectly possible for the
_, cl
~~FIE
t~
__
' " (o ( _,1 ~_.,
~ (b)
It_
"' (c I
~
"' "'
r:7
1: . ~
4 2
(d )
7.e)
c
0
0
(f)
ljl
;::;;;
y
( kl
z
1 P,h 2
Mp
Figure ll.S
bending moment at B to exceed that at E. The reactant bending-moment diagram

due to the loading at c is drawn at e. The two bending-moment diagrams are
superimposed and the resultant shaded in for g.
Next it is necessary to decide how many plastic hinges are required for collapse
and where these are located. It is fairly obvious in this case that two hinges are
required-these would turn the portal into a four-bar chain. For the bending-
moment diagram f the hinges will occur at E and C since the resultant bending
moment at E is greater than at B. For g the hinges occur at Band C.
From the bending-moment diagrams we can see that either
2 M = P 1 l + P2 h {11.2)
p 4 2
or
{11.3)
The two different collapse modes corresponding to these two equations are
shown at hand j. It is possible to plot the equations in a particular way. This is
known as an interaction diagram, sometimes referred to as a yield-surface diagram,
shown at k. Equation 11.2 is represented by the line XY and equation 11.3 by the
line YZ. So that for any point on the line XYZ the portal is just collapsing. Any
point inside the lines represents a safe combination of P 1 and P2 •
A point outside the line represents an impossible combination of loads since
collapse will already have taken place. At Y both modes of collapse occur
simultaneously and three hinges form.
(d) (e) (f)
(g) (h)
Figure 11.6
The solution for the pitched-roof portal with built-in feet (figure 11.6a) will
be more complicated. The portal is made statically determinate by removing the
connection at C (b). The reactant forces are shown at c. We shall assume that a
positive bending moment produces tension on the inside edges of the members.
The bending moments can now be written down for possible hinge points.
HZ
M 8 =M+ VI+--P 11
2
Mc=M
Hl
M 0 =M- Vl+-
2
3HZ
ME=M- Vl+-
2
A total of 4 plastic hinges will be required to turn the portal into a mechanism,
again a four-bar chain. There are five possible modes of collapse (d to h). If the
loads are restricted such that P 1 can only act downwards and P 2 only to the right,
two of the possible modes d and h are unlikely to occur. For a vertical downward
displacement of C the displacement of B would be to the left, in the opposite
direction to that in which the load is applied.
We shall investigate mode 2 (e) with hinges at A, C, D and E.
MA =M0 = -Mp Me= ME =Mp
It should be noted that once the sign of one moment has been obtained, the
others can be written down, since consecutive hinges have opposite signs.
AtC
M=Mp
AtD
HI
M p -VI+-
2 =-Mp
AtE
At A
3Hl
M p +VI + 2
- - P11- P2 1= -Mp
FromM0 andME,m = 2Mp. FromME andMA

Vl= Ptl + P2l -M
2 2 p
Substituting in M 0 gives
For mode 4
AtC
AtD
m
Mp
M p -Vl+-=-
2
AtE
3m
Mp- Vl+l=Mp
AtB
m
Mp + Vl+2-P 1 l=-Mp
FromMo andME,H/ = 2Mp. FromM0 andMB, Vl =P 1 l/2. Substituting inM0

gives
Ptl = 6Mp
It is possible to investigate all modes in this way and the following equations
will result
Mode 1 P 1 l- P2 l = 6Mp
Mode 2 P 1 l + P 2 l = 8Mp
Mode 3 P 2 l = 4Mp
Mode 4 P 1 l =6Mp
Mode 5 P1 l- 2P2 l = 8Mp
The interaction diagram is drawn in figure 11.6j. It can be seen at once that
modes 1 and 5 will not be of interest for positive values of P1 andP2 -this was
seen earlier. At X modes 2 and 4 occur at the same time and at Y modes 2 and 3,
there being five hinges in each case.
The analysis by this method is somewhat tedious, particularly if the framework
is at all complicated. The method of virtual work together with combined mech-
anisms is generally to be preferred. This will be discussed in the following sections.
11.6 Virtual-work approach

An equilibrium equation using virtual work could be written down relating MA,
M B, and Me to the load P, for the beam problem in figure 11.7a. It would be neces-
sary to use a compatible set of deformations, and that shown at b would be
suitable.
1-A ~-lcP
~
~
. t 8~
~
----/-----1
(a)
(b)
Figure 11.7
The moment sign-convention is the same as that used in the previous section.
A joint rotation will be considered positive if the joint appears to open if viewed
from the underside of a beam or the inside of a frame.
For the deformations suggested
8A=-8 8e=28 8B=-8 d=l8/2
The virtual-work equation can be written
Pl
MA( -8) + Mc(28) + MB(-8) = - 8
2
or
Pl
-MA + 2Me -MB = - (11.4)
2
If the beam is allowed to reach collapse the values of the bending moments at
A, B, and Care known,MA =MB = -Mp; Me =Mp.
So equation 11.4 simplifies to Mp = Pl/8 and this will be the collapse equation
for the beam.
It should be noted that when the virtual-work equation is formed with values
of Mp substituted, all terms on the left-hand side become positive, that is a posi-
tive bending moment is associated with a positive joint rotation.
The same collapse equation could be written down using real work, as long as
all elastic deformations are disregarded as small compared with plastic deformations.
So far we have not considered any loading apart from concentrated loads. If a
uniformly distributed load of value p per unit length is applied in figure 11.7
instead of P, the hinges which form at the points of maximum bending moment
would occur in identical positions. It is only necessary to determine the work
done by the external load: this is equivalent to taking the total load on the beam
through a distance equal to one half the maximum displacement. The right-hand
side of the virtual-work equation would be
pix ~(~e)
and at collapse
pl2
M =-
P 16
The case shown in figure 11.8-that of a cantilever with a prop-will cause a
little more trouble. Two hinges will form at collapse. It is obvious that one will be
at A, the built-in end. Presumably the other will occur at the point of maximum
I I /p 'I
~ (0 )
(b)
Figure 11.8
bending moment. This is unfortunately unknown. We shall assume it to be at a

distance X from c. If the hinge at A rotates through e, that at B will rotate
through 18 jx.
By virtual work
I
Mp8 + Mp- 8 =pix !{I- x)8
X
Note that all terms on the left-hand side of the equation have been written
down as positive.
plx(l- x)
M =..:........~_..:.. (ll.S)
P 2(1 +x)
The value of Mp that we are looking for is the maximum. Knowing this we
could determine the value of Zp and hence the size of section that is capable of
carrying this maximum bending moment. This means that we have a safe design.
dMp
-=0
dx
which gives
X2 + 2x/ - / 2 = 0
or
x=(y2-1)/
Substituting this value into equation 11.5 will give the value of Mp
Mp = 0·086pl 2
As a further example of the virtual-work approach we shall endeavour to find
the collapse load-factor A for the beam in figure 11.9a, where the fully plastic
moment isMP. This implies that the loads applied are working loads and that we
must scale then up by A to investigate collapse.
c~
(a)
(b)
Figure 11.9
It is possible for the beam to collapse in one of two ways-failure in span AB

with three hinges at A, B, and D; or failure in span BC with two hinges at Band
E. The possible failure mechanisms are shown in figure 11.9b.
For collapse of AB
For collapse of BC
Hence span RC will collapse before span AB and A2 is the correct load factor.
The portal frame in figure ll.Sa can be solved very easily by virtual work. The
frame is redrawn in figure 11.1 Oa.
Possible hinge points are at A, B, C, D, and under load the P 1 at E. Four hinges
would be required for the side sway mechanism of figure 11.1 Ob. Also four are
(a) (b) (c) (d)
Figure 11.10
required for the 'combined' mechanism (d). Partial collapse can occur by failure of
the beam with three hinges (c).
For the sidesway mode
MABA +MaBa +MeBe +MoBo =P2 d
The values of the hinge rotations can all be set equal to ±Band the moments
to ±Mp.
For beam collapse

1
4MpB =P1 2_B 8Mp =P1 1
For the combined mode

MABA +MEBE +MeBe +M0 B0 =P1 d 4 +P2 d 3
If BA= -Bit will be found that BE= 2B; Be= -2B; B0 = B;d 3 = hB; d 4 = lB/2,
therefore
or
The equations obtained are identical to those derived previously from bending-
moment diagrams.
To make sure that the correct mode of collapse has been chosen for a problem
it is necessary to show that the value of Mp has not been exceeded at any point,
in other words the yield condition is satisfied. For the combined mode it would
only be necessary to show thatMa.;;;;. Mp· Probably the easiest way of determining
the bending moment at B is by virtual work. A suitable set of displacements would
be those in figure 11.1 Ob or c. Choosing the beam mechanism
Ba =Be =-B BE= 2B
ME=Mp Me=-Mp
therefore
Substituting for Mp from 6Mp = (P1 1/2) +P 2 h gives
_ P2h
M B- Ptl
-
2 4
The reader should satisfy himself that the sidesway displacement diagram
gives an identical result.
11.7 Combination of mechanisms

For a particular framework and loading system it would be perfectly possible to
sketch every mechanism. Using virtual-work principles a load factor could then
be found for each mechanism. The lowest value of A. would then be the true col-
lapse load-factor for the system. This approach could become somewhat tedious
and it is always possible to overlook a particular mechanism which might be the
vital one. The combination of mechanisms is a method that is more systematic in
its approach.
If we consider the portal frame again {figure 11.5a or figure 11.10a) and
introduce load factors
for beam collapse
for sidesway collapse
Now it is possible to combine the mechanisms of figures 11.1 Ob and c into a

further mechanism (d). This is formed by the elimination of the hinge at B which
has different directions of rotation for the two basic mechanisms. The total
external work equals the sum of the values of the external work for each of the
basic mechanisms; and the total internal work equals the sum of the values of
internal work. The value, however, will be too large since it contains the work for
both a positive and a negative rotation of joint B, which have cancelled each other
out. This means that we should have to subtract an amount of work 2Mp0 from
the total.
Adding the two equations
1
"A3P1 28 + "A3P2h8 = 8Mp8
-2Mp8
6Mp
"A 3 = -P-1(-1/-2)--'+'-P-2-h
To make the example more specific the following values have been assumed for
the variables
1 = 4 m; h =3m; P 1 = 40 kN; P 2 = 20 kN; Mp = 40 kNm
8x40 4x40
Al = 40 X 4 = 2 A2 = 20 X 3 = 2·67
So beam collapse would occur before sidesway collapse. For the combined
mechanism
6 X 40
"A3 = - - = 1·714
20x 3
This is lower than either of the other load factors so the combined mode will
be the one to occur. Before we finally accept this result we ought to make quite
sure that the yield criterion is not invalidated. This would involve finding the
bending moment at B. An expression for M 8 was found at the end of the last
section.
MB = ( 2 -;f-
P2h - P 1) "A = (30- 40)1·714 = -17·14 kNm
3
The final bending-moment diagram would consist of a series of straight lines

and, as the value of M B is very much less than the value of M P, the correct mode
of collapse has been assumed.
It can be seen that if the method of combined mechanisms is to be used to
advantage, then it is first necessary to determine the number of independent
mechanisms. Let n be the number of bending moments that are required to specify
completely the distribution of bending moments throughout a structure. If the
structure is statically determinate the values of all the bending moments can be
found. One method of doing this would be by using virtual work to form equations
of equilibrium. Suitable mechanisms would have to be chosen. The number of
mechanisms m would be equal to the number of equations of equilibrium which
in turn would equal n. If the structure is redundant and d is the degree of statical
indeterminacy of the structure, we could make use of the m mechanisms but the
structure would not be determinate unless d bending moments were specified in
some way or another. Thus we may write
n=m +d
This means that the numberof basic mechanisms m will be given by (n - d).
We may easily check this for the portal. The value of n is 5. The complete
bending-moment distribution would be known given the values of the bending
moments at A, B, C, D, and E. The frame is three times statically indeterminate.
Thus n - d = 2, giving two basic mechanisms.
~
(a) (c)
(f)+jolnt rot.= (g) (g)+(b)=(h)
Figure 11.11
For the two-bay portal frame in figure ll.lla the complete distribution of
bending moments would be known if the bending moments could be determined
at all the points marked with a cross. This means that n is equal to 10. Note that
we must know the three bending moments at the junction of the three beams. We
are not allowed to state that if two of the bending moments are known then the
third can be found by summing all the bending moments to zero. If this is done an
equation of equilibrium has been used. The framework will be found to be indeter-
minate to the sixth degree: so there should be four independent mechanisms. Three
of these are obvious, collapse of the left-hand beam, collapse of the right-hand
beam, and sidesway of the whole framework.
Each of these mechanisms corresponds to an equilibrium equation. The equi-
librium equation for the tee junction of the three members is that the sum of the
bending moments should be zero. This corresponds to a rotation of the joint. This
is in fact the fourth mechanism. Taken by itself it would not mean a great deal
but it can be combined with certain basic mechanisms to form other mechanisms.
For the sake of simplicity it will be assumed that the value of the fully plastic
moments for the beams and the columns have the same value.
Three of the basic mechanisms are shown in figures 11.11 b, c and d. The com-
bination of sidesway and beam collapse requires no further comment. These are
shown at e and f. f can be combined with a joint rotation of the tee joint. This
will mean that the two joints, one in a beam and the other in a column, can be
replaced by a single joint in the other beam, resulting in a lower value of the load
factor. A load factor would be obtained for every mechanism and the lowest
value selected. All that would remain would be to check that the yield criterion
had not been violated.
As a numerical example we shall consider the slightly more difficult case of a
two-storey portal-type structure. The problem is presented in a slightly different
manner. The dimensions of the members are given, and so also are the relative values
of Zp or Mp for each member. These are shown in circles on the line diagram in
figure 11.12a. The structure has to be designed for a load factor of 2. This will
mean that we would have to find the value of Mp for each member, from which
the value of Zp would be determined and a suitable cross-section could be
selected for the members.
To specify the complete distribution of bending moments in the structure it
would be necessary to know the bending moments at twelve points (b). The frame
is statically indeterminate to the sixth degree so there should be six basic mech-
anisms. The basic frame is shown again at c with the loads increased by the value
of the load factor.
Using virtual work the value of Mp has been calculated for the four basic mech-
anisms d to g; the working is shown under each diagram. We shall of course be
looking for the highest value of M P, instead of the smallest value of the load
factor. The worst case so far is the collapse of the lower beam. Note that for col-
lapse of the top beam the hinges form in the columns rather than at the ends of
the beam.
e and g are combined together at h. The easiest way of finding the value of Mp
is to sum the two virtual-work equations. This leads to a value of Mp of 25·7. We
can, however, modify this collapse mechanism if joint B is allowed to rotate form-
ing the mode at h'. This would reduce the internal work done by an amount 5 Mp·
A value of 31·3 is obtained for Mp, the highest so far. The rest of the diagrams are
self-explanatory, and the worst case is at k giving a value of Mp = 33·1.
It is quite possible to form other mechanisms, for example by combining d
and f, but since the value of Mp for these mechanisms is on the low side it is
unlikely that the combination will approach the maximum found.
It is essential to carry out a check to find out if the yield condition is satisfied
before accepting the value of Mp·
Assuming a value of Mp =33·1 kNm and attaching the correct signs M AB =
-99·3; MFE = 99·3; MG = 99·3; MH = 132·4; MoE= -66·2; MEB = -132·4.
The original structure was six times redundant. We now have plastic hinges at
five points, or five equations of condition. This means that unfortunately the
~-Sm-----+1 20kN t40kN
R
10~
ro~F9
t
-t
--
C @ G D
® ® Sm 20kN
10~6
t40kN
E_j
~ H
t
Fl
~12m 6m
A
~ (a) ~ (b) (c)
RRRR (d) (e) (f) (g)
RRRR
6Mp•4Mp=40x6 16Mp=B0x6 BMp=20x5 12Mp=20x6+ 20x6
Mp=24 Mp=30 Mp=12·5 Mp=20
(e)•(Ql=(hl (h') (h')•(fl= (J) (j')

16Mp=4BO 28Mp=720 23Mp=720 21Mp=B20
12Me=240 -SMP BMp=100 -Me
2BMp=720 23Mp=720 31Mp=B20 26Mp=B20
Mp= 25 7 Mp= 313 -4Me Mp= 31-6
27Mp=B20
Mp= 30·4
(j')+(d)=(k)
26Mp=B20
10Mp=240
36Mp=1060
-4Mp (l)
32Mp=1060
Mp= 331
Figure 11.12
portal is still singly redundant. However, it should be possible to find bending
moments in terms of one unknown. We shall proceed using virtual work.
For a displacement mode corresponding to collapse of beam CD
2MG - McB - MnE = -! x 40 x 12
198·6 - McB + 66'2 = 240
McB = 24·8
For collapse of beam BE

2MH - MBE - MEB = 480
MBE = -82·8
All the bending moments so far determined have been entered on the line
diagram in figure 11.121 where the frame is split into two parts. If we now assume
that the bending moment MEn isM, thenMEF can be written down.
Using virtual work with a sidesway displacement system for the top storey
MBc(-0) + 24·80- 66·2(-0) + MO = 100
MBc =M- 9·0
MBA follows at once from equilibrium of joint B.

The limits are known for the value of the bending moment at any point. For
example -66·2 ~MEn ~ 66·2.
The following inequalitites may be written down
-66·2 ~ M- 9·0 ~ 66·2 or -57·2 ~ M ~ 75·2
-99·3 ~ M- 91·8 ~ 99·3 or 7·5 ~ M~ 191·1
-99·3~M-132·4~99·3 or 33·1~M~231·7
Also
-66·2 ~M~ 66·2
The limits of M can now be written down

33·1 ~ M ~ 66·2 (11.6)
This means that as long as M lies anywhere between these limits the value of
Mp will not be exceeded at any point on the structure.
If a system of bending moments can be found that is in equilibrium with the
applied loading and the yield condition is satisfied, and if a mechanism has also
formed, then the solution is unique. Most certainly we have equilibrium as is
shown by the set of moments in figure 11.121, and we also have a mechanism.
Any value of M can be chosen between 66·2 and 33·1 kNm and we have a unique
solution. This means that the value of Mp is the correct one. If the incorrect
mechanism had been chosen it would be found impossible to obtain a value of M
to satisfy the equivalent expression to equation 11.6.
It is not always possible to write down the geometric relationships between the
hinge angles in all cases. The pitched-roof portal in figure 11.6a is somewhat more
complicated. Usually a sketch of the displacement diagram for a small displace-
ment will provide the relationships quite easily. If the reader has a knowledge of
instantaneous centres from applied mechanics, this can often provide a rapid
solution.
We shall conclude with the case of a portal with a uniformly distributed load
on the top member, together with a side load (figure 11.13). Values of Mp are
shown in circles.
~ e
(b l
49
4-x
~
4-x
(a l
Figure 11.13
For a pure sidesway collapse

At X 40 X 3=4 X 60
A.t =2
For collapse of the beam with a hinge distant x from B (figure 11 . 13b) the
hinges will form in the top of the column rather than at the ends of the beam.
4x8 60x8 3208
:\2 X 40 X - = 608 + - - + - -
2 4 - x 4-x
For a pure beam collapse the hinge would occur at the centre, that is x = 2m
A.2 = 1·75
For a combined collapse it is not known where the hinge will form in the beam.
Adding the solutions for sidesway and for a hinge distant x from the left-hand
side and correcting for the positive and negative rotation at B
60x 320
A. 3(120 + 80x) = 240 + 60 + - - + - - - 2 x 60
4-x 4- x
A. _ 26 - 3x
3 - 12 + Sx - 2x 2
3x 2 - 52x + 83 = 0
x = 1·8 m
The value of A3 is 1·42. The hinge forms to the left of the centre of BC and as
this value of A is lower than the other two, failure will occur by a combined
mechanism. The error introduced by assuming a hinge in the middle of BC is not
very large, and A= 1·43.
It will be noted that in all the problems covered in this chapter the loading has
been of a proportional type, the loads in effect being steadily applied such that
they all reach their maximum values just as the structure collapses. It should be
pointed out that this does not occur in practice; it may well be that loads can
vary between certain limits and may well be applied in a random fashion to a
structure. The methods that we have been using for the determination of a load
factor will not cater for this type of loading and the reader who is interested will
need to read further in this subject.
Problems
11.1 A beam of uniform cross-section with plastic momentMP and length 2/
rests on simple supports at its ends and on a central prop. Equal concentrated
loads are applied at the centre of each span. Find the value of the collapse load.
Find also the collapse load if the ends had been built in.
11.2 A continuous beam oflength I is of uniform section; it is simply supported

at each end and at two intermediate points such that the end spans are of the
same length. A uniformly distributed load of p per unit length is applied over the
entire length. For the most economic plastic design locate the position of the
intermediate supports.
11.3 A continuous beam ABC is freely supported at the ends A and C, and also
at B, where AB =L and BC = 2L. The maximum plastic moment in AB isM and
in BC l·SM The beam carries a uniformly distributed load p over its entire length.
Find using plastic theory the maximum value of p.
11.4 The beam shown at P 11.4 is to be designed according to the principles of

plastic design. The length of the beam and the positions of the built-in end and the
applied forces are prescribed, but the uniform plastic moment of resistance Mp
P11.4
and the position of the simple support defined by the length h, are at the designer's
disposal. Determine the values of Mp and h such that Mp shall be as small as
possible, consistent with the safety of the structure.
11.5 A portal frame made from members of equal plastic moment of resistance
Pl/2 is loaded as shown in P 11.5. Determine the greatest value of A for which the
portal is just safe and find the bending moments at B, C, and E.
p 11.5
11.6 The symmetrical frame in P 11.6 has each leg with moment of resistance
Mp and a beam of value 2Mp. Determine how the collapse load and the collapse
mechanism vary with k.
p 11.6
11.7 All the members of the frame P 11.7 are of the same cross-section, and the
vertical loads are applied at the mid-points of the beams. If Mp = 150 kNm
determine the collapse load-factor.
p 11.7
11.8 For the framework in P 11.8 the relative values of Mp for the members are
shown in circles. Show that suitable values for the plastic moment of resistance
of the various members are given by (2 or 3) x 1320/111.
10kNt'-4m-j-4m--it 10 kN
lOkN I
~·r-~~~~r-~®~~-r
® ® -+4m
®
® ® 4m
_l
0
I-sm .,: 8m~ 0

p 11.8
SUGGESTIONS FOR FURTHER READING
Chapter 6 J. H. Argyris and S. Kelsey. Energy Theorems and Structural

Analysis, Butterworth, London, 1968
Chapter 7 E. Lightfoot. Moment Distribution, Spon, London, 1961.

J. Rygol. Structural Analysis by Direct Moment Distribution, Crosby
Lockwood, St Albans, 1968
Chapter 8 R. K. Livesley. Matrix Methods of Structural Analysis, Pergamon,

Oxford, 1964
Chapter 9 T. M. Charlton. Model Analysis of Plane Structures, Pergamon,

Oxford, 1966
W. J. Larnach. Influence Lines for Statically Indeterminate Plane
Structures, Macmillan, London, 1964
Chapter 10 F. Bleich. Buckling Strength of Metal Structures, McGraw-Hill, New

York, 1952
M. R. Horne and W. Merchant. The Stability of Frames, Pergamon,
Oxford, 1965
S. P. Timoshenko and J. M. Gere. Theory of Elastic Stability,
McGraw-Hill, New York, 1961
Chapter 11 J. Baker and J. Heyman. Plastic Design of Frames Vol. 1, Cambridge

University Press, 1969
J. Heyman. Plastic Design of Frames Vol. 2, Cambridge University
Press, 1971
M. R. Horne. Plastic Theory of Structures, Nelson, London, 1971.
344
ANSWERS TO PROBLEMS
Chapter 1
1.1 VA= 19·5 kN; Vs = 25·5 kN; Q = -3·5 and -6 kN; M= 46 and 10 kNm
I p/2
1. 2 jp; y12; 36y3
1.4 14i kN; 22* kN; 30* kN; 22! kN
1.6 Pl/8
1.7 X f/8 f/4 3//4 7//8
y 0·2081 0·3791 0·4721 0·2731
1.9 13·5 kNm; 4·5 kN
1.10 8·46 m; 192·5 kNm
1.11 22·5 kN; 186·7 kNm; 35·8 kN; 210·1 kNm
Chapter 2
2.1 AC: -5/6 kN; AD: 35y2/6 kN; CD: 35/3 kN
2.2 DE: 12·5 kN; AB: -12·5 kN; CA: 35/12 kN; CD: -35/12 kN;
CB: -35yl3/12; CE: 35y13/12
2.3 -P/4
2.4 DG: 2·12P; DE: -0·72P; AF: -0·6P
2.5 V0 = t P; VE =HP; VF =*P; HF = 2P
2.6 AD= DF = FB = -40 kN; DC= 40 kN; CB = 20y3 kN
2.7 41·2 kN
Chapter 3 ( )
P Px a!
3.1 -exp - ; -
a a E
3.2 as= 30·8 N/mm 2 ; as= 64·7 N/mm 2 ; as= as= 43·3 N/mm 2 ;
as= 12·5 N/mm 2 ; as= 108 N/mm 2
345
3.3 a1 = ~; a = ~; a=~; ~1tr3 [T(3a- 13) +P(2:E (1- v) + ~}

8 l
p + (3a- 13)TK2tE
2tE + 3rK(l - v)
13 is the coefficient of cubical expansion of water.
3.4 169, -79 N/mm 2 ; (53·3, 81·5, - 23·5) x 10-s
3.5 240° gauge
3.6 5·2kN/m
E!' 2 d
3.7 £1'2 +£!'2
1·34P/3
3.8 1·65 td 3 • 0·56td 3 · - ----=--
' ' Etd 3
__s_a· P= 15y3R 3 am
3·9 N ' lOR +48z
Chapter4
4.1 r = 56·8 mm; 0·024 rad/m
1tG 2R 3 IG
4 ·2 lOA ; 20AR
4.4 2·05d 3 Ty
4.5 o·598T; 0·96r
4.6 7·81 N/mm 2
4.8 2r from centre
4.10 16/7t 2
4.11 Ga 4 /8
4.12 28·4GtR 3
Chapter 5
p 1 p
5.1 - 48 £/ (6pl + 5P); 24 (6pl + 5P); 12£/ (pl + P)
p/4 p/3 pl
5"2 - 1536£/; 256Ei centre 32
23 PZ 3
5"3 324 EI
ANSWERS TO PROBLEMS 347
5.4 ~points·
3 pl
--
2 z - 3-pl- ( xl-x 2
- ( x 2 ) x<-·
2 2)
--
z z z
-<x<-
<J ' 64EI ' 4' 64EI 8 ' 4 2
6pl 4
5.5
Ebhg
5.6 8Mjl 2 ; 11·6Mjl 2
5.7 12·7 per cent; 6 per cent; 3·5 per cent
Chapter 6
6.1 0
6.2 0·19 mm; 0·024 mm
6.3 0·324P
6.4 20° 18'.
6.6 0·62P
6.7 -~pd2
6.8 23·3 kN; 10·6 mm
6.9 4P/21; 10P/21
Tl
6.10 2 2
(EI cos o: + GJ sin o:)
6.11 H= f
PR 3
6.12 0·46P· 0·11PR· 0·124-
, ' EI
Chapter 7
7.1 MAa = -27 kNm; Mac= -68·4 kNm; VA= 29·1 kN; Va = 93·7 kN;
Vc = 77·2kN
7.2 40mm
7.3 MaE= -84 kNm
7.4 13·3 kN
7.5 MAE= -0·012Pl; May= -1·069Pl; MeG= -1·962Pl
7.6 MAa = 19 kNm; MaA =47·5 kNm; Men= -41·2 kNm; Moe= -25·3 kNm
7.7 0·301P; 0·463P; 0·236P
5 Ml
7.8
72 EI
7.9 MAa = 200 kNm; MaA = 133 kNm; Men =Moe = 67 kNm
7.10 MAa = -11·5 kNm; Mac =- 20·8 kNm; Mea= -10·4 kNm
Chapter 8
l l
8.1 4·43 AE to right; 18·1 AE down; F 80 = 4·45 kN
pz2
8.4 0·048-
EI
PZ PZ 2
8·5 v = 2AE ; e =64EI
Chapter 9
1
9.1 (2 (2x 2 l- x 3 - l 2 x)
ot.l x
9.2 4(1-a 2 ); a=z,a:1> 1; Mnmax =~l
Px 2 (3l- x)
9.3 213
9.4 See Table 9.1
51
9.5 8 h (a- 2a + a ), a= xfl
3 4
Chapter 10
al al
10.1 tan-= 2
2 +a 212
1t 2 EI 21t 2 Elvo
10.6 P= [2; F= /3[(2/1t) tanh (1t/2) -1]
1t 2EI k/ 2
10.8 y+ 1t2
10.9 13·9 kN
Chapter 11
11.2 End span 0·315/
11.3 P= 3·9:MP
I
y's -1 Z· M = ( 3 - yS) PI
11.4 h = 2 • p 2
11.5 X=2; M8 =Pl/6
ANSWERS TO PROBLEMS 349
11.6 BeamP=- 6Mp (k<~;

1) SideswayP= 2y3MP (k>-
y3) ;
l 3v3 kl 2
CombinedP= (
2
J
14M
+ k 3)!
( 1
3y3
y3)
~<k<-
2
11.7 1·865
INDEX
Antisymmetry, stiffness factor 224 plastic bending 109

Arches, funicular polygon 25 plastic bending with axial force 114
influence line 40 principal axis 99
rib shortening 208 second moment of area 99
temperature effects 208 section modulus 101
three-hinge 7 strain energy 107
two-hinge 206 unsymmetric section 104
Area-moment method 146 with axial force 108
Axes, local 244 with torsion and axial force 122
member 246 Betti's reciprocal theorem 289
system 245 Bow's notation 22, 50
Axial force, composite rods 80 Buckling of struts 299
sign convention 12 Bulk modulus 93
with bending 108
Carry-over factor 215
Beams, bending deflection 138 Castig1iano's theorem 177
bending stresses 97 applied to non-linear members 183
collapse of redundant 319 applied to redundant trusses 21 0
composite 102 Centre of shear 131
influence lines 29, 287 Collapse, of beams 319
moment distribution 220 of frames 325
plastic bending 109 Combination of mechanisms 334
shear centre 131 Combined, bending with axial force 108
shear deflection 153 plastic bending with axial force 114
shear stress 124 torsion, bending and axial force 122
shear stress in flanges 129 Compatibility 164
slope deflection 237 Engesser's theorem of 192
stiffness matrix 261 Complementary energy 157
unsymmetrical section I 04 applied to beams 182
Bending moment, calculation of 12 applied to frames 200
envelope 39 applied to rings and arches 204
graphical method 23 applied to trusses 180
influence lines 30 Complementary shear stress 76
rigid frames 62 Complex trusses 53
sign convention 12 Composite beams I 02
Bending of beams, composite 102 Composite rods, axial forces 80
Mohr circle 1OS temperature effects 81
neutral axis 98 Compound trusses 53
350
INDEX 351
Conjugate beam 150 Form factor 110

Core of section I 08 Frames, energy methods 199
Criticalload 299 flexibility matrix method 279
inclined members 233
Deflection of beams 13 8 moment-distribution method 230
complementary energy 182 multi-storey rectangular 235
conjugate beam 150 slope-deflection method 239
direct integration 139 stiffness-matrix method 264
duetoshear 153 virtual-work method 199
Macaulay's method 141 Free-body diagram 3
moment area 146 Fully plastic moment 110
singularity functions 141 Funicular polygon, bending moment 23
standard cases 151 description 21
variable section 150, 182 differential equation 27
virtual work I 7 5 reactions 22
Deflection of trusses, Castigliano's three-pin arch 25
theorem 183
complementary energy 179 Graphical analysis, plastic collapse 325
non-linear members 176 Grillages 269
real work 168
virtual work 171 Hooke's law 78
Williot-Mohr diagram 185 generalised 80
Determinacy of reactions 4 Hydrostatic pressure 93
Direct integration 139
Direction cosines 259 Influence lines, applications of Betti's
Direct stress 7 5 theorem 290
Displacement diagram 56 beams with two spans 287
Distribution factor 217 bending moments 30, 33, 36
Double modulus theory 307 envelope 36, 39
girders with floor beams 39
Electrical resistance strain gauge 89 multi-load systems 35
Energy methods applied to struts 313 multi-span beams 294
Energy theorems, Castigliano's theorem reactions 29, 32
177 shear force 30, 33, 35
complementary energy theorem three-pin arch 40
179 trusses 66
Engesser's theorem 192 virtual work 3 2
Engesser's theorem 192 Initial lack of fit 192
Equation of condition 5, 8 Instability 298
Equilibrium equations 2 Interaction diagram 115, 327
Equivalent section I 03
Euler critical load 299 Kinematic indeterminacy 165
Fixed-end moments 217 Lack of fit 192

Flexibility coefficients 159 Load factor 3 24
Flexibility matrix for space member Local coordinates 244
274 Lower-bound theorem 325
Flexibility method, outline of 270
plane frame 279 Macaulay's method 141
redundant plane truss 275 Matrix analysis of structures 244
Flexural rigidity 139 Maximum bending moment 36
Force polygon 20 Maximum shear force 35
352 INDEX
Maxwell's reciprocal theorem 151, 161 Principal strain 87

Modified stiffness, antisymmetry 224 Principal stress 83
pinned end 222 Principle of superposition 16
symmetry 223 Principle of virtual work 9
Modular ratio 103 Product second moment of area 99
Modulus, bulk 93
elasticity 78 Rayleigh method 314
elastic section 101 Reactions, influence lines 29, 32
plastic section 11 0 moments and resolution 6
rigidity 80 statical determinacy 4
Mohr circle, second moment of area virtual work I 0
105 Real work, method of 168
strain 89 Reciprocal theorem, Betti's 289
stress 85 Maxwell's 161
Mohr equation of virtual work 170 Relationships between elastic constants
Moment-area method, 146 92
Moment distribution, deflection of Rib shortening 208
supports 227 Rigidity, modulus of 78
distribution factor 217 Rigid jointed structures 60
frames with inclined members 233 Rings 203
fixed-end moments 21 7
modified stiffness 222 Safe theorem 325
rectangular multi-storey frames 23~ St Vernant's principle 79
sign convention 215 Second moment of area 99
stiffness and carry-over factor 215 Section modulus 101
Miiller-Breslau principle 287 Sections, method of 51
Shape factor 110
Neutral axis 98 Shear centre 131
Non-linear members 176 Shear deflection 153
Normal strain 77 Shear force, calculation 12
Normal stress 7 5 influence lines 30, 33, 35
No-sidesway distribution 232 sign 12
Shear strain 77
Permanent set 112 Shear stress, complementary 76
Perry-Robertson formula 311 in beams 124
Plane stress 83 in flanges 129
Plastic bending of beams 109 Sidesway 227
Plastic bending with axial force 114 Sign convention, general 73
Plastic collapse, combination of moment distribution 215
mechanisms 334 shear force and bending moment 12
deflections 3 21 Singularity functions 141
graphical analysis 325 Slenderness ratio 305
interaction diagram 327 Slope-deflection method 236
load factor 324 Southwell method for struts 311
redundant beams 319 Space truss 58
theorems 325 Squash load 114
virtual-work method 330 Stability of columns, double modulus
Plastic section modulus 110 theory 307
Plastic torsion of circular rods 122 Euler critical loads 299
Poisson's ratio 78 ideal elasto-plastic material 305
Principal axes 99 initially deformed strut 304
Principal plane 84 practical strut formulae 310
INDEX 353
Stability of columns-con tinued Tension coefficients 58
Rayleigh method 314 Three-pin arch, determinatio n of
Southwell method 311 reactions 7
tangent modulus theory 309 funicular polygon 25
Timoshenko method 314 influence line 40
Statical determinacy, reactions 4 Timoshenko method 314
rigid jointed frames 61 Torsion. circular rods 118
space trusses 57 multi-cell tubes 133
trusses 46 plastic 122
Stiffness coefficients 167 strain energy 121
Stiffness factor 215 thin rectangular section 134
Stiffness matrix, continuous beams 261 thin-walled tubes 132
grillages 269 with bending and axial force 122
pinned members 250 Transformati on matrix 250
plane frames 264 Transformed section, beams 103
space structure member 259 composite rods 81
transformatio n matrix 251 Trussed beam 198
Stiffness method 244 Trusses, complex 53
Strain 76 compound 53
Mohr circle for 89 deflections, see Deflection of trusses
normal 77 matrix methods applied 250
principal 87 redundant 189
shear 77 simple plane 45
Strain energy 9 5 space 57
axial load 9 5
bending 107
torsion 121 Uniqueness theorem 325
Stress 74 Unit-load method 171
complementa ry shear 76 Unsafe theorem 325
hoop 94 Upper-bound theorem 325
Mohr circle for 85
normal 75 Virtual displacement 10
principal 83 Virtual work, applied to plastic
shear 75 collapse 330
String polygon, see Funicular polygon deflections by 170
Struts, see Stability of columns influence lines by 32, 68
Superposition , principle of 16 Mohr equation of 170
Supports 2 principle of 9
Symmetry, stiffness factor 223
System coordinates 245
Warping 127, 154
Tangent modulus theory 309 Williot-Mohr displacement diagram
Temperature effects, arches 208 185
composite rods 81
redundant trusses 192 Young's modulus 78

Todd J. D., 1974, Structural Theory and Analysis

Uploaded by

Copyright:

Available Formats

Todd J. D., 1974, Structural Theory and Analysis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Todd J. D., 1974, Structural Theory and Analysis

Uploaded by

Copyright:

Available Formats

STRUCTURAL THEORY

All rights reserved. No part of this publication may be

First published 1974 by

2 Statically determinate structures 45

2.6 Compound trusses 53

3 Elementary elasticity, plasticity and bending of beams 73

4 Torsion and shear effects 118

5 Deflection of beams 138

8.12 Member flexibility matrix 274

9 Influence lines for statically indeterminate beams 287

10 Stability of columns 298

II Plastic analysis of beams and frames 318

Answers to problems 345

Oxford 1974 J.D. TODD

1.2 Equations of equilibrium

Several special cases arise for a coplanar system

1.3 Supports, reactions and free-body diagrams

1.4 Stability and determinacy of reactions

1.5 Calculation of reactions

concentration. However, for the purpose of calculation it will be sufficiently

Resolving forces horizontally

1: ME for portion ECD

1.6 Equation of condition

first degree. There is however an equation of condition which gives VA = Vc and it

1. 7 Principle of virtual work

1.8 Shear force and bending moment

8V8 = 150 + 240 + (40 x 9)

1.9 Relations between load, shear force and bending moment

M2 -M1 = Jx,x, -Qdx

1.1 0 Principle of superposition

A small knowledge of elasticity is assumed in the following. Let the deflection

Making use of Hooke's law

¥~'$< ' "'~ ?f'.&" -~!Y"&""&.:l 'IJD.' !!I (a)

A further example is that of a propped cantilever (figure 1.18a). The value of

1.11 The force polygon

1.12 The funicular polygon

1.13 Graphical determination of reactions

1.14 Graphical construction of bending-moment diagrams

VAx =H(y + xld) + 'L mx (1.12)

1.15 Graphics applied to a three-pin arch

1.16 The differential equation for a vertical load funicular polygon

1.17 Introduction to influence lines

1.19 Loading systems

It is comparatively simple to find the maximum value of a function for a single

~~1/c P """" 1.---

A bridge girder (figure 1.33a) has a total length of 100m. It is supported at

Vc 0 5/4 1 3/4 0 -1/4 0

1.21 Multi-load systems

where Ll VB represents the difference in the value of VB when Pn and Pn + 1 are

A~f':1-----a...;....:_ ~--+i -~~

dM=( R:d- R~d)dx

If x is the distance of P4 from the left-hand support

The value of the bending moment under P4 is

1.22 Influence lines for girders with floor beams

1.23 Influence line for three-pin arch

1.4 A beam is simply supported at A, B, C and D. AB =CD= 8 m, BC =12m.

Horizontal distance from l I 31 7/ 3/

2.1 Simple plane trusses

N\N\ Warren Truss Warren Truss With Verticals

Pratt Truss Howe Truss