Sei sulla pagina 1di 295

Contents

CHAPTER 1 Airplane and ship structures 1


Structures and Engineering 1
Principal structural units 2
Design 4
Loads 6
Function of flight vehicle structural members 6
Ships’ structures 9
Key words and concepts from Chapter 1 16
References 16

CHAPTER 2 Bars Subjected to Axial Loads 17


Axially loaded bar 17
The tensile test 19
Effect of temperature on strain 23
Bar reference axis 24
Linear elastic response 26
EXAMPLE 2.1 Axial bar with a specified uniform distributed load and specified end
displacements 27
EXAMPLE 2.2 A bar with fixed ends and subjected to an axial point force. 28
Work and energy methods 30
Concept of virtual displacement 30
Principle virtual work 32
EXAMPLE 2.3 Approximating the response of a bar using PVW. 34

Thin-Walled Structures v
Contents

Strain energy density 36


EXAMPLE 2.4 Strain energy of a bar with fixed ends and subjected to an axial point
force. 38
EXAMPLE 2.5 An elastic bar subjected to two forces and a thermal load 40
Castigliano’s first theorem 41
EXAMPLE 2.6 Response of a stepped bar by Castigliano’s first theorem 43
Complementary virtual work 45
Complementary strain energy 47
Relationship between the complementary strain energy and the strain energy
densities 49
EXAMPLE 2.7 Application of complementary virtual work to an elastic bar 50
Generalized form of Castigliano’s second theorem 52
EXAMPLE 2.8 Stepped bar response by Castigliano’s second theorem 54
EXAMPLE 2.9 A suspended bar subjected to self weight 55
Trusses 56
EXAMPLE 2.10 Three bar planar truss 60
EXAMPLE 2.11 Three bar truss with lack of fit 61
References 62
Problems 63

CHAPTER 3 Axial Normal Stress in Pure Bending and Extension 67


Pure bending 67
Geometry of deformation 68
Bending normal stress — flexure formula 75
EXAMPLE 3.1 Bending normal stress distribution in a cantilever beam with a thin-
walled zee section. 78
EXAMPLE 3.2 Lateral displacements of the zee section beam 80
Moments of areas 83
EXAMPLE 3.3 Thin-walled zee section properties by the composite body technique 86
EXAMPLE 3.4 Semicircular section with two stringers 88
Extension, pure bending, and thermal effects for multi-material beams 89
EXAMPLE 3.5 A multi-material beam with a symmetric cross section 92
Problems 96

CHAPTER 4 Axial Force, Shear Force and Bending Moment


Diagrams 99
Method of sections 99
Differential equation method 101
EXAMPLE 4.1 Cantilever wing with tip tank 105
EXAMPLE 4.2 The air load acting on a wing given as discrete data. 109
Semi-graphical method 116
EXAMPLE 4.3 Uniform barge with symmetric load 116
Buoyancy Force Distribution on Ships 118

vi Thin-Walled Structures
Contents

References 121
Problems 121

CHAPTER 5 Bending of Beams under Transverse Loads 125


Approximations for slender beams 125
Beam displacements 126
EXAMPLE 5.1 A statically indeterminate beam with an unsymmetrical cross section.
131
EXAMPLE 5.2 Contact between two cantilever beams. 135
Complementary virtual work and complementary energy 139
Complementary virtual work 140
Complementary strain energy 142
Beam displacement by Castigliano’s second theorem 145
EXAMPLE 5.3 End rotations of a simply supported beam subject to an end moment 145
EXAMPLE 5.4 Tip displacement of a cantilever wing spar a under distributed load 147
EXAMPLE 5.5 Strut-braced spar 149
Problems 153

CHAPTER 6 Shear Flow due to Shear Forces 155


Shear flows and shear stresses due to bending in a rectangular section beam 155
Shear flows due to transverse shear forces in open section beams 158
EXAMPLE 6.1 Shear flow distribution in a tee beam 160
Shear center of a thin-walled open section 163
EXAMPLE 6.2 Shear center location in an unsymmetrical section 167
Skin-stringer idealization 170
EXAMPLE 6.3 Shear flows in a stringer-stiffened C-section 173
Influence of transverse shear deformations on bending 177
Transverse shear strains, forces, and complementary energy density 178
Complementary energy density obtained from a two-dimensional element of the
wall 180
Determination of the transverse shear compliances 183
EXAMPLE 6.4 Shear compliances of a stiffened blade section 184
EXAMPLE 6.5 Deflection of a cantilevered beam due to bending and shear
deformation. 186
Problems 188

CHAPTER 7 Bars Subjected to Torsional Loads 191


Uniform torsion of a circular tube 191
Uniform torsion of an open section 197
EXAMPLE 7.1 Torsional response of a thin-walled open section and an equivalent
closed section 201
Non-uniform torsion; governing boundary value problem 202

Thin-Walled Structures vii


Contents

EXAMPLE 7.2 A uniform distributed torque acting on a bar with fixed ends. 203
EXAMPLE 7.3 A point torque acting on a bar with fixed ends. 204
Virtual work and strain energy 205
Strain energy density 207
Complementary virtual work and energy 208
Complementary strain energy 209
Unit twist of a single cell beam due to shear flow 210
Hooke’s law 211
Shear strain-displacement relation 212
Tangential displacement of a typical point on the contour 212
Relation between the shear flow and unit twist 214
Uniform torsion of a thin-walled closed section with a contour of arbitrary
shape 215
EXAMPLE 7.4 Torsion of a circular, bi-material section 218
Shear center of a closed section 219
EXAMPLE 7.5 Shear center location of a single-cell, closed section having one axis of
symmetry. 219
Uniform torsion of multi-cell closed sections 222
EXAMPLE 7.6 Uniform torsion of a two-cell section 225
Resultant of a constant shear flow in a curved branch 226
EXAMPLE 7.7 Torsion of a five cell closed section; circuit shear flow 229
Torsion of hybrid sections 231
References 232
Problems 233

CHAPTER 8 Criteria for Initial Yielding 237


Ductile and brittle behavior 237
Criteria for initial yielding of ductile materials 238
Stress transformation equations for generalized plane stress 241
Principal stresses and maximum shear stress 243
EXAMPLE 8.1 Maximum shear stress in tensile test 246
EXAMPLE 8.2 Mohr’s circle for hydrostatic stress state 246
EXAMPLE 8.3 Principal stresses and maximum shear stress 247
Octahedral shear stress 249
Mises criterion for initiation of yielding 252
Maximum shear-stress criterion 254
EXAMPLE 8.4 Factor of safety against initial yielding 256
EXAMPLE 8.5 Stress responses of a stringer-stiffened, single cell beam. 257
EXAMPLE 8.6 Minimum weight design of the beam in Example 8.5 subject to a
constraint on initial yielding 261
References 264

viii Thin-Walled Structures


Contents

CHAPTER 9 Buckling 265


One-degree of freedom model 265
Static equilibrium 266
Stability analysis 267
Perfect Columns 269
EXAMPLE 9.1 Critical load for clamped-free boundary conditions (B) 272
Imperfect columns 275
Eccentric load 275
Geometric imperfection 277
Column Design Curve 279
Inelastic buckling 280
Bending of thin plates 283

Thin-Walled Structures ix
Contents

x Thin-Walled Structures
CHAPTER 1 Airplane and ship
structures

The objective of Thin-Walled Structures is to provide an understanding of the basic concepts of stress and defor-
mation of stiffened-shell structures with applications to aerospace and ocean vehicles. In this chapter we discuss
why a structure carries load, the types of loads acting on airplane and ship structures, the definitons and functions
of principal structural units, and the relation of structural design to the overall design.

1.1 Structures and Engineering


A structure may be defined as any assemblage of materials which is intended to sustain loads. Structures function
to protect people and things, and are so common and familiar to us that when we are informed of their use and
form we conceive of ourselves as knowledgeable. After all, it does not take a structural engineering degree to
build an ordinary shed. However, when asked to build an airplane or a ship we would probably be more hesitant,
since if an airplane or a ship breaks many people are likely to be killed. The crux of the issue in the design of
complex structures, like aircraft, is to not only know the use and form of the structure, but also to know why a
structure is able to carry load.

Although Galileo (1564-1642) began an inquiry into the strength of materials, it is Robert Hooke (1635-
1702) who is credited with providing the answer as to why structures carry load. A historical account of Hooke’s
discoveries is discussed in chapter two of an informative and readable book on structures by Gordon (1978). Gor-
don’s text is the source for what is written here. By Newton’s law of action and reaction we know that isolated
forces do not exist in nature. A force acting on an inanimate solid is reacted by a force produced by the solid. But
how does the solid produce such a reaction force? Hooke in 1676 recognized that every kind of solid changes its
shape when subjected to a mechanical force and it is the change in shape which enables the solid to provide the
reaction force. This shape change extends to the very fine scale of molecules where there is a large resistance to
stretching and compressing of chemical bonds. Hooke’s measurements also showed that many solid materials
recovered their original shape when they are unloaded; i.e, they are elastic. The science of elasticity is about the
interactions between forces and deflections in materials and structures. The formulation of the mathematical the-
ory of elasticity for a solid continuum is well established; e.g., see Sokolnikoff (1956). Exact elasticity solutions
are known for a small, but important, number of structural problems. However, the mathematical solutions to
elasticity problems are usually challenging. Approximations to elasticity theory which exploit the geometry and

Thin-Walled Structures 1
Airplane and ship structures

material construction of practical structures is called structural mechanics. For example, the structural mechanics
of beams, plates, and shells approximate elasticity by exploiting the fact that one or two dimensions of the solid
body are very small with respect to the remaining dimensions. Structural mechanics reduces the mathematical
complexity somewhat relative to three-dimensional elasticity, and it is the theory most often used by the practic-
ing engineer.

Two important properties common to the analysis of any structure are stiffness and strength. Stiffness is a
measure of the force required to produce a given deflection, and strength refers to the force, or force intensity,
necessary to cause failure. A criterion for failure is required in order to determine the strength of a structure, and
this depends upon the particular application. For example, failure can be defined when a stress (internal force
intensity) exceeds the yield stress of the material, or failure can mean excessive displacements which occur dur-
ing buckling. The stiffness and strength of a structure depend on its geometrical configuration, connections, and
the stiffness and strength of the materials from which it is made.

It is important to recognize that structures are made from materials, and that the history of structures follows
the development of materials and the development of tools to fabricate the materials. The evolution of the air-
frame, for example, is tied closely to the introduction of materials and cost-effective means for their fabrication.
For example, early aircraft were constructed of wire-braced wood frames with fabric covers. Currently, advanced
composite materials are very attractive for weight-sensitive structures, like aircraft, because of their high stiff-
ness-to-weight and strength-to-weight ratios.

The distinction between structures and materials is not always clear. It may be said that the forward-swept
wing on Grumman's X-29 demonstrator airplane is a structure, and the material it is made from is an advanced
composite. However, advanced fiber-reinforced composites are made from stiff, strong, continuous fibers embed-
ded in a pliant matrix. The complex constitution of an advanced composite, therefore, may be considered either
as materials or structures.

1.2 Principal structural units


The principal structural units of fixed-wing airplanes are the fuselage, wings, stabilizers, control surfaces, land-
ing gear, nacelle and engine mounts. Light airplanes are shown in Fig. 1.1 and a heavier airplane (Douglas DC-3)
is shown in Fig. 1.2. A cargo ship is depicted in Fig. 1.3, and its principal structural units will be discussed in
more detail in Section 1.6. After some study of the structures shown in these figures, it is reasonable to suggest
that some principal structural units such as the fuselage, wings, and ship hull have the features commonly attrib-
utable to a beam. That is, two dimensions of the overall component, or the cross-sectional dimensions, are small
with respect to the third, or longitudinal, dimension. Indeed, to simplify the analysis of such complex structures,
we can approximate them as slender, built-up beams!

Hence, the basic conceptualizing we make for complex vehicle structures such as aircraft and ships is that a
fuselage, a wing, or a hull is a thin-walled beam. That is, a vehicle structure as a whole is assumed to be a one-
dimensional structural element in the mathematical sense that its response under load can be described by ordi-
nary differential equations. Aero-hydrodynamic and other loads that act on the structure cause extensional, bend-
ing, and torsional deformations of the structure. The cross section of the structure is built from many actual
structural elements such as spars, frames, and panels. This beam assumption is particularly suited for the analysis
required in preliminary design. Of course not all principal structural units can be modeled as a beam. In constrast
to a high aspect ratio wing, a delta wing whose span and chord are of comparable value (low aspect ratio) is not
modeled very well by using the beam assumption.

2 Thin-Walled Structures
Principal structural units

(a) Global Swift

(b) Taylorcraft airplane


Fig. 1.1 Principal structural units of light airplanes (from Aircraft Basic Science, 1948)

Thin-Walled Structures 3
Airplane and ship structures

Fig. 1.2 Principal structural units of a Douglas airplane (from Aircraft Basic Science, 1948)

1.3 Design
An aircraft or ship may be considered a system consisting of several subsystems; a structural subsystem, a con-
trol subsystem, a propulsion subsystem, cargo handling subsystem, etc. Vehicle design must consider the total, or
integrated, system to achieve optimal performance. An important contribution to the overall vehicle performance
is to minimize weight in the structural subsystem design. A minimum weight vehicle structure can carry the same
payload with less fuel consumption. In addition, a lighter structure can reduce operating costs through less main-
tenance, and also may reduce initial cost by requiring less labor for fabrication. Modern engineering design has
been revolutionized by the development of high-speed computers combined with optimization theory. As a math-
ematical problem in optimization, structural design may be considered as the development of a computational
algorithm for choosing member spacing and dimensions such that weight is minimized (objective) subject to
constraints on strength and stiffness. The role of structural mechanics in this design process is to provide a
description of the state of stress and deformation throughout the structure for a given structural configuration,
such that the constraints can be evaluated. Structural mechanics provides the theory for the analysis of the struc-
tural response (state of stress and deformation).

4 Thin-Walled Structures
Design

Fig. 1.3 Levels of structural analysis for a large ship (from Hughes, 1983)

The overall dimensions of a vehicle structure are usually determined by more general requirements rather
than for structural considerations. Thus, the structural design begins with the overall dimensions given, and two
levels of design may be distinguished. The first level is called preliminary design, and at this level the locations
and dimensions of the principal structural members are determined. The second level is called detail design, and

Thin-Walled Structures 5
Airplane and ship structures

at this level the geometry and dimensions of the local structure like joints, cutouts, reinforcements, etc. are deter-
mined.

1.4 Loads
The first step in preliminary design is to determine the external loads acting on the structure. Maneuvering flight
vehicles are subjected to gravity, aerodynamic forces, and inertial loads. In addition, the landing loads, and wind
gust loads during turbulent weather must be considered. Ships are subjected to gravity, buoyancy forces, and
inertial loads. Also, wave loads and other hydrodynamic loads such as slamming, sloshing of liquid cargos, etc.,
must be considered in ship design. The calculation of aerodynamic and hydrodynamic forces are sufficiently
complex such that their determination is done by specialists rather than by designers.

Loads on a vehicle structure may be classified as static or dynamic, and either deterministic or probabilistic.
Gust loading conditions for aircraft and wave loading conditions for ships are not known with absolute certainty,
so that these load magnitudes are estimated on a statistical basis using probability theory together with past expe-
rience. The type of loading has a direct influence on the type of structural response analysis required. For exam-
ple, dynamic loading requires a structural dynamics analysis.

In traditional structural design most of these loads are not affected by the structural configuration or dimen-
sions of the members. They are a function of the wing shape, or hull shape, and other nonstructural factors.
Hence, the determination of the loads, a very crucial aspect of the design process, is essentially a separate task
typically performed by an aerodynamicist or hydrodynamicists. In modern flexible vehicles the loads greatly
influence the shape and so aeroelastic or hydroelastic load analysis must be performed. That is, the structural and
aero/hydrodynamic analysis must be combined to obtain the correct loads. This interaction between the structure
and the shape is so important for flexible vehicles made from composite materials that it is expected that in the
future the shape and structural design will be combined.

1.5 Function of flight vehicle structural members


The following description of the functions of flight vehicle structures is excerpted from Rivello (1969, Section 7-
6).

The structure of a flight vehicle usually has a dual function: it transmits and resists the
forces which are applied to the vehicle, and it acts a cover which provides the aerodynamics
shape and protects the contents of the vehicle from the environment. This combination of roles
is fortunate since, from the standpoint of structural weight, the most efficient location for the
structural material is at the outer surface of the vehicle. As a result, the structures of most flight
vehicles are essentially thin shells. If these shells are not supported by stiffening members,
they are referred to as monocoque. When the cross-sectional dimensions are large, the wall of
a monocoque structure must be relatively thick to resist bending, compressive, and torsional
loads without buckling. In such cases a more efficient type of construction is one which con-
tains stiffening members that permit a thinner covering shell. Stiffening members may also be
required to diffuse concentrated loads into the cover. Constructions of this type are called
semimonocoque. Typical examples of semimonocoque body structures are shown in Fig. 7-5.
While at first glance these structures appear to differ considerably, functionally there are simi-
larities. Both have thin-sheet coverings, longitudinal stiffening members, and transverse sup-

6 Thin-Walled Structures
Function of flight vehicle structural members

Longitudinal
stringers

Cover Skin
Transverse
frames
(a)

Cover skin

Transverse
Spar web rib

Spar cap Longitudinal


stringers
(b)

Fig. 7-5 Typical semimonocoque construction. (a) Body


structures: (b) aerodynamic surface structures.

porting elements which play similar structural roles.


In semimonocoque structures the cover, or skin, has the following functions:
1. It transmits aerodynamic forces to the longitudinal and transverse supporting members by
plate and membrane action (Chap. 13).
2. It develops shearing stresses which react the applied torsional moments (Chap. 8) and
shear forces (Chap. 9).
3. It acts with the longitudinal members in resisting the applied bending and axial loads
(Chaps. 7, 15, and 16).
4. It acts with the longitudinals in resisting the axial load and with the transverse members in
reacting the hoop, or circumferential, load when the structure is pressurized.
In addition to these structural functions, it provides an aerodynamic surface and cover for
the content of the vehicle. Spar webs (Fig. 7-5b) play a role that is similar to function 2 of the
skin.
The longitudinal members are known as longitudinals, stringers, or stiffeners. Longitudi-
nals which have large cross-sectional areas are referred to as longerons. These members
serve the following purposes:
1. They resist bending and axial loads along with the skin (Chap. 7).
2. They divide the skin into small panels and thereby increase its buckling and failing stresses
(Chaps. 15 and 16).
3. They act with the skin in resisting axial loads caused by pressurization.

Thin-Walled Structures 7
Airplane and ship structures

The spar caps in aerodynamic surface perform functions 1 and 2.


The transverse members in body structures are called frames, rings, or if they cover all or
most of the cross-sectional area, bulkheads. In aerodynamic surfaces they are referred to as
ribs. These members are used to:
1. Maintain the cross-sectional shape.
2. Distribute concentrated loads into the structure and redistribute stresses around structural
discontinuities (Chap. 9).
3. Establish the column length and provide end restraint for the longitudinals to increase their
column buckling stress (Chap. 14).
4. Provide edge restraint for the skin panels and thereby increase the plate buckling stress of
these elements (Chap. 16).
5. Act with the skin in resisting the circumferential loads due to pressurization.
The behavior of these structural elements is often idealized to simplify the analysis of the
assembled component. The following assumptions are usually made:
1. The longitudinals carry only axial stress.
2. The webs (skin and spar webs) carry only shearing stresses.
3. The axial stress is constant over the cross section of each of the longitudinals, and the
shearing stress is uniform through the thickness of the webs.
4. The transverse frames and ribs are rigid within their own planes, so that the cross section is
maintained unchanged during loading. However, they are assumed to possess no rigid-
ity normal to their plane, so that they offer no restraint to warping deformations out of
their plane.
When the cross-sectional dimensions of the longitudinals are very small compare to the
cross-sectional dimensions of the assembly, assumptions 1 and 3 result in little error. The
webs in an actual structure carry significant axial stresses as well as shearing stresses, and it
is therefore necessary to use an analytical model of the structure which includes this load-car-
rying ability. This is done by combining the effective areas of the webs adjacent to a longitudi-
nal with the area of the longitudinal into a total effective area of material which is capable of
resisting bending moments and axial forces. A method for determining this effective area is
given in Sec. 15-7. In the illustrative examples and problems on stiffened shells in this and suc-
ceeding chapters it may be assumed that his idealization has already been made and that
areas given for the longitudinals are the total effective areas. The fact that the cross-sectional
dimensions of most longitudinals are small when compared with those of the stiffened-shell
cross section makes it possible to assume without serious error that the area of the effective
longitudinal is concentrated at a point on the midline of the skin where it joins the longitudinal.
The locations of these idealized longitudinals will be indicated by small circles, as shown in
Fig. 7-6b. In thin aerodynamic surfaces the depth of the longitudinals may not be small com-
pared to the thickness of the cross section of the assembly, and more elaborate idealized
model of the structure may be required.
The fewer the number of longitudinals, the simpler the analysis, and in some cases several
longitudinal may be lumped into a single effective longitudinal to shorten computations. (Fig. 7-
6). On the other hand, it is sometimes convenient to idealize a monocoque shell into an ideal-
ized stiffened shell by lumping the shell wall into idealized longitudinals, as shown in Fig.7-
7,and assuming that the skin between these longitudinals carries only shearing stresses.. The
simplification of an actual structure into an analytical model represents a compromise, since
elaborate models which nearly simulate the actual structure are usually difficult to analyze. A
more complete discussion of the idealization of shell structures will be found in Ref. 4
Once the idealization is made, the stresses in the longitudinals due to bending moments,

8 Thin-Walled Structures
Ships’ structures

Actual skin and web


carries axial and
shear stresses

(a) Effective longitudinals


(axial stress only)
Idealized webs
(shear stress only)

(b)
Fig. 7-6 Idealization of semimonocoque structure. (a) Actual
structure; (b) idealized structure

Effective longitudinals
(axial stress only)

Wall carries axial


Idealized web
and shear stresses
(shear stress only)
(a) (b)
Fig. 7-7 Idealization of a monocoque shell. (a) Mono-
coque shell; (b) idealization

axial load, and thermal gradients can be computed from the equations of this chapter if the
structure is long compared to its cross-sectional dimensions and if there are no significant
structural or loading discontinuities in the region where the stresses are computed. In many
flight structures the cross section tapers; the effects of this taper upon the stresses are dis-
cussed in Chap. 9. When discontinuities or other conditions arise which violate the analytical
assumptions made in the Bernoulli-Euler theory, it is necessary to analyze the stiffened shell
as an indeterminate structure (Chaps. 11 and 12).

1.6 Ships’ structures


The following description of the distortion and functions of ship structures is excerpted from Muckle (1967).
The Distortion of the ship’s structure
The study of the static forces on the ship has shown that the ship can bend in a longitudi-
nal vertical plane like a beam. This is one of the most important types of distortion to which the
ship is subjected, and is one in which the entire structure of the ship takes part. While consid-
ering this longitudinal bending of the structure it should be mentioned that it is also possible for
the ship to bend in horizontal plane. Consider a ship moving diagonally across a regular wave
system as in Fig. 4. The crests are not perpendicular to the centre line of the ship and Fig. 5

Thin-Walled Structures 9
Airplane and ship structures

shows that the slope of the waves at various points in the length of the ship varies, being
sometimes positive and sometimes negative. This means that there are sideways forces acting
on the ship which will not only cause swaying, but also bending in the horizontal plane. This
bending has in the past been neglected and it is safe to say that the forces and moments gen-
erated are likely to be of small amount.
Wave crest

Wave crest

Wave crest

Wave crest
Fig. 4 Ship moving diagonally across waves

Referring again to Fig. 5, it will be evident that, because of the variation in the wave slope
at different sections in the length, not only will sideways forces be generated but there will also
be moments applied at the various sections. As these may change sign along the length of the
ship, twisting is possible with the consequent generation of torsional stresses. Once again it is
perhaps doubtful whether this type of distortion is important from the point of view of the
strength of the structure. The problem has been, partially investigated in the past, and at the
present there appears to be some interest in it in view of the tendency to increase the size of
hatch openings, thus reducing the torsional rigidity of the structure.

A.P. F.P.
1/4 length 3 length
/4

Fig. 5 Wave surface at various


Amidships positions in length

Consider now a transverse section of a ship as shown in Fig. 6. This section is subject first
of all to static pressure due to the surrounding water. It will also be subjected to internal load-
ing due to the weight of the structure itself and the weight of the cargo etc. which is carried.
The effect of these static forces is to cause transverse distortion of the section, as shown by
dotted lines in Fig. 6. It is worthy of note that this type of distortion would take place regardless
of whether there was bending in the longitudinal direction. It is possible therefore to recognise
an entirely independent study dealing with the transverse deformation of the ship’s structure.

10 Thin-Walled Structures
Ships’ structures

Cargo load

Cargo load

Water pressure load

Fig. 6 Distortion of transverse section due to static loading

Not only do the water pressure and the local internal loads cause transverse bending but it
is possible to have local deformation of the structure due to these forces. A typical example of
this is the bottom plating of a ship between floors or longitudinals. Fig. 7 shows a strip of such
plating between two floors or longitudinals. The tendency is for the plating to bend as a beam

Inner bottom

Floors

Outer bottom

Water pressure
Fig. 7 Distortion of bottom plating due to water pressure

in between these members. Other parts of the structure which could be deformed under local
loads are tank top plating, bulkheads, girders under heavy loads such as machinery etc. In this
way it will be seen that there is another aspect of the strength of the structure which may be
defined as local deformation.
Summarising this section, it is clear that the overall problem of the strength of the ship’s
structure may be conveniently divided into three sections:
(1) Longitudinal strength (2) Transverse strength (3) Local strength
Since any given part of the structure of the ship may be subjected to one or more of the
modes of distortion discussed, it will be seen that the resultant state of stress in that part could

Thin-Walled Structures 11
Airplane and ship structures

be very complex. It is for this reason that, in a first study at least of the strength of the ship’s
structure, longitudinal bending, transverse bending and local bending are treated entirely inde-
pendent, so that each of the three divisions of the subject of strength of ships quoted above
can be investigated separately. This is the only realistic way of tackling the problem.

Function of the ship’s structure

It has been shown that the ship is capable of bending in a longitudinal vertical plane and it
follows therefore that there must be material in the ship’s structure which will resist this bend-
ing; or in other words there must be material distributed in the fore and after direction to fulfil
this purpose. It follows that any material distributed over a considerable portion of the length of
the ship will contribute to the longitudinal strength. Items which come into this category are the
side and bottom shell plating, inner bottom plating and any decks which there may be. Fig. 8 is
an outline section showing these items. As far as decks are concerned, it is usual to consider
only the material abreast the line of openings, such as hatches and engine casings.

Upper deck
plating

2nd deck
plating

Side
shell

Inner bottom
plating
Margin Centre
plate girder
Bottom shell

Fig. 8 Section through ship showing material resisting longitudinal bending

It will be clear that this longitudinal material forms a box girder of very large dimensions in
relation to its thickness. Consequently, unless the plating was stiffened in some way it would
be incapable of with standing compressive loads. For this reason therefore it becomes neces-
sary to fit transverse rings of material spaced from 2 ft. to 3 ft. apart throughout the length of
the ship. This is the procedure which is adopted in what is usually called a transversely framed
ship. The transverse stiffening consists of three parts; in the bottom between the outer and
inner bottoms there are several vertical plates called floors which have lightening and access
holes cut in them as shown in Fig. 9; in the sides of the ship rolled sections called side frames,
are welded to the plating (see Fig. 9); the decks are also supported by rolled sections welded
to the plating, called beams. The floors, side frames and beams at the various decks are con-
nected by means of brackets so that a continuous transverse ring of material is provided. As
stated earlier, the spacing of these transverse rings, usually called the frame spacing, is
between 2 ft. and 3 ft. and depends upon the length of the ship. It will be seen that the effect of

12 Thin-Walled Structures
Ships’ structures

supporting the plating in this way is to reduce the unsupported span and hence to raise the
buckling strength of the plating, to enable it to carry compressive loads.

Tween deck Upper deck


frame Tween deck
beam
pillar

Second deck
beam Hold
pillar

Floor
plate
Tank side Centre
bracket girder
Side girder

Fig. 9 Section through ship showing transverse structure

Another function of these transverse rings is to prevent transverse distortion of the struc-
ture, so that the floors, side frames and beams are the main items contributing to the trans-
verse strength of the structure of the ship. The main force involved here is that due to water
pressure and, as this will be greatest on the bottom of the ship, the bottom structure should be
very heavy. This is in fact so, a very heavy girder being provide by the floor plate in conjunction
with its associated inner and outer bottom plating. The side of the ship is also subjected to
water pressure of rather lesser magnitude, and in this case adequate stiffening is provided by
the girder consisting of the side frame welded to the side shell plating. As far as decks are con-
cerned, here again the beam with its associated deck plating forms an effective built-up girder.
The main factor determining the sizes of the beams is the load which they have to carry. This
load may be a cargo load, a load due to passengers or, in the case of a weather deck some
weather load.
Other items of the structure which contribute to transverse strength are watertight bulk-
heads. Their primary object is, of course, to divide the ship into a series of watertight compart-
ments, but since they consist of transverse sheets of plating they have very considerable
transverse rigidity and hence contribute greatly to the prevention of transverse deformation of
the structure.
The structure shown in Fig. 9 is typical of a transversely framed ship. It is common practice
nowadays to adopt a different form of construction in which the sides of the ship are stiffened
transversely whilst the decks and bottom are stiffened by means of longitudinals. This type of
construction is shown in Fig. 10. As will be shown later, the effect of stiffening the deck and
bottom by longitudinal members instead of transverse members is to increase very greatly the
buckling strength of the plating, and it is largely for this reason that this method of construction
has been adopted.Since these longitudinals are effectively attached to the plating they contrib-
ute also to the general longitudinal strength of the structures. The longitudinals have to carry
cargo and water pressure loads and so, in order to reduce their scantlings, they must be sup-
ported at positions other than at bulkheads. This is achieved by introducing deep transverse
beams in the decks spaced some 6 to 12 ft. apart and by having transverse plate floors in the

Thin-Walled Structures 13
Airplane and ship structures

Tween deck
frame
Upper deck
longitudinals

Deep transverses
2nd deck spaced 6 to 12 ft.
longitudinals apart
Side frame

Inner bottom
longitudinals

Outer bottom
longitudinals
Fig. 10 Section through ship with longitudinally stiffened decks and bottom

bottom at the same spacing. These widely spaced transverse members, in conjunction with
closely spaced side framing, then provide the transverse strength of the structure.
The longitudinal system of framing has often also been extended to the sides of the ship
as well as the decks and bottom. In fact when initially developed for use in oil tankers this was
the method which was adopted. This was called the Isherwood System. At a later stage in the

Flat bar deck


longitudinals

Transverses spaced Side Centre girder


about 10 ft. apart girder
Wing bulkhead
Flat bar side longitudinals Flat bar longitudinals
Side girder
Center girder

Flat bar bottom


longitudinals
Fig. 11 Section through large modern oil tanker

development of the tanker the combined system of longitudinals in the bottom and deck with
transverse side framing was employed. In many of the larger oil tankers of the present day,
however, the complete longitudinal framing system has been used. Figure 11 shows the mid-
ship section of such a tanker.
Where transverse beams are employed in the decks of ships it would be impracticable to

14 Thin-Walled Structures
Ships’ structures

run these from side to side of the ship without intermediate support. It is therefore necessary
to introduce pillars to support the beams. In the early development of the iron and steel ship
these pillars were closely spaced, generally being on alternate beams with a longitudinal angle
runner fitted under the beams to spread the load to those beams not supported by pillars. This
meant that access to the sides of cargo holds could only be made between two pillars, so that
the available space was only about 5 ft. The later development was to support the deck beams
by one or more heavy longitudinal girders and to support these girders by means of wide-
spaced pillars. With this arrangement there would be probably two girders in the breadth of the
ship each supported by two pillars in the length of the hold. Such an arrangement is shown in
Fig. 12. By supporting the deck beams with lines of pillars or heavy longitudinals, the scant-

Hatch coaming
Girder
Pillars
Hatch coaming

Girder
Bulkhead

Bulkhead
Pillars

Inner bottom

Outer bottom
Elevation through hold

Pillars
Girder

Deck
beams

Girder
Pillars

Plan at deck
Fig. 12 Wide-spaced pillar and girder arrangement in transversely framed ship

lings of the beams are greatly reduced and, further, by the use of wide-spaced pillars access
to the holds is made easy. When longitudinal stiffening of decks is used, the system of con-
struction just described can be imagined to have been turned around, The longitudinals
replace the beams and the deep transverse beams replace the longitudinal deck girders in the
transversely framed ship.
In addition to their functions in resisting longitudinal and transverse bending, many of the
parts of the structure referred to in this section have also to support local loads. Thus beams
and girders will often be subjected to loads due to machinery and loads produced by lifting

Thin-Walled Structures 15
Airplane and ship structures

equipment such as derricks and the like. The outside plating of the ship has also to withstand
water pressure, and this could produce local bending of the plating between the stiffening
members such as floors and frames. In general it could be said that nearly every structural
member in the ship is a local strength member.
The foregoing discussion has shown briefly the functions which the various parts of the
ship’s structure have to perform. It can be seen that particular part of the structure may have to
perform several functions at the same time. In succeeding chapters methods for determining
the stresses in the various parts will be dealt with in detail.

1.7 Key words and concepts from Chapter 1


structure
elasticity
structural mechanics
stiffness
strength
preliminary design
types of loads
monocoque & semimonocoque
spar, spar caps, spar web
bulkheads, ribs, rings
structural functions of the skin, longitudinals, and frames
idealization of semimonocoque structure
stresses due to bending and torsion
longitudinal, transverse, and local strength of ship structures
transversely framed, longitudinally framed, and Isherwood system of framing of ships
girder, pillar, beam, floor plate
hatch, hatch coaming
buckling strength
scantlings

1.8 References
Anon., Aircraft Basic Science, 1948, First Edition, Northrop Aeronautical Institute, Charles E. Chapel, Chief
Editor, McGraw-Hill Book Company, Inc, p. 59 & 60.

Gordon, J.E., 1978, Structures: or, Why things don’t fall down, (A Da Capo paperback) Reprint. Originally
published by Harmondsworth: Penguin Books, pp. 33-44.

Hughes, O.F., 1983, Ship Structural Design, John Wiley and Sons, New York, N.Y., p. 8.

Muckle, W., 1967, Strength of Ships' Structures, E. Arnold Inc., pp. 5-12.

Rivello, R. M., 1969, Theory and Analysis of Flight Structures, McGraw-Hill, pp. 143-147.

Sokolnikoff, I.S., 1956, Mathematical Theory of Elasticity, Second Edition, McGraw-Hill Book Company,
New York.

16 Thin-Walled Structures
CHAPTER 2 Bars Subjected to Axial
Loads

A bar is a structural member that is relatively long along one axis and relatively compact in cross section in
planes perpendicular to the axis. Bars can be straight or curved. Bars are among the most widely use structural
elements. In this chapter only straight bars are considered that are subjected to loads directed along the reference
axis of the bar. The reference axis is parallel to the long axis of the bar and will be defined in Section 2.4. Axial
loads applied along the reference axis of a straight bar cause extensional and/or compressive deformations. A
slender bar in compression is likely to buckle and in that case the bar is called a column. Buckling results in a
combination of bending and compressive deformations of the column. Loads applied perpendicular to the refer-
ence axis cause the bar to bend, and in that case the bar is called a beam. Beams are the subject of the next chap-
ter.

The three basic steps to analyzed the static response of any structure are discussed for a bar in Section 2.1 to
Section 2.5. These three fundamental steps of static structural mechanics are
• equilibrium conditions,
• strain-displacement conditions, or conditions of geometric fit, and
• a material law, or constitutive behavior.
Work and energy methods are presented in Section 2.6 to Section 2.11, which includes the topics of virtual work,
strain energy, complementary virtual work, complementary strain energy, and Castigliano’s theorems. Applica-
tions of the energy method to trusses is presented in Section 2.12.

2.1 Axially loaded bar


Consider a straight bar of length L, whose cross section is uniform along its length with its cross-sectional area
denoted by A. The bar is referred to a Cartesian coordinate system x, y, and z with the z-axis parallel to the length
and the x and y axes in a plane parallel to the cross section. The origin of the z-axis is taken at the left end of bar,
so 0 ≤ z ≤ L . The bar is subjected to the following loads: a distributed force per unit length of intensity p z ( z ) ,
either an axial force Q 1 or axial displacement q 1 at the left end, and to either an axial force Q 2 or axial dis-

Thin-Walled Structures 17
Bars Subjected to Axial Loads

placement q 2 at the right end. The distributed force intensity p z ( z ) , forces Q1 and Q2, and the corresponding
displacements q1 and q2, respectively, are defined positive if they act in the positive in the positive z-direction.
See Fig. 2.1. Under the imposed loads, the bar is in tension and/or compression.

y
y
pz ( z )
Q 1, q 1 Q 2, q 2
z, w x

L
Cross section
Fig. 2.1 Axially loaded bar

Equilibrium The internal normal force, or axial force, acting in the z-direction is denoted by function N ( z ) ,
and N is positive if tensile and negative if compressive. See Fig. 2.2. If we consider an interior element of the bar

p z ( z * )dz
Q1 N (ε) N N + dN N (L – ε) Q2

ε→0 z dz
ε ε ε→0
z L
z < z * < z + dz
left end right end

Fig. 2.2 Free body diagrams for equilibrium of the bar

as shown in the center sketch of Fig. 2.2 and a positive normal force is defined to act in the positive z-direction
on a positive z-face, then the action-reaction law requires a positive normal force acting on the negative z-face to
act in the negative z-direction. A positive z-face of this interior element is the face whose normal pointing away
from the material inside the element is in the positive z-direction. Conversely, a negative z-face of this interior
element is the face whose normal pointing away from the material inside the element is in the negative z-direc-
tion. Force equilibrium in the z-direction of the differential interior element of the bar shown in the figure, in the
limit as dz → 0 , gives the following differential equation of equilibrium.

dN
------- + p z ( z ) = 0 0<z<L (2.1)
dz
(In Fig. 2.2 note that coordinate z* where the distributed load intensity is evaluated on the differential element
approaches z in the limit as the element length goes to zero; i.e., z * → z in the limit.)

Let the axial displacement function be denoted by w(z). The function w(z) is the displacement in the z-direc-
tion of a particle located at z in the undeformed bar due to the imposed loads as is shown in Fig. 2.3. The axial

18 Thin-Walled Structures
The tensile test

displacement of all material points in the cross section is assumed to be the same, but the displacement changes
from cross section to cross section. Hence, the axial displacement is a function of z and not of x and y. The
boundary conditions at z = 0 are

either w ( 0 ) = q 1 or N ( 0 ) = – Q 1 but not both (2.2)

and the boundary conditions at z = L are

either w ( L ) = q 2 or N ( L ) = Q 2 but not both (2.3)

Note that we cannot specify both the displacement and its corresponding force at a point on the body. If we do
not specify the displacement then equilibrium applied separately at the boundaries of the bar (at either z = 0 or z
= L) relates the external force to the internal bar force at the boundary point. See the free body diagrams of the
left and right ends of the bar in Fig. 2.2.

Strain-displacement Extensional deformation of an element of the bar is shown in Fig. 2.3. The axial normal

y
z z + dz z + w(z) z + dz + w + dw
dz dz + dw
z - axis
w + dw
w(z)

Fig. 2.3 Extensional deformation of an element of the bar

strain due to extension is given by

( dz + dw ) – dz dw
ε z = Limit ------------------------------------ = -------
dz dz
dz → 0 (2.4)

The axial normal strain εz is a function of z, and it is dimensionless – it is the change in length divided by the
original length. If the axial displacement function w is a constant value for 0 ≤ z ≤ L, then the bar displaces as a
rigid body and the normal strain is zero, as is evident from eq. (2.4).

To complete the analysis for the response of the bar, we need to relate the internal axial force N to the axial
normal strain εz, and this requires a material law. The material law is obtained from material characterization
tests of carefully designed specimens as discussed in the following section.

2.2 The tensile test


Material characterization involves devising experiments for a very simple loading situation from which a mate-
rial law can be developed to predict behavior in general loading situations. The tensile test is an important mate-
rial characterization test in which a slender member is pulled parallel to its axis. Metallic tensile test specimens
are usually circular section bars that are designed to achieve as nearly as possible a uniform state of axial normal
stress and strain in portion of the specimen called the gage length. The test is usually conducted in a universal
load frame and the with an attached load cell to measure the axial force applied to the specimen and an exten-
someter, or electrical resistance strain gage, to measure the elongation/strain of the gage length. The load frame
Thin-Walled Structures 19
Bars Subjected to Axial Loads

may be servo-hydraulically controlled using a feedback system to maintain specified load magnitude or elonga-
tion as a function of time. The loading rate is slow in static testing so that inertia effects are negligible, and usu-
ally the load is increased monotonically in time until fracture of the specimen. In addition, servo-hydraulic load
frames can apply cyclic loading at frequencies of around10Hz, which is used for fatigue testing. Electrical sig-
nals from the load cell and strain transducers are conditioned and converted to digital form. The digital response
data can be stored and then plotted using personal computers and laboratory software. The American Society of
Testing Methods (ASTM) publishes standards for testing materials. The standard governing the tensile test of
ductile metals is ASTM E8 – Standard Test Methods for Tension Testing of Metallic Materials.

The engineering stress σ z in the gage section is defined as the axial force N divided by the original cross-
sectional area A of the specimen. The engineering normal strain ε z is defined as the elongation ∆L of the mate-
rial in the gage length divided by the original gage length L, and for uniform extension in the gage section it is the
same as the point-wise definition given in eq. (2.4). Typical engineering stress-strain plots for a low carbon steel
and an aluminum alloy are shown in Fig. 2.4. For most engineering materials there is a linear relationship

L N
σ z = ----
A
N N

σz L ( 1 + εz ) σz

σu
σu
σ yu
fracture σy fracture
σp σ yl (ε f , σ f ) σp (ε f , σ f )

E
ε 0.2 = 0.002
1
εz εz
0 0
(a) some steels (b) aluminum alloy

Fig. 2.4 Typical stress-strain curves for steel and aluminum alloy from tensile tests

between the axial normal stress σ z and normal strain ε z near the origin of the plot as is shown in the figure. The
slope of the linear portion is a material property called the modulus of elasticity, and is denoted by E. The value
of the stress where the stress-strain plot deviates from a straight line is called the proportional limit, and is
denoted by σ p . The proportional limit is difficult to measure since under test conditions the deviation from a
straight line is subject to some judgement. The deformation of the material in the linear region is usually elastic.
Elastic deformation is defined as the deformation that disappears on removal of the load. The largest stress for
which elastic deformation occurs is called the elastic limit. The elastic limit of a material is also difficult to mea-
sure precisely since the specimen must be unloaded and reloaded to determine it. Just after the linear region of
the stress-strain plot for some low carbon steels, there is a relative maximum stress followed by a relative mini-
mum, and the deformation, or strain, begins to increase rapidly for small changes in the load. (To observe this
behavior, the elongation or displacement of the specimen is controlled and the load is measured.) The relative

20 Thin-Walled Structures
The tensile test

maximum value of the stress is called the upper yield point σ yu , and the stress at the subsequent relative mini-
mum value is called the lower yield point σ yl . Values of the upper yield point for metals are sensitive to the load-
ing rate and accidental bending stresses. Yielding is associated with plastic deformation of the material. Plastic
deformation is defined as deformation which is independent of time and remains on release of the load. The prin-
cipal physical mechanism causing plastic deformation in metals is slippage between planes of atoms in the crys-
tal grains of the material (Dowling, 1993).

Loading, unloading, and reloading of a metallic specimen beyond its yield stress is depicted in Fig. 2.5. The

σz

σy ε 0B = ε 0 A + ε AB

ε 0B = total strain
ε 0 A = plastic strain
ε AB = elastic strain

0 εz
A B
Fig. 2.5 Loading and unloading a tensile test specimen beyond the yield stress

unloading slope is nearly the same as the slope of the linear elastic portion, or E. The total strain at load is the
sum of the plastic strain (permanent portion upon removal of the load) and the elastic strain (recoverable por-
tion). Since plastic deformation of the material results in a change in size and shape of the structural component,
it is undesirable in design. Hence, yielding of the material is an important phenomena to quantify.

Under service loads, it is required that a structural component not be stressed beyond the yield stress. In
design, the condition of no yielding under service loads is called a limit state.

Aluminum alloys do not exhibit the abrupt yield point of low carbon
σz
steels; i.e., there is no stress just after the linear elastic portion where the σ yield
stress-strain curve has a zero slope. See Fig. 2.4b. Instead, following the lin-
ear elastic region, the slope of the stress-strain curve continuously decreases
until a relative maximum engineering stress occurs deep into the response
regime where plastic deformation is dominate. For such material behavior we
E E
define an offset yield stress. A straight line is drawn parallel to the linear elas-
1 1
tic portion of the stress-strain curve starting from a strain ε = ε 0.2 = 0.002 εz
on the strain axis. The stress at the intersection of this straight line with the 0 0.002 ε yield
stress-strain curve is defined to be the yield strength σ yield of the material. Fig. 2.6 0.2% offset
Note that the strain of 0.002, or 0.2% (percent strain is defined as 100ε ), is yield strength
plastic strain, since unloading the specimen from the point ( ε yield, σ yield ) on
the stress strain-curve would follow the straight dashed line in Fig. 2.6 and the strain of 0.002 would not be
recovered. However, a permanent strain of 0.2% is not considered detrimental for most structural components,
and the 0.2% offset yield strength has the advantage of being a precisely defined quantity. The offset yield stress
is generally the most satisfactory means of defining the yielding event for engineering materials.

Thin-Walled Structures 21
Bars Subjected to Axial Loads

After yielding the load may have to increase to cause further plastic defor-
mation of the specimen. (Also, the elastic deformation increases.) The increase
N N
in load required for further plastic deformation after yielding is called strain
hardening. The maximum tensile load carried by the specimen occurs during
strain hardening, and this maximum load divided by the original cross-sectional
necking deformation
area is called the ultimate tensile strength, or just the tensile strength, of the
material and is denoted by σ u . At the maximum load the deformation of most
metal specimens becomes localized in the form of an abrupt reduction in cross section along a small length in the
gage section. Prior to the maximum load the deformation is spatially uniform in the gage section. Plastic defor-
mation becomes concentrated in the reduced cross section after the maximum load. The non-uniform deforma-
tion is called necking, and its location in the gage section depends on imperfections in the particular specimen
which are random in nature. The load decreases after necking commences. The tensile test reaches its conclusion
when a small crack develops at the center of the neck and spreads outward to complete the fracture. The engi-
neering stress at fracture is denoted by σ f and the engineering strain at fracture is denoted by ε f . Note that the
engineering stress is defined with respect to the original cross-sectional area.

In addition to the axial elongation of the bar in the tensile test, most engi-
ε x or ε y σ
ε z = -----z neering materials exhibit a lateral contraction of the cross section. In axial
E compression, the cross section expands. In the linear elastic range of mate-
0 rial behavior, the ratio of the lateral normal strain to the axial normal strain
εz
1 is found to be a constant called Poisson’s ratio. Poisson’s ratio is denoted
ν by ν, and it is a dimensionless quantity. See Fig. 2.7. It has the same value
in tension and compression. The x and y axes of the Cartesian system
(x,y,z) lie in the cross section, so the lateral normal strains are denoted by
Fig. 2.7 Lateral normal ε x and ε y . That is, measurements of the diameter of the test specimen in
strains in the tension test
the tensile test reveal that the normal strain in the x- and y-directions are
ε x = ε y = – νε z

The negative sign is introduced to make the Poisson’s ratio positive for lateral contraction under uniaxial exten-
sion. It is implicit in the above expression that the Poisson’s ratio is the same along the x- and y-directions. If a
material is isotropic, then the Poisson’s ratio is necessarily the same in the x- and y-directions. An isotropic mate-
rial is one for which the material properties are independent of direction. In general, a material in which the prop-
erties vary with direction is called anisotropic. Materials whose properties vary from point to point are said to be
heterogeneous, and a material whose properties are uniform from point to point are said to be homogeneous.

For the linear elastic range of material behavior, the material law relating the axial stress σ z to the three nor-
mal strains is
ε x = – νσ z ⁄ E ε y = – νσ z ⁄ E εz = σz ⁄ E (uniaxial loading only) (2.5)

22 Thin-Walled Structures
Effect of temperature on strain

Typical values of the modulus of elasticity and Poisson’s ratio are listed in the table below for selected metals.

Room temperature modulus of elasticity and Poisson’s ratios for selected metals and metal
alloys (from Callister, 1997)

Modulus of elasticity
Poisson’s ratio,
Material 106 psi GPa dimensionless
Aluminum and aluminum alloys 10 69 0.33
Copper 16 110 0.34
Steels, plain carbon 30 207 0.30
Titanium and titanium alloys 15-16.8 104-116 0.34

2.3 Effect of temperature on strain


In the elastic region of material behavior changes in temperature can cause two effects:
• The elastic constants (E and ν) of the material can change with temperature.
• The temperature change causes the material to strain in the absence of stress.
For many structural materials a change in temperature of a few hundred degrees Fahrenheit does not result in
much change in the elastic constants. We will neglect the effect of temperature on the elastic constants in this text
and consider only the second effect. The strain caused by a temperature change in the absence of stress is called
thermal strain, and thermal strain is denoted by ε t . For an isotropic material, symmetry arguments show that a
rectangular parallelepiped of material deforms into rectangular parallelepiped of larger dimensions as the tem-
perature increases. That is, the thermal strain must be a pure expansion or contraction of the material with no dis-
tortion or shear. For the axially loaded bar, the important thermal strain component is the axial component ε zt .
Although the thermal strain is not exactly a linear function of the temperature change, for temperature changes of
a few hundred degrees Fahrenheit the actual thermal strain is nearly linear with the change in temperature. If the
temperature of the material is changed from T 0 to T , then the thermal strain is approximated by

ε zt = α ( T – T 0 ) (2.6)

in which α is called the linear coefficient of thermal expansion. The coefficient of thermal expansion is
0 0
abbreviated as CTE. The dimensional units of α are 1 ⁄ F or 1 ⁄ C , since strain is a dimensionless quantity. Val-
ues of the linear coefficient of thermal expansion for selected metals and metal alloys are listed in the table
below. The total strain in the bar is the sum of the strains due to mechanical stress and thermal strain. That is,
σ
ε z = -----z + α ( T – T 0 )
E





{

mechanical thermal (2.7)

Thin-Walled Structures 23
Bars Subjected to Axial Loads

Room temperature linear coefficients of thermal expansion (from Callister, 1997)

Coefficients of thermal expansion

0 0
Material 10 –6 ⁄ F 10 –6 ⁄ C
Aluminum 13.1 23.6
Aluminum alloys (cast) 10.0-13.6 18.0-24.5
Aluminum alloys (wrought) 10.8-13.4 19.5-24.2
Copper 9.4 17.0
Steel (low alloy) 6.2-7.1 11.1-12.8
Steel (plain carbon) 6.1-6.7 11.0-12.0
Titanium and titanium alloys 4.2-5.4 7.6-9.8

2.4 Bar reference axis


Consider the bar to be made from a homogeneous, isotropic material. Under the assumption of uniform exten-
sional deformation of the bar over its cross section, the distribution of the axial normal stress σ z is uniform over
the cross section.With respect to an arbitrary coordinate system x , y in the cross section, this uniform normal
stress distribution is statically equivalent to an axial force N and a moment with components M x and M y as
shown in Fig. 2.8.

y y
y

yc
Mx C

σ z dA
statically equivalent
N
N x
xc
My
x x

Fig. 2.8 Statical equivalence of the normal stress to the bar resultants.

The relationships of static equivalence are

N =
∫ ∫ σ dA
A
z Mx =
∫ ∫ yσ dA
A
z My =
∫ ∫ xσ dA
A
z (2.8)

24 Thin-Walled Structures
Bar reference axis

Substitute the material law for the axial normal stress from eq. (2.7) into these conditions of statical equiva-
lence to get

N =
∫ ∫ [E(ε
A
z – α( T – T 0 ) ) ]dA

Mx =
∫ ∫ y[E (ε
A
z – α( T – T 0 ) ) ]dA My =
∫ ∫ x[E (ε
A
z – α( T – T 0 ) ) ]dA

Since the bar is homogenous, the material properties are independent of the cross-sectional coordinates. Also, the
extensional strain ε z is independent of the cross-sectional coordinates, and the change in temperature is uniform
over the cross section, in pure extensional deformation. Hence, the previous equations become

N = EA [ ε z – α ( T – T 0 ) ]
(2.9)
M x = Q x [ E ( εz – α ( T – T 0 ) ) ] M y = Q y [ E ( εz – α ( T – T 0 ) ) ]

where Q x denotes the first moment of the cross-sectional area about the x -axis and Q y denotes the first area
moment about the y -axis. These first area moments are defined by the integrals

Qx =
∫ ∫ ydA
A
Qy =
∫ ∫ xdA
A
(2.10)

For pure extension of the bar with no bending, the distribution of the axial normal stress must be equivalent to an
axial force and no bending moments. The bending moments can be made to vanish by moving the resultants to a
special point in the cross section. The coordinates ( x c, y c ) of the point in the cross section where the moments
due to a uniform distribution of axial normal strain vanish is called the centroid. Static equivalence yields the
resultants at the centroid as

N = N
M x = M x – yc N = 0 (2.11)

M y = M y – xc N = 0
Solving the last two of eqs. (2.11) for the coordinates of the centroid, and using eqs. (2.9) we get

xc = Q y ⁄ A yc = Q x ⁄ A (2.12)

The location of the centroid depends on distribution of the area of the section about the x - y coordinate system.
The origin of the parallel Cartesian system x, y is at the centroid of the cross section as is shown in Fig. 2.8. The
two coordinate systems are related by
x = xc + x y = yc + y (2.13)

The first area moments have dimensional units of L3. Equations (2.12) establish the location of where the z-
axis passes through the cross section. That is, the reference axis of the bar, or the z-axis, passes through the cen-
troid of each cross section. If the bar is uniform along its length, then the reference axis is a straight line. Also,
eqs. (2.12) imply the first moments of the cross-sectional area about the x and y axis system vanish; i.e.,

Qx =
∫ ∫ ydA = 0
A
Qy =
∫ ∫ xdA = 0
A

Thin-Walled Structures 25
Bars Subjected to Axial Loads

From the first of eqs. (2.9) and the first of eqs. (2.11), the material law for the axial force is given by
t
N = EA ( ε z – ε zt ) = EAε z – N (2.14)

where the product the modulus of elasticity and the cross-sectional are (EA) is called the axial stiffness , or exten-
sinal stiffness, of the bar, and N t denotes the so-called thermal force. The thermal axial force is defined by
t
N = EAε zt = EAα ( T – T 0 ) (2.15)

Note that eq. (2.14) shows that the magnitude of the thermal force is equal to the magnitude of the actual internal
bar force only if the strain vanishes. For example, if the bar is constrained from changing its length
( q 1 = q 2 = 0 ), the distributed load intensity vanishes ( p z ( z ) = 0 ), and there is a change in temperature from
the stress free state, then N = – N t .

2.5 Linear elastic response


As stated at the beginning of this chapter, there are three basic steps to determine the static response of a structure
to specified loads: equilibrium conditions, strain-displacement relations, and material laws. For the uniaxially
loaded bar problem, we assume the it is made of a linear elastic, isotropic, homogeneous material, and summa-
rize these fundamental equations as follows.
dN
– equilibrium, eq. (2.1): ------- + p z ( z ) = 0 0<z<L
dz
dw
– strain-displacement, eq. (2.4): ε z = -------
dz
t
– material law, eq. (2.14): N = EAε z – N

Further lets assume the temperature is independent of the axial coordinate z as well as being uniform over
the cross section; i.e., the bar is subjected to a spatially uniform change in temperature. Substitute eq. (2.4) for
the strain in eq. (2.14), and in turn substitute the resulting equation into the equilibrium eq. (2.1) for the axial
force to get

d  dw
EA = – pz ( z ) 0<z<L (2.16)
dz dz 
where we used d N t ⁄ dz = 0 for a spatially uniform change in temperature. If the axial stiffness EA is indepen-
dent of z, that is, the bar is uniform along its length, then eq. (2.16) becomes
2
dw
EA 2 = – p z ( z ) 0<z<L (2.17)
dz

Equation (2.16) is the governing ordinary differential equation for the displacement function w(z), 0 ≤ z ≤ L,
and is of second order. For a second order differential equation we need two boundary conditions to determine
the two constants that arise in the general solution of eq. (2.16). These boundary conditions are given in eqs. (2.2)
and (2.3). Using the material law and the strain displacement equations these boundary conditions become

26 Thin-Walled Structures
Linear elastic response

dw
either w ( 0 ) = q 1 or EA ------- – N t = – Q 1 but not both (2.18)
dz z=0

dw
either w ( L ) = q 2 or EA ------- – N t = Q 2 but not both (2.19)
dz z=L

EXAMPLE 2.1 Axial bar with a specified uniform distributed load and specified end displacements

Consider a linear elastic, uniform bar subjected to a uniformly dis-


T – T0 ≠ 0
tributed load with intensity p z = p 0 , specified end displacements q1 p0
and q2, and a uniform change in temperature T - T0 from the stress free
state as shown in Fig. 2.9.Determine the axial displacement, normal
strain, and normal force of the bar. z, w L

Solution The governing differential equation (2.17) for the displace- w ( 0 ) = q1 w ( L ) = q2


ment reduces to Fig. 2.9 Bar of Example 2.1
2
dw p
2
= – ------0- 0<z<L
dz EA
Integrate this equation twice with respect to z to get
p z2
w ( z ) = – ------0-  ---- + c 1 z + c 2
EA  2 
where c1 and c2 are constants obtained in the indefinite integration. Substitute this expression into the boundary
condition at the left end and right end to get
p 02 p L2
– ------0-  ----- + c 1 0 + c 2 = q 1 – ------0-  ----- + c 1 L + c 2 = q 2
EA  2  EA  2 
Solve these last two equation for constants c1 and c2 to find

q2 – q1 p0 L
c 1 = ---------------
- + ----------- c2 = q1
L 2EA
Substitute these results for the constants of integration in the solution for the axial displacement to get

z z p0
w ( z ) = q 1  1 – --- + q 2 --- – ----------- ( z 2 – Lz ) (2.20)
 L L 2EA
The axial strain in the bar is defined by eq. (2.4), and substituting eq. (2.20) for w(z) we get

( q2 – q1 ) p0
- – ----------- ( 2z – L )
ε z = -------------------- (2.21)
L 2EA
Finally, substitute the strain given by eq. (2.21) into eq. (2.14) to find the distribution of the axial force as

EA L
N ( z ) = ------- ( q 2 – q 1 ) – p 0  z – --- – N t (2.22)
L  2

Thin-Walled Structures 27
Bars Subjected to Axial Loads

where N t = EAα ( T – T 0 ) . Equations (2.20) to (2.22) constitute the complete solution for the linear thermoelas-
tic response of the uniform bar subject to a uniform distributed load intensity p0, uniform temperature change
T – T 0 , and end displacement q1 and q2.

We can use eq. (2.22) and the relations N ( 0 ) = – Q 1 and N ( L ) = Q 2 (refer to eqs. (2.2) and (2.3)), to get

EA p0 L t
Q 1 = – ------- ( q 2 – q 1 ) – --------
-+N (2.23)
L 2

EA p0 L
- – Nt
Q 2 = ------- ( q 2 – q 1 ) – -------- (2.24)
L 2
Addition of these results gives
Q1 + Q2 + p0 L = 0

which is the condition of overall axial force balance for the bar.

EXAMPLE 2.2 A bar with fixed ends and subjected to an axial point force.

F Determine the axial displacement and normal force for the uniform bar
subjected to a point force F as shown in Fig. 2.10. Each end of the bar is
z fixed to a rigid support. There is no change in temperature from the stress
a b free state, and the specified data are, the axial stiffness EA, dimensions a, b,
L and L, and F.
Fig. 2.10 Point force acting on
a clamped bar Solution This is a statically indeterminate problem so all three steps (equi-
librium, strain-displacement, and a material law) are needed to solve for the
response of the bar. The governing differential equation (2.17) is valid in
the open intervals 0 < z < a and a < z < L, but is not valid at the location of the point force since a point force is
mathematically equivalent to an infinite value of the distributed load intensity pz acting over a zero length.
Hence, we will need to solve the governing differential equation separately in each open interval. That is
2
dw
EA = 0 0<z<a and a<z<L (2.25)
dz2
Two constants of integration will occur in the solution of this differential equation in each open interval for a total
of four unknown constants. Thus, we need four boundary conditions. Since the ends of the bar are fixed, we have
w(0) = 0 w( L) = 0 (2.26)

The displacement of the bar must be continuous at z = a, so

w( a– ) = w( a+ ) (2.27)

where a – implies the limit of z → a for values of z < a , and a + implies the limit of z → a for values of z > a .
The last boundary condition comes from axial force equilibrium at the point of application of the force.
dw dw
– N ( a– ) + N ( a+ ) + F = 0 or – EA ------- + EA ------- +F = 0 (2.28)
dz z→ a–
dz z → a+

28 Thin-Walled Structures
Linear elastic response

The solution to eqs. (2.25) is


w ( z ) = c1 z + c2 0<z<a
w ( z ) = c3 z + c4 a<z<L

where c1, c2, c3, and c4 are constants to be determined from the four boundary conditions. From the two bound-
ary conditions given by eqs. (2.26), we find c 2 = 0 and c 4 = – Lc 3 . Using these results in the continuity condi-
tion given by eq. (2.27) we get
c1 a = c3 ( a – L )

and in the force balance given by eq. (2.28) we get


– EAc 1 + EAc 3 + F = 0

These last two equations can be solved for the constants c1 and c3, and the solution is c 1 = ( bF ) ⁄ ( EAL )
c 3 = – ( aF ) ⁄ ( EAL ) , where we used b = L - a. Hence the displacement function is

bF
 ---------- -z 0≤z≤a
 EAL
w(z) =  (2.29)
aF
 ----------
-(L – z) a≤z≤L
 EAL

The normal force in each interval is obtained from N = EA ( dw ⁄ dz ) , so we have

 --b- F 0≤z≤a
 L
N (z) =  (2.30)
 – --a- F a≤z≤L
 L
The displacement and normal force are plotted with respect to the axial coordinate in Fig. 2.11. Note that the dis-

w
abF
-----------
EAL

z
0
a L
N Fig. 2.11 Displacement and normal force
b distributions for the clamped bar
--- F subjected to a point force.
L

z
0 a L
a
– --- F
L

placement is continuous in z, and that the normal force has a jump discontinuity at z = a. The value of the discon-

Thin-Walled Structures 29
Bars Subjected to Axial Loads

a b
tinuity in the normal force is N ( a + ) – N ( a – ) =  – --- F –  --- F = – F , which satisfies the first of eqs. (2.28).
 L  L 

2.6 Work and energy methods


In particle mechanics the incremental work of a force acting on a particle is the product of the force and the com-
ponent of the incremental displacement of the particle in the direction of the force. If we have a system of parti-
cles, then the incremental work of the forces acting on their respective particles can be defined as the sum the
incremental work of the forces acting on their respective particles. For the problems of solid mechanics in which
we assume the body is made of a continuum of particles, each infinitesimal particle, or material point, is identi-
fied by its coordinates in the undeformed or reference state of the body. The displacement components of the par-
ticle are continuous, single valued functions of its coordinates in the reference state so that the body in the
deformed configuration remains a continuum; i.e., the deformation is such that no holes or overlapping of mate-
rial points occur in the deformed configuration.

Hence, to define incremental work in continuum structures we need to account for small variations in the
displacement variables which are themselves continuous functions of the coordinates. In the mathematical sense,
a bar is called a one-dimensional structure, since the dependent variables describing the displacement, strain, and
the internal normal force are functions of a single independent variable z, 0 ≤ z ≤ L, where L is the length of the
bar. The coordinate z locates a material point in the undeformed bar.

2.6.1 Concept of virtual displacement


Assume the bar is in equilibrium under external loads Q 1, Q 2, p z ( z ) , and a temperature change ( T – T 0 ) . The
displacement w ( z ) , the strain ε z ( z ) , and the axial force N ( z ) are those of this equilibrium state. Now give the
end displacements an infinitesimal change in magnitude, denoted by δq 1 and δq 2 . These infinitesimal changes in
the displacements are called virtual displacements. The word virtual means in essence and not in fact. That is, the
virtual displacements are possible displacements, but do not necessarily coincide with the actual displacements
of the bar. Since the bar is assumed to be a continuum, the displacements of the material points along the bar will
also change, and the change in axial displacement is a function denoted by δw ( z ) . The virtual displacement
function δw ( z ) must be a continuous, single valued function of z for the bar not to fracture, and its first deriva-
tive, or virtual strain δε z ( z ) , must be piecewise continuous so that its integral is computable. Also, the virtual
displacement function must satisfy the end conditions δw ( 0 ) = δq 1 and δw ( L ) = δq 2 .

A displacement function is said to be kinematically admissible if it satisfies continuity conditions and speci-
fied boundary conditions on the displacement, if any. A kinematically admissible function may not satisfy the
governing boundary value problem for the response of the bar. However, within the set of all kinematically
admissible displacement functions one will satisfy the governing differential equation of equilibrium, eq. (2.16),
and the specified boundary conditions on the axial force, if any, which are given in eqs. (2.2) and (2.3). That is,
the exact displacement function of the equilibrium state is in the set of kinematically admissible displacement
functions.

The virtual displacement function is a new infinitesimal function, and to emphasize this fact it is also repre-
sented as

30 Thin-Walled Structures
Work and energy methods

δw ( z ) = εη ( z ) (2.31)

where ε is an infinitesimal scalar parameter and the function η ( z ) is an arbitrary, but kinematically admissible,
function. The varied displacement function is defined as
w̃ ( z ) = w ( z ) + εη ( z ) (2.32)

and the varied displacement equals the displacement


of the equilibrium state of the deformed bar as y
ε → 0 . This displacement situation is depicted in P P* P̃
Fig. 2.12. The particle at position P having the coor-
0 z w(z) δw ( z )
dinate z in the undeformed bar is displaced to posi-
tion P * having the coordinate z + w ( z ) in the w̃ ( z )
equilibrium configuration of the deformed bar.
Imposition of the infinitesimal virtual displacement Fig. 2.12 Displacements of a particle originally a
point P.
δw ( z ) from the equilibrium state of the deformed
bar causes the particle to move to the position P̃
having the coordinate z + w̃ ( z ) . The point of all this is that the virtual displacement is a change in the displace-
ment following a particle, and the particle is identified by its position z in the undeformed bar. That is, from the
mathematical viewpoint, the virtual displacement is an infinitesimal change in the displacement holding the inde-
pendent variable z fixed in value. In mechanics, incremental work is defined as product of the force in the direc-
tion of the displacement times an infinitesimal change in the displacement of the particle on which the force acts.
Hence, for a continuum, like our bar example, we have to introduce the concept of an infinitesimal change in dis-
placement holding the independent variable fixed in value in order to compute the incremental work of a force
acting on a particle.

Although the virtual displacement δw ( z ) is a new displacement function, the use of the prefix greek letter δ
is meant to convey a change in the function w ( z ) in analogy to the use of letter “d” used in calculus. However,
the greek letter "δ" means an infinitesimal change in the displacement for a fixed value of z, where the symbol
"d" in calculus means the change in a function with respect to a change in its independent variable. The differ-
ence between the variation of a function and the differential of a function is depicted in Fig. 2.13. Since in the

w̃ ( z )

w(z)
δw ( z )
q 1 + δq 1
dw
dz q 2 + δq 2
q2
q1

0 z z + dz L
z

Fig. 2.13 The variation of a function and the differential of a function

ordinary calculus the differential of the function w(z) is dw = w ( z + dz ) – w ( z ) , the physical interpretation of
dw is the difference in the displacements of two particles, one particle originally at z + dz and the second one

Thin-Walled Structures 31
Bars Subjected to Axial Loads

originally at z in the undeformed body. The physical interpretation of the infinitesimal function δw(z) is change in
the displacement of a single particle originally a z in the undeformed body. The "δ" is considered an operator in a
sense similar to the differential operator "d" in calculus. The "δ" operator means the variation in a function and
not the differential of a function with respect to a change in its independent variable. The distinction between the
variation of a function and the differential of a function is essential in formulating the concept of work in a con-
tinuous system of particles, and the mathematics dealing with the variation of functions is called the calculus of
variations. That is, the calculus of variations is a mathematical tool by which work and energy methods are
applied to systems with infinitely many degrees of freedom. A more complete description of the calculus of vari-
ations applied to structural problems is presented in the text by Langhaar (1962), and Courant and Hilbert (1953)
present the essential mathematical framework. However, the reader may be interested in the recent text book by
Dym (1997) which presents energy methods for continnum models with limited use of the calculus of variations.

We will encounter the variation of the derivative of a function and the variation of the definite integral of a
function in the discussions that follow. These are defined as (refer to Fig. 2.13)
L L L
dw dw̃ dw  
δ  ------- ≡ ------- – -------
 dz  dz dz ∫  ∫
δ  w dz ≡ w̃ dz – w dz
∫ (2.33)
0 0 0

Substitute the for the total displacement w̃ ( z ) the relation w̃ ( z ) = w ( z ) + δw ( z ) into these definitions to get
L L
dw d  
δ  ------- = ----- ( δw )
 dz  dz  ∫
δ  w dz =
 ∫ δw dz (2.34)
0 0

These relations show that the variational operator and the derivative, and the variational operator and the integral
are interchangeable.

2.6.2 Principle virtual work


The incremental work of the external forces acting through the virtual displacements is denoted by δW ext , and
this quantity is called the external virtual work. Internal and external forces are held constant during the (infini-
tesimal) virtual displacements, so that from definition of incremental work in mechanics, we write the external
virtual work as
L


δW ext = Q 1 δq 1 + Q 2 δq 2 + p z ( z )δw ( z ) dz (2.35)
0

Since it is assumed that the bar is in an equilibrium state prior to the application of the virtual displacements, we
have from the boundary conditions on the axial force, see eqs. (2.2) and (2.3), that
L
d
Q 1 δq 1 + Q 2 δq 2 = – N ( 0 )δw ( 0 ) + N ( L )δw ( L ) =
∫ 0 z
d
( Nδw ) dz

Hence, eq. (2.35) can be written as


L
d
δW ext =
∫ dz
( Nδw ) + p z ( z )δw ( z ) dz
0

Differentiating the first term in the integral on the right hand side of this equation, and using the interchange of

32 Thin-Walled Structures
Work and energy methods

the derivative and variation, as shown by the first of eqs. (2.34), we get
L
 dN + p  δw + Nδ  dw dz
δW ext =
∫ dz z dz 
0

The first term in the integral on the right-hand side of this equation vanishes via the equilibrium condition given
in eq. (2.1), and the second term contains the variation in the strain as defined in eq. (2.4) Hence,
L

δW ext =
∫ Nδε dz
z (2.36)
0

Since the integral on the right-hand side of this equation is determined by the internal bar force and the incremen-
tal strain, it is defined as the internal virtual work δW int . The internal virtual work is defined as


δW int ≡ Nδε z dz (2.37)
0

What the manipulations from eq. (2.35) to (2.37) show is that for a bar in equilibrium, the external virtual work is
equal to the internal virtual work for a kinematically admissible variation in the displacement function w ( z ) .
This proves the necessary condition that if the body is in equilibrium then the external virtual work equals the
internal virtual work for a kinematically admissible variation in the displacement. In problem solving we assume
that equating the external to the internal virtual work for every kinematically admissible displacement is suffi-
cient for equilibrium of the body. We state this principle as follows.

Principle of Virtual Work (PVW)


If the external virtual work (δWext) equals the internal virtual work (δWint) for every
kinematically admissible variation in the displacement, then the body is in equilibrium.

For the axially loaded bar, the principle of virtual work states that the bar is in equilibrium if
δW ext = δW int for every kinematically admissible δw ( z ) (2.38)

where the external virtual work is given by eq. (2.35) and the internal virtual work is given by eq. (2.37). The
principle of virtual work is an integral form of the equations of equilibrium for the bar. The mathematical condi-
tions of kinematic admissibility of the virtual displacement function δw(z) are that it is continuous, its first deriv-
ative is piecewise continuous, and that it vanishes at points where the displacement w(z) is prescribed. That is, the
virtual displacement, or variation in the displacement, must be a possible displacement of the bar.

If no restrictions are placed on the virtual displacement function δw(z) other than kinematic admissibility,
then the PVW leads to the exact differential equation of equilibrium for the bar, eq. (2.1), and the associated
boundary conditions, eqs. (2.2) and (2.3). Moreover, it will be shown in Example 2.5 that the PVW for a linear
elastic bar gives the exact solution if the assumed displacement function has the flexibility to represent the exact
displacement. Hence, it may seem that the PVW yields nothing new. However, in complex structures, an exact,
closed form, mathematical solution is rarely known. We must resort to approximate solutions, and the PVW is a
powerful technique to develop approximate solutions based on assumed displacement functions containing
unknown parameters. In fact, the PVW is the basis of many finite element procedures used in structural mechan-
ics. The use of the PVW to obtain an approximate solution is illustrated in the following example.

Thin-Walled Structures 33
Bars Subjected to Axial Loads

EXAMPLE 2.3 Approximating the response of a bar using PVW.

Consider the bar of Example 2.2 again, and take length a equal to b so that a = b = L/2; that is, the point force F
z
acts at the center of the bar. Assume the axial displacement of the bar of the form w ( z ) = q 1 sin  π --- , in which
 L
the unknown parameter q1 represents the axial displacement at z = L/2.
a) Show that the assumed displacement function is kinematically admissible.
b) Use the principle of virtual work to determine the equilibrium equation associated with the assumed dis-
placement function.
c) Assume the material is linear elastic and solve for the unknown parameter q1.
d) Calculate the percentage error in the maximum displacement and maximum stress with respect to the
exact solution.

Solution (a) To be kinematically admissible the displacement function must be continuous in the interval 0 ≤ z ≤
L, and satisfy prescribed boundary conditions. The sine function is continuous in the interval, and its derivative
with respect to z, or the strain, is also continuous. The prescribed boundary conditions on the axial displacement
are w(0) = 0 and w(L) = 0. Since the function sin ( πz ⁄ L ) is zero at z = 0 and z = L, the prescribed displacement
boundary conditions are satisfied. Thus, the assumed displacement is kinematically admissible.

(b) The variation in the assumed displacement, or virtual displacement, is δw ( z ) = δq 1 sin ( πz ⁄ L ) , where
δq1 is the variation in the unknown parameter q1. The only external force acting on the bar that performs work is
the point force F, so the external virtual work is
δW ext = Fδw ( L ⁄ 2 ) = Fδq 1

The internal virtual work is given by eq. (2.37), which requires the virtual strain. The virtual strain is merely the
d π z
derivative with respect to z of the virtual displacement function; i.e., δε z = ----- ( δw ) = δq 1 --- cos  π --- . Hence
dz L  L
the internal virtual work is
L L
π z
Nδε z dz = δq 1 --- N cos  π --- dz
δW int =
∫ L ∫  L
0 0

Now equate the external virtual work to the internal virtual work, and rearrange the result to get
L
π z
--- N cos  π --- dz – F δq 1 = 0
L ∫  L
∀δq 1
0

The PVW gives the equilibrium equation as


L
π z
--- N cos  π --- dz – F = 0
L∫  L
(2.39)
0

Note that at this point we cannot carry out the integration in this last equation since we do not know how the axial
force depends on the displacement. The relationship that is missing is the material law; i.e., the PVW gives the

34 Thin-Walled Structures
Work and energy methods

equilibrium condition related to the kinematically admissible displacement independent of the material law.

(c) For a linear elastic material and no change in temperature from the stress free state, the material law is
N = EAε z and the strain-displacement relation is ε z = dw ⁄ dz . Substitute the approximate displacement into
the strain-displacement relation, and then substitute this result into the material law to get

π z
N = EAq 1 --- cos  π ---
L  L

Substitute this expression for the axial force into the equilibrium condition, eq. (2.39), and rearrange it to get
L
2
π z
EAq 1  ---
 L ∫ cos  π --L- dz = F
2

Performing the integration on the left-hand side of this expression we have


L
π 2 z L z
EAq 1  --- --- + ------ sin  2π --- = F
 L 2 4π  L
0

Evaluating the limits in this equation we find

π2 2 FL
EAq 1 ------ = F solving for q 1 q 1 = ----2-  -------
2L π  EA
We have determined the unknown parameter q1 in the assumed displacement function terms of the applied force,
length of the bar, and the extensional stiffness of the bar. Finally, the displacement approximation is

2 FL z
w ( z ) = ----2-  ------- sin  π --- 0≤z≤L (2.40)
π  EA  L

The normal force is obtained from N = EA ( dw ⁄ dz ) , so it is approximated as

2 z
N = --- F cos  π --- (2.41)
π  L

(d) For a = b = L/2, the maximum displacement of the exact solution, eq. (2.29), occurs at the center of the
bar and is w exact = ( FL ) ⁄ ( 4EA ) . The maximum displacement in the approximate solution, eq. (2.40), also
occurs at the center of the bar and is w approx = ( 2FL ) ⁄ ( π 2 EA ) . The percentage error in the approximate solu-
tion is given by
w exact – w approx 1 ⁄ 4 – ( 2 ⁄ π2 )
----------------------------------- × 100 = --------------------------------- × 100 = 18.9%
w exact 1⁄4

So the approximate displacement at the center of the bar is 18.9% less than the exact value. The fact that the
approximate displacement is less than the exact displacement means our assumed displacement function is too
restrictive and results in a stiffer response than what the exact solution gives.

Thin-Walled Structures 35
Bars Subjected to Axial Loads

The maximum value of the normal force in the exact solution, eq. (2.30), occurs in the left half of the bar and
is N exact = F ⁄ 2 . The maximum value of the normal force in the approximate solution, eq. (2.41), occurs at z =
0 and is N approx = ( 2F ) ⁄ π . The percentage error in the approximate solution for the normal force is

N exact – N approx 1⁄2–2⁄π


-----------------------------------
- × 100 = -------------------------- × 100 = – 27.3%
N exact 1⁄2

The maximum normal force in the approximate solution exceeds the exact value by 27.3%.

We can improve our approximate solution in Example 2.3 by assuming

z z
w ( z ) = q 1 sin  π --- + q 3 sin  3π --- 0≤z≤L (2.42)
 L  L

in which parameters q1 and q3 are unknown and are independent. Note that this approximate displacement func-
tion is continuous and satisfies the prescribed displacement boundary conditions w(0) = 0 and w(L) = 0; that is,
eq. (2.42) is a kinematically admissible displacement function for the bar with fixed ends. In fact, each of the
individual sine functions in this equation are kinematically admissible, so there sum is kinematically admissible.
Also note that axial displacement at the center of the bar is w(L/2) = q1 - q3, and the axial displacements at w(L/
3) = w(2L/3) = q1. We are approximating the axial displacement by the superposition of two sine functions of dif-
ferent frequencies as is depicted in Fig. 2.14. The principle of virtual work will lead to two equilibrium condi-
tions: one condition determined by the independent virtual displacement δq1 and the second condition
determined from the independent virtual displacement δq3. Explicit determination of these equilibrium condi-
tions is one of the problems in Section 2.13.

1
0.75
z
0.5 sin  π ---
0.25  L
zêL
0.2 0.4 0.6 0.8 1
-0.25
-0.5
-0.75
Fig. 2.14 Sine functions of a two term -1
approximation to the axial
displacement z
sin  3π ---
1  L

0.5

zêL
0.2 0.4 0.6 0.8 1
-0.5

-1

2.6.3 Strain energy density


Let’s denote the integrand of the internal virtual work expression in eq. (2.37) as the incremental quantity δU 0 .
That is,

36 Thin-Walled Structures
Work and energy methods

dw
δU 0 = Nδε z = Nδ  ------- (2.43)
 dz 

For a linear elastic material, we can eliminate the axial force in this equation using the material law in eq. (2.14)
to get

dw dw
δU 0 = EA ------- – ε zt δ  ------- (2.44)
dz  dz 
where we used eq. (2.4) to eliminate the strain in terms of the displacement derivative. The form of this incre-
mental quantity suggests there is a function of the derivative of the displacement, or strain, such that its variation
is given by eq. (2.44). If such a function exists, it is actually a function of a function (the displacement deriva-
tive), and it is denoted by U 0 [ w′ ( z ) ] , where the prime means ordinary derivative; i.e., w′ = dw ⁄ dz . Functions
that depend on a functions are called functionals. But what is meant by the variation of the functional
U 0 [ w′ ( z ) ] ? To answer this question we consider evaluating the functional for the varied function w̃ ( z ) . Since
the varied function contains the scalar parameter ε , and w̃ ( z ) → w ( z ) as ε → 0 (refer to eq. (2.32)), the change
in U 0 with respect to the function w ( z ) can be examined by considering U 0 [ w̃ ( z ) ] as an ordinary function of
the parameter ε . Thus, the variation of the functional is defined as

 U 0 [ w′ + εη′ ] – U o [ w′ ]  d 
δU 0 ≡  lim -------------------------------------------------------- ε =  d ε U 0 [ w̃′ ] ε (2.45)
 ε → 0 ε   ε = 0

Regard the quantity w̃′ as a simple variable in the functional U 0 , and by using the chain rule for differentiation,
we write the previous equation as

∂U 0 ∂ ∂U 0
δU 0 = ( w̃′ ) ε = η′ε
∂ w̃′ ∂ ε ε=0
∂ w′

since w̃′ → w′ as ε → 0 , and where ∂w̃′ ⁄ ∂ε = η′ . But eq. (2.31) gives δw′ = εη′ . So the variation of func-
tional U 0 is

∂U 0
δU 0 = δw′ (2.46)
∂ w′
Equate this definition to eq. (2.44) and we have

∂U 0
δU 0 = δw′ = EA [ w′ – ε zt ]δw′
∂ w′

which must old for every kinematically admissible δw′ . Consequently,

∂U 0
= EA ( w′ – ε zt )
∂ w′

Integrate this relation this relation with respect to w′ , treating w′ as a simple variable as in the ordinary calculus,
to find
1
U 0 [ w′ ( z ) ] = --- EA ( w′ ) 2 – EAε zt w′ (2.47)
2

Thin-Walled Structures 37
Bars Subjected to Axial Loads

in which the constant of integration is zero since U 0 is taken be zero if the strain is zero. We have also assumed
in this integration that the thermal strain is independent of the strain. The functional U 0 [ w′ ( z ) ] or U 0 [ ε z ] , given
by eq. (2.47) is called the strain energy density. The dimensional units of the strain energy density in this case are
(F-L)/L. Notice that the partial derivative of the strain energy density with respect to the strain gives the axial
force; i.e.,

∂U 0
N = = EA ( ε z – ε zt ) (2.48)
∂ εz

The fact that N = ∂U 0 ⁄ ∂ε z is an important property of the strain energy density. In fact, an elastic material can
be defined as one for which a strain energy density function exists.

Return to the internal virtual work expression given in eq. (2.37) and identify the integrand on the right-hand
side as the variation in the strain energy density as given in eq. (2.47). We write the internal virtual work as
L

δW int =
∫ δU dz = δU
0 (2.49)
0

where we interchanged the variational operator and the integral per the second of eqs. (2.34), and then defined
the strain energy as
L L
1
∫ U dz = ∫ EA  --2- ε – ε z ε z dz
2 t
U = 0 z (2.50)

0 0

The strain energy represents the energy stored in the bar due to elastic deformation. It has the dimensional units
of F-L. Note that the temperature, which is contained in the thermal strain, enters the strain energy as a parame-
ter.

EXAMPLE 2.4 Strain energy of a bar with fixed ends and subjected to an axial point force.

Consider the bar with fixed ends subjected to the axial point force as discussed in Example 2.2 on page 28 .
(a) Determine the strain energy in the bar.
(b) Determine the strain energy in the approximate solution of the bar given in Example 2.3.
(c) Determine the percentage error in the approximate strain energy with respect to its exact value.

Solution (a) The displacement function of the exact solution is given by eq. (2.29). Differentiating this expres-
sion with respect to z determines the strain, and this strain is

bF
 ----------- 0≤z≤a
 EAL
εz = 
aF
 – ----------
- a≤z≤L
 EAL

Substitute this result into the expression for the strain energy given by eq. (2.50), and note that in this example
the thermal strain vanishes, to find

38 Thin-Walled Structures
Work and energy methods

a L
1 bF 2 1 aF 2
U = --- EA  ----------- dz + --- EA  – ----------- dz
2 ∫ EAL  2  EAL∫
0 a

Since the bar is uniform, the extensional stiffness EA is uniform in z. So, after integration, the strain energy
becomes
ab 2 F 2 a 2 bF 2 abF 2
U = ----------------2- + ----------------2- = --------------
2EAL 2EAL 2EAL

where we used the fact that b = L – a . For the case of a = b = L/2, the strain energy in the exact solution
reduces to
LF 2
U = ----------- (a = b = L ⁄ 2)
8EA

(b) The approximate solution for the displacement is given by eq. (2.40), which is repeated below.

2 FL z
w ( z ) = ----2-  ------- sin  π --- 0≤z≤L
π  EA  L
Differentiating this expression to find the strain we get

2 F z
ε z = ---  ------- cos  π --- 0≤z≤L
π  EA  L

Substitute this result into the expression for the strain energy given by eq. (2.50), and again note there is no ther-
mal strain, to find
L
2
1 2 F z
U = --- EA ---  ------- cos  π ---
2 ∫
π  EA  L
dz
0

Evaluating this result we get


L L
2F 2 z 2F 2 z L z 1 F2L
- --- + ------ sin  2π ---
- cos2  π --- dz = ------------ = ----2-  ----------
U = ------------
π EA2 ∫  L π 2 EA 2 4π  L
0
π  EA 
0

LF 2
(c) The exact strain energy is determined in part (a) as U exact = ----------- , and the strain energy in the approxi-
8EA
1 F2L
mate solution is determined in part (b) as U approx = ----2-  ---------- . The percentage error in the approximate solution
π  EA 
with respect to the exact solution is given by
U exact – U approx 1 ⁄ 8 – 1 ⁄ π2
------------------------------------ × 100 = ---------------------------- × 100 = 18.9%
U exact 1⁄8

The next example illustrates that the PVW as applied to a linear elastic bar gives the exact solution, if the
assumed displacement function can represent the exact displacement. Of course, in solving for the response of
more complex structures the exact form of the displacement function is not known. It is reassuring, however, that

Thin-Walled Structures 39
Bars Subjected to Axial Loads

if the exact displacement is substituted into the PVW, the exact equilibrium equations are obtained from the prin-
ciple.

EXAMPLE 2.5 An elastic bar subjected to two forces and a thermal load

Show that virtual work eq. (2.38) leads to equilibrium equations for the displacements of the linear elastic bar of
Example 2.1 on page 27. For simplicity, take the distributed load p z ( z ) = 0 for all z. The bar is subjected to
specified end forces Q 1 and Q 2 , and a spatially uniform temperature change ( T – T 0 ) . The end displacements
q 1 and q 2 are unknown, since the end forces are prescribed. Take the kinematically admissible displacement to
be

z z
w ( z ) = q 1  1 – --- + q 2 --- (2.51)
 L L
Parameters q1 and q2 are unknown and independent. Parameters q1 and q2 physically represent the end displac-
ments of the bar since w ( 0 ) = q 1 and w ( L ) = q 2 .

Solution Since end displacements are not known the variation in the displacement function, or virtual displace-
ment, is

z z
δw ( z ) = δq 1  1 – --- + δq 2  ---
 L  L

and the variation in the derivative of the displacement, or virtual strain, is


( δq 2 – δq 1 )
δw′ = ---------------------------
- (2.52)
L
From eq. (2.49) the internal virtual work is given by the variation of the strain energy. That is,
L

δW int =
∫ δU dz 0
0

From eq. (2.47) the strain energy density is


1
U 0 [ w′ ( z ) ] = --- EA ( w′ ) 2 – EAε zt w′ ,
2
and its variation, where the variation of a functional is defined by eq. (2.46), is

∂U 0
δU 0 = δw′ = EA ( w′ – ε zt )δw′
∂ w′
Obtain the derivative of the displacement from eq. (2.51) and substitute it, and eq. (2.52) for the variation of the
derivative, into the previous equation to get
( q2 – q1 ) ( δq 2 – δq 1 )
- – ε zt ----------------------------
δU 0 = EA --------------------
L L
Substitute this variation of the strain energy density into the internal virtual work to get

40 Thin-Walled Structures
Castigliano’s first theorem

( q2 – q1 )
- – ε zt ( δq 2 – δq 1 )
δW int = EA --------------------
L
The external virtual work for no distributed loading, see eq. (2.35), becomes
δW ext = Q 1 δq 1 + Q 2 δq 2

The principle of virtual work, eq. (2.38), is


( q2 – q1 )
- – ε zt ( δq 2 – δq 1 )
Q 1 δq 1 + Q 2 δq 2 = EA -------------------- ∀δq 1 and δq 2
L
Rearrange this result into the form

EA t EA t
Q 1 + ------- ( q 2 – q 1 ) – N δq 1 + Q 2 – ------- ( q 2 – q 1 ) + N δq 2 = 0 ∀δq 1 and δq 2 (2.53)
L L

where the thermal force N t = EAε zt . Since the virtual displacements are independent, we can take δq 1 ≠ 0 and
δq 2 = 0 , so that eq. (2.53) leads to

EA t
Q 1 + ------- ( q 2 – q 1 ) – N δq 1 = 0 ∀δq 1 ≠ 0
L
which in turn gives
EA t
Q 1 = – ------- ( q 2 – q 1 ) + N (2.54)
L
Now take δq 1 = 0 and δq 2 ≠ 0 and then eq. (2.53) leads to

EA t
Q 2 – ------- ( q 2 – q 1 ) + N δq 2 = 0 ∀δq 2 ≠ 0
L
which in turn gives
EA t
Q 2 = ------- ( q 2 – q 1 ) – N (2.55)
L
Equations (2.54) and (2.55) agree with the exact result as derived in Section 2.5 and specified by eqs. (2.23) and
(2.24) if the distributed load intensity p 0 is set to zero. We get the exact result because the given displacement
function, eq. (2.51), is the exact displacement for this case, as can be verified from eq. (2.20).

2.7 Castigliano’s first theorem


Example 2.5 showed that the exact relationship between the end forces, the corresponding end displacements,
and the thermal force, can be determine from virtual work for a uniform bar with no distributed load acting on it
if the displacement function specified is exact. That is, if the displacement function specified corresponds to the
displacement of the equilibrium state of the elastic bar, then virtual work gives the exact result for the point
forces acting on the bar. In the example, the displacement function given by eq. (2.51) is continuous in z, can rep-
resent a state of uniform strain, and the end displacements appear as parameters in the function. Of course, the
displacement function corresponding to the equilibrium state is not known without solving the governing differ-

Thin-Walled Structures 41
Bars Subjected to Axial Loads

ential equation (2.16) subject to boundary conditions (2.2) and (2.3). But the procedure followed in this example
can be extended to structures built-up of bars subjected to end forces and a uniform temperature change from the
stress free state, since the displacement of each bar is of the form of eq. (2.51). These observations motivate Cas-
tigliano’s first theorem.

The virtual work eq. (2.38) is extended to an elastic structural system subjected to many external point forces
Q n , n = 1, 2, …, N , but no distributed loads. The displacements at the points of application of the forces, in the
direction of the forces, are denoted q n , n = 1, 2, …, N , and are assumed to be independent. Forces Q n and dis-
placements q n are said to be corresponding forces and displacements, since they are defined to act at the same
point on the body’s surface and in the same direction. We consider the external point forces as specified, and the
corresponding displacements as unknown. The virtual work eq. (2.38) becomes
N

∑ Q δq n n = δU (2.56)
n=1

In principle, it is possible to write the displacement functions for the structural system such that the independent
q n ‘s appear as parameters in the functions. Then the strains in terms of the q n ‘s are determined from the strain-
displacement equations similar to eq. (2.4). Finally, the strains are substituted into the strain energy of the system
and integration over the volume of the system is carried out explicitly. Hence, the strain energy of the structural
system becomes a function of the q n ‘s; i.e.,

U = U ( q 1, q 2, …, q N )

Hence, the variation in the strain energy is


N
∂U
δU = ∑ ∂ q δq n
n (2.57)
n=1

and the virtual work, eq. (2.56), becomes


N
∂U 
∑  Q n – δq = 0
∂ q n n
∀δq n n = 1, 2, …, N (2.58)
n=1

Since the virtual displacements are independent, we conclude from eq. (2.58) that
∂U
Q n = -------- n = 1, 2, …, N (2.59)
∂q n
This is the equation of Castigliano’s first theorem, which is also called Castigliano’s theorem part I. Note the sim-
ilarity of eq. (2.59) to the property of the strain energy density that the internal bar force is equal to the derivative
of the strain energy density with respect to the strain; or, N = ∂U 0 ⁄ ∂ε z . This theorem is stated for the general
case as follows.

42 Thin-Walled Structures
Castigliano’s first theorem

Castigliano’s first theorem


If the strain energy of an elastic structure is expressed in terms of the independent
displacement components q n , n = 1, 2, …, N , in the direction of the prescribed point
forces Q 1, Q 2, …Q N , and there are no distributed loads, then the first partial derivative
of the strain energy with respect to the displacement q n is equal to the corresponding
force Q n or

∂U
Q n = -------- n = 1, 2, …, N
∂q n

To illustrate the application of Castigliano’s first theorem consider the two-force member in Example 2.5 on
page 40 again. Substitute the displacement function given in eq. (2.51) into the strain-displacement relation of eq.
(2.4) to determine the strain as
( q2 – q1 )
ε z = --------------------
-
L
Substitute this expression for the strain into the strain energy given by eq. (2.50) to get
L
1 q2 – q1 2 q2 – q1
EA ---  ---------------- – ε zt  ---------------- dz
U =
∫ 2 L   L 
0

Perform the integral over the length of the bar to find


1 EA t
U = --- ------- ( q 2 – q 1 ) 2 – N ( q 2 – q 1 )
2 L

where the thermal force is defined as N t = EAε zt . From eq. (2.59) we find

∂U EA t
Q1 = = ------- ( q 2 – q 1 ) ( – 1 ) + N
∂ q1 L

∂U EA t
Q2 = = ------- ( q 2 – q 1 ) – N
∂ q2 L

which coincide with the exact solution given in Example 2.1 on page 27 in eqs. (2.23) and (2.24) when the dis-
tributed load intensity set to zero.

EXAMPLE 2.6 Response of a stepped bar by Castigliano’s first theorem

Determine the relationship for between the external forces and the corresponding displacements for equilibrium
of the axially loaded bar containing a step change in cross section as shown below. There is no change in temper-
ature from the stress-free state. Use Castigliano’s first theorem.

Thin-Walled Structures 43
Bars Subjected to Axial Loads

( EA ) 1 ( EA ) 2

q 1, Q 1
z, w q 2, Q 2

L1 L2

Fig. 2.15 Stepped bar

Solution The deformation of the bar is a state of uniform strain in each section. Using the displacement function
corresponding to uniform strain given by eq. (2.51) in Example 2.5 on page 40 as a guide, we write the axial dis-
placement functions for each section of the bar as
z
w ( z ) = q 1 ----- 0 ≤ z ≤ L1
L1

z – L1 z – L1
w ( z ) = q 1 1 –  -------------- + q 2  -------------- L1 ≤ z ≤ L1 + L2
 L2   L2 

Note that these displacement functions are continuous, w(0) = 0, w(L1) = q1 for both functions, and w(L1 + L2) =
q2. From the strain-displacement relation, eq. (2.4), these displacements yield the strains in each section as

q q2 – q1
( ε z ) 1 = ----1- ( ε z ) 2 = ---------------
-
L1 L2

The displacement is continuous at z = L1, but the strain is discontinuous at z = L 1 unless the external forces are
applied in such a manner that ( L 1 + L 2 )q 1 – L 2 q 2 = 0 . The strain energy is the sum of the strain energies in
each section as determined from eq. (2.50) with the thermal strain set equal to zero; i.e.,
L1 ( L1 + L2 )
1 1
U = --- ( EA ) 1 ( ε z ) 12 dz + --- ( EA ) 2 ( ε z ) 22 dz
2 ∫ 2 ∫
0 L1

Substituting for the strains in terms of the displacements we get

1 EA 1 EA
U = ---  ------- q 12 + ---  ------- ( q 2 – q 1 ) 2
2 L  1 2 L  2
Castigliano’s theorem gives

∂U EA EA
Q1 = =  ------- q 1 +  ------- ( q 2 – q 1 ) ( – 1 )
∂ q1  L 1  L 2

∂U EA
Q2 = = 0 +  ------- ( q 2 – q 1 )
∂ q2  L 2

Writing these results in matrix form, we have

44 Thin-Walled Structures
Trusses

Q1 k 1 + k 2 –k 2 q1 EA EA
= where k 1 =  ------- k 2 =  -------
Q2 –k 2 k2 q2  L 1  L 2

The 2 x 2 matrix

( k 1 + k 2 ) –k 2
(2.60)
–k 2 k2

is called the stiffness matrix of the structure. Note that the stiffness matrix is symmetric. A symmetric stiffness
matrix results for any linear elastic structural system undergoing small displacements.

2.8 Trusses
Trusses are idealized as an assemblage of two-force members connected by smooth ball-and-socket joints in
three-dimensional trusses, or by smooth hinge joints in a planar truss. External forces are assumed to act only at
the joints. The line connecting the joints at the end of the bar is assumed to coincide with the reference axis of the
bar. Hence, the axial force and strain in each bar is uniform along its length, and the bar is either in tension or
compression. A planar truss consisting of fifteen bars and eight joints is shown in Fig. 2.16. Each joint in a pla-

4 8 12 16
2 4 6 8 3 7 11 15

2 6 10 14

1 3 5 7 1 5 9 13
joint numbering degree of freedom numbering

Fig. 2.16 A 15 bar truss

nar truss has two degrees of freedom, one horizontal and the other vertical. Hence, there are sixteen degrees of
freedom for this truss. At joint i, which is often called node i, i = 1, 2,..., 8, the horizontal displacement is denoted
by q 2i – 1 and the vertical displacement is denoted by q 2i . The positive directions for the displacements and cor-
responding forces in the fifteen bar truss are defined as shown in Fig. 2.16. The original coordinates of the joints
and the sixteen displacements completely define the configuration of the truss in the deformed state. Cas-
tigliano’s first theorem is particularly useful in the analysis of the static structural response of trusses. The dis-
placements q n and the corresponding forces Q n , n = 1, 2, …, 16 , used in the formulation of Castigliano’s
theorem are the displacements and forces at the joints, or nodes.

To use Castigliano’s first theorem, we need the strain energy of the truss in terms of the nodal displacements.
Begin with a generic bar, say, the i-th bar whose original length is L i and elongation in the deformed state is ∆ i .
Since the strain is uniform along the bar the axial strain of the i-th bar is

Thin-Walled Structures 45
Bars Subjected to Axial Loads


ε i = -----i (2.61)
Li
The bar may also be subjected to a uniform temperature change from the stress free state, and the thermal strain
due to temperature for the i-th bar is denoted as ε it . From eq. (2.50) the strain energy in the i-th bar is

Li
1 2
--- ε i – ε it ε i dz
Ui =
∫ ( EA ) i
2
(2.62)
0

The integrand in this equation is independent of the z-coordinate along the reference axis of the bar, and the
result of the integration is simply the length of the bar times the integrand. Substitute eq. (2.61) for the strain, and
denote the thermal force in the i-th bar as N it = ( EA ) i ε it , so that the strain energy becomes

1 EA
U i = ---  ------- ∆ i2 – N it ∆ i (2.63)
2 L  i

The strain energy of the assemblage is simply the sum of the strain energies in each bar; i.e.,

1  EA 2
U = ∑
all bars
--- ------- ∆ i – N it ∆ i
2 L  i
(2.64)

We have assumed the temperature change is spatially uniform in each bar, but it can be different from bar to bar.
The elongation of a truss bar depends on the nodal displacements at its end points. Hence, Castigliano’s first the-
orem for the truss shown in Fig. 2.16 gives
15
EA
 ------  ∂∆ i 
Qn = ∑ - ∆ – N it
 L i i  
 ∂ q n
n = 1, 2, …, 16 (2.65)
i=1

Since the directions of the nodal displacements at the end of a typical bar do not coincide with the reference
axis of the bar, it is necessary to express the elongation of the bar in terms of the nodal displacements in order to
use Castigliano’s theorem.

Geometry of deformation To relate the elongation of a typical bar to its nodal displacements, requires a study
of the geometry of the deformation of the bar. The end nodes, or joints, of the bar in the undeformed state are
labeled by integers i and j as shown in Fig. 2.17. In the x-y plane of the truss, the coordinates of the beginning
node i are ( x i, y i ) , and the coordinates of the end node j are ( x j, y j ) . The square of the length of the undeformed
bar is determined by the pythagorean theorem as

L 2 = ( x j – xi ) 2 + ( y j – yi ) 2 (2.66)

and the direction cosines of the undeformed bar are


x j – xi y j – yi
cos α = --------------
- cos ( 90 – α ) = sin α = -------------- (2.67)
L L

As is depicted in Fig. 2.17, the bar displaces and rotates to a new position in the plane. The coordinates of
the beginning node in the deformed state are ( x i + q 2i – 1, y i + q 2i ) , and the coordinates of the end node in the
deformed state are ( x j + q 2 j – 1, y j + q 2 j ) . The counterclockwise rotation of the bar is denoted by the angle Ω .

46 Thin-Walled Structures
Trusses

j*
y
L*
q2 j
i* α+Ω
j
yj
q2 j – 1
L q 2i
i α
yi
q 2i – 1

0 xi xj x

Fig. 2.17 Displacement, rotation, and elongation of a truss bar

Let L * denote the length of the bar in the deformed state. The square of the length of the bar in the deformed
configuration is

( L * ) 2 = [ ( x j + q 2 j – 1 ) – ( x i + q 2i – 1 ) ] 2 + [ ( y j + q 2 j ) – ( y i + q 2i ) ] 2 (2.68)

and the direction cosines of the bar in the deformed configuration are
( x j + q 2 j – 1 ) – ( x i + q 2i – 1 )
cos ( α + Ω ) = ---------------------------------------------------------------
- (2.69)
L*
( y j + q 2 j ) – ( y i + q 2i )
sin ( α + Ω ) = --------------------------------------------------
- (2.70)
L*
Define the relative elongation of the bar in the x-direction as
( q 2 j – 1 – q 2i – 1 )
ε x ≡ ------------------------------------- (2.71)
L
and the relative elongation of the bar in the y-direction as
( q 2 j – q 2i )
ε y = ------------------------
- (2.72)
L
Rearranging eq. (2.68) and using the above definitions we get

( L * ) 2 = L 2 [ ( cos α + ε x ) 2 + ( sin α + ε y ) 2 ] (2.73)

The left-hand sides of eqs. (2.69) and (2.70) are expanded by the trigonometric identity for the sum of angles,
and the definitions given by eqs. (2.71) and (2.72) are used in the right-hand sides of eqs. (2.69) and (2.70) to get
L
cos α cos Ω – sin α sin Ω = ------ ( cos α + ε x ) (2.74)
L*
L
sin α cos Ω + cos α sin Ω = -----* ( sin α + ε y ) (2.75)
L

Thin-Walled Structures 47
Bars Subjected to Axial Loads

If ∆ denotes the elongation of the bar, then L * = L + ∆ . The latter expression can be written as
L * = L ( 1 + ε ) where the strain of the bar is denoted by ε = ∆ ⁄ L . Expand the right-hand side of eq. (2.73), use
the trigonometric identity cos2 α + sin2 α = 1 , and let L * = L ( 1 + ε ) to get

( 1 + ε ) 2 = 1 + 2 ( cos α )ε x + 2 ( sin α )ε y + ε x2 + ε y2 (2.76)

Expand the left-hand side of this result, subtract one from each side, and the divide each side by two to get

1 1
ε  1 + --- ε = ( cos α )ε x + ( sin α )ε y + --- ( ε x2 + ε y2 ) (2.77)
 2  2

To find the sine of the rotation angle, multiply eq. (2.74) by – sin α , multiply eq. (2.75) cos α , and then add the
resulting equations. We obtain
1
sin Ω = ---------------- [ ( cos α )ε y – ( sin α )ε x ] (2.78)
(1 + ε)
Given the relative x-direction displacements and the relative y-direction displacements of the nodes, the strain of
the bar is determined from eq. (2.77) and the rotation of the bar is determined by eq. (2.78). For most trusses, the
strains and rotation of the bars are very small. For infinitesimal deformations, the following quantities are very
small
0≤ ε «1 0 ≤ εx « 1 0 ≤ εy « 1

Hence, eq. (2.77) yields the following approximation for the bar strain
ε ≈ ( cos α )ε x + ( sin α )ε y (2.79)

and eq. (2.78) yields the following approximation for the rotation
Ω ≈ ( cos α )ε y – ( sin α )ε x (2.80)

Substitute eqs. (2.71) and (2.72) for the relative elongations into eq. (2.79) to get

∆ q 2 j – 1 – q 2i – 1 q 2 j – q 2i
--- ≈ cos α  -------------------------------- + sin α  --------------------
L  L   L 

Hence, the elongation is


( q 2 j – q 2i )
∆ = ( cos α ) ( q 2 j – 1 – q 2i – 1 ) + ( sin α ) ( q 2 j – q 2i ) (2.81) α
j α
For small strain and small rotations of the bar, eq. (2.81) shows that the elonga-
tion is the sum of the projections of the relative displacements onto the refer- ( q 2 j – 1 – q 2i – 1 )
i
ence axis of the undeformed bar, as is shown in Fig. 2.18. α
Fig. 2.18
EXAMPLE 2.7 Three bar planar truss

The planar truss shown in Fig. 2.19 consists of three bars and one movable joint. Take the thermal strains to be
zero. Determine the 2 x 2 stiffness matrix using Castigliano’s first theorem.

Solution The elongation of each bar as determined from eq. (2.81) is


∆ i = cos ( α i ) q 1 + sin ( α i ) q 2 i = 1, 2, 3 (2.82)

48 Thin-Walled Structures
Trusses

q 2, Q 2

q 1, Q 1

EA
 ------ EA
 ------ EA
 ------
- - -
 L 1  L 2  L 3
α1 α2 α3

Fig. 2.19 Three bar truss

Equation (2.65) results from the application of Castigliano’s theorem, and applied to this example gives
3
EA
Q1 = ∑  ------L- [ cos ( α ) q
i
i 1 + sin ( α i ) q 2 ] [ cos ( α i ) ]
i=1

3
EA
Q2 = ∑  ------L- [ cos ( α ) q
i
i 1 + sin ( α i ) q 2 ] [ sin ( α i ) ]
i=1

These results are written in the matrix form

Q1 k 11 k 12 q 1
= (2.83)
Q2 k 21 k 22 q 2

where the elements of the stiffness matrix are


3 3
EA
 ------ EA
k 11 = ∑  L i
- cos2 ( α i ) k 22 = ∑  ------L- sin ( α )
i
2
i
i=1 i=1
(2.84)
3
EA
 ------
k 12 = k 21 = ∑  L i
- cos ( α i ) sin ( α i )
i=1

Note that this example is statically indeterminate, since we have only two equilibrium equations at the movable
joint for three unknown bar forces. Given the applied forces Q 1 and Q 2 , matrix eq. (2.83) is solved for the nodal
displacements q 1 and q 2 . From eq. (2.82) the elongation of each bar is then computed, and from these elonga-
tions the bar forces are determined from

EA
N i =  ------- ∆ i i = 1, 2, 3 (2.85)
 L i

EXAMPLE 2.8 Three bar truss with lack of fit

Thin-Walled Structures 49
Bars Subjected to Axial Loads

Consider the same three bar truss of Example 2.7, but now assume that bar 1 was too short and had to be
stretched an amount ∆ 1 in order to connect to the joint. This is a case of lack of fit, and lack of fit is common in
the fabrication of structures. That is, before the external loads are applied ( Q 1 = Q 2 = 0 ), the truss bars experi-
ence initial forces due to the lack of fit of bar 1. Determine the initial forces in the bars using Castigliano’s first
theorem.

Solution The strain energy for bar 1 is determined by the total elongation of the bar, which is the sum of initial
elongation due to lack of fit plus the elongation experienced by bar 1 after the connection to the joint is made.
That is, the total elongation of bar 1 is ∆ 1 + ∆ 1 , where ∆ 1 denotes the additional elongation of bar 1 after the
connection of the bar to the joint is made. Hence, the total strain energy for the system in this case is
3
1 EA 1 EA
U = ---  ------- ( ∆ 1 + ∆ 1 ) 2 + ∑ --2-  ------L- ∆ 2
i
2 L  1 i
i=2

The elongation of each bar after the connection is made is


∆ i = cos ( α i ) q 1 + sin ( α i ) q 2 i = 1, 2, 3

Use Castigliano’s first theorem to get


3
EA ∂∆ 1 EA ∂∆ i
Q n =  ------- ( ∆ 1 + ∆ 1 ) + ∑  ------L- ∆ ∂ q
i n = 1, 2
 L 1 ∂ qn i n
i=2

or
3
EA ∂∆ i EA  ∂∆ 1
Qn = ∑  ------L- ∆ ∂ q
i
i
n
+  ------- 
 ∆
L  1  ∂ q n 1
n = 1, 2
i=1

Substitute into this equation the relation between the nodal displacements and the elongation of each bar, eq.
(2.82), after the connection is made to get

Q1 k 11 k 12 q 1 EA cos ( α 1 )
= +  ------- ∆ 1
Q2 k 12 k 22 q2  L 1  sin ( α 1 )

where the elements of the stiffness matrix are the same as given in Example 2.7. Setting Q 1 = 0 and Q 2 = 0 ,
since no external forces are applied to the joint just after assembly, we can solve the above matrix equation for
the joint displacements to get

q1 1 k – k 12 cos ( α 1 )  EA
- 22
= -------------------------------- ------- ∆ 1
q2 ( k 11 k 22 – k 12 ) – k 12 k 11 sin ( α 1 )  L  1
2

From this solution we can calculate the elongation of each bar after assembly from
∆ i = cos ( α i ) q 1 + sin ( α i ) q 2 i = 1, 2, 3

and finally calculate the initial bar forces from

50 Thin-Walled Structures
Complementary virtual work

EA EA EA
N 1 =  ------- ( ∆ 1 + ∆ 1 ) N 2 =  ------- ∆ 2 N 3 =  ------- ∆ 3
 L 1  L 2  L 3

2.9 Complementary virtual work


For the bar shown in Fig. 2.1 on page 18 consider the loading situation where end displacements q 1 and q 2 and
the distributed load intensity p z ( z ) are prescribed. Also, we assume that the axial displacement function w ( z ) is
continuous and single valued with w ( 0 ) = q 1 and w ( L ) = q 2 ; i.e., kinematically admissible. The external
forces Q 1 and Q 2 are interpreted as reactions to the prescribed loading. Now consider an arbitrary, infinitesimal
change in the external forces Q 1 and Q 2 from this deformation state with the distributed load intensity function
unchanged. These infinitesimal changes in the forces are called a virtual forces, and are denoted as δQ 1 and
δQ 2 . The virtual forces cause a change in the internal normal force in the bar. The change in the internal normal
force is denoted by the function δN ( z ) . Take the internal virtual force to satisfy the equilibrium conditions for
this case; i.e.,
d
----- ( δN ) = 0 0<z<L (2.86)
dz
δN ( z = 0 ) = – δQ 1 δN ( z = L ) = δQ 2 (2.87)

Note that the distributed load does not enter into the equilibrium differential equation (2.86) since it is not part of
the virtual force system (δQ1, δQ2, δN(z)). The virtual force system is required to satisfy equilibrium conditions
separately from the actual force system. If the eq. (2.86) is integrated over the length of the bar we get
L
d
∫ d z( δN ) dz = 0 or δN ( L ) – δN ( 0 ) = 0
0

Substitute into this last expression the boundary conditions from eqs. (2.87) to get
δQ 1 + δQ 2 = 0 (2.88)

The last condition, eq. (2.88), is necessary for overall equilibrium of the bar under the application of the external
virtual forces. The solution to eqs. (2.86) to (2.88) is that the internal virtual force δN ( z ) is uniform in z with

δN = – δQ 1 = δQ 2 (2.89)

Note that eqs. (2.89) are satisfied by the overall equilibrium condition (2.88) of the bar. Only one of the external
virtual forces, either δQ 1 or δQ 2 , is independent and permitted to be varied arbitrarily.

Internal actions, like function δN ( z ) above, that satisfy the equilibrium differential equations and those
boundary conditions where the forces are specified are called statically admissible functions. A statically admis-
sible function may not satisfy the governing boundary value problem for the response of the bar. However, within
the set of all statically admissible functions one will lead to strains determined from the material law and the dis-
placements determined from the strain-displacement relations such that the specified boundary conditions on the
displacements are satisfied.

Thin-Walled Structures 51
Bars Subjected to Axial Loads

Define the product of the prescribed displacements q 1 and q 2 with there corresponding virtual forces δQ 1
* ; i.e.,
and δQ 2 , respectively, as the external complementary virtual work δW ext

*
δW ext = q 1 δQ 1 + q 2 δQ 2 (2.90)

From the solution for the virtual force given in eq. (2.89), the complementary external virtual work can be written
as
*
δW ext = ( q 2 – q 1 )δN (2.91)

Now we relate the strain to the end displacements by integrating the strain-displacement relation given by eq.
(2.4) over the length of the bar to get
L L
dw
∫ ε dz = ∫ d z dz = w ( L ) – w ( 0 ) = q
z 2 – q1 (2.92)
0 0

Equation (2.92) is a condition of compatibility1; it relates the strain in the deformed state of the bar to the speci-
fied end displacements. Substitute eq. (2.92) for q 2 – q 1 in eq. (2.91) to get
L
*
 
δW ext

=  ε z dz δN
 
0

The virtual force is spatially uniform via equilibrium eq. (2.86), so that this last equation can be written as
L
*
δW ext =
∫ ε δN dz
z (2.93)
0

The integrand of this equation contains the strain and the virtual force, and these quantities are defined internal to
the bar. Therefore, we define the right-hand side of eq. (2.93) as the internal complementary virtual work δW int * ;

i.e.,
L
*
δW int

≡ ε z δN dz (2.94)
0

The internal complementary virtual work is the integral over the length of the bar of the product of the strain in
the deformed state with the virtual force.

We have shown in the process of going from eq. (2.90) to (2.94) that, for compatible displacements and
strains satisfying eq. (2.92), and for virtual forces satisfying the equilibrium conditions given in eqs. (2.86) to
(2.88), the external complementary virtual work is equal to the internal complementary virtual work. That is, we
have proved the necessary condition that if the displacements and strains are compatible, and the virtual force
system is statically admissible, then the external complementary virtual work equals the internal complementary
virtual work. Compatibility means that the displacements and strains satisfy the strain-displacement equations

1. Compatibility for a deformed body means that the displacements are continuous and single-valued functions such that
strains are computable and that no gaps or discontinuities occur in the deforming body. More generally, compatibility
means that the strains have to satisfy certain mathematical conditions such that, when integrated, the strain functions lead
to continuous, single-valued displacements.

52 Thin-Walled Structures
Complementary virtual work

and prescribed displacement boundary conditions. In problem solving we assume that equating external comple-
mentary virtual to the internal complementary virtual work for every statically admissible system of virtual
forces is sufficient for the displacements and strains of the body to be compatible. We state this principle as fol-
lows.

Principle of Complementary Virtual Work (PCVW)


* ) equals the internal complementary virtual work
If the external complementary virtual work ( δW ext
* ) for every statically admissible variation in the forces, then the displacements and strains of the
( δW int
body are compatible.

Note that the phrase “variation in the forces” means the virtual forces. For the axially loaded bar, the PCVW
means the displacement function w(z) and strain function εz(z) are compatible if
*
δW ext *
= δW int for every statically admissible virtual force δN ( z ) (2.95)

where the external complementary virtual work is given by eq. (2.90) and the internal complementary virtual
work is given by eq. (2.94). The principle of complementary virtual work is independent of the material behavior.

Apply the principle of complementary virtual work to our bar example. Substitute eq. (2.90) for the external
virtual work and eq. (2.94) for the internal virtual work to get
L

q 1 δQ 1 + q 2 δQ 2 =
∫ ε δN dz
z
0

Requiring the virtual forces to satisfy equilibrium conditions given by eqs. (2.86) to (2.88) results in the solution
given by eq. (2.89). Hence, we can eliminate two of the virtual forces in terms of the third in the third. Say we
take δN = δQ 2 and δQ 1 = – δQ 2 , so that the principle becomes
L


q 2 – q 1 – ε z dz δQ 2 = 0 ∀δQ 2
0

This last equation is satisfied if and only if


L


q 2 – q 1 – ε z dz = 0
0

which coincides with the compatibility condition, eq. (2.92).

2.9.1 Complementary strain energy


Now assume the material of the bar is elastic and linear. Solve the material law given by eq. (2.14) for the strain
and then substitute it into the internal complementary virtual work of eq. (2.94) to get
L
* = 1 t
δW int
∫ ------
EA
- ( N + N )δN dz (2.96)
0

Let the integrand of this equation be denoted by the incremental quantity δU 0* . So

Thin-Walled Structures 53
Bars Subjected to Axial Loads

1
δU 0* = ------- ( N + N t )δN (2.97)
EA
This form of the incremental quantity suggests there is a functional of the axial force whose variation is given by
eq. (2.97). This functional is called the complementary strain energy density, and is denoted by U 0* [ N ( z ) ] . Fol-
lowing the developments we used to find the strain energy density functional (refer to eqs. (2.45) to (2.47)) we
would find in a similar manner that the variation of the complementary strain energy density is given by

δU 0* = ( U * )δN (2.98)
∂N 0
Equating eq. (2.97) and eq. (2.98) for every statically admissible virtual force gives

∂ 1
( U * ) = ------- ( N + N t ) (2.99)
∂N 0 EA
Integrating this expression with respect to N, and taking the constant of integration to be zero when N = 0, we get
the complementary strain energy density as

1 1
U 0* = -------  --- N 2 + N N t (2.100)
EA  2 

The thermal force appears as a parameter in the complementary strain energy density, since it assumed indepen-
dent of the normal force N. The important property of the complementary strain energy density is

εz = (U *) (2.101)
∂N 0
which can be seen from eqs. (2.99) and (2.14). A comparison of eq. (2.101) with eq. (2.48) for the strain energy
density shows the dual attributes that these energies possess: the derivative of the complementary strain energy
density with respect to the axial force gives the strain, and the derivative of the strain energy density with respect
to the strain gives the axial force.

Return to the internal complementary virtual work given by eq. (2.96), and use eq. (2.97) to write it as
L
* *
δW int =
∫ δU 0 dz
0

Now interchange the variational operator and the integral operator in accordance with the second of eqs. (2.34),
and write this equation as
* = δU *
δW int

where the complementary strain energy for the bar is defined as


L L
1
U* U 0* dz 2 + 2N t N ) dz
=
∫ =
∫ ----------
2EA
-(N (2.102)
0 0

54 Thin-Walled Structures
Relationship between the complementary strain energy and the strain energy densities

2.10 Relationship between the complementary strain energy and the


strain energy densities
To get a clearer understanding of the complementary strain energy density, consider a plot of the axial force ver-
sus axial strain for the uniaxially loaded bar with no change in temperature from the stress free state as is shown
in Fig. 2.20. For a linear elastic material under these isothermal conditions, the material law for the bar is

U 0* U 0*
N N

EA
1 U0 U0

0 εz 0 εz
(a) linear elastic (b) nonlinear elastic

Fig. 2.20 Strain energy density and complementary strain energy density
under isothermal conditions for a uniaxially loaded bar.

N = EAε z which plots as a straight line with slope of EA to one (Fig. 2.20a). The area between the straight line
1
and the strain axis is --- ( ε z ) ( EAε z ) . From eq. (2.47) the strain energy density is given as U 0 = ( EAε z2 ) ⁄ 2 .
2
Hence, the area between the force-strain curve and the strain axis is the strain energy density. For isothermal
deformation, the complementary strain energy density from eq. (2.100) is U o* = N 2 ⁄ ( 2EA ) . The area between
1 N
the force-strain curve and the force axis is --- ( N )  ------- . Hence, the area between the force-strain curve and the
2  EA
force axis is the complementary strain energy density. For the linear elastic material under isothermal conditions,
the strain energy density and complementary strain energy density are equal in value, U 0 = U 0* . However, it is
preferred to express the strain energy density as a functional of the strain and the complementary strain energy as
a functional of the axial force, since the former was obtained by considering virtual displacements and the latter
was obtained by considering virtual forces.

Examination of Fig. 2.20a shows that the relationship between the strain energy density and complementary
strain energy density can also be expressed as

U 0* = – U 0 + N ε z (2.103)

Actually, this relationship is more general than for a linear elastic material under isothermal deformation. Equa-
tion (2.103) is valid for nonlinear elastic material behavior as is depicted in Fig. 2.20b, and is valid if the defor-
mations is caused in part by a temperature change from the stress free state. For a nonlinear elastic material law
and/or in the non-isothermal deformation the complementary strain energy density is not equal in value to the
strain energy density ( U 0* ≠ U 0 ), but eq. (2.103) remains valid.

Thin-Walled Structures 55
Bars Subjected to Axial Loads

We derive the complementary strain energy density for a linear elastic material under non-isothermal defor-
mation using eq. (2.103). Start with the strain energy density as given by eq. (2.47), which is
1
U 0 = --- EAε z2 – EAε zt ε z
2
Use the material law, eq. (2.14), to eliminate the strain in this equation to express the strain energy in terms of the
axial force, and get
t 2 t
1 N N N N
U 0 = --- EA  ------- + ------- – N t  ------- + -------
2  EA EA  EA EA
which upon simplification becomes
t 2
1 N2 1(N )
U 0 = --- ------- – --- ------------- (2.104)
2 EA 2 EA
Now substitute this result for the strain energy density in eq. (2.103), and substitute the strain-force form of the
material law for the strain in eq. (2.103), to get
t 2 t
1 N2 1(N ) N N
U 0* = –  --- ------- – --- ------------- + N  ------- + -------
 2 EA 2 EA   EA EA
After simplification this equation reduces to
1 t 2
U 0* = ----------- [ N 2 + 2N N + ( N t ) ] (2.105)
2EA
Compare this expression to the complementary strain energy density given by eq. (2.100) and we see that the
t 2
expressions differ by the thermal force term ( N ) ⁄ ( 2EA ) . However, this term is irrelevant since it does not con-
tribute to the property of the complementary strain energy density that

εz = (U *) (2.106)
∂N 0
That is, both eqs. (2.100) and (2.105) will yield the same resulting form of the strain-force form of the material
t 2
law if substituted into eq. (2.106). Hence, we can drop the term ( N ) ⁄ ( 2EA ) in eq. (2.105) without loss of gen-
erality since it is independent of the axial force.

Comparing the complementary strain energy density given by eq. (2.100) to the strain energy density written
in terms of the axial force, eq. (2.104), it is clear that U 0* ≠ U 0 because of the temperature effect.

EXAMPLE 2.9 Application of complementary virtual work to an elastic bar

Consider the uniform bar fixed against rigid body displacement by a support at z = 0, as is shown in the figure
below. That is, the displacement q 1 = 0 . Take the end displacement q 2 to be specified, but not necessarily zero.
End forces Q 1 and Q 2 are regarded as reactions to the prescribed displacements. In addition, the bar is subject to
a uniformly distributed load of intensity p 0 , and a uniform change in temperature ( T – T 0 ) from the stress free
state. Assume the bar is in a state of equilibrium with the displacements and strains being compatible under the

56 Thin-Walled Structures
Relationship between the complementary strain energy and the strain energy densities

imposed loads. Apply the principle of complementary virtual work and use, in addition, that the material of the
bar is linear elastic to relate the end displacement q 2 to the applied loads and the reaction Q 2 .

p0
q1 = 0 q 2, Q 2
z L EA = constant

Solution From eq. (2.90) the external complementary virtual work is q 2 δQ 2 , since q 1 = 0 . The internal virtual
* = δU * . The
work is the variation of complementary strain energy, since the bar material is elastic; i.e., δW int
complementary strain energy is given by eq. (2.102) for a linear elastic material. Hence, the principle of comple-
mentary virtual work for this linear elastic case is
L
1 2 t
N NN 
q 2 δQ 2 = δ  ---
2 ∫ ------- + 2 ---------- dz
EA EA 
0

Interchange the variational operator and the integral operator on the right-hand side of this equation in accor-
dance with the second of eqs. (2.34), then determine the variation of the integrand per the definition of the varia-
tion of a functional given by eq. (2.46). Mathematically these operations are
L L
2 t
1 N NN   ∂ N2 NtN 
δ  ---
2 ∫ ------- + 2 ---------- dz =
EA EA  ∫  2EA EA 
- + ---------- δN dz
 ∂ N ----------
0 0

Perform the differentiation in the integrand to get


L
N Nt
q 2 δQ 2 =
∫ ------- + ------- δN dz
EA EA
(2.107)
0

Following the principle of complementary virtual work, the virtual forces are to satisfy equilibrium conditions
given in eqs. (2.86) to (2.88), which have the solution given in eq. (2.89). Hence, on the basis of virtual force
equilibrium we can take δN = δQ 2 , and then eq. (2.107) becomes
L
 N Nt 
 q2 –
 ∫ ------- + ------- dz δQ 2 = 0
EA EA 
∀δQ 2
0

This equation is satisfied if and only if


L
N Nt
q2 =
∫ ------- + ------- dz
EA EA
(2.108)
0

Equation (2.108) can be recognized as the condition of compatibility for the bar made of a linear elastic material
by substituting the strain from the material law given in eq. (2.14) into the integrated form of the strain-displace-
ment relation given in eq. (2.92).

Thin-Walled Structures 57
Bars Subjected to Axial Loads

The bar was assumed to be in equilibrium prior to application of the p0


virtual force δQ 2 . Hence, the axial force N ( z ) must satisfy the equilib- Q2
N (z)
rium eq. (2.1) with p z ( z ) = p 0 . Alternatively we can impose equilibrium z
of the free body diagram of the right portion of the bar cut at z as shown in ζ
Fig. 2.21. Equilibrium gives L

L Fig. 2.21


N ( z ) = Q 2 + p 0 dζ = Q 2 + p 0 ( L – z ) (2.109)
z

Note that the force Q 2 is unknown at this point since the displacement q 2 was specified instead of Q 2 . Substi-
tute eq. (2.109) for the axial force in eq. (2.108) and perform the integration to get

L L p0 L 2
q 2 = -------Q 2 + -------N t + ----------- (2.110)
EA EA 2EA

The exact solution was obtained Section 2.5 as eq. (2.24). Solving eq. (2.24) for displacement q 2 and setting
q 1 = 0 we arrive at eq. (2.110). Equation (2.110) can be used to find the force Q 2 in terms of the prescribed
quantities q 2 , p 0 and N t .

In summary, we have related the displacement q 2 to the corresponding point force Q 2 by first using the
principle of complementary virtual work to establish compatibility between the displacement q 2 and the strain
ε z ; second, we used the linear elastic material law to relate the strain to the axial force N , and finally equilibrium
was used to determine the axial force N such that the unknown force appeared as parameter.

2.11 Generalized form of Castigliano’s second theorem


Based on the results for the elastic bar subjected only to a variation in the point force Q 2 given in Example 2.9,
the complementary virtual work eq. (2.95) is extended to an elastic structural system restrained against rigid
body displacements, and subjected to many external point forces Q n , n = 1, 2, …, N . Distributed loads and
thermal strains can act on the system as well. The point forces may be independently varied since the variations
in the reactions at the support points compensate to keep the structure in equilibrium. The corresponding dis-
placements at the points of application of the forces, in the direction of the forces, are denoted q n ,
n = 1, 2, …, N . The complementary virtual work eq. (2.95) becomes
N

∑ q δQ n n = δU * (2.111)
n=1

From the differential equilibrium equations and the prescribed boundary conditions on the forces, it is possible,
in principle, to write the internal actions (axial force, bending moments, torque, etc.) of the structure in terms of
the independent external point forces Q n . Since the complementary strain energy of the structure is determined
from integrals of the internal actions over the domain of the structure, it becomes, after integration, a function of

58 Thin-Walled Structures
Generalized form of Castigliano’s second theorem

the Q n ‘s; i.e.,

U * = U * ( Q 1, Q 2, …, Q N )

Hence, the variation in the complementary strain energy is


N

δU * = ∑ ∂Q (U n
* )δQ n (2.112)
n=1

and the complementary virtual work eq. (2.111) becomes


N

∑  q n –
∂ Qn
( U * ) δQ n = 0

∀δQ n n = 1, 2, …, N (2.113)
n=1

Since the virtual forces are independent, we conclude from eq. (2.113) that

∂U *
q n = ---------- n = 1, 2, …, N (2.114)
∂Q n
Equation (2.114) is the general form of Castigliano’s second theorem. Note the similarity of eq. (2.114) to the
property of the complementary strain energy density that the strain is equal to the derivative of the complemen-
tary strain energy density with respect to the internal bar force; or, ε z = ∂U * 0 ⁄ ∂N . This theorem is stated for
the general case as (Langhaar, 1962):

Generalized form of Castigliano’s second theorem


If an elastic structure is mounted such that rigid body displacements are impossible
and certain point forces Q 1, Q 2, …, Q N act on the structure, in addition to distributed
loads and thermal strains, the displacement component q n , n = 1, 2, …, N of the
point of application of force Q n in the direction of Q n is determined by the equation

∂U *
q n = ----------
∂Q n
Although this theorem is derived from complementary virtual work where we regarded the displacement as pre-
scribed and the corresponding point forces as unknown, it is usually used in practice to determine the displace-
ment corresponding to a point force applied to the structure. The example problems to follow will illustrate this
point.

If the material of the bar is linear elastic and there are no thermal strains, we showed in Section 2.10 that the
strain energy density and the complementary strain energy density have the same value. The strain energy and
complementary strain energy have the same value as well since they are defined as integrals over the domain of
the structure of their respective densities. Hence, in this restricted case we can replace U * by U in eq. (2.114) to
get

∂U
qn = (linear elastic, no thermal strain) (2.115)
∂ Qn
Equation (2.115) is Castigliano’s second theorem, which is also called Castigliano’s theorem part II. Note that
eq. (2.115) implies that the strain energy is written in terms of the internal actions of the structure rather than in

Thin-Walled Structures 59
Bars Subjected to Axial Loads

its natural form, which is in terms of the strains. Since we prefer to write the strain energy in terms of the strains
and the complementary strain energy in terms of the internal forces or stresses, the implementation of Cas-
tigliano’s second theorem in this text will use the complementary strain energy on the right-hand side of eq.
(2.115) instead of the strain energy.

EXAMPLE 2.10 Stepped bar response by Castigliano’s second theorem

Determine the relationship between the displacements and the corresponding forces of the stepped bar in Exam-
ple 2.6 on page 43 using Castigliano’s second theorem. There is no change in temperature from the stress free
state.

Solution From the free body diagrams shown in Fig. 2.22,


equilibrium givesthe axial force as N (z) Q1
Q2
N ( z ) = Q1 + Q2 0 < z < L1 z

N ( z ) = Q2 L1 < z < L1 + L2 L1

The complementary strain energy stored in the structure due N (z) Q2


z
to elastic deformation is the sum of the energies from each
segment. (Note: the complementary strain energy and strain L1 + L2
energy are the same in this case.) That is,
Fig. 2.22
L1 ( L1 + L2 )
1 N2 1 N2
U * = --- -------------- dz + ---

2 ( EA ) 1 2 ∫ -------------- dz
( EA ) 2
0 L1

From Castigliano’s second theorem


L1 ( L1 + L2 )
∂ N ∂N N ∂N
(U * ) =
q1 =
∂ Q1 ∫ --------------
( EA ) 1 ∂ Q 1
dz +
∫ --------------
( EA ) 2 ∂ Q 1
dz
0 L1

L1 ( L1 + L2 )
∂ N ∂N N ∂N
(U * ) =
q2 =
∂ Q2 ∫ --------------
( EA ) 1 ∂ Q 2
dz +
∫ --------------
( EA ) 2 ∂ Q 2
dz
0 L1

Substituting for the axial forces determined from equilibrium we get


L1 ( L1 + L2 )
( Q1 + Q2 ) Q2
q1 =
∫ ------------------------ ( 1 ) dz +
( EA ) 1 ∫ --------------
( EA ) 2
( 0 ) dz
0 L1

L1 ( L1 + L2 )
( Q1 + Q2 ) Q2
q2 =
∫ ------------------------ ( 1 ) dz +
( EA ) 1 ∫ --------------
( EA ) 2
( 1 ) dz
0 L1

Performing the integrations gives

L
q 1 =  ------- ( Q 1 + Q 2 )
 EA 1

60 Thin-Walled Structures
Generalized form of Castigliano’s second theorem

L L
q 2 =  ------- ( Q 1 + Q 2 ) +  ------- Q 2
 EA 1  EA 2

Finally, write these results in matrix form as

1 1
----- -----
q1 k1 k1 Q1 EA EA
= k 1 =  ------- k 2 =  -------
q2 1 1 1 Q2  L 1  L 2
----- ----- + -----
k1 k1 k2

The 2 x 2 matrix

1 1
----- -----
k1 k1
(2.116)
1 1 1
----- ----- + -----
k1 k1 k2

is called the flexibility matrix. It is the inverse of the stiffness matrix obtained in Example 2.6, eq. (2.60); i.e.,

1 1 k1 + k2 k2 k1 + k2 1 1
----- ----- ---------------- – ----- ---------------- – k 2  ----- + -----
k 1 + k 2 –k 2 k 1 k1 k1 k1 k1  k 1 k 2
= = 10
–k 2 k2 1 1 1 k 1 1 01
----- ----- + -----
k1 k1 k2 0 – ----2- + k 2  ----- + -----
k1  k 1 k 2

EXAMPLE 2.11 A suspended bar subjected to self weight

A a uniform bar is suspended from the ceiling and is subjected to self weight and a tip force Q 2 as shown in Fig.
2.23. There is no change in temperature from the stress free state, and the material is linear elastic. The material
has a specific weight denoted by γ which has dimensional units of F ⁄ L 3 . Determine the displacement at the tip
of the bar using Castigliano’s second theorem.

z
N
EA
g L dz
Fig. 2.23 Suspended bar under self γAdz
weight and a tip force

N + dN

q 2, Q 2

Thin-Walled Structures 61
Bars Subjected to Axial Loads

Solution The bar is subjected to a distributed load intensity caused by self weight. From the free body diagram
of an element of the bar shown in Fig. 2.23, the equilibrium differential equation is

dN
+ γA = 0 0<z<L
dz
The force prescribed at the tip gives the boundary condition for the axial force as
N ( L ) = Q2

Integrating the differential equation with respect to z from z to L gives


L L
dN
∫ dz
dz = N ( L ) – N ( z ) = – γA dz

z z

and from the boundary condition prescribed on the axial force, we find the axial force is
N ( z ) = Q 2 + γA ( L – z )

The complementary strain energy for no thermal force, eq. (2.102), is


L
1 N2
U*

= --- ------- dz
2 EA
0

The tip displacement is determined from Castigliano’s second theorem as


L L
∂ 1 ∂  N 2 N ∂N
( U * ) = --- -  dz
q2 =
∂ Q2 ∫ ------- dz =
2 ∂ Q 2  EA ∫ ------
EA  ∂ Q 
2
0 0

Now substitute for the axial force this equation to get


L
( Q 2 + γA ( L – z ) )
q2 =
∫ ----------------------------------------- ( 1 ) dz
EA
0

Thus, after integration, we find the tip displacement is


L γ L2
q 2 = -------Q 2 + ---------
EA 2E

This result for the displacement can be evaluated for Q 2 = 0 , of course, which suggests a method to compute
the displacement at a point on the structure where no point force acts. Merely, put a force at the point in question,
apply Castigliano’s theorem, then set the point force to zero.

2.12 References
Callister, W.D., 1997, Materials Science and Engineering, Fourth Edition, John Wiley & Sons, Inc., New York,
pp. 777-779, 788, 789.

Courant, R. and Hilbert, D., 1953, Methods of Mathematical Physics, Volume I, Interscience Publishers, Inc.,

62 Thin-Walled Structures
Problems

New York, pp. 164-274.

Dym, C.L.,1997, Structural Modeling and Analysis, Cambridge University Press, Cambridge, United King-
dom, pp. 73-95.

Dowling, N.E., 1993, Mechanical Behavior of Materials, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, p.
110.

Langhaar, H.L., 1962, Energy Methods in Applied Mechanics, John Wiley and Sons, Inc., New York, pp. 75-
103, 133-136.

Oden, J.T., and Ripperger, E.A., 1981, Mechanics of Elastic Structures, Second Edition, Hemisphere Publish-
ing Corporation, New York, pp. 241-257.

2.13 Problems
1. Take the bar of Example 2.1 to have a solid circular cross section with a diameter of 0.50 inches, and take its
6
length L = 20 inches. The material is aluminum alloy with a modulus of elasticity E = 10 ×10 psi , yield
3
strength in tension and compression of magnitude σ yield = 47 ×10 psi , and thermal coefficient of expansion
–6
α = 13 ×10 /°F .

a) For the bar subjected to the loads p 0 = 265 lb/in. , q 1 = 0 in. , q 2 = 0.0060 in. , and
T – T 0 = 100 °F , determine the maximum axial displacement, the maximum tensile normal stress,
and the maximum magnitude of the compressive normal stress. (Partial answer: maximum tensile nor-
mal stress is 3,494 psi.)
b) Assume proportional loading such that p 0 = ( 265 lb/in. ) λ , q 1 = 0 in. , q 2 = ( 0.0060 in. )λ , and
T – T 0 = ( 100 °F )λ , where λ is a dimensionless load factor. Determine the largest positive value of λ
such that normal stress σ z is between the limits – σ yield ≤ σ z ≤ σ yield and the maximum displacement
w max is less than 0.0180 inches.

2. A bolt is threaded through a tubular sleeve, and the nut is turned up just tight by hand as shown. Using
wrenches, the nut is then turned further, the bolt being put in tension and the sleeve in compression. If the bolt
has 16 threads per inch, and the nut is given an extra quarter turn (900) by the wrenches, estimate the tensile force
in the bolt if the bolt and sleeve are of steel and the cross-sectional areas are: Bolt area = 1.00 in2, Sleeve area =
0.60 in2.

6 in.

Thin-Walled Structures 63
Bars Subjected to Axial Loads

3. Consider the two term approximation for the displacement of the bar in Example 2.3 on page 34 as given by
eq. (2.42), which is repeated below.

z z
w ( z ) = q 1 sin  π --- + q 3 sin  3π --- 0≤z≤L
 L  L

This is a kinematically admissible displacement function with parameters q1 and q3 unknown. (Refer to Fig. 2.14
on page 36 and the discussion proceeding it.)
a) Use the principle of virtual work to determine the two equilibrium equations associated with the virtual
displacements δq1 and δq3. Do not assume a material law has been specified.
b) Now assume the material is linear elastic and solve for the unknown parameters q1 and q3. Note: the fol-
lowing definite integral, with m and n integers, is encountered in the solution.

L L
 z  z  --- m = n
cos mπ --- cos nπ --- dz =  2
∫  L  L
0 m≠n
0

c) Calculate the percentage error in the strain energy, maximum displacement, and maximum normal force
with respect to the exact solution. The exact solution is given in Example 2.2 on page 28.

4. A uniform bar with axial stiffness EA and length L is fixed at each end, and it is subjected to an axial distrib-
2πz
uted load p z ( z ) = p 0 sin  --------- , 0 ≤ z ≤ L . We seek an approximate solution using the principle of virtual work.
 L 

pz ( z )

Q1 Q2

EA
z, w
L

Assume the kinematically admissible axial displacement function

32 z z z
w ( z ) = ------ q  ---  1 – ---  1 – 2 ---
3  L  L  L

where the unknown parameter q represents the displacement at z = L ⁄ 4 and at z = 3L ⁄ 4 . Use the principle
of virtual work, and assume the material of the bar is linear elastic, to determine
a) The value of parameter q in terms of the given data.
b) The strain energy stored in the bar and the percentage difference with respect to the strain energy for the
exact solution. The strain energy in the exact solution is U = ( p 02 L 3 ) ⁄ ( 16π 2 EA ) .

c) The end forces Q 1 and Q 2 and the percentage difference with respect to the exact solution. The end
forces in the exact solution are Q 1 = – Q 2 = – ( p o L ) ⁄ ( 2π ) .

64 Thin-Walled Structures
Problems

5. Each bar in the truss shown has a cross-sectional


2 100 in. 1
area of 1.0 in.2, and a modulus of elasticity of 107 psi.
There is no change in temperature. Use Castigliano’s 45°
first theorem to find 60°

a) the horizontal and vertical displacements of


3 10,000 lb.
node 1, 45°
y
b) the stress in psi in each bar, and
4 x
c) the horizontal and vertical support reactions
at node 5. 30°

5
6. The bars in the truss shown have the following
cross-sectional areas: A 1 = 1.0 in. 2 , 1
A 2 = A 4 = 2.0 in. 2 , A 3 = 1 ⁄ 2 in. 2 ,
A 5 = A 6 = 3 ⁄ 2 in. 2 . The modulus of elasticity of 2 3 4
40 in.
each bar is 107 psi. Compute the vertical displacement
of the right-hand joint using Castigliano’s second theo-
rem. Note this truss is statically determinate and all bar 5 6
forces can be determined in terms of external load Q.
30 in. 30 in.
7. Use Castigliano’s’ second theorem to compute the
horizontal displacement of the right-hand joint of the Q = 30, 000 lb.
previous problem.

8. The plane truss shown represents a single bay of a


–6
wing spar truss. For all bars: E = 75 GPa and α = 23.0 ×10 ⁄ °C . The cross-sectional areas of the bars are:
2580 mm2 for the horizontal bars, 387 mm2 for the vertical bars, and 2690 mm2 for the diagonal bars. The upper
horizontal bar is heated to 250°C above the zero stress temperature, and all other bars remain at the zero stress
temperature. Two 45 kN lift forces act at joints 1 and 2.

45kN
250°C
4 2

810 mm

3 1
1080 mm 45kN
Use Castigliano’s first theorem to find
a) stiffness matrix in kN/mm,
b) displacement of all joints in mm,

Thin-Walled Structures 65
Bars Subjected to Axial Loads

c) all boundary reactions in kN, and


d) the stresses in MPa in each bar.

66 Thin-Walled Structures
CHAPTER 3 Axial Normal Stress in
Pure Bending and
Extension

In the first four sections the flexure formula for straight, prismatic beams made from a single homogenous, iso-
tropic, and linear elastic material is developed. The cross section can be asymmetric or symmetric, and the exam-
ples emphasize thin-walled approximations. The flexure formula (eq. 3.27) relates the axial normal stress acting
on the beam's cross section to the components of the bending moment. In the last section we consider combined
extension, pure bending, and thermal effects for multi-material beams. The axial normal stress in this latter case
is given by eqs. (3.66-3.68).

3.1 Pure bending


The beam is assumed to be initially straight with uniform cross section and uniform material properties along its
length. The z-axis is taken along the length of the beam and the x- and y-axes are in the cross section. Each end of
the beam is subjected to equal and opposite couples, which are equivalent to a moment denoted by M. These cou-
ples act in the plane, called the plane of loading, which is inclined to the y-z plane by the angle α as shown Fig.
3.1. The moment of each end couple is directed perpendicular to the plane of loading. Equilibrium dictates that
each cross section is subjected to the applied moment M. This loading condition is called pure bending because
the shear force vanishes along the entire length. The bending moment M projects on the x- and y-axes as
M x = M cos α and M y = M sin α . The sign convention for positive bending moments Mx and My is such that
they tend to cause elongation of the fibers parallel to the z-axis that have positive x and y coordinates (first quad-
rant of the x-y plane in the cross section). Thus Mx is positive if directed along the positive x-axis by the right-
hand screw rule, and My is positive if directed along the negative y-axis by the right-hand screw rule.

Thin-Walled Structures 67
Axial Normal Stress in Pure Bending and Extension

y
α
plane of loading
y

M sinα A x Mx
M cosα
z α
A
M cosα
M sinα
M
My
Section A-A
Fig. 3.1 A uniform beam subjected to equal and opposite couples at each end.

3.2 Geometry of deformation


Under the special conditions of pure bending described above, the assumptions of the deformed geometry of the
beam are listed in the table below.

Geometry of deformation in pure bending


1. The beam deforms into a plane circular arc, and
2. That plane sections originally perpendicular to the z-axis remain plane and perpendicular
to the z-axis in the deformed beam.

Plane of bending The plane in which the beam bends is denoted the y – z plane where the y -axis is inclined
by the angle β from the y-axis as shown in Fig. 3.2. Note that the plane of bending is not assumed to coincide
with the plane of loading; i.e., α ≠ β. Coordinate axes x and y are rotated through the angle β relative to the x
and y coordinate axes as shown in the figure. A point in the cross section can be located by x and y coordinates
and by x and y coordinates. Consider a typical point A with coordinates x and y. The x and y coordinates of
point A are determined by the geometry shown in Fig. 3.3. From Fig. 3.3, the relationship between these coordi-
nates is

x = x cos β – y sin β
(3.1)
y = x sin β + y cos β
The cosines of the angles between the positive coordinate axes, or direction cosines, can be written in tabular
form as shown in Fig. 3.2. For example, the direction cosine of the angle between the x -axis and the y-axis is
cos ( 90° + β ) = – sin ( β ) . Note that the sum of the squares of the direction cosines in a row or column equals
unity. Equations (3.1) show that the rows of the direction cosine table are associated with the x - y coordinates.
The inverse of eqs. (3.1) is

68 Thin-Walled Structures
Geometry of deformation

y
α
y
β direction cosines
x x y
α x cos β – sin β
plane of bending
β y sin β cos β
M
plane of loading
x

Fig. 3.2 Projection of the plane of loading and the plane of bending onto
the cross section of the beam.

y x sin β
90°

y cos β β A
y Fig. 3.3 Point A located in two Cartesian
90° β coordinate systems, where one system is
rotated through angle β relative to the other.
β
90°

β x

x 90°

y sin β
x cos β

x = x cos β + y sin β
y = – x sin β + y cos β

Hence, the columns of the direction cosine table are associated with the x -y coordinates

Axial normal strain The particles on the z-axis in the undeformed beam displace to a circular arc of radius ρ in
the plane of bending, which we call the z-curve. See Fig. 3.4. If the beam is considered to be made up of longitu-
dinal fibers, then fibers parallel to the z-axis in the undeformed beam deform into concentric circular arcs. Some
of these parallel fibers are elongated and some are shortened. Consequently, the fibers on some surface between
the top and bottom surfaces of the beam, called the neutral surface, will remain the same length. Although we do
not know the location of the neutral surface at this point, we take the z-axis to lie in the neutral surface of the

Thin-Walled Structures 69
Axial Normal Stress in Pure Bending and Extension

undeformed beam. We will show later that the neutral surface in the undeformed beam coincides with the x -z
plane.

z-curve in the neutral surface.

z
ρ

Fig. 3.4 Beam deformed into a circular arc in the plane of bending.

)
An element in the undeformed beam of length ∆z bounded by two parallel line segments PQ and RS is
)

shown in Fig. 3.5. Line PQ is part of the z-axis and line RS is characterized by a constant y -coordinate value.

–θ
– θ + ∆θ

y R* y
P* S*
R S ρ ∆θ
θ y Q*
z P Q
z ∆z

Fig. 3.5 Deformation of an element in the plane of bending.


)

In the deformed geometry these line segments map to concentric circular arcs P * Q * and R * S * . Let θ denote the
rotation of a line element originally parallel to the y -direction in the undeformed beam, positive clockwise.
Since cross sections remain perpendicular to the neutral surface in the deformed beam, the angle θ is also equal
)

to the rotation of the line element coinciding with z-axis. In Fig. 3.5, the line P * R * is shown to rotate counter-
)

clockwise which corresponds to a – θ , and line Q * S * is shown to rotate clockwise through angle – θ + ∆θ .
)

Then, the included angle between sections P * R * and Q * S * is ∆θ. The radius of the arc P * Q * is ρ as described
in the previous paragraph. It is assumed that the distortion of the cross section in the plane of bending is negligi-

70 Thin-Walled Structures
Geometry of deformation

)
ble, so that the distance between the two circular arcs P * Q * and R * S * remains y . Hence, the radius of arc
)

)
R * S * is ρ + y . The strain of line element RS is defined as

)
R * S * – RS
ε z = ------------------------- (3.2)

)
RS
From Fig. 3.5

)
R * S * = ( ρ + y )∆θ (3.3)
)

)
and since line element P * Q * is in the neutral surface P * Q * = PQ = R S = ρ∆θ . Thus, eq. (3.2) becomes
)

)
R*S* – P*Q* ( ρ + y )∆θ – ρ∆θ y
ε z = -------------------------------- = ----------------------------------------- = --- (3.4)
ρ∆θ ρ
)

P*Q*
)

)
Also, we have P * Q * = ρ∆θ = ∆z , since the line element P * Q * is in the neutral surface. From this relation we
have in the limit as ∆z → 0 and ∆θ → 0 , the following equation for the curvature of the z-curve in the deformed
beam
1 dθ
--- = ------ = θ′ (3.5)
ρ dz
where the prime denotes a derivative with respect to z. Equation (3.5) is consistent with the definition of curva-
ture as the change in the slope angle of the curve with respect to its arc length. Hence,

ε z = yθ′ (3.6)

Equation (3.6) shows that the normal strain of a line element parallel to the z-axis is proportional to the distance
from the z-axis times the curvature. Fibers above the z-axis ( y > 0 ) are stretched ( ε z > 0 ) if the beam is bent
convex side up ( 1 ⁄ ρ > 0 ). Fibers below the z-axis are shortened ( ε z < 0 ) if the beam is bent convex side up (Fig.
3.4).

Projection of the z-curve onto the x-z and y-z planes Now we need to express the normal strain given by eq.
(3.6) in terms of the x-y coordinate system. This is accomplished by projecting the z-curve in the deformed beam,
which lies in the y – z plane, onto the coordinate planes x-z and y-z. The z-curve is represented in the x-y-z coor-
dinate system by introducing displacement functions u ( z ), v ( z ), and w ( z ) that represent the x-, y-, and z-direc-
tion displacements, respectively, of a particle at point P on the z-axis. Under the deformation the particle
originally at P with coordinates ( 0, 0, z ) displaces to the point P* with coordinates ( u ( z ), v ( z ), z + w ( z ) ) . See
Fig. 3.6.

Also the displacements of the particle at P can be represented in the x - y -z coordinate system. Let u ( z )
v ( z ) denote the displacements in the x - and y -directions, respectively, of the particle at point at P. These dis-
placement components are related to components u ( z ) and v ( z ) by the considering the position of the projec-
tion of P* onto the x-y plane. The projection of P* onto the x-y plane is labeled point P 3* in Fig. 3.6 and Fig.

Thin-Walled Structures 71
Axial Normal Stress in Pure Bending and Extension

v(z)

P 1*
P: ( 0, 0, z ) P 3*
θx x

P*
w(z)
u(z)

Q*
z P 2*
θy

Fig. 3.6 Displacements of the particle from P to P*, and the rotations of the
projections of a line element P*Q*.

3.7. From Fig. 3.7, the displacement components in the x - y coordinates are related to those in the x-y coordi-

y v

v P 3*
y
β

x u
P
plane of bending
u

Fig. 3.7 Lateral displacement components of the particle at point P.

nates by

u = u cos β – v sin β
(3.7)
v = u sin β + v cos β
which is of the same form as the coordinate transformations given by eq. (3.1) with the coordinates replaced by
their respective displacements. Since the z-curve lies in the y – z plane, displacement u = 0 for all z. That is,
point P 3* is in the y -z plane. Also, the angle β is independent of the z-coordinate, since the beam is assumed to
deform into a plane circular arc. Next we relate the rotation, or slope, of an element of the z-curve in the plane of
bending to the rotations its projections onto the x-z and y-z planes.

72 Thin-Walled Structures
Geometry of deformation

)
)
The displacement and rotation of an infinitesimal line element PQ on the z-axis to P * Q * on the z-curve is
shown in Fig. 3.8. Since the line element does not change length we have by the Pythagorean theorem in the

P * dz + dw
v w θ
– dv
dz
Q*
v + dv w + dw

P Q
z
z z + dz z* z * + dz *

Fig. 3.8 Displacement and rotation of an infinitesimal line element on the z-axis
in the plane of bending.

deformed state that ( dz ) 2 = ( dz + dw ) 2 + ( – dv ) 2 . Hence, in the limit as dz → 0 , the derivative of the axial dis-
placement is related to the derivative of the lateral displacement by

( 1 + w′ ) 2 + ( – v′ ) 2 = 1 (3.8)

As is shown in Fig. 3.8, the rotation, or slope, of the z-curve is related to displacement v by

dv
θ ≅ sin θ = – ------ = – v′ (3.9)
dz

where we assume the angle θ is small so that the sine of it can be replaced by the angle itself in radians. Differ-
entiating eqs. (3.7) with respect to z, and noting that u = 0 and the angle β is constant for all z, we get

0 = u′ cos β – v′ sin β
(3.10)
– θ = u′ sin β + v′ cos β

The derivatives of the displacement components in the x - and y -coordinate directions are directly related to
the rotations of the projections of the z-curve onto the x-z and y-z planes. The projection of point P* on the onto
the y-z plane is labeled P 1* , and its projection onto the x-z plane is labeled P 2* , as is shown in Fig. 3.6. Points P 1*
and P 2* and the projections of the differential line element along the z-curve onto the coordinate planes shown in
Fig. 3.6 are also shown in Fig. 3.9. From the differential geometry of Fig. 3.9, we have in the limit as dz → 0
that
– v′ – u′
tan θ x = --------------- tan θ y = --------------- (3.11)
1 + w′ 1 + w′
Since v′ is assumed small with respect to unity, eq. (3.8) yields that w′ is also small with respect to unity. Thus,
we neglect w′ with respect to one in the denominators of eqs. (3.11), and approximate the rotations of the projec-
tions as
θ x = – v′ θ y = – u′ (3.12)

Thin-Walled Structures 73
Axial Normal Stress in Pure Bending and Extension

y x

P 1* dz + dw P 2* dz + dw
θ x – dv θ y – du
v u
v + dv u + du
z z
z* z * + dz * z* z * + dz *

projection in y-z plane projection in x-z plane

Fig. 3.9 Rotations of the projections of the z-curve onto the cartesian planes.

where the tangent of a small angle is approximated by the angle itself in radians. Thus, eqs. (3.10) become
0 = – θ y cos β + θ x sin β
(3.13)
θ = θ y sin β + θ x cos β
which are the relations between the rotation of the z-curve to the rotations of its projections onto the coordinate
planes. Solving these equations for the rotations in the x-y coordinate system, we get
θ y = θ sin β θ x = θ cos β (3.14)

Equations (3.14) show that for small rotations, the rotations of the projected curves are simply the components
of the rotation of the z-curve .

Normal strain in the x-y-z system Finally, the strain given by eq. (3.6) is written in terms of the x-y system by
substituting the second of eqs. (3.1) for y , and the derivative of the second of eqs. (3.13) for θ′ , to get

ε z = ( x sin β + y cos β ) ( θ′ y sin β + θ′ x cos β )

Rearrange this equation to


ε z = x ( θ′ y sin2 β + θ′ x sin β cos β ) + y ( θ′ y cos β sin β + θ′ x cos2 β )

Differentiating the first of eqs. (3.13), recalling that angle β is independent of z, results in the relation
θ′ y cos β = θ′ x sin β

Now use this result in the previous equation to get the strain expressed as
ε z = xθ′ y + yθ′ x (3.15)

where the curvatures of the projected curves are determined from eqs. (3.12) to be
θ′ y = – u″ θ′ x = – v″ (3.16)

74 Thin-Walled Structures
Bending normal stress — flexure formula

3.3 Bending normal stress — flexure formula


The strain ε z is related to the normal stress σ z acting on the cross section by the y
material law, and we consider the beam to be made from a linear elastic, isotropic σy
material. For uniaxial loading only, this material law is given by eq. (2.5) on p. 22 . σx
Under multi-axial loading the material law is obtained by superpostion of the normal
strains ε x, ε y, and ε z for line elements parallel to the x-, y-, and z-directions due to σz x
z
the combined action of normal stresses σ x, σ y, and σ z acting on faces of an infinites-
imal element normal to to the x-, y-, and z-directions. Superposition of normal strains using eq. (2.5) on p. 22
yields

1 ν ν
--- – --- – ---
εx E E E σ
x
ν
= – --- 1 ν (3.17)
εy --- – --- σ y
E E E
εz σz
ν ν 1
– --- – --- ---
E E E
where E is the modulus of elasticity and ν is Possion’s ratio. The material law for a linear elastic, isotropic mate-
rial is called Hooke's law. Material properties E and ν which vary from point to point in the cross section (x-y
plane) are permissible, but the material properties are assumed uniform along the z-axis. Thus, a laminated beam
of more than one material may be considered, which is addressed in Section 3.5. However, in this section we
limit consideration to a beam made of a single homogeneous material, so that E and ν are independent of the spa-
tial coordinates x, y and z. Now we make the third basic assumption of beam theory:

Assumption regarding the material law


3. Lateral normal stresses σ x and σ y are assumed small with respect to the axial normal
stress σ z , and hence are neglected in Hooke’s law.

Neglecting stress components σ y and σ y , Hooke's law becomes

σ z = Eε z (3.18)

Eliminating the strain in eq. (3.18) via eq. (3.15) we get


σ z = E ( xθ′ y + yθ′ x ) (3.19)

The distribution of normal stresses given by eq. (3.19) over the cross section is, in general, statically equiva-
lent to the resultants N, Mx, and My as shown in Fig. 3.10. The axial load N, and bending moments Mx and My
are related to the normal stress by

N =
∫ ∫ σ dA
A
z Mx =
∫ ∫ yσ dA
A
z My =
∫ ∫ xσ dA
A
z (3.20)

where dA = dxdy is an area element in the cross-sectional area denoted by A. Substituting eq. (3.19) into eq.
(3.20) and integrating over the cross section we obtain

Thin-Walled Structures 75
Axial Normal Stress in Pure Bending and Extension

y
dA ⇔ Mx
x N
z σz z
My
x

Fig. 3.10 Statical equivalence of the normal stress to the beam resultants.

N = EQ x θ′ x + EQ y θ′ y
M x = EI xx θ′ x + EI xy θ′ y (3.21)

M y = EI xy θ′ x + EI yy θ′ y

in which we defined

Qx =
∫ ∫ ydA
A
Qy =
∫ ∫ xdA
A
(3.22)

and

2 2
I xx =
∫ ∫ y dA
A
I yy =
∫ ∫ x dA
A
I xy =
∫ ∫ xy dA
A
(3.23)

In obtaining the results shown in eqs. (3.21), the modulus of elasticity E was brought outside the area inte-
grals since it is independent of x and y for a homogeneous material cross section. The integrals in eqs. (3.22) and
(3.23) are purely geometric quantities. They reflect how the area is distributed with respect to the x- and y-coordi-
nate axes. The quantities Qx and Qy in eqs. (3.22) are called first area moments since the power of the coordi-
nate in the integrand is unity, while the Ixx and Iyy in eqs. (3.23) are called second area moments since the power
of the coordinates in the integrand is two. The quantity Ixy is called the product area moment.

For the case of pure bending, only moments Mx and My are non-zero, so the axial force N must vanish. Set-
ting N = 0 in the first of eqs. (3.21) implies that
( N = 0) ⇒ Qx = Qy = 0 (3.24)

The first area moments vanish if the x-axis and the y-axis pass through the centroid of the cross-sectional area.
This locates the z-axis; i.e., the z-axis of the beam passes through the centroid of each cross section. For centroi-
dal coordinates in the cross section, the moment-curvature relationship of eq. (3.21) can be written in the matrix
form

76 Thin-Walled Structures
Bending normal stress — flexure formula

Mx I xx I xy θ′ x
= E (3.25)
My I xy I yy θ′ y

where the second are moments are computed for centroidal coordinates in the cross section. The curvature-
moment relationship is the inverse of this, or

θ′ x 1 I yy – I xy M x
= -----------------------------------
2
(3.26)
θ′ y E ( I xx I yy – I xy ) – I xy I xx M y

Equations (3.25) and (3.26) are regarded as the material law for beam bending, and these equations are valid only
if the x- and y-axes pass through the centroid of the cross section.

The formula for the bending normal stress σz in eq. (3.19) can be written in terms of the bending moments
Mx and My by using the curvature-moment relation in eq. (3.26). Substituting the curvatures from eq. (3.26) into
eq. (3.19) we get
– I xy M x + I xx M y I yy M x – I xy M y
σ z =  -------------------------------------- x +  ----------------------------------
- y (3.27)
 I xx I yy – I xy  2  I xx I yy – I xy 2 

or we can re-write this as

I yy y – I xy x  I xx x – I xy y 
σ z =  -------------------------- M x +  -------------------------- My (3.28)
 I xx I yy – I xy 2   I xx I yy – I xy 2 

Equation (3.27), or its equivalent eq. (3.28), is called the flexure formula, since it determines the normal stress at
cross-sectional coordinates x and y due to bending. Notice that the bending normal stress is zero at the origin,
which coincides with the centroid of the cross section, and that the bending normal stress varies linearly in x and
y over the cross section. Thus, the extreme values of the bending normal stress occur somewhere on the boundary
of the cross section where coordinates x and y can attain their largest values.

The plane of bending of the beam can be determined in terms of the bending moment components Mx and
My by first differentiating eqs. (3.14) with respect to z and recognizing that the z-curve lies in the plane of bend-
ing (angle β is independent of z for pure bending). Second, solve the resulting equations for β to get
θ′
tan β = -------y (3.29)
θ′ x
Now use eq. (3.26) to eliminate the curvatures in eq. (3.29) to find

– I xy M x + I xx M y
tan β = ---------------------------------------
- (3.30)
I yy M x – I xy M y

Thus, knowing the bending moment components and the second area moments of the cross section, the angle β,
and hence the plane of bending, is determined by eq. (3.30). The axis normal to the plane of bending through the
centroid is labeled as the x axis in Fig. 3.2. On the x axis y = 0 , so that from strain-curvature relation given
by eq. (3.6) the normal strain ε z = 0 on the x axis. Also, from Hooke’s law this means that the bending normal

Thin-Walled Structures 77
Axial Normal Stress in Pure Bending and Extension

stress σ z = 0 on the x axis. Since ε z = 0 on the x axis, the x axis is called the neutral axis . The particles in
x -z plane map to the neutral surface in the deformed beam. The equation for the neutral axis in the cross section
is determined from the condition that y = 0 in the first of eqs. (3.1). We get

0 = ( x sin β + y cos β ) NA
or y NA = – x NA tan β (3.31)

where the tangent of β is determined by eq. (3.30).

The neutral axis passes through the centroid of the cross section, since the centroid was used as the origin of
the x-y system. If the axial force N ≠ 0 in addition to non-zero bending moments, then the axis of zero stress, or
the neutral axis, will no longer pass through the centroid, but will be parallel to the line obtained from eq. (3.31).
From Fig. 3.1, recall that the plane of loading is located by
M
tan α = -------y
Mx

Using this result in eq. (3.30), we get


– I xy + I xx tan α
tan β = ------------------------------------ (3.32)
I yy – I xy tan α

Equation (3.32) shows that, in general, the plane of loading and the plane of bending do not concide ( α ≠ β ). The
plane of loading and the plane of bending coincide if I xy = 0 and M x = 0 , or I xy = 0 and M y = 0 , or
I xy = 0 and I xx = I yy .

EXAMPLE 3.1 Bending normal stress distribution in a cantilever beam with a thin-walled zee section.

The cantilever beam shown in Fig. 3.11 is subjected to a bending moment M at its tip. The cross-sectional
dimensions a and t are considered known with 0 < t/a << 1; i.e., this is a thin-walled section. Note that only the
center line of the wall is drawn in the sketch of the cross section and not the thickness, since the thickness is
small. The wall center line is called the contour of the cross section. Given that the second area moments about
the centroidal axes x and y are

8 2
I xx = --- a 3 t I yy = --- a 3 t I xy = – a 3 t (3.33)
3 3
determine the neutral axis in the cross section and the distribution of the bending normal stress.

Solution First, note that the components of the bending moment are M x = – M and M y = 0 . Thus, the plane
of the couple whose moment is M is the y-z plane, or the angle α = π for the plane of loading shown in Fig.
3.1. From eq. (3.30)
– I xy –( –a 3 t ) 3
tan β = --------- = ------------------ = --- so β = 56.31°
I yy 2 3 2
--- a t
3

3
and from eq. (3.31) the equation of the neutral axis in the cross section is y NA = – --- x NA .
2

78 Thin-Walled Structures
Bending normal stress — flexure formula

y y
a

M
z centroid
a
C x

t, typical a

a
Fig. 3.11 Pure bending of a cantilevered beam with a zee cross section
Cross Section

The flexure formula, eq. (3.28), for this example becomes

 2--- a 3 t y – ( – a 3 t )x
3  M
σ z = --------------------------------------------------------- ( – M ) = – ( 6y + 9x ) ----------
3t
8 2
 --- a 3 t  --- a 3 t – ( – a 3 t ) 2 7a
3  3 

We plot the bending normal stress on the contour only. Coordinates x and y are related on the contour. That is,

top flange: –a ≤ x ≤ 0 y = a

web: x = 0 –a ≤ y ≤ a
bottom flange: 0≤x≤a y = –a

The neutral axis and the bending normal stress distribution are shown in Fig. 3.12.

Thin-Walled Structures 79
Axial Normal Stress in Pure Bending and Extension

6M
– --- ------
-
3M 7 a2t
--- ------
-
7 a2t
y

x Fig. 3.12 Bending normal stress


distribution along the contour of the
M zee section

σz
3M
– --- ------
-
7 a2t

6M 3
--- ------
- y NA = – --- x NA
7 a2t 2

EXAMPLE 3.2 Lateral displacements of the zee section beam

Determine the lateral displacement functions u(z) and v(z), 0 ≤ z ≤ L, for the zee section beam of Example 3.1.

Solution First note that this a statically determinate problem. So the equilibrium equations are satisfied for
M x = – M and M y = 0 for all z ∈ ( 0, L ) , where M denotes the specified end moment. To find the lateral dis-
placements u ( z ) and v ( z ) , 0 ≤ z ≤ L , we begin with the curvature-moment relations given by eqs. (3.26). These
equations are repeated below.

θ′ x 1 I yy – I xy M x
= -----------------------------------
2 )
θ′ y E ( I xx I yy – I xy – I xy I xx M y

From equilibrium we have M x = – M and M y = 0 for 0 ≤ z ≤ L . From the given data for the second area
8 2
moments we know I xx = --- a 3 t , I yy = --- a 3 t , and I xy = – a 3 t . Thus,
3 3

2 = 8
det ( I ) = I xx I yy – I xy
2 7
--- × --- – ( – 1 ) 2 a 6 t 2 = --- a 6 t 2 (3.34)
3 3 9
Substitute these data into curvature-moment relations above to get
6 M 9 M
θ′ x = – --- ----------
- θ′ y = – --- ----------
- (3.35)
7 Ea 3 t 7 Ea 3 t

Integrate these equations with respect to z once to get

80 Thin-Walled Structures
Bending normal stress — flexure formula

6 M  9 M 
θ x = – ---  ----------
- z + c1 θ y = – ---  ----------
- z + c2 (3.36)
7  Ea 3 t 7  Ea 3 t
The constants due to indefinite integration are determined from the boundary conditions. At z = 0 the beam is
clamped to the rigid wall. Hence the cross section at z = 0 is prevented from rotation, and this means θ x ( 0 ) = 0
and θ y ( 0 ) = 0 . Substituting these boundary conditions into eqs. (3.36), we find c 1 = 0 and c 2 = 0 . Hence,
the rotations are

6 M  9 M 
θ x = – ---  ----------
- z θ y = – ---  ----------
- z 0≤z≤L (3.37)
7  Ea 3 t 7  Ea 3 t

The rotation-displacement relations are given in eqs. (3.12), and are repeated below.
θ x = – v′ θ y = – u′

Substitute these relations into eqs. (3.37) to get

6 M  9 M 
v′ = ---  ----------
- z u′ = ---  ----------
- z (3.38)
7  Ea 3 t 7  Ea 3 t
Integrate these equations with respect to z to get

3 M  2 9 M  2
v = ---  ----------
- z + c3 u = ------  ----------
- z + c4 (3.39)
7  Ea 3 t 14  Ea 3 t
The constants of indefinite integration are evaluated by the boundary conditions. Since the beam is fixed to the
rigid wall at z = 0, we have v ( 0 ) = 0 and u ( 0 ) = 0 . Substitute these boundary conditions into eqs. (3.39) to
find that c 3 = 0 and c 4 = 0 . Hence the lateral displacements of the beam are

3 M  2 9 M  2
v = ---  ----------
- z u = ------  ----------
- z 0≤z≤L (3.40)
7  Ea 3 t 14  Ea 3 t

A plot of the z-axis in the deformed beam, or z-curve, is shown in Fig. 3.13. In the plot the scaled lateral dis-
placements are defined by

Thin-Walled Structures 81
Axial Normal Stress in Pure Bending and Extension

Ea 3 t Ea 3 t
v = v  ----------2- u = u  ----------2-
ML  ML 

z-curve

0.4

ê0.3
v
0.2
0.1
0 0.6
0
0.4
0.25
ê
u
0.5 0.2
zêL 0.75
0
1

Fig. 3.13 The z-axis in the deformed, zee section beam .

The view of the lateral displacements at the end of the beam are shown in Fig. 3.14.

y
y
M
β = 56.3°

v( L)

x
u( L)

Fig. 3.14 Displacement of the tip of the cantilevered, zee section beam.

82 Thin-Walled Structures
Moments of areas

3.4 Moments of areas


First and second area moments of the cross-sectional area need to be determined before using the flexure for-
mula. The most frequently used methods in the computation of the moments of areas are the parallel axis theo-
rem and the composite body technique. Also, the geometric approximations appropriate for thin-walled sections
are introduced and discussed in the examples of this section.

Parallel Axis Theorem Consider two parallel axes systems


y in the cross section. The origin of the cartesian axes x and y
y y = yc + y coincide with the centroid of the cross-sectional area, which
x = xc + x is labeled C in Fig. 3.15. The second cartesian system x and
C y has its origin at an arbitrary point O, the x axis is parallel
yc x to the x axis, and the y axis is parallel to the y axis. The x
and y system should not be confused the rotated system
introduced earlier in the discussion of the plane of bending
O (Fig. 3.2). The x and y system in this section is not rotated
x relative to the x and y system. The location of the centroid in
xc
the x and y system is denoted by coordinate values
( x c, y c ) . Usually the x and y system is selected as some-
Fig. 3.15 Parallel cartesian axes systems .
thing convenient to start with, and the first and second area
moments with respect to the x and y system are computed or
looked-up in tables. Then the ( x c, y c ) coordinates of the centroid are computed and the parallel axis theorem is
used to find the second area moments in the x and y system.

In the x and y system, the first area moments are defined as

Qx =
∫ ∫ ydA
A
Qy =
∫ ∫ xdA
A
(3.41)

in which the area element d A = d xd y . The relationship between the two parallel coordinate systems is deter-
mined from the location of a generic point in the plane in each system. That is

x = xc + x y = yc + y (3.42)

The relationship between the area elements is d A = dA where dA = dxdy , since the ( x c, y c ) values are
fixed. If eqs. (3.42) are substituted into eqs. (3.41), we get

Q x = yc A + Q x Q y = xc A + Q y (3.43)

where

A =
∫∫
A
dA Qx =
∫ ∫ ydA
A
Qy =
∫ ∫ xdA
A
(3.44)

Since the origin of the x and y system is at the centroid, the first moments Qx and Qy are zero by definition. Set-
ting Q x = 0 and Q y = 0 in eqs. (3.43), we can solve to find the location of the centroid as

Thin-Walled Structures 83
Axial Normal Stress in Pure Bending and Extension

Q Q
x c = ------y y c = ------x (3.45)
A A

In the x and y coordinate system the second area moments are

2 2
I xx =
∫ ∫ y dA
A
I yy =
∫ ∫ x dA
A
I xy =
∫ ∫ xy dA
A
(3.46)

Second area moments are often called moments of inertia in analogy to moments of inertia of mass elements
used in rigid body dynamics. The fact that eqs. (3.46) are second moments of area elements and not mass ele-
ments should be kept in mind even if the terminology “moments of inertia” is used in the context of beam bend-
ing. Now substitute eqs. (3.42) for the x and y coordinates into eqs. (3.46) to get

I xx = y c2 A + 2y c Q x + I xx
I yy = x c2 A + 2x c Q y + I yy (3.47)

I xy = x c y c A + x c Q x + y c Q y + I xy

where the second area moments in the x and y system are defined as

2 2
I xx =
∫ ∫ y dA
A
I yy =
∫ ∫ x dA
A
I xy =
∫ ∫ xy dA
A
(3.48)

Since the x and y coordinates are centroidal, Qx = Qy = 0, and eqs. (3.47) reduce to

I xx = y c2 A + I xx I yy = x c2 A + I yy I xy = x c y c A + I xy (3.49)

Equations (3.43) and (3.47) are the generalized parallel axis theorem, but in problem solving we use eqs. (3.43)
to find the centroid and then the parallel axis theorem reduces to the use of eqs. (3.49). Note that eqs. (3.48) show
that the Ixx and Iyy are always positive in value with dimensional units of L4. The product area moment Ixy can be
positive, zero, or negative in value. The product area moment Ixy is zero if either the x axis or y axis is a axis of
symmetry of the cross section.

Radius of gyration It is common with respect to the second area moments Ixx and Iyy to define radii of gyration
by the definitions

rx = I xx ⁄ A ry = I yy ⁄ A (3.50)

The radii of gyration rx and ry have dimensional units of length. However the radii of gyration do not locate a
physically significant point in the cross section. For example, r x ≠ y c + r x , where r x it the radius of gyration
with respect to the x -axis. (Using the parallel axis theorem, the relation between the radius of gyration about the
x -axis to the x-axis is r x2 = y c2 + r x2 .)

Approximations in thin-walled sections For thin-walled sections, the thickness t of the wall is much smaller
than the largest dimension in the cross section. This geometry permits simplifications with respect to computing
the first and second area moments without loss of significant accuracy with respect to an exact computation. As
an example we model the thin-walled zee section with thin rectangular elements for the web and flanges as is

84 Thin-Walled Structures
Moments of areas

shown in Fig. 3.16 The gap and overlap at the corners are ignored. Instead of an actual drawing of the zee sec-

t t
0 < --- « 1
h

Fig. 3.16 Modeling a thin-walled zee section with rectangular elements for the web and flanges

tion, or a sketch of the mathematical representation as shown in Fig. 3.16, we merely draw the contour of the
section to represent it. The contour consists of the locus of points in the middle of the wall for each branch of the
cross section. The contour for the zee section is illustrated in Fig. 3.11. The term “branch” is used to mean either
a flange or web of the zee section. The slope of the contour is continuous within the branch. At the junctions
between branches, like the web-flange junctions of the zee section, the contour usually exhibits a discontinuous
slope. The contour may be a curve in the x-y plane for more complex sections.

For branches that can be represented by a thin-walled


y ds rectangular area, we can obtain simple formulas for the
s second area moments. Consider a thin rectangular area,
where 0 < t << l, represented by a straight line contour
φ inclined at a angle φ as is shown in Fig. 3.17. The con-
x
tour coordinate is denoted by s, and the area element is
dA = tds . The x and y coordinates of the point s on
t the contour are given by x = s cos φ and y = s sin φ .
l/2 l Hence, the second area moments are computed from
l⁄2
l3t
( s sin φ ) 2 t ds = ------ sin2 φ
Fig. 3.17 A thin rectangular area inclined with
I xx =
∫ 12
–l ⁄ 2
respect to the centroidal coordinates. l⁄2
l3t
( s cos φ ) 2 t ds = ------ cos2 φ
I yy =
∫ 12
(3.51)
–l ⁄ 2
l⁄2
l3t
I xy =
∫ ( s sin φ ) ( s cos φ )t ds = ------ sin φ cos φ
12
–l ⁄ 2

Thin-Walled Structures 85
Axial Normal Stress in Pure Bending and Extension

Composite area technique The composite body technique for


y
1 computing the centroid and second area moments is a method
N applicable to cross-sectional areas that can be subdivided into
simple geometric shapes whose properties are known. For our
yi 2
example of the thin-walled zee section, we can subdivide it into
yi
xi three rectangular areas and use the formulas in eqs. (3.51) for
Ci
the second area moments about the local centroidal axis system
i-th sub-area in each rectangle. With the centroid and second area moments
known for each sub-area, the parallel axis theorem is used to
x transfer their properties to a parallel reference system. Then a
xi
summation of the first and second moments for each sub-area
about the reference axis system determines the total properties
Fig. 3.18 Composite body technique for the cross section. The pertinent equations for an area subdi-
vided into N sub-areas labeled i = 1, 2, …, N (see Fig. 3.18)
are

N N N

A = ∑ Ai Ay c = ∑ yi Ai Ax c = ∑ xi Ai (3.52)
i = 1, 2, … i = 1, 2, … i = 1, 2, …

I xx = ∑ ( I xx + y 2 A ) i (3.53)
i = 1, 2, …

I yy = ∑ ( I yy + x 2 A ) i (3.54)
i = 1, 2, …

I xy = ∑ ( I xy + xy A ) i (3.55)
i = 1, 2, …

where A i, x i, y i, ( I xx ) i, ( I yy ) i, ( I xy ) i are the known properties of the i-th sub-area. The overall centroid of the area
is computed from the last two formulas in eqs. (3.52). The second area moments for the reference system deter-
mined from eqs. (3.53) to (3.55) are then used with the coordinates of the overall centroid to compute the second
area moments about the centroidal x and y axes by the parallel axis theorem. This method is best illustrated by
example.

EXAMPLE 3.3 Thin-walled zee section properties by the composite body technique

Determine the centroid and the second area moments for the thin-walled zee section of Example 3.1. The section
is sub-divided into three rectangular branches as shown in Fig. 3.19. One branch corresponds to the web and two
branches correspond to the flanges.

Solution First we find the centroid. Equations (3.52) are represented in the table shown below. Summation of the
appropriate columns gives

A = 4at xc A = 0 y c A = 4a 2 t

86 Thin-Walled Structures
Moments of areas

y y, y
y3
t = thickness of all branches
x3

a⁄2
a
a
y2
C
x2 x

a y1 yc = a

x1
x x
O a⁄2 O
a

Fig. 3.19 Zee section approximated by three rectangular areas.

so that the centroid has coordinates x c = 0 and y c = a .

The second area moments are computed for the reference coordinate system ( x, y ) using the second table
shown below. Note that for the local centroidal coordinate systems in each branch we can identify the angle φ in

i Ai xi yi xi Ai yi Ai
1 at a/2 0 a2t/2 0
2 2at 0 a 0 2a2t
3 at – a/2 2a – a2t/2 2a2t
Sum 4at 0 4a2t
eqs. (3.51) as φ 1 = 0° , φ 2 = 90° , and φ 3 = 0° . These values of the angle φ in each branch are used to com-
pute the local second area moments in each rectangular branch via eqs. (3.51). From the summation of the col-

i y i2 A i ( I xx ) i x i2 A i ( I yy ) i xi yi Ai ( I xy ) i
1 0 0 a3t/4 a3t/12 0 0
2 (a2)2at (2a)3t/12 0 0 0 0
3 (2a)2at 0 a3t/4 a3t/12 – a3t 0
Sum 6a3t 2a3t/3 a3t/2 a3t/6 – a3t 0

umns, the second area moments in the ( x, y ) system via eqs. (3.53) to (3.55) are

Thin-Walled Structures 87
Axial Normal Stress in Pure Bending and Extension

I xx = 6a 3 t + 2a 3 t ⁄ 3 = ( 20a 3 t ) ⁄ 3
I yy = ( a 3 t ) ⁄ 2 + ( a 3 t ) ⁄ 6 = ( 2a 3 t ) ⁄ 3
I xy = – a 3 t
Now we use the parallel axis theorem to transfer these moments to the centroidal system. Equations (3.49) give

20 8
I xx = I xx – y c2 A = ------ a 3 t – a 2 ( 4at ) = --- a 3 t
3 3

2 2
I yy = I yy – x c2 A = --- a 3 t – ( 0 ) ( 4at ) = --- a 3 t
3 3

I xy = I xy – x c y c A = – a 3 t – ( 0 ) ( a ) ( 4at ) = – a 3 t

These are the second area moments given the problem statement of Example 3.1.

EXAMPLE 3.4 Semicircular section with two stringers

The beam cross section consists of a thin-walled semicircular web of radius a and thickness t, and two stringers
each with area A s = ( πat ) ⁄ 2 . The section and the reference coordinate system are shown Fig. 3.20. Determine

y y y

As As
ds
s
θ O O
x x, x
C
a

t As As
Fig. 3.20 Stringer-stiffened, semicircular section.

the location of the centroid and the second area moments.

Solution Since this cross section is symmetric with respect to the x axis, the centroid is located on this axis, and
the product area moment I xy is zero. In thin-walled construction the stringer’s cross-sectional dimensions are
small with respect to the largest dimension of the cross section. Hence, the stringer is modeled by its area As con-
centrated at the stringer’s centroid. Also, the second area moments of the stringer area are neglected with respect
to the transfer terms in the parallel axis theorem. It is convenient to use the polar angle θ in the moment compu-
tations for the web, where – π ⁄ 2 ≤ θ ≤ π ⁄ 2 , as is shown in the figure. The differential area of the web is
dA = tds = tadθ , and the coordinates to this differential area are x = – acos θ and y = asin θ . The
cross-sectional area is

88 Thin-Walled Structures
Extension, pure bending, and thermal effects for multi-material beams

π
---
2
π
= πat + 2  --- at = 2πat
A =
∫ ta dθ + 2 A s 2 
π
– ---
2

From the second of eqs. (3.41), the first area moment about the y axis is
π⁄2

( – acos θ )ta dθ = – 2a 2 t
Qy =

–π ⁄ 2

and from the second of eqs. (3.43) with Q y = 0 by the definition of the centroid we find x c A = – 2a 2 t . Thus,
the centroid is located at
– 2a 2 t a
x c = -------------- = – --- yc = 0
2πat π

The second area moment about the x axis is (see the first of eqs. (3.46))

π⁄2 π⁄2
θ 1 3
( asin θ ) 2 ta dθ + a 2 A s + ( – a ) 2 A s = a 3 t  --- – --- sin 2θ + πa 3 t = --- πa 3 t
I xx =
∫ 2 4 
–π ⁄ 2
2
–π ⁄ 2

and the second area moment about the y axis is

π⁄2 π⁄2
θ 1 π
( – acos θ ) 2 ta dθ + 2 ( 0 ) 2 A s = a 3 t  --- + --- sin 2θ = --- a 3 t
I yy =
∫ 2 4 
–π ⁄ 2
2
–π ⁄ 2

Now we use the parallel axis theorem, eqs. (3.49), to get the second area moments about the centroidal system.

3 3
I xx = I xx – y c2 A = --- πa 3 t – ( 0 ) 2 A = --- πa 3 t
2 2

π a 2 π 2
I yy = I yy – x c2 A = --- a 3 t –  – --- 2πat = a 3 t  --- – ---
2  π  2 π

To summarize, the second area moments about the centroidal system are

I xx = 4.712a 3 t I yy = 0.934a 3 t I xy = 0

3.5 Extension, pure bending, and thermal effects for multi-material


beams
Strain-displacement Extensional deformation of an element of the beam is shown in Fig. 3.21. The axial nor-
mal strain due to extension is given by

Thin-Walled Structures 89
Axial Normal Stress in Pure Bending and Extension

dz dz + dw
z
z w + dw
w(z)

Fig. 3.21 Extensional deformation of a element of the beam

( dz + dw ) – dz
ε z = ------------------------------------ which in the limit as dz → 0 gives ε z = w′
dz
For combined bending and extension, we superpose the extensional strain and the bending strain, eq. (3.15), to
obtained the total strain. Hence,

ε z = w′ + xθ′ y + yθ′ x
{






extension bending (3.56)

Material law Consider Hooke’s law for the case of thermal strain. A change in temperature causes an expan-
sion of an unrestrained beam element without an associated mechanical stress. If mechanical stresses are present
then the total strain is composed of the sum of the mechanical strain and the thermal strain. That is,
εz = σz ⁄ E + α ( T – T 0 )






mechanical thermal (3.57)

in which T denotes the temperature, T 0 the stress-free temperature, α the coefficient of linear thermal expan-
sion (CTE) of the material, and E denotes the modulus of elasticity of the material. If the temperature is given in
degrees fahrenheit ( °F ), then the CTE has dimensional units 1 ⁄ °F . In the material law for the beam given by
eq. (3.57), we have assumed that the lateral stresses σ x and σ y are negligible with respect to the axial normal
stress σ z and that the thermal strains in the x- and y-directions can be neglected. Solving eq. (3.57) for the axial
normal stress we get
σ z = Eε z – Eα∆T (3.58)

where
∆T = T – T 0 (3.59)

In general the material properties E and α can be a function of the temperature as well. However, we assume
that over moderate temperature changes these material properties do not vary significantly from an average value.

In addition to the thermal effects, we assume that the beam cross section is hetrogeneous; i.e., the material
properties E and α are functions of coordinates x and y. In the case of heterogeneity, we define a reference mod-
ulus of elasticity E 0 , which is usually selected as some convenient positive value in problem solving. The refer-
ence modulus is independent of x and y. We write eq. (3.58) as

90 Thin-Walled Structures
Extension, pure bending, and thermal effects for multi-material beams

E
σ z = E 0  ------ ε z – Eα∆T (3.60)
 E 0

Resultants We substitute the strain-displacement relation,eq. (3.56), into eq. (3.60) for the axial normal strain to
get

E
σ z = E 0  ------ ( w′ + xθ′ y + yθ′ x ) – Eα∆T (3.61)
 E 0

Recall that the beam resultants, first given by eqs. (3.20), are

N =
∫ ∫ σ dA
A
z Mx =
∫ ∫ yσ dA
A
z My =
∫ ∫ xσ dA
A
z

Now substitute eq. (3.61) for the axial normal stress in these beam resultants to get the material law for the beam.
We write the final expression of the material law for the resultants in the matrix form

N A * Q x* Q y* w′ Nt
M x = E 0 Q x* I xx
* I*
xy
θ′ x – M xt (3.62)

My Q y* I xy
* I*
yy
θ′ y M yt

where the modulus weighted area is


E
A* =
∫ ∫ -----
A
E
- dA
0
(3.63)

the modulus weighted first area moments are

E E
Q x* = Q y* =
∫ ∫ -----
A
E
- ydA
0
∫ ∫ -----
A
E 0
- xdA (3.64)

the modulus weighted second area moments are

* = E 2 * = E 2
I xx
∫ ∫ -----
A
E 0
- y dA I yy
∫ ∫ -----
A
E 0
- x dA (3.65)

the modulus weighted product area moment is

* = E
I xy
∫ ∫ -----
A
E
- xy d A
0
(3.66)

the thermal axial force is

Nt =
∫ ∫ Eα∆T dA
A
(3.67)

and the thermal bending moments are

M xt = M yt =
∫ ∫ y ( Eα∆T ) dA
A
∫ ∫ x ( Eα∆T ) dA
A
(3.68)

Thin-Walled Structures 91
Axial Normal Stress in Pure Bending and Extension

We take the origin of the x-y coordinate system at the modulus-weighted centroid. The modulus-weighted
centroid is characterized by the fact that the modulus-weighted first area moments vanish; i.e.,

Q x* = Q y* = 0 (3.69)

We will show how to use this property to find the modulus-weighted centroid in the next example. With the ori-
gin of the x-y coordinates at the modulus-weighted centroid, the material law for the resultants, eq. (3.62),
reduces to

N A * 0 0 w′ Nt
* I*
M x = E 0 0 I xx t
xy θ′ x – M x (3.70)

My * I*
0 I xy yy
θ′ y M yt

By locating the origin at the modulus-weighted centroid we de-couple the extensional response of the beam from
its bending response.

Lastly we write the normal stress σ z given by eq. (3.61) in terms of the resultants. Invert eq. (3.70)) and
solve for the extensional strain w′ , and curvatures θ′ x and θ′ y . to get

1
w′ = ------------* ( N + N t ) (3.71)
E0 A

θ′ x * –I *
I yy t
1 xy M x + M x
= ---------------------------------------
* I * – I * 2)
- (3.72)
θ′ y E 0 ( I xx yy xy * I*
– I xy M y + M yt
xx

Now substitute eq. (3.71) for w′ and eqs. (3.72) for θ′ x and θ′ y into eq. (3.61) to eliminate these terms in the
equation for σ z . The equation for the normal stress in the form

σ z = σ zm + σ zt (3.73)

where σ zm is the mechanical portion of normal stress and σ zt is the thermal portion of the normal stress. The for-
mulas for these components are
* M – I * M )x
( I xx * M – I * M )y
( I yy
E y xy x x xy y
σ zm = ------ N + -------------------------------------------
- + ------------------------------------------- (3.74)
E0 * *
( I xx I yy – I xy ) *2 * I * – I *2 )
( I xx yy xy

* M t – I * M t )x
( I xx * M t – I * M t )y
( I yy
E y xy x x xy y
σ zt = ------ N t + -------------------------------------------
- + -------------------------------------------
- – Eα∆T (3.75)
E0 * *
( I xx I yy – I xy ) *2 * I * – I *2 )
( I xx yy xy

In eqs. (3.74) and (3.75) the modulus E and coefficient of thermal expansion α are functions of x and y in gen-
eral.

EXAMPLE 3.5 A multi-material beam with a symmetric cross section

A beam of three materials is subjected to a positive bending moment M x only. The maximum tensile normal
stress is 35 ksi. Find the maximum compressive normal stress.

92 Thin-Walled Structures
Extension, pure bending, and thermal effects for multi-material beams

2in 1
Mx E 1 = 35msi
x
E 2 = 20msi
2in 2
E 3 = 5msi
y c*
y 1msi = 10 6 psi
2in 3
x
1in

Fig. 3.22 Bending of a multi-material beam with a symmtric cross section

Solution In this example we take the reference modulus E 0 = 10Msi , so that E 1 ⁄ E 0 = 3.5 , E 2 ⁄ E 0 = 2 , and
E 3 ⁄ E 0 = 0.5 . The modulus weighted area from eq. (3.63) is

3
E Ei
A* =
∫∫ ------ d A =
E0 ∑ -----
E
-A
0
i = 3.5 × 2in 2 + 2 × 2in 2 + 0.5 × 2in 2 = 12in 2
A i=1

Note that the modulus weighted area is twice the geometric area. We choose the x - y system centered at the bot-
tom of the beam as shown in Fig. 3.22. Since the y -axis is an axis of symmetry, both in geometry and in mate-
rial properties, the modulus weighted centroid lies on this axis. We need only to find the location on the y -axis of
the modulus weighted centroid; i.e., y c* . Begin by calculating the modulus weighted first area moment about the
x -axis. See eqs. (3.64). That is,

3
E Ei
Q x* =
∫∫ ------ y d A =
E0 ∑ -----
E
-y A
0
i i = 3.5 × 5in × 2in 2 + 2 × 3in × 2in 2 + 0.5 × 1in × 2in 2
A i=1

so

Q x* = 48in 3
We now use the parallel axis theorem for the first area moment as given by eqs. (3.43), but generalized to the
multi-material case to write

Q x* = Q x* – y c* A * = 0 for the modulus weighted centroid

Hence

Q* 48in 3
y c* = ------*x = ------------2- = 4in
A 12in
Note that the modulus weighted centroid is located at the intersection of materials 1 and 2, which does not coin-

Thin-Walled Structures 93
Axial Normal Stress in Pure Bending and Extension

cide with the geometric centroid of the cross-sectional area.

Now that the modulus weighted centroidal coordinates x-y are located, the modulus weighted second area
moments are calculated. Again, the y-axis is an axis of symmetry, both in geometry and in material properties, so
* = 0 . Since
that the modulus weighted product area moment about the x-y axes vanishes; i.e., I xy
* = 0 , the formula for the axial normal stress, eqs. (3.73) and (3.74), reduces to
N = M y = I xy

E Mx
σ z = ------ ------
*
-y 2in ≤ y ≤ – 4in
E 0 I xx

* . This computation proceeds as follows


Thus, we only need to compute I xx

3 3 3
E E E Ei
* = ------ y 2 d A = ∑ ∫∫ ------i y 2 d A = ∑ ∫∫ ------i y2dA = ∑ ----- + y 2 A )i
I xx
∫∫
A
E0
i = 1 Ai
Eo
i=1
Eo
Ai
E
i=1
-(I
o
xx

In achieving this result we used the fact that the (geometric) second area moment of the i-th area element about
2 + y i2 A i , which is based on the composite body technique and parallel axis theorem.
the x-axis is
∫ ∫ y dA = I
Ai
xxi

See eq. (3.53). Performing the computations we have

1
* = 3.5 ×  ----- 1
I xx - × 1 × 2 3 + ( 1 ) 2 × 2 + 2 ×  ------ × 1 × 2 3 + ( – 1 ) 2 × 2 +
 12   12 
1
0.5 ×  ------ × 1 × 2 3 + ( – 3 ) 2 × 2 = 24in 4
 12 

Only material 1 is in tension for M x > 0 . So the axial normal stress is a maximum in material 1 at
y = 2in ;i.e., σ z = 35ksi Thus,

Mx
35ksi = 3.5 × ------------4- × 2in
24in
3
Hence M x = 120 ×10 lb-in . The maximum compressive normal stress in material 2 is

E2 M x
- ( – 2in ) = 2 × 5000lb/in 3 × ( – 2in ) = – 20ksi
σ z = ------ ------
*
E 0 I xx

The maximum compressive normal stress in material 3 is


E3 M x
σ z = ------ ------
*
- ( – 4in ) = 0.5 × 5000 × ( – 4 ) = – 10ksi
E 0 I xx

Therefore, the magnitude of the maximum compressive normal stress is 20ksi.

As a function of y, the bending normal stress is given by

94 Thin-Walled Structures
Extension, pure bending, and thermal effects for multi-material beams

3
 17, 500 lb/in × y 0 < y ≤ 2in
σ z =  10, 000 lb/in 3 × y – 2in < y < 0

 2, 500 lb/in 3 × y – 4in ≤ y < – 2in
and this is plotted in Fig. 3.23. Note that the bending normal stress jumps at the material interfaces since the
modulus is discontinuous, but is linear in y within a material layer. The axial normal strain in terms of the kine-
matic quantities w′ , θ x ′ , and θ y ′ is given by eq. (3.56). For this problem N = N t = 0 , so the w′ = 0 from eq.
(3.71); i.e., there is no axial extensional component to the strain. Also, M y = M yt = M xt = I xy
* = 0 , which

* ) and θ ′ = 0 . From eq. (3.56), the axial strain distribution in y -


gives from eq. (3.72) that θ x ′ = M x ⁄ ( E 0 I xx y
coordinate is

Mx 5000 lb/in 3
ε z = yθ x ′ = y ------------
* 6
- = ( 500 × 10 –6 in –1 )y
- = y ------------------------------ – 4in ≤ y ≤ 2in
E 0 I xx 10 ×10 lb/in 2

The maximum tensile strain is 1000 × 10 –6 in/in = 1000µε at y = 2 in, and the maximum compressive strain is
– 2000 × 10 –6 in/in = – 2000µε at y = -4 in. (The quantity µε is read as microstrain.) The strain is continuous in
the coordinate y as is shown in Fig. 3.23.

y y, in
y, in

2 2
2in 1 Mx > 0
x σ z, ksi ε z, µε
– 20 – 10 – 5 – 2000
35 0 1000
2in 2
–2 –2
2in 3
–4 –4
1in

Fig. 3.23 Distribution of normal stress and strain through the thickness for the multi-material beam

Thin-Walled Structures 95
Axial Normal Stress in Pure Bending and Extension

3.6 Problems
y, y 1.For the thin-walled Y-section shown at right, the
branch thickness t is much smaller that the branch
45° 45° length l.

a)Determine the location y c of the centroid.

C x b)Determine the second area moment I xx of the


cross section about the x axis.
yc
O c)If the bending moment components Mx > 0 and My
x
l l = 0, determine the magnitudes of the largest tensile
and compressive normal stress due to bending. State
Each of the five branches: length l and thickness t the coordinates where they occur.

2. For the thin-walled S-section shown at right the


radius a = 5mm, the wall thickness t = 0.64mm, and
y the bending moment components are Mx = 3500
Nmm, My = 0. Determine the angle β of the neutral
t axis in the cross section, and the magnitude of the
a
maximum bending normal stress.

4a C
x
3.Half of the cross section of a ship is shown on the
next page. Only the material that is effective in the
β
longitudinal bending is illustrated in the figure.
a neutral axis in the cross Determine the area A, location of the centroid y c ,
t section
and the second area moment about the x-axis (
I xx )for the full section. Use the tabular format for
the computations as given in Example 3.3. All plat-
ing has a thickness t = 14 mm unless other wise noted. The descriptions of the numbered structural elements
shown in the figure are given in the table.
item # Description
1 outer bottom
2 Inner bottom
3 Center girder
4&5 Side girders
6 Bilge (curved portion)
7 Side plating
8 Second deck plating
9 & 11 Hatch side girders L500 × 400 × 25
10 Strength deck plating

96 Thin-Walled Structures
Problems

4m

10

11

7
9m

8
9

x
C
5.5 m

yc
2
5 4 3
6
t/2 x
R=1m 1 3.5 m

6.5 m

L500 × 400 × 25

π
R A = --- Rt
2R ⁄ π 2 500mm
C π 1 4 C x
t x I xx = --- R 3 t  --- – -----
2  2 π 2 139mm

400mm
A = 0.0225m 2
–4
I xx = 6.077 ×10 m4

Thin-Walled Structures 97
Axial Normal Stress in Pure Bending and Extension

98 Thin-Walled Structures
CHAPTER 4 Axial Force, Shear Force
and Bending Moment
Diagrams

The following three methods to construct equilibrium shear force and bending moment diagrams for beams are
described
• the method of sections
• the differential equation method, and
• the semi-graphical approach.

In actual practice the distinction between these methods becomes blurred, because they may be used simul-
taneously by the analyst. In the last section of this chapter the buoyancy force distribution on ships is described
so that the longitudinal bending response can be computed.

4.1 Method of sections


If an aerospace or ocean vehicle structure is assumed to behave as a slender bar built-up from many members,
then in order to determine stresses in the members we follow the approach in mechanics of materials and first
determine the distribution of the internal forces and moments along the length. The objective of this section is to
review how to determine these internal actions and plot them along the length of the structure. The relationship
between the internal forces and moments to the stresses will be discussed in subsequent chapters. The review is
limited to statically determinate problems such that internal forces and moments may be obtained by the equa-
tions of statics alone.

Consider a straight slender bar with uniform cross section, which has a length 3c, and is simply supported at
two locations as shown in Fig. 4.1. In a right-handed Cartesian coordinate system (x,y,z), take the z-axis is paral-
lel to the length, the y-axis in the plane of the figure, and the x-axis perpendicular to the plane of the figure. The
origin is at the left end of the bar; 0 ≤ z ≤ 3c . The bar is in equilibrium subject to the external loads F and p y ( z )
which act in the z-y plane. Neglect the weight of the bar relative to the loads F and p y ( z ) , as is frequently done
in structures where the applied loads are much greater than the weight of the structure. The load F is a point
force acting at the left end of the bar, and is replaced by its horizontal and vertical components Fz and Fy. The

Thin-Walled Structures 99
Axial Force, Shear Force and Bending Moment Diagrams

F Fy py( z)

z
Fz

c 2c

Fig. 4.1 A slender bar simply supported at two locations and subjected to force F and
distributed load p y ( z ) .

load p y ( z ) is a distributed load and has units of force per unit length. It is assumed positive if it acts vertically
upward (positive y-direction), and negative if it acts downward. At a typical value for z we want to find the inter-
nal axial force N(z), shear force Vy(z), and bending moment Mx(z). We write the internal actions as N(z), Vy(z),
and Mx(z), since they are mathematically functions of the coordinate z. The basis for their determination is equi-
librium.

Free-body diagrams of the bar removed from its supports and by imagining the bar is cut at some value of z
are shown in Fig. 4.2. The internal actions N, Vy, and Mx are shown as well as the unknown support reactions Ay,
Bz, and By. Let us assume the distributed load p y ( z ) vanishes in this discussion. Internal actions N, Vy, and Mx
are shown in their assumed positive senses. A positive z-face has its outward normal in the positive z-direction,
and a negative z-face has its outward normal in the negative z-direction. That is, on a positive z-face a positive
axial force N acts in a positive z-direction, a positive shear force Vy acts in the positive y-direction, and the posi-
tive moment Mx acts clockwise, or as a vector in the positive x-direction by the right-hand screw rule. By New-
ton's third law on action/reaction, positive values of N, Vy, and Mx acting on a negative z-face have a sense
opposite to their positive values on the positive z-face. The sign convention for the internal forces and moments
needs be carefully followed.

The usual procedure to determine N(z), Vy(z), and Mx(z), is to first draw an overall free-body diagram of the
bar removed from its supports and find the unknown support reactions, and then section the bar at various z-loca-
tions to find N, Vy, and Mx. The reader should verify that the support reactions in Fig. 4.2 are A y = ( – 3 ⁄ 2 )F y ,
B z = F z , and B y = F y ⁄ 2 , where Fy and Fz are the known applied loads. Note that the force Ay has a sense
opposite to what was originally assumed.

The overall equilibrium free-body diagram of the bar is shown at the top of Fig. 4.3. The diagrams shown
from top to bottom directly below the overall free-body diagram are the axial force diagram, the shear force dia-
gram, and the bending moment diagram, respectively. The axial force diagram is obtained by sectioning the bar
at any value of z, 0 < z < 3c, placing a positive internal force N on the cut faces, and then summing forces in the
z-direction for one of the two free-body diagrams. Thus, N(z) is a constant, for all z, 0 < z < 3c. In a similar man-
ner one obtains Vy(z) and Mx(z). Note that the shear force has a discontinuity at z = c since a point force acts
there. It is necessary to consider free-body diagrams in two separate ranges of z to draw the shear force diagram;
0 < z < c, and c < z < 3c. The magnitude of the jump in the shear force is equal to the magnitude of the point

100 Thin-Walled Structures


Differential equation method

Fy

Fz Bz

Ay By
c 2c

overall FBD
Fy
Vy

Fz N Bz

z Mx Mx
Ay By
Vy
z 2c

Left and right FBD’s for cuts in the range 0 < z < c

Fig. 4.2 Free body diagrams (FBDs) of the slender bar with p y ( z ) = 0.

force: V y ( c + ) – V y ( c - ) = ( 3 ⁄ 2 )F y . The shear force is a piecewise constant for this problem. The bending
moment is piecewise linear; i.e., Mx(z) = z Fy, 0 < z < c, and Mx(z) = (3c - z)(Fy/2), c < z < 3c. The bending
moment is continuous at z = c, but has a discontinuity in slope M' x ( c + ) – M' x ( c - ) = ( 3 ⁄ 2 )F y , where
M' x = d M x ⁄ dz . As illustrated in Fig. 4.2, there are two free-body diagrams for each cut. Either one may be
used to find N, Vy, and Mx. If the left free-body diagram is used to find N, Vy, and Mx, then equilibrium of the
right free-body diagram will give the same values of N, Vy, and Mx. If it does not, there is either a math error, the
sign convention is violated, or overall equilibrium is in error. Sketching the axial force, shear force, and bending
moment diagrams by the method illustrated in this section is called the method of sections.

4.2 Differential equation method


The distributed load intensity p y ( z ) , the shear force Vy(z), and bending moment Mx(z) are related by simple dif-
ferential equations at z, if no point forces or concentrated couples act at z. These differential equations are useful
in the construction of the shear force and bending moment diagrams.

Thin-Walled Structures 101


Axial Force, Shear Force and Bending Moment Diagrams

Fy Fy/2
Fz z
c 2c
3/2 Fy
N
Fz

0 z

Vy

Fy/2
0 z

- Fy Vy

N + N
Mx Mx Mx
Vy

0 z

- cFy

Fig. 4.3 Axial force N, shear force Vy, and bending moment Mx diagrams
for the slender bar with p y ( z ) = 0,

Consider a portion of a straight beam subjected to distributed load whose intensity is p y ( z ) as shown in Fig.
4.4. By convention p y ( z ) is positive upwards and negative downwards. A free-body diagram of a portion ∆z-
long of the beam is shown in Fig. 4.5. The shear force and bending moment change with z, and so their values at
z + ∆z are different from their values at z. The distributed load acting on the segment ∆z is replaced by single
force of magnitude p y ∆z acting long a line of action given by z = z*, where z < z* < z + ∆z. Mathematically this
is permissible by the mean value theorem (for integrals) for continuous functions of a single variable, and from
this theorem we have

102 Thin-Walled Structures


Differential equation method

( z + ∆z )
1
p y = ------
∆z ∫ p y ( ζ ) dζ
z

Note that in the limit as ∆z → 0 , z * → z , and p y → p y .

py( z)
y

Fig. 4.4 Distributed load intensity shown acting in the positive sense

py( z)

Vy + ∆Vy

Fig. 4.5 Mx

Vy Mx + ∆Mx
z
z + ∆z

p y ( z * )∆z

Vy + ∆Vy
Fig. 4.6
Mx
Vy Mx + ∆Mx
z
z*
z + ∆z

In the free-body diagram of Fig. 4.6 we sum forces vertically, divide by ∆z, and take the limit as ∆z → 0 to
get

dV y
= – py (4.1)
dz
Often it is more convenient to use an integrated form of eq. (4.1). Integrating it from z1 to z we obtain

Thin-Walled Structures 103


Axial Force, Shear Force and Bending Moment Diagrams


V y ( z ) = V y ( z 1 ) – p y ( ζ ) dζ (4.2)
z1

The integrated form is valid only if there are no point loads between z1 and z. If we sum moments at z in the free-
body diagram, divide by ∆z, and take the limit as ∆z → 0 we get

dM x
= Vy (4.3)
dz
The integrated form of eq. (4.3) is
z


M x ( z ) = M x ( z 1 ) + V y ( ζ ) dζ (4.4)
z1

which is valid if no point couples act between z1 and z. Equations (4.1) and (4.3) are the differential equations of
equilibrium for the beam. Equation (4.1) is valid at z if no point force acts there, and eq. (4.3) is valid at z if no
point couple acts there. Considering separately the equilibrium of a point force F0 and a point couple with
moment magnitude C0 acting at z = z0, as shown in Fig. 4.7, we obtain

F0

y C0

V y ( z 0+ )
z V y ( z 0- ) ε

M x ( z 0+ )
M x ( z 0- )
z0

Fig. 4.7 Point force and couple acting at z0 on an inifintesimal beam element

V y ( z 0+ ) – V y ( z 0- ) = – F 0 M x ( z 0+ ) – M x ( z 0- ) = – C 0 (4.5)

Distributed loads may be replaced by a resultant force acting at the center of pressure. This procedure is con-
venient in many situations. For example, the air load distribution on a wing may be replaced by lift and drag
forces acting at the center of pressure. A segment of a beam from z = z1 to a typical value of z, z > z1, which has a
distributed load with intensity p y ( z ) acting on it is shown in Fig. 4.8. Using ζ as a dummy variable to measure
the axial position, the resultant force F y ( z ) is

F y(z) =
∫ p ( ζ ) dζ
y (4.6)
z1

and the center of pressure z p ( z ) is given by

104 Thin-Walled Structures


Differential equation method

py(z)
F y(z)
V y ( z1 ) V y(z) V y ( z1 ) V y(z)

M x ( z1 ) M x(z) M x ( z1 ) M x(z)
z1 z1

ζ
z p(z)
z z

Fig. 4.8 Resultant force at the center of pressure statically equivalent to a distributed load.

z p ( z )F y =
∫ ζ p ( ζ ) dζ
y (4.7)
z1

Both the resultant force F y ( z ) and center of pressure z p ( z ) are functions of z. Equations (4.6) and (4.7) are con-
ditions of statical equivalence for the distributed load p y ( z ) .

EXAMPLE 4.1 Cantilever wing with tip tank

Consider the cantilever wing with tip tank as shown in Fig. 4.9. Given the weight of the tip tank and its contents

z
p y ( z ) = p 0 ---
L

z
e

L
W

Fig. 4.9 Cantilever wing with tip tank.

W, the distance e of the weight W from the wing tip, the wing span L, and the value of the distributed load inten-
sity p 0 at the wing root, determine the shear force and bending moment along the span. The solution to this
problem is given by a Mathematica 4.0 program listed below.

Thin-Walled Structures 105


Axial Force, Shear Force and Bending Moment Diagrams

1 2 3 4 5 6

ü Example 4.1
Shear force and bending moment diagrams for a cantilever wing with tip tank.

Input the distributed load function.


z
py = p0 ÅÅÅÅ ;
L
The shear force and bending moment distributions are determined from eqs. (4.2) and
(4.4) with z1 = 0. The shear force and bending moment at the wing tip (z = 0) are
denoted by Vy0 and Mx0 , respectively.

Vy = Vy0 - ‡ py „ z

z2 p0
- ÅÅÅÅÅÅÅÅÅÅÅÅ
Å + Vy0
2L

Mx = Mx0 + ‡ Vy „ z

z3 p0
Mx0 - ÅÅÅÅÅÅÅÅÅÅÅÅ
Å + z Vy0
6L
Boundary conditions at the wing tip, obtained from equilibrium of the tip tank, deter-
mined the shear force Vy0 and moment Mx0 .
bc1 = HVy ê. z Ø 0L - W
bc2 = HMx ê. z Ø 0L - e W
-W + Vy0
-e W + Mx0
slv1 = Solve@bc1 == 0, Vy0 D
88Vy0 Ø W<<
Vy0 = Vy0 ê. slv1@@1DD
W
slv2 = Solve@bc2 == 0, Mx0 D
88Mx0 Ø e W<<
Mx0 = Mx0 ê. slv2@@1DD
eW

106 Thin-Walled Structures


Differential equation method

Print@ "Shear force Vy HzL = ", Vy D


Print@"Bending moment Mx HzL = " Mx D

Shear force Vy HzL = W - ÅÅÅÅÅÅÅÅÅÅÅÅ


z2 p0
Å

Bending moment Mx HzL = i


z3 p0 y
2L

j Åz
je W + W z - ÅÅÅÅÅÅÅÅÅÅÅÅ z
k 6L {
Plot the shear force and bending moment diagrams for the following parameter
values: L = 144 in, p0 = 70 lb/in, W = 500 lbs, and e = 6 in. (Plots labled p1 and
p2 have been suppressed using the DisplayFunction option.)

Plot@HVy ê. 8 L Ø 144, p0 Ø 70, W Ø 500, e Ø 6<L,


p1 =

8z, 0, 144<,
PlotRange Ø 81000, -5000<,

AxesLabel Ø 8"z,inches", "Vy , lbs"<,


GridLines Ø Automatic,

PlotLabel Ø
StyleForm@"Shear force diagram",
"Section"D,
DisplayFunction -> IdentityD
Ü Graphics Ü

Plot@HMx ê. 8 L Ø 144, p0 Ø 70, W Ø 500, e Ø 6<L,


p2 =

8z, 0, 144<,

AxesLabel Ø 8"z,inches", " Mx , lb-in"<,


GridLines Ø Automatic,

PlotLabel Ø
StyleForm@"Bending moment diagram",
"Section"D,
DisplayFunction -> IdentityD
Ü Graphics Ü
Show@GraphicsArray@88p1<, 8p2<<DD
(The plots are are shown on the next page.)
1 2 3 4 5 6

The magnitude of the bending moment is largest at the wing root, and this vlaue is
important for wing structural design. Its value in lb-in is
Mx ê. 8 L Ø 144, p0 Ø 70, W Ø 500, e Ø 6, z Ø 144<
-166920

Thin-Walled Structures 107


Axial Force, Shear Force and Bending Moment Diagrams

1 2 3 4 5 6

Vy , lbs Shear force diagram


1000

z,inches
20 40 60 80 100 120 140
-1000

-2000

-3000

-4000

-5000

Mx , lb-in Bending moment diagram

10000

z,inches
20 40 60 80 100 120 140
-10000
-20000
-30000
-40000
-50000

Shear force and bending moment diagrams for the cantilever wing with tip tank

108 Thin-Walled Structures


Differential equation method

The shear force and bending moment at the wing tip are determined from equilibrium of the tip tank, which
gives Vy(0) = W and Mx(0) = eW. Note that eq. (4.1) shows that the slope on the shear diagram is equal to the
distributed load intensity. For example at z = 0, p y ( 0 ) = 0 , so that the slope of the shear diagram is zero at z = 0.
Similarly, the slope on the moment diagram d M x ⁄ dz is equal to the shear force as given by eq. (4.3). In particu-
lar, at z = 45.36 inches the shear force is zero. Thus, in the bending moment diagram the moment is stationary at
z = 45.36 in (i.e., it has a horizontal slope), and Mx may be either a local maximum, minimum, or a value corre-
sponding to horizontal inflection point. It is important to compute the largest magnitude of the bending moment,
and this is accomplished by checking the bending moments where Vy = 0, at the end points of the beam, and the
locations where the bending moment is discontinuous. Note that the maximum bending moment magnitude
occurs at the wing root, where M x = – 166920 lb-in . The bending moment changes sign, and hence vanishes, at
z = 81.4 inches. In a plot of the beam deflection versus z, which is not shown above, the location z = 81.4 inches
is called an inflection point because the curvature is changing from concave down (positive Mx) to concave up
(negative Mx) as z increases through z = 81.4 inches.

EXAMPLE 4.2 The air load acting on a wing given as discrete data.

The problem statement and data for this example is taken from the aircraft structures text by Peery (1950). How-
ever, the notation is changed to that of the this text, and the solution is given in terms of a Mathematica 3.0 pro-
gram. Since the air load on the wing is given at discrete spanwise locations and not as a mathematical function, it
is useful to use Mathematica’s list manipulation capabilities to effect the solution. Before giving the problem
statement, we will discuss some aspects of list manipulations.

Lists provide a mechanism for representing arrays, vectors, matrices, and for grouping together objects such
as data, variables, or expressions. A list is a collection of objects whose symbols are enclosed in braces, {}, and
separated by commas, as in { item 1, item 2, item 3, …, item n } . It is usually more efficient to do operations on
lists rather than to do operations on individual items in the list. A function is applied separately to each element in
the list if it has the attribute “Listable”. For example, addition, multiplication, and the logarithm have the attribute
Listable, so that
{ 5, 8, 11 } + { 2, – 3, – 6 } = { 7, 5, 5 }
{ 5, 8, 11 }* { 2, – 3, – 6 } = { 10, – 24, – 66 }
Log [ { 5, 8, 11 } ] = { Log [ 5 ], Log [ 8 ] ], Log [ 11 ] ] }
That is, Listable functions in Mathematica are automatically distributed or “threaded” over lists that appear as its
arguments. The number of elements in a list is given by the built-in function Length [ ]; e.g.,
Length [ { 5, 8, 11 } ] = 3 . A summary of Mathematica’s built-in functions used for list manipulation is given in
the table below.

Summary of list manipulation functions in Mathematica (taken from Blachman, 1992)

Function Description
Range [min, max, step] Generates the list {min, .., max} using step (arithmetic progression)
Table [expr, {imax}] Generates a list of imax copies of expr (more general)
Array [s, dim] Generates a list of length dim with elements s[i]
Sort [list] Sorts elements of list into canonical order

Thin-Walled Structures 109


Axial Force, Shear Force and Bending Moment Diagrams

Summary of list manipulation functions in Mathematica (taken from Blachman, 1992)

Function Description
Reverse [list] Reverses elements in list
RotateLeft [ list, n] Cycles the elements n positions to the left
RotateRight [ list,n] Cycles the elements n positions to the right
Permutations [list] Generates a list of all possible permutations of the elements of list
Drop [list, n] Drops the first n elements from list
Take [list, n] Takes the first n elements from list
First [list] Give the first element of list
Last [list] Gives the last element of list
list [[n]] or Part [list, n] Gives the nth element
Rest [list] Returns all but the first element of list
Select [list, crit] Picks out elements in list which meet the criterion of crit
Append [list, elem] Returns a list with elem appended to the end of list
AppendTo [list, elem] Changes list by appending elem to the end
Prepend [list, elem] Returns a list with elem added to the from of list
PrependTo [list, elem] Changes list by adding elem to the front
Insert [list, elem, n] Inserts elem at position n in list
Length [list] Gives the number of elements in list
Dimensions [list] Gives the dimensions of a list or expression
Complement [ list1, list2, ... ] Gives the complement, i.e., those elements in list1 but not in list 2, ...
Intersection [ list1, list2, ... ] Gives a sorted list of all the elements common to all list1, list2, ...
Union [list1, list2, ... ] Gives a sorted list of the distinct elements
Join [list1, list2, ... ] Joins or concatenates lists together
Partition [list, n] Partition list into sublists of length n
Flatten [list] Flattens out nested lists, i.e., eliminates nested lists
Transpose [list] Transpose
Apply [f, list] Replaces the head of list with f
Map [f,list] Applies f to each element in list
Listable An attribute, if set, automatically maps a functions onto a list
ColumnForm[list] Prints list as a column
MatrixForm[ list] Prints elements in list in a regular array

Problem statement: The aerodynamic loads on an airplane wing cannot be represented by a simple equa-
tion. The load per inch of span, p y ( z ) = p ( z ) , of the airplane wing shown in Fig. 4.10is tabulated in column
two of the Table printed at line “Out[14]” in the Mathematica program below. Find the shear force and bending
moment diagrams for the wing.

Solution: The values of the shear force and bending moment at various points along the wing are calculated
in the Table (see Out[14] in the code). The points are called stations and are designated by their distances from
110 Thin-Walled Structures
Differential equation method

Fig. 4.10

the centerline of the airplane, as shown in Fig. 4.10. These distances are measured along the wing rather than
horizontally, since the air loads are perpendicular to the wing. The distances between stations, ∆z , are computed
as a list in the code. The value of the shear at any point is obtained as the area under the load curve from that
point out to the wing tip. The load curve is assumed to be a series of straight lines between the known points, and
the area is obtained as the sum of the areas of the trapezoids. The area of the trapezoids are obtained as the prod-
uct of the average height p ave and the base ∆z . The change in the shear ∆V y between two stations is equal to
the area of the load curve between the stations. The shear V y is then obtained by summation of the ∆V y -values.
The change in the bending moment ∆M x between two stations is equal to the area under the shear curve. This
area is also assumed trapezoidal and is obtained by multiplying the sum of the shears at the adjacent stations by
one-half the distance between the stations. The bending moments are obtained by a summation of the ∆M x -val-
ues. Plots of the air load, shear force, and bending moment distributions are shown at the end of the Mathematica
program

Example 4.2: Numerical quadrature for the shear force and bending moment in a wing
(Peery, 1950, pp. 107-109)

In[1]:= Off@General::spell1D

Input the airload intensity at each z-station and each z-station coordinate as two separate lists. Dimensional
units: z-list, inches; p-list, lb/in.

In[2]:= z = 80, 20, 40, 60, 80, 100, 120, 140, 160, 180,

p = 8125, 123, 120, 116, 111, 105, 98, 89, 80, 71,
200, 220, 225<;

58, 35, 0<;

Compute distances between stations.

In[3]:= Dz = Take@HRotateLeft@zD - zL, Length@zD - 1D


Out[3]= 820, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 5<

Thin-Walled Structures 111


Axial Force, Shear Force and Bending Moment Diagrams

1. 2 3 4 5 6 7

Compute average airload intensity in each interval between stations.

In[4]:= pave = Take@N@HRotateLeft@pD + pL ê 2D, Length@pD - 1D


Out[4]= 8124., 121.5, 118., 113.5, 108.,
101.5, 93.5, 84.5, 75.5, 64.5, 46.5, 17.5<

The trapezoidal rule of numerical integration is used to compute the change in the shear force over each
interval from eq. (4.2). A list of DVy - values is computed by a direct multiplication of lists Dz and pave .

In[5]:= DVy = - Dz * pave

Out[5]= 8-2480., -2430., -2360., -2270., -2160.,


-2030., -1870., -1690., -1510., -1290., -930., -87.5<

Since Vy HtipL - Vy HrootL = Ÿroot


tip HdV ê dzL „ z, and V HtipL = 0, the shear force at the root is the negative of
y y
the sum the elements in the DVy -list. A simple way to sum elements in a list is to change the Head of the list
from "List" to "Plus" by using the Apply function.

In[6]:= Head@DVy D

Out[6]= List

In[7]:= Vy0 = - Apply@Plus, DVy D

Out[7]= 21107.5

Vy HiL = ⁄ij=1 DVy H jL + Vy0 . We use the Table function to generate a list of the shear force values at each
The shear force at station i is the the partial sum of the of the DVy - values over the intevals from 1 to i; i.e.,

station.

In[8]:= Vy = Table@HSum@DVy @@jDD, 8j, 1, i<D + Vy0 L, 8i, 1, Length@DVy D<D

Out[8]= 818627.5, 16197.5, 13837.5, 11567.5, 9407.5,


7377.5, 5507.5, 3817.5, 2307.5, 1017.5, 87.5, 0.<

Add the value of the shear force at the root to the beginning of this list in order to have the shear force at each
zi - station, including the root.

In[9]:= Vy = PrependTo@Vy , Vy0 D

Out[9]= 821107.5, 18627.5, 16197.5, 13837.5, 11567.5, 9407.5,


7377.5, 5507.5, 3817.5, 2307.5, 1017.5, 87.5, 0.<

112 Thin-Walled Structures


Differential equation method

1 2 3 4 5 6

Compute the average force in each interval from the list of shear force values.

In[10]:= Vave = Take@N@HRotateLeft@Vy D + Vy L ê 2D, Length@Vy D - 1D

Out[10]= 819867.5, 17412.5, 15017.5, 12702.5, 10487.5,


8392.5, 6442.5, 4662.5, 3062.5, 1662.5, 552.5, 43.75<

The change in the bending moment is computed from eq. (4.4) using the trapezoidal rule of numerical
integration.

In[11]:= DMx = Dz * Vave


Out[11]= 8397350., 348250., 300350., 254050., 209750.,
167850., 128850., 93250., 61250., 33250., 11050., 218.75<

Since Mx HtipL - Mx HrootL = Ÿroot


tip HdM ê dzL „ z, and M HtipL = 0, the bending moment at the root is the
x x
negative of the sum the elements in the DMx -list.

In[12]:= Mx0 = -Apply@Plus, DMx D

Out[12]= -2.00547 µ 106

Compute the bending moment at each station.

In[13]:= Mx = Table@HSum@DMx @@jDD, 8j, 1, i<D + Mx0 L,


8i, 1, Length@DMx D<D;
Mx = PrependTo@Mx , Mx0 D
Out[13]= 9-2.00547 µ 106 , -1.60812 µ 106 , -1.25987 µ 106 ,
-959519., -705469., -495719., -327869., -199019.,
-105769., -44518.7, -11268.8, -218.75, 0.=

In[14]:= TableForm@Transpose@8z, p, Vy , Mx <D,

8None, 8"z,in", "p,lbêin",


TableHeadings Ø

"Vy ,lb", "Mx ,lb-in"<<D

Thin-Walled Structures 113


Axial Force, Shear Force and Bending Moment Diagrams

1 2 3 4 5 6 7

Out[14]//TableForm=

z,in p,lbêin Vy ,lb Mx ,lb-in


0 125 21107.5 -2.00547 µ 106
20 123 18627.5 -1.60812 µ 106
40 120 16197.5 -1.25987 µ 106
60 116 13837.5 -959519.
80 111 11567.5 -705469.
100 105 9407.5 -495719.
120 98 7377.5 -327869.
140 89 5507.5 -199019.
160 80 3817.5 -105769.
180 71 2307.5 -44518.7
200 58 1017.5 -11268.8
220 35 87.5 -218.75
225 0 0. 0.

Plots of the airload, shear force, and bending moment distributions. (Intermediate plots have been suppressed.)

PlotStyle Ø 8PointSize@0.02D<,
In[15]:= p11 = ListPlot@Transpose@8z, p<D,

DisplayFunction -> IdentityD


p12 = ListPlot@Transpose@8z, p<D,
PlotJoined -> True,
DisplayFunction -> IdentityD

AxesLabel Ø 8"z, in.", "lbêin"<,


p1 = Show@p11, p12,

PlotLabel Ø "Airload distribution",

p21 = ListPlot@Transpose@8z, Vy <D,


DisplayFunction -> IdentityD

PlotStyle -> 8PointSize@0.02D<,

p22 = ListPlot@Transpose@8z, Vy <D,


DisplayFunction -> IdentityD

PlotJoined -> True,


DisplayFunction -> IdentityD

AxesLabel Ø 8"z, in.", "lb"<,


p2 = Show@p21, p22,

PlotLabel Ø "Shear force",

p31 = ListPlot@Transpose@8z, Mx <D,


DisplayFunction -> IdentityD

PlotStyle -> 8PointSize@0.02D<,


DisplayFunction -> IdentityD

114 Thin-Walled Structures


Differential equation method

In[24]:= Show@GraphicsArray@88p1<, 8p2<, 8p3<<DD

lbêin Airload distribution


120

100

80

60

40

20

z, in.
50 100 150 200

lb Shear force
20000

15000

10000

5000

z, in.
50 100 150 200

lb-in Bending Moment


z,in
50 100 150 200

-500000

6
-1µ10

6
-1.5µ10

6
-2µ10

Thin-Walled Structures 115


Axial Force, Shear Force and Bending Moment Diagrams

4.3 Semi-graphical method


The semi-graphical method to draw the shear force and bending moment diagrams is best illustrated by doing an
example.

EXAMPLE 4.3 Uniform barge with symmetric load

Consider a barge at rest in still water with a uniform immersed cross section, and subjected to the symmetrical
loads shown in Fig. 4.11. This is an example of a structure with no boundary supports, and is typical of aero-

5m 5m 5m 5m
15kN (total) 15kN (total)
10kN

Fig. 4.11 Uniform section barge in still water with symmetric load.

space and ocean vehicle structures. We wish to sketch the shear force and bending moment diagrams for the
barge. In this example there is a distributed load acting on the barge due to buoyancy forces produced by displac-
ing the water. Let p b represent the distributed load intensity due to buoyancy, and p b is a constant along the
barge because the immersed cross section is uniform and the water is still.

Solution: A semi-graphical method is used to sketch the shear force and bending moment diagrams. In this
approach we first sketch the distributive load p y ( z ) , then the shear force Vy(z), and finally the bending moment
Mx(z). Equations (4.1) and (4.3) are used to note that the slope of shear diagram at z is the negative of the distrib-
uted load intensity at z, and the slope of the moment diagram at z is the shear force at z. In addition, eqs. (4.2) and
(4.4) set at z = z2 give
z2


V y ( z 2 ) – V y ( z 1 ) = – p y ( z ) dz (4.8)
z1

z2

M x ( z2 ) – M x ( z1 ) =
∫ V ( z ) dz
y (4.9)
z1

Equation (4.8) is interpreted in a graphical sense to mean that the difference in the shear force between z2
and z1 is the area under the distributed loading diagram from z1 to z2. This is not geometrical area. The area
between the p y ( z ) curve and the z-axis has units of force, and may be positive, zero, or negative. Similarly eq.
(4.9) is interpreted to mean the difference in the bending moment is the area under the shear force diagram.

Vertical equilibrium of the entire barge requires the buoyant upthrust equals 40kN, so that p b = ( 2kN ) ⁄ m .
The total distributed load intensity is the difference between p b and the magnitude of the downward acting

116 Thin-Walled Structures


Semi-graphical method

applied loading intensity. The distributed loading intensity diagram is constructed in this manner as shown in Fig.
4.13.

py
10 kN
2

kN/m 15
0 z, m
5 10 20
-1

kN Shear force diagram

z, m
5 10 15 20

-2

-4

kN-m Bending moment diagram

17.5
15
12.5
10
7.5
5
2.5
z, m
5 10 15 20

Fig. 4.12 Shear force and bending moment diagrams for the barge in still water

Thin-Walled Structures 117


Axial Force, Shear Force and Bending Moment Diagrams

The point force of 10kN acting at z = 10m is shown schematically in the p y ( z ) -diagram as a downward
pointing arrow. Actually, p y → – ∞ as z → 10m , because a point force is a finite load acting over zero length.
Point forces are idealizations to actual loads and introduce discontinuities in the mathematical descriptions of
some of the dependent variables. The reader should verify the distributive loading intensity diagram of Fig.
4.13.The shear force diagram is drawn below the loading intensity diagram in Fig. 4.13. Equilibrium at z = 0
requires Vy(0) = 0, and the slope dVy/dz at z = 0 is equal to 1kN/m. The slope is constant between 0 < z < 5m ,
thus Vy(z) is a straight line in this range of z. The difference in the shear force between z = 5m and z = 0 is equal
to the negative of the area under the p y ( z ) curve which is 5kN. Thus Vy(5) = 5kN since Vy(0) = 0. At z = 5+m the
loading intensity jumps to +2kN/m. The slope of the shear force jumps from 1kN/m to -2kN/m at z = 5m, but the
shear force is itself continuous. The difference Vy(10) - Vy(5) is equal to the negative of the area between the
p y ( z ) -curve and the z-axis between z = 5m and z = 10m. Thus Vy(10) - Vy(5) = -10kN, so Vy(10) = -5kN. Note
the shear force is zero at z = 7.5m. At z = 10m the point force of 10kN acts. According to the first of eqs. (4.5)
Vy(10+) - Vy(10-) = 10kN, so that Vy(10+) = 5kN. The slope of the shear at z = 10m is +2kN/m, and remains con-
stant until z = 15m. The difference Vy(15) - Vy(10+) = -10kN, so that Vy(15) = -5kN. Finally, the slope changes to
-1kN/m at z = 15+m and remains constant in the range 15 < z < 20. The difference Vy(20) - Vy(15) = 5kN, so that
Vy(20) = 0. Checking vertical equilibrium at z = 20m verifies that Vy(20) should be zero.
Moment equilibrium at z = 0 shows Mx(0) = 0. The slope of Mx at z = 0 is equal to the shear force at z= 0.
Hence ( d M x ) ⁄ ( dz ) = 0 at z = 0, as shown in Fig. 4.13. The slope on the moment diagram increases linearly
from zero at z = 0 to 5kN at z = 5m. Thus Mx(z) is parabolic from z= 0 to z = 5. The difference Mx(5) - Mx(0) is
equal to the area under shear diagram from z = 0 to z = 5m. Hence, Mx(5) - Mx(0) = 12.5 kNm, and Mx(5) = 12.5
kNm since Mx(0) = 0. From z = 5 to z = 7.5 the slope of the moment decreases from 5kN to zero. At z = 7.5, Mx
is a local maximum with a magnitude of 18.75 kNm. The slope of Mx(z) for 7.5 < z < 10 is negative, decreasing
linearly from zero to -5kN. The difference Mx(10) - Mx(7.5) = -6.25 kNm, so that Mx(10) = 12.50 kNm. The
slope of Mx(z) at z = 10m jumps from a -5kN to a +5kN as shown in Fig. 4.13, but the moment itself is continu-
ous. The bending moment diagram in the range 10 < z < 20 is completed in a manner similar to the description of
its construction in the range 0 < z < 10.
In this example the shear force diagram is antisymmetric about z = 10m and the bending moment is symmet-
ric about z = 10m. This follows from the symmetrical loading on the barge and equilibrium eqs. (4.1) and (4.3).

4.4 Buoyancy Force Distribution on Ships


The simple uniform buoyancy distribution acting on the barge in Example 4.3 is an exception to the buoyancy
distributions found in practice. It is true that equilibrium requires the total buoyant upthrust to equal the weight of
the ship and its contents. However, the distribution of the buoyancy and weight along the length of the ship is not
necessarily the same. The difference in the magnitudes of the buoyancy and weight distribution intensities is the
applied load intensity p y ( z ) . In ship design three conditions are recognized to compute p y ( z ) for the same ship.
These conditions are called
• the still water condition,
• sagging condition, and
• the hogging condition.

118 Thin-Walled Structures


Buoyancy Force Distribution on Ships

A more detailed account of these conditions on the longitudinal bending of the ship is given by Muckle
(1967) and Zubaly (1996), and here we only summarize the basic ideas.

A ship in still water is shown in Fig. 4.13, and a section between z and z + dz is shown in Fig. 4.14.

Fig. 4.13

A ( z )dz
dF b

Fig. 4.14

Archimedes' principle asserts that the buoyant upthrust is equal to the weight of the fluid displaced. Let A(z)
denote the submerged cross section at z, and let γ denote the specific weight (force per volume) of the fluid. The
differential buoyancy force dFb acting on the ship over a differential length dz is

dF b = γA ( z )dz (4.10)

Consequently, the buoyant upthrust per unit ship length, which we designate p b , is equal to γA(z); i.e.,

dF
p b = --------b- = γA ( z ) (4.11)
dz
A curve of p b for a ship as well as the weight per unit length is shown in Fig. 4.15. Overall equilibrium requires
weight/length

buoyancy/length

Fig. 4.15

the area under these curves to have the same magnitude. If the submerged cross section is uniform in z, as is the

Thin-Walled Structures 119


Axial Force, Shear Force and Bending Moment Diagrams

case for the barge in Example 4.3, the distribution of the buoyancy per unit length p b is a constant.

At sea a ship is subjected to waves, and this alters the buoyancy distribution. For longitudinal bending of the
ship two extreme static conditions are assumed: sagging and hogging. In each condition, the length of the wave is
assumed to be the length of the ship. This is an “accepted” assumption for the worst buoyancy distribution caus-
ing the most severe bending of the ship.

The sagging condition is shown in Fig. 4.16. (Also see Fig. 1.3 on page 5.) The wave crests are at the bow
and stern, and the wave trough is amidships. A schematic of the buoyancy per unit length is shown below the ship
in Fig. 4.16. The immersed cross section is the largest at or near the wave crests, and is least near the trough. The
intensity of the buoyancy distribution reflects this. In this condition the deck sags and is in compression while the
bottom is in tension. The worst location to concentrate the cargo in the ship is amidships, as this will result in the
largest bending moment.

concentrated weight

pb

Fig. 4.16 Sagging

The hogging condition is depicted in Fig. 4.17. Here the wave troughs are at bow and stern, and the crest is
amidships. The immersed cross section is greatest near amidships and is least near bow and stern. The distribu-
tion of the buoyancy per unit length p b , shown in Fig. 4.17, reflects this situation. In hogging the deck is in ten-
sion and the bottom is in compression. The worst possible locations to concentrate cargo is fore and aft, as this
will produce the greatest bending moment in the ship.

concentrated weights

pb

Fig. 4.17 Hogging

120 Thin-Walled Structures


References

4.5 References
Blachman, N.R., 1992, Mathematica: A Practical Approach, Prentice Hall, Englewood Cliffs, New Jersey,
p. 133.

Muckle, W., 1967, Strength of Ship’s Structures, Edward Arnold Ltd., London, pp. 27-69.

Peery, D.J., 1950, Aircraft Structures, McGraw-Hill, New York, pp. 107-108.

Zubaly, R.M., 1996, Applied Naval Architecture, The Society of Naval Architects and Marine Engineers,
Cornell Maritime Press, Inc., Centreville, Maryland, pp. 195-237.

4.6 Problems
2L
1. The cantilever wing is subjected to a distributed air load p y ( z ) = ------------- 1 – ( z ) 2 , where the total lift (2
πz max
wings) L = 20, 000lbs at cruise, wing length z max = 32.5 ft., and z = z ⁄ z max . Also, the wing supports an
engine weighing 1000 lbs. Plot the loading diagram, shear force diagram V y ( z ) , and bending moment diagram
M x ( z ) as functions of z for 0 ≤ z ≤ 32.5 ft. Partial answer: V y ( 0 ) = 9, 000 lb. and M x ( 0 ) = – 131934 lb-ft.

y
py( z)

py Vy

z
Mx Vy
+ Mx

6 ft. 1000 lb engine

32.5 ft.

fuselage

2. A proposed solar airplane called Centurion is being designed to achieve semi-perpetual flight (Aviation
Week & Space Technology, May 4, 1998, p.54). Centurion is a flying wing with a span of 206 ft., an 8-ft. chord,
and no taper or sweep. The wing has five sections, one center, two mid-span, and two tips. It is supported by four
landing pods. The tip sections have a dihedral to assist in turning and washout twist to prevent tip stall. The
empty weight is predicted to be 1,105 lb., comprising 630 lb. for structure, 160 lb. for engines and propellers, 150
lb. for avionics, 75 lbs for batteries, 20 lb. of miscellaneous and 70 lb. for 7% growth. The aircraft should be able
to take a 100 lb payload to 100,000 ft. It is powered by 14 electric motors producing a maximum of 2 hp. each.
Assume the following: The span-wise airload distribution acting on the wing is as given in problem 1. Each
engine is modeled as a concentrated weight acting its location on the wing. The payload, avionics, batteries, etc.,
lumped are together as a concentrated load at the center, and that the structural weight is uniformly distributed

Thin-Walled Structures 121


Axial Force, Shear Force and Bending Moment Diagrams

along the span. For steady level flight, determine the shear force and bending moment diagrams from the center-
line of the wing to its tip, and show them in a sketch. Label significant points. The front view of half of the Cen-
turion is shown below.

z
z1 4.0 ft.
z2 12.0 ft. z1
z3 19.2 ft. z2
z4 32.7 ft. z3
z4
z5 46.3 ft. z5
z6 60.0 ft. z6
z7 67.1 ft. z7
z8
z8 80.6 ft.
z9
z9 93.4 ft. z 10
z10 103.0 ft. Centerline

3. The barge shown below has a uniform cross section along its length and is subjected to a uniformly distrib-
uted load of intensity p y ( z ) = – P , in which P has dimensional units of force/length. Also it is subjected to

z
buoyancy for the extreme hogging condition py( z) = P 1 – cos  π --- . Draw the shear force,
buoyancy  L

V y ( z ) , and bending moment, M x ( z ) , diagrams. Label significant points. Note that M x = 2P ( L ⁄ π ) 2 .


max

z
L L

waterline

4. A barge has a plan view as shown. All waterplanes are identical. Cargo is loaded evenly in the four rectangu-
lar holds as shown. Neglecting the weight of the barge itself, construct curves of weight, buoyancy, load, shear,
and bending moment for the loaded barge in still sea water. Label the values of each curve at each bulkhead, and
identify the maximum shear and bending moment. (Zubaly, 1996)

122 Thin-Walled Structures


Problems

40 ft 40 ft 40 ft 40 ft 40 ft 40 ft

empty 400 950 950 400 empty


tons tons tons tons 30 ft

5. A barge of uniform rectangular construction has a length of 30m, breadth of 10m, depth of 5m, floats at an
even keel in fresh water at a draft of 2m when unloaded. The barge is transversely divided into three equal com-
partments. These compartments are uniformly loaded as follows:
No. 1 hold, 200 tonne; No. 2 hold, 155 tonne; No. 3 hold, 245 tonne
(Note: one metric ton, or tonne, is equal to 1000 kg, and the mass density of fresh water is 1 tonne/m3)

You will plot the loading intensity diagram, shear force diagram, and the bending moment diagram for the
loaded barge in a column format. Do not neglect the weight of the barge itself.
a) Since the moments of the weight about amidships are not equal for the loaded barge, the barge trims.
Assume the trim angle is small. Show from overall equilibrium that the draft at z = 0 is d 0 = 3.7 m , and
that the draft at z = L = 30m is d L = 4.3m

b) Plot the loading intensity diagram, p y ( z ) , where the loading intensity is in N/m, Newton/meter.
(The specific weight in N/m3 is g times the mass density in kg/m3. If for simplicity g is taken as 10
(instead of 9.8) then specific weight of fresh water is 10m/sec 2 × 1000kg/m 3 = 10, 000N/m 3 .)

c) Determine the shear force, and plot it directly below the loading intensity diagram. Note; the dimen-
sional unit of the shear force is N, or Newtons.
d) Determine the bending moment, and plot it directly below the shear diagram. Note; the dimensional
units of the bending moment are Nm, or Newton-meters.

trim angle
z
5m
2m
d0 200t 245t dL
30m 155t

unloaded loaded

Thin-Walled Structures 123


Axial Force, Shear Force and Bending Moment Diagrams

124 Thin-Walled Structures


CHAPTER 5 Bending Deflections of
Beams under Transverse
Loads

5.1 Approximations for slender beams


In CHAPTER 3 we discussed the special case of pure bending of a beam caused equal and opposite couples
applied to the ends of the beam. In pure bending, the bending moment interior to the beam is equal to the moment
of the applied end couples and the shear force in the beam vanishes. In this chapter, forces and/or distributed
loads act perpendicular to the reference axis causing transverse shear forces to develop interior to the beam. For
example, if a beam is subjected to a transverse distributed load of intensity p y ( z ) , then the shear force Vy(z) is
non-zero by equilibrium eq. (4.1) on p. 103 . Thus, moment equilibrium, eq. (4.3) on p. 104 , requires the bend-
ing moment to change with z along the beam. The presence of the transverse shear force causes plane cross sec-
tions not to remain plane in the deformed beam. Cross section of the beam will warp out of plane due to the
action of transverse shear. This invalidates the assumptions used for the deformation analysis of the beam under
pure bending (no shear) in Section 3.2. The analysis of the deformations associated with the presence of both
bending and transverse shear is best approached from the theory of elasticity. However, the elasticity analysis is
too complex for routine calculations, and in structural mechanics a simpler theory is used. The simpler theory to
compute stresses in beams under transverse loading is called engineering theory, or technical theory, or Ber-
noulli-Euler theory.

Assumption in the engineering theory of beams


In the engineering theory of beams it is assumed that the distribution of normal stress σz
given by the flexure formula, eq. (3.27) on p. 77 , or equivalently eq. (3.28), is essentially
correct even in the presence of a nonuniform bending moment: i.e., when the shear force is
non-zero. This implies the deformations associated with shear are small, and consequently
neglected, with respect to deformations associated with bending.

The engineering theory is known from actual practice, and from the few exact elasticity solutions available,
to provide reasonable estimates for long slender beams. As a general guide, the length of the beam should be
greater than ten times the largest cross-sectional dimension for the engineering theory to be applicable.

In the 5.2 we discuss the governing boundary value problem for the displacements of a beam subject to
transverse loads. Following this, complementary virtual work and complementary strain energy equations for

Thin-Walled Structures 125


Bending Deflections of Beams under Transverse Loads

beam bending are derived. The complementary energy method leads to the generalized form of Castigliano’s sec-
ond theorem for beams. Application of Castigliano’s second theorem to compute beam displacements and rota-
tions at discrete points is illustrated in several examples.

5.2 Beam displacements


Neglecting the displacements due to transverse shear, the formulation of the pure bending problem discussed in
Section 3.1 and Section (3.2) is used to estimate beam displacements. Computing displacements is necessary in
design situations where there are limits on the maximum displacements. Also, in statically indeterminate beam
problems it is necessary to consider all three steps of the structural analysis to solve the problem: equilibrium,
strain-displacement (geometry of deformation), and the material law. For beam bending, derivatives of the dis-
placements of the material points on the z-axis determine the curvature components of the z-curve in the
deformed beam. The curvatures of the z-curve, in turn, determine the normal strain distribution of the fibers orig-
inally parallel to the z-axis in the undeformed beam.

Equilibrium Let p x ( z ) and p y ( z ) , denote the lateral load intensities (F/L) in the x-direction and y-direction,
respectively, acting on the beam. These external load intensities are defined positive if they act in their respective
positive coordinate directions as shown in Fig. 5.1. Also shown in the figure, are the positive shear forces, V x
py y, v
Vy

px x, u
Vx
z
Mx

My

Fig. 5.1 External distributed loads and internal actions

and V y , and the positive bending moments, M x and M y , acting on the positive z-face of the beam. Consider a
free body diagram of a differential element of the beam dz-long obtained by cutting the beam parallel to the x-y
plane at axial locations z and z + dz. Equilibrium conditions for bending in the y-z plane involve the sum of
forces in the y-direction and sum of moments about the x-axis of the differential element. In the limit as dz → 0 ,
the sum of forces leads to
dV
---------y + p y = 0 (5.1)
dz
and the sum of the moments leads to
dMx
----------- – V y = 0 (5.2)
dz

126 Thin-Walled Structures


Beam displacements

Equations (5.1) and (5.2) were derived in Chapter 2 as eqs. (2.1) and (2.3), respectively. Equilibrium conditions
for bending in the x-z plane involves the sum of forces in the x-direction and sum of moments about the y-axis of
the differential element. In the limit as dz → 0 , the sum of forces in the x-direction leads to

dV x
--------- + p x = 0 (5.3)
dz
and the sum of moments about the negative y-direction leads to
dMy
----------- – V x = 0 (5.4)
dz
Differentiating eq. (5.2) with respect to z, and then substituting eq. (5.1) into this result for the derivative of the
shear force gives
2
d Mx
= – py( z) (5.5)
dz2
Similarly, differentiating eq. (5.4) with respect to z, and then substituting eq. (5.3) into this result for the deriva-
tive of the shear force gives
2
d My
= – px(z) (5.6)
dz2

Strain-displacement The normal strain of line elements parallel to the z-axis in the undeformed beam is deter-
mined in Section 3.2, and is given by eq. (3.15) on p. 74 . Repeating this result we have

dθ y dθ x
ε z = x  -------- + y  -------- (5.7)
 dz   dz 

in which the rotation of the projection of the z-curve into the y-z plane (Fig. 3.9) is
dv
θ x = – ------ (5.8)
dz
and the rotation of the projection of the z-curve into the x-z plane is
du
θ y = – ------ (5.9)
dz
These rotations are depicted in Fig. 3.9 on page 74. In eq. (5.8) the displacement component of the material
points on the z-axis in the y-direction is denoted by function v(z), and in eq. (5.9) the displacement in the x-direc-
tion is denoted by function u(z). The derivatives of the rotations in eq. (5.7) are the curvature components of the
projections of the z-curve onto the x-z and y-z coordinate planes. Combining eqs. (5.7) to (5.9), we get

2 2
 d u  d v
ε z = x  – 2 + y  – 2 (5.10)
 dz   dz 

Material law For a linear elastic, isotropic material the normal stress-strain relation is σ z = Eε z , where E is the
modulus of elasticity of the material. The normal strain distribution given in eq. (5.7) can be substituted into this
material law to get the normal stress distribution over the cross section due to bending. For a beam made of a

Thin-Walled Structures 127


Bending Deflections of Beams under Transverse Loads

homogenous material, it was shown in eq. (3.25) on p. 77 that statical equivalence of the normal stress distribu-
tion to the bending moments leads to

Mx I xx I xy θ′ x
= E (5.11)
My I xy I yy θ′ y

where Ixx, Iyy, and Ixy are the second area moments about the centroidal x-y coordinates in the cross section, and
where the prime denotes an ordinary derivative with respect to z. The inverse of eq. (5.11) is

θ′ x 1 I yy – I xy M x
= -----------------------------------
2 )
(5.12)
θ′ y E ( I xx I yy – I xy – I xy I xx M y

Now substitute the rotation-displacement relations, eqs. (5.8) and (5.9), into the material law, eq. (5.11), and
in turn substitute the moments from this result into equilibrium eqs. (5.5) and (5.6) to get

2 2 2 2
d  d v d  d u
2
 EI xx 2  + 2
 EI xy 2  = p y ( z ) 0<z<L (5.13)
dz  dz  dz  dz 

2 2 2 2
d  d v d  d u
2
 EI xy 2  + 2
 EI yy 2  = p x ( z ) 0<z<L (5.14)
dz  dz  dz  dz 

Equations (5.13) and (5.14) are the governing ordinary differential equations for the displacement functions u(z)
and v(z) in the open interval 0 < z < L , where L is the length of the beam. These equations are coupled in the
sense that u(z) and v(z) appear in both equations if the product area moment Ixy is non-zero. If the product area
moment is zero, then eq. (5.13) governs function v(z) alone and eq. (5.14) governs function u(z) alone. If the
bending stiffness terms EI xx, EI yy , and EI xy are independent of coordinate z, then the governing equations
reduce to
4 4
dv du
EI xx 4 + EI xy 4 = p y 0<z<L (5.15)
dz dz
4 4
dv du
EI xy + EI yy 4 = p x 0<z<L (5.16)
dz4 dz

The governing ordinary differential equations (5.15) and (5.16)are fourth order in each displacement func-
tion, which means that their solution will contain four arbitrary constants for each displacement function. These
constants are determined from the boundary conditions at z = 0 and z = L. Boundary conditions specify how the
beam is supported at each end. For bending in the y-z, plane specify at z = 0 and z = L
a) either displacement v or shear force V y , but not both

b) either rotation θ x or bending moment M x , but not both

The particular choices for the so-called standard boundary conditions for bending in the y-z plane are given in
Fig. 5.2. Other, more complex, boundary conditions exist in practice.

For bending in the x-z plane, specify at z = 0 and z = L

128 Thin-Walled Structures


Beam displacements

y, v, V y

θ x, M x 2 2
 d v  d u
z v = 0 M x = EI xx  –  + EI xy  –  = 0
2
 dz   d z 2
(1) simple support

z v = 0 θ x = – v′ = 0
(2) clamped

2 2
d  d v  d u 
V y = -----  EI xx  –  + EI xy  –   = 0
dz  2
 dz   d z 2 
(3) free z 2 2
 d v  d u
M x = EI xx  –  + EI xy  –  = 0
 d z 2  d z 2

2 2
z d  d v  d u 
(4) free in y V y = -----  EI xx  –  + EI xy  –   = 0 θx = 0
dz  2
 dz   d z 2 
clamped about x

Fig. 5.2 Standard boundary conditions for bending in the y-z plane.

a) either displacement u or shear force V x , but not both

b) either rotation θ y or bending moment M y , but not both

The so-called standard boundary conditions for bending in the x-z plane are shown in Fig. 5.3.

Thin-Walled Structures 129


Bending Deflections of Beams under Transverse Loads

x , u, V x

θ y, M y 2 2
 d v  d u
z u = 0 M y = EI xy  –  + EI yy  –  = 0
2
 d z   d z 2
(1) simple support

z u = 0 θ y = – u′ = 0
(2) clamped

2 2
d  d v  d u 
x V x = -----  EI xy  –  + EI yy  –   = 0
dz   d z 2
  d z 2 
(3) free z
2 2
 d v  d u
M y = EI xy  –  + EI yy  –  = 0
2
 dz   d z 2

2 2
(4) free in x z d  d v  d u 
V x = -----  EI xy  –  + EI yy  –   = 0 θy = 0
clamped about y dz  2
 dz   d z 2 

Fig. 5.3 Standard boundary conditions for bending in the x-z plane.

It is possible to de-couple the displacements in the governing differential equations, but the problem may not
totally de-couple because the boundary conditions can involve both displacements. To get the de-coupling of the
differential equations, consider the case were the bending stiffnesses are uniform in the z-coordinate. Substitute
the rotation-displacement relations, eqs. (5.8) and (5.9), into the inverse form of the material law, eq. (5.12), to
get

1 1
------------ – ------------
– v″ = ER yy ER xy M x
(5.17)
– u″ 1 1 My
– ------------ ------------
ER xy ER xx

where the second area moment ratios are defined by

I xx I yy – I xy 2 I xx I yy – I xy 2 I xx I yy – I xy 2
R xx = -------------------------- R yy = -------------------------- R xy = -------------------------- (5.18)
I xx I yy I xy

130 Thin-Walled Structures


Beam displacements

Now differentiate eqs.(5.17) twice with respect to z, and substitute eqs. (5.5) and (5.6) into this result for the sec-
ond derivatives of the moments to get
4
dv py( z) px( z)
4
= ------------
- – ------------- 0<z<L (5.19)
dz ER yy ER xy

4
du py( z) px( z)
4
= – ------------
- + ------------- 0<z<L (5.20)
dz ER xy ER xx

EXAMPLE 5.1 A statically indeterminate beam with an unsymmetrical cross section.

The uniform beam with a zee cross section is subjected to a uniformly distributed load of intensity p 0 in the pos-
itive y-coordinate direction as shown in Fig. 5.4. The beam is clamped in both the y-z plane and the x-z plane at z
= 0. At z = L the beam is simply supported in the y-z plane and clamped in the x-z plane. (The mathematical con-
ditions of support at z = L approximate a door butt hinge with the hinge axis parallel to the x-direction.) The sec-
8 2
ond area moments for the cross section are I xx = --- a 3 t , I yy = --- a 3 t , and I xy = – a 3 t , where dimension a is as
3 3
shown in the figure and t denotes the wall thickness. Determine the displacements u ( z ) and v ( z ) , and the bend-
ing moments Mx(z) and My(z).

y,v x,u y
p0
a
a x
z z
a
L L
a

Fig. 5.4 Statically indeterminate beam with a zee cross section.

Solution For this cross section the second area moment ratios defined by eqs. (5.18) are

7 7 7
R xx = ------ a 3 t R yy = --- a 3 t R xy =  – --- a 3 t
24 6  9

The governing equation for bending in the y-z plane is obtained from eq. (5.19) as
4
dv p0
4
= -----------
- 0<z<L (5.21)
dz ER yy
and the boundary conditions are
v(0) = 0 (5.22)

Thin-Walled Structures 131


Bending Deflections of Beams under Transverse Loads

θ x ( 0 ) = – v′ ( 0 ) = 0 (5.23)

v( L) = 0 (5.24)

M x ( L ) = EI xx ( – v″ ) + EI xy ( – u″ ) = 0 (5.25)
z=L

Note that boundary condition (5.25) requires that the displacement response u(z) be known. The governing equa-
tion for bending in the x-z plane is obtained from eq. (5.20) as
4
du p0
4
= – -----------
- 0<z<L (5.26)
dz ER xy
and the boundary conditions are
u = u′ = 0 at z = 0 and z = L (5.27)

Note the boundary value problem for u(z) given by eqs. (5.26) and (5.27) is independent of the displacement
response v(z). Thus, we can solve eqs. (5.26) and (5.27) independent of the solution of eqs. (5.21) to (5.25). The
general solution to eq. (5.26) is
p0 z 4 z3 z2
u = – ------------ ------+ c 1 ---- + c 2 ---- + c 3 z + c 4
ER xy 24 6 2

where c1,..., c4 are arbitrary constants. The boundary conditions, eq. (5.27) at z = 0 require that constants c3 = c4
= 0. The boundary conditions at z = L lead to

L3 L2 p0 L 4
( u ( L ) = 0 ) → -----c 1 + -----c 2 = -----------------
-
6 2 24ER xy
L2 po L 3
( u′ ( L ) = 0 ) → -----c 1 + Lc 2 = --------------
-
2 6ER xy

These simultaneous equations are solved for constants c1 and c2 to get

p0 L p0 L 2
c 1 = --------------
- c 2 = – -----------------
-
2ER xy 12ER xy

Hence, the solution for u(z), after substitution for Rxy, is

3 p0 2
- z ( L – z )2
u ( z ) = ----------------- (5.28)
56Ea 3 t

The general solution of eq. (5.21) for v(z) is


p0 z 4 z3 z2
v = ------------ ------ + c 5 ---- + c 6 ---- + c 7 z + c 8
ER yy 24 6 2

where c5,..., c8 are arbitrary constants. Satisfaction of boundary conditions (5.22) and (5.23) requires c7 = c8 = 0.
Boundary condition (5.25)) can be re-written as
R xx
v″ ( L ) + -------- u″ ( L ) = 0
R xy

Substituting for v(z) and u(z), eq. (5.28), this boundary condition leads to

132 Thin-Walled Structures


Beam displacements

p 0 L 2 R xx p 0 L 2
Lc 5 + c 6 = -------------------
2
- – ---------------
12ER xy 2ER yy

Boundary condition (5.24) leads to

L3 L2 p0 L 4
-----c 5 + -----c 6 = – -----------------
-
6 2 24ER yy

Solving these last two equations for c5 and c6, we get

L ( – 5R xy 2 + R R )p L 2 ( 3R xy 2 – R R )p
xx yy 0 xx yy 0
c 5 = -----------------------------------------------------
2 R
- c 6 = --------------------------------------------------
2 R
-
8ER xy yy 24ER xy yy

Substituting for the second area moment ratios gives


p0
- z 2 ( 39L 2 – 71Lz + 32z 2 )
v ( z ) = -------------------- (5.29)
896Ea 3 t
Plots of the displacement distributions, eqs. (5.28) and (5.29), are shown in Fig. 5.5. The response of the beam in
the plane of loading (y-z plane) is represented by the lateral displacement v ( z ) . The out-of-plane lateral displace-
ment u ( z ) is non zero in this problem since the product area moment I xy ≠ 0 . If I xy = 0 , then u ( z ) = 0
∀z ∈ ( 0, L ) . See eqs. (5.18) and (5.26). That is, the beam would not bend out of the plane of loading if
I xy = 0 . However, as can be seen in Fig. 5.5, the displacement in the plane of loading v ( z ) is much larger than
the out of plane displacement u ( z ) in this example.

The bending moments are determined from the moment-curvature relations given by eqs. eq. (5.11). Substi-
tuting the solutions from eqs. (5.28) and (5.29) into these moment-curvature relations gives
p0
M x = ----- ( – L 2 + 5Lz – 4z 2 )
8
p0 L
M y = --------- ( L – 3z )
64
The bending moment diagrams for Mx and My are shown in Fig. 5.6. Note that the moment for bending in the y-
z plane, or Mx, is larger than the moment for bending in the x-z plane, or My. Moment My would vanish if
I xy = 0 .

Thin-Walled Structures 133


Bending Deflections of Beams under Transverse Loads

Ea3 t vêp0 L4

2.5
2

1.5
1
0.5

zêL
0.2 0.4 0.6 0.8 1

Ea3 t uêp0 L4

0.003
0.0025
0.002
0.0015
0.001
0.0005
zêL
0.2 0.4 0.6 0.8 1

Fig. 5.5 Lateral displacements of the unsymmetrical section beam

134 Thin-Walled Structures


Beam displacements

Mx êp0 L2

0.05

zêL
0.2 0.4 0.6 0.8 1

-0.05

-0.1

My êp0 L2

0.01

zêL
0.2 0.4 0.6 0.8 1
-0.01

-0.02

-0.03
Fig. 5.6 Bending moment diagrams for example 4.1

Fig. 5.6 Bending moment diagrams for the unsymmetrical section beam.

Fi 5 6 B di t di f l 41
EXAMPLE 5.2 Contact between two cantilever beams.

Consider two symmetrical section beams, one resting on top of the other, both clamped to a wall at z = 0.
The top beam is twice as long as the bottom beam, and the top beam is subjected to a downward force P at its tip.
Both beams have the same bending stiffness EIxx, which we abbreviate as EI here. See Fig. 5.7. Determine the
displacements due to bending for both beams assuming that they contact only at the tip of the shorter beam. Also
determine the bending moments and shear forces in each beam.

Solution Since the product area moment for each beam is zero, and neither beam is subjected to loads in the x-
direction, the response for each is bending in the y-z plane. Also, the distributed load intensity p y ( z ) is zero for
each beam, because we have assumed that the contact between the two beams occurs only at one point. The gov-
erning equation for the displacement of the top beam is

Thin-Walled Structures 135


Bending Deflections of Beams under Transverse Loads

y, v

z P

L/2 L/2
P

R
Fig. 5.7 Two cantilever beams in contact with
each other.

4
dv L L
= 0 0 < z < --- --- < z < L
dz4 2 2
Consequently, the displacement function in the top beam is

z3 z2 L
v ( z ) = v 11 ( z ) = c 1 ---- + c 2 ---- + c 3 z + c 4 0 ≤ z ≤ --- (5.30)
6 2 2
z3 z2 L
v ( z ) = v 12 ( z ) = c 5 ---- + c 6 ---- + c 7 z + c 8 --- ≤ z ≤ L (5.31)
6 2 2
The boundary conditions are
v ( 0 ) = v′ ( 0 ) = 0 M x ( L ) = – EIv″ ( L ) = 0 V y ( L ) = – EIv′′′ ( L ) = – P (5.32)

and the conditions of continuity of displacement, rotation, and bending moment are
L
v 11 = v 12 v′ 11 = v′ 12 v″ 11 = v″ 12 at z = --- (5.33)
2
There is a jump in the shear force at z = L/2 due to contact with the lower beam given by

 V +  --L- – V -  --L- + R = 0 → – EI v′′′  --L- + EI v′′′  --L- + R = 0 (5.34)


 y  2 y 
2  12  
2 11  
2

For the lower beam, the governing equation is


4
dv L
= 0 0 < z < ---
dz4 2
and the general solution of this equation is

z3 z2 L
v ( z ) = v 2 ( z ) = c 9 ---- + c 10 ---- + c 11 z + c 12 0 ≤ z ≤ --- (5.35)
6 2 2
The boundary conditions for the lower beam are

136 Thin-Walled Structures


Beam displacements

L L L L
v ( 0 ) = v′ ( 0 ) = 0 M x  --- = – EIv″  --- = 0 V y  --- = – EIv′′′  --- = – R (5.36)
 2  2  2  2

The condition that the beams are in contact at z = L/2 is

L L
v 11  --- = v 2  --- (5.37)
 2  2

The unknown reaction force R can be eliminated between eq. (5.34) and the third of eqs. (5.36) to get
v′′′ 11 – v′′′ 12 + v″′ 2 = 0 (5.38)
L
z = ---
2

where we have divided by the common factor EI.

For displacement v11, eq. (5.30), to satisfy the first two clamped boundary conditions of eqs. (5.32) constants
c4 = c3 = 0. For displacement v12, eq. (5.31), to satisfy the last two boundary conditions of eqs.(5.32) constants
c5 = P/EI and c6 = L P/EI. For the displacement of the lower beam v2, eq. (5.35), to satisfy the first two clamped
end conditions of eqs. (5.36) constants c11 = c12 =0. For v2 to satisfy the zero moment condition, third of eqs.
(5.36), constant c10 = - c9 L/2. Thus, to this point the displacements are

z3 z2 P LP
v 11 = c 1 ---- + c 2 ---- v 12 = ---------z 3 – ---------z 2 + c 7 z + c 8 (5.39)
6 2 6EI 2EI

z 3 Lz 2
v 2 = c 9  ---- – -------- (5.40)
6 4 
Substitute both of eqs. (5.39) into continuity of moments, second of eqs. (5.33), to get
L LP
c 1 --- + c 2 + --------- = 0
2 2EI
Substitute the first of eqs. (5.39) and eq. (5.40) into the contact condition, eq. (5.37), to get
L3 L2 L3
------ c 1 + -----c 2 + ------ c 9 = 0
48 8 24
Substitute the displacements eqs. (5.39) and (5.40) into the contact force continuity condition, eq. (5.38), to get
P
c 1 + c 9 – ------ = 0
EI

These last three equations are solved for c1, c2, and c9. We get c 1 = – P ⁄ ( 4EI ) , c 2 = – ( 3LP ) ⁄ ( 8EI ) , and
c 9 = ( 5P ) ⁄ ( 4EI ) .

Only c7 and c8 remain unknown. The conditions to determine these constants are the continuity conditions
of displacement and rotation, the first and second of eqs. eq. (5.33). Using the results for c1, c2, and c9 these con-
tinuity conditions become

L 5L 3 P 5L 2 P
--- c 7 + c 8 = ------------- – c 7 + ------------- = 0
2 96EI 32EI

Thin-Walled Structures 137


Bending Deflections of Beams under Transverse Loads

Thus, we find that c 7 = ( 5L 2 P ) ⁄ ( 3EI ) and c 8 = – ( 5L 3 P ) ⁄ ( 192EI ) .

The solution is complete since all of the constants appearing in eqs. (5.30), (5.31), and (5.35) have been
determined. The displacement of the upper beam is

 3LP P L
– ------------z 2 – ------------z 3 0 ≤ z ≤ ---
 16EI 24EI 2
v1 ( z ) =  (5.41)
3
5L P 5L P 2 LP P L
 – -------------- - z – ---------z 2 + ---------z 3
 192EI- + ------------ 32EI 2EI 6EI
--- ≤ z ≤ L
2
and the displacement of the lower beam is
5LP 5P
v 2 ( z ) = – ------------z 2 + ------------z 3 (5.42)
16EI 24EI

The assumption that the lower beam contacts the upper beam only at its tip needs to be verified. The condi-
tion of no penetration of the contact between the upper and lower beam is
L
v2 ≤ v1 0 ≤ z ≤ ---
2
Substitute eqs. (4.41) and (4.42) into this inequality to get
5LP 5P 3LP P
– ------------z 2 + ------------z 3 ≤ – ------------z 2 – ------------z 3
16EI 24EI 16EI 24EI
Multiply this equation by the positive factor EI/P, and after rearrangement we find

z2 L
----  – --- + z ≤ 0 which is valid for 0 ≤ z ≤ L ⁄ 2
4 2 

Thus, the assumption on the contact condition between the two beams is correct. A plot of the displacements of
the two beams is shown in Fig. 5.8. The bending moment in each beam is determined by the formula
M x ( z ) = – EIv″ ( z ) . Using eqs. (5.41), we find for the upper beam

P--- ( 3L + 2z )
L
0 ≤ z ≤ ---
8 2
M x1 = 
 P( L – z) L
--- ≤ z ≤ L
 2
and using eq. (5.42) we find for the lower beam
5P
M x2 = ------- ( L – 2z ) 0≤z≤L⁄2
8

138 Thin-Walled Structures


Complementary virtual work and complementary energy

The bending moment distribution in each beam is shown in Fig. 5.9.

v EI/(P L^3)

z/L
0.2 0.4 0.6 0.8 1

-0.05

-0.1

-0.15

-0.2
Fig. 5.8 The displacement functions of the upper beam (solid line) and the lower beam (dashed line)
for the two contacting cantilever beams.

M/(P L)

0.6

0.5

0.4

0.3

0.2

0.1

z/L
0.2 0.4 0.6 0.8 1
Fig. 5.9 The bending moment distributions in the upper beam (solid line) and the lower beam
(dashed line) for the two contacting cantilever beams.

5.3 Complementary virtual work and complementary energy


Beam displacements can also be determined using complementary energy. In particular, the generalized form of
Castigliano’s second theorem will be used in the examples in the following section. However, we first establish
the complementary virtual work principle for the beam. Then for a beam made of a linear elastic material the
complementary strain energy expression is obtained.

Thin-Walled Structures 139


Bending Deflections of Beams under Transverse Loads

5.3.1 Complementary virtual work


Consider combined bending of the beam in both the x-z plane as shown in Fig. 5.10, and in the y-z plane as
shown in Fig. 5.11. In these figures the end displacements and rotations are designated by q n , n = 1, 2, …, 8 .
The set of end displacements and rotations are called generalized displacements. The word “generalized” in gen-
eralized displacements is used to represent collectively both displacements and rotations. Associated with the
generalized displacements are the corresponding generalized forces Q n , n = 1, 2, …, 8 . Generalized forces rep-
resent a set of point forces and moments. The generalized displacement q n is measured at the same point on the
beam as the corresponding generalized force, and both are measured positive in the same direction. The product
q n Q n has the dimensional units of work, i.e., F-L. So if q n is a displacement, then Q n is point force. If q n is a
rotation in radians, then Q n is a point moment with dimensional units of F-L. The generalize displacements are
related to the beams displacements u and v , and beam rotations θ x and θ y as shown in the figures. Also, the

x
px
q 1, Q 1 q 5, Q 5
u ( 0 ) = q1 Vx u ( L ) = q5
θ y ( 0 ) = q2 θ y ( L ) = q6
z My
V x ( 0 ) = –Q1 q 2, Q 2 q 6, Q 6 V x ( L ) = Q5
M y ( 0 ) = –Q2 L M y ( L ) = Q6

Fig. 5.10 Generalized displacements and forces for bending in the x-z plane.

generalized forces are related to the beam shear forces V x and V y , and beam moments M x and M y , as shown in
the figures. From the possible boundary conditions for the beam, we either prescribe the generalized displace-

y, v
py
q 3, Q 3 q 7, Q 7
v ( 0 ) = q3 Vy v ( L ) = q7
θ x ( 0 ) = q4 θ x ( L ) = q8
z Mx
V y ( 0 ) = –Q3 q 4, Q 4 q 8, Q 8 V y ( L ) = Q7
M x ( 0 ) = –Q4 L M x ( L ) = Q8

Fig. 5.11 Generalized displacements and forces for bending in the y-z plane.

ment q n or the corresponding generalized force Q n , but not both.

140 Thin-Walled Structures


Complementary virtual work and complementary energy

To derive the complementary virtual work, we consider the generalized displacements to be prescribed.
Assume the beam displacements u ( z ) and v ( z ) , and rotations θ x ( z ) and θ y ( z ) , to be continuous, satisfying
compatibility eqs.(5.8) and (5.9), and to be consistent with the prescribed generalized displacements at z = 0 and
z = L. Known distributed load intensities p x ( z ) and p y ( z ) can act on the beam. Consider a variation in the gen-
eralized forces δQ n with the distributed loads unchanged. Virtual forces δQ n cause a variation in the internal
actions δV x , δM y , δV y , and δM x of the beam. Take these variations to satisfy the differential equations of
equilibrium for bending in the x-z plane given by eqs. (5.3) and (5.4); i.e.,
d d
( δV x ) = 0 ( δM y ) – δV x = 0 0<z<L (5.43)
dz dz
with the boundary conditions on the virtual forces for bending in the x-z plane given by
δV x ( 0 ) = – δQ 1 δV x ( L ) = δQ 5
(5.44)
δM y ( 0 ) = – δQ 2 δM y ( L ) = δQ 6
Similarly, from eqs. (5.1) and (5.2) the differential equilibrium conditions for bending in the y-z plane are
d d
( δV y ) = 0 ( δM x ) – δV y = 0 0<z<L (5.45)
dz dz
and the boundary conditions for bending in the y-z plane are
δV y ( 0 ) = – δQ 3 δV y ( L ) = δQ 7
(5.46)
δM x ( 0 ) = – δQ 4 δM x ( L ) = δQ 8

The external complementary virtual work performed by displacements acting through the virtual forces is
8
*
δW ext = ∑ q δQn n (5.47)
n=1

Substitute for the generalized displacements in eq. (5.47) the relationships to the beam displacements listed in
Fig. 5.10 and Fig. 5.11. Also substitute for the virtual forces in eq. (5.47) the boundary conditions given by eqs.
(5.44) and (5.46). After these substitutions we obtain
*
δW ext = – u ( 0 )δV x ( 0 ) – θ y ( 0 )δM y ( 0 ) – v ( 0 )δV y ( 0 ) – θ x ( 0 )δM x ( 0 ) +
u ( L )δV x ( L ) + θ y ( L )δM y ( L ) + v ( L )δV y ( L ) + θ x ( L )δM x ( L )

Write this last expression as


* L
δW ext = [ uδV x + θ y δM y + vδV y + θ x δM x ] ,
0

which is mathematically identical to


L
* d
δW ext =
∫ d z[ uδV x + θ y δM y + vδV y + θ x δM x ] dz
0

In the integrand of this equation, distribute the derivative of the sum to the sum of the derivatives of each term,
and then differentiate the product composing each term to get

Thin-Walled Structures 141


Bending Deflections of Beams under Transverse Loads

L
* du dθ y dv dθ x
δW ext =
∫ dz
δV x +
dz
δM y + δV y +
dz dz
δM x +
0
d d d d
u ( δV x ) + θ y ( δM y ) + v ( δV y ) + θ x ( δM x ) dz
dz dz dz dz
(5.48)

From the differential equations of equilibrium for the virtual forces, eqs. (5.43) and (5.45), eq. (5.48) reduces to
L
*  du + θ  δV + dθ y δM +  dv + θ  δV + dθ x δM dz
δW ext =
∫ dz y x
dz y dz x y
dz x (5.49)
0

It was assumed that the displacements and rotations satisfied the compatibility conditions given by eqs. (5.8) and
(5.9), so the terms multiplying the virtual shear forces in eq. (5.49) vanish for all z. Hence, we get
L
*
dθ y dθ x
δW ext =
∫ dz
δM y +
dz
δM x dz (5.50)
0

The right-hand side of this equation contains quantities defined internal to the beam, so we define the integral on
the right-hand-sid of eq. (5.50) as the internal complementary virtual work. Thus,
L
*
δW int

≡ [ θ x ′δM x + θ y ′δM y ] dz (5.51)
0

in which the prime denotes the derivative with respect to z.

The development from eq. (5.43) to (5.51) shows the external complementary virtual work equals the inter-
nal complementary virtual work for the beam having compatible displacements and rotations if the virtual com-
plementary forces are statically admissible. The principle of complementary virtual work asserts that the
displacements and rotations of the beam are compatible if the external complementary virtual work equals the
internal complementary virtual work for every statically admissible variation of the forces. If we were to apply
the principle of complementary virtual work to the beam problem, then we would obtain the compatibility condi-
tions for the displacements and rotations given by eqs. (5.8) and (5.9).

5.3.2 Complementary strain energy


The principle of complementary virtual work is independent of the material behavior. Now assume the material
of the beam is linear elastic and isotropic. Define the integrand of the internal complementary virtual work, eq.
(5.51), as the incremental quantity δU 0* ; i.e.,

δU 0* ≡ θ x ′δM x + θ y ′δM y (5.52)

For the linear elastic beam, the rotation gradients, or curvatures, were related to the bending moments by eq.
(3.72) on p. 92 . This equation is repeated below as eq. (5.53).

θ′ x * –I *
I yy t
1 xy M x + M x
= ---------------------------------------
* I * – I * 2)
- (5.53)
θ′ y E 0 ( I xx yy xy * I*
– I xy M y + M yt
xx

* , I * , and I * denote modulus-weighted second area moments, E is a reference modulus of


In eq. (5.53) I xx yy xy 0

142 Thin-Walled Structures


Complementary virtual work and complementary energy

elasticity, and M xt and M yt are thermal moments due to a linearly varying thermal strain distribution over the
cross section. If the beam is made of a homogeneous material and the reference modulus is taken as the modulus
of the homogeneous material, then the modulus-weighted second area moments reduce to the geometric second
area moments of the cross-sectional area. If the change in temperature from the stress free state is spatially uni-
form over the cross section for the homogeneous material beam, then the thermal moments vanish, but the ther-
mal axial force N t is non-zero. Substitute eq. (5.53) for the rotation gradients in eq. (5.52) to get
* (M + M t ) – I * (M + M t )
I yy x x xy y y
δU 0* = -------------------------------------------------------------------------
- δM x +
E 0 det ( I * )
– I xy* (M + M t ) + I * (M + M t )
x x xx y y
----------------------------------------------------------------------------- δM y
E 0 det ( I * )
(5.54)
in which
* I * – ( I * )2
det ( I * ) = I xx (5.55)
yy xy

is the determinate of the matrix of modulus-weighted second area moments. (This determinate is positive for real
areas and materials.) The form of eq. (5.54) suggests there is a functional U 0* [ M x, M y ] whose variation is pro-
vided by eq. (5.54). Assuming this functional exists, it is called the complementary strain energy density of the
beam having dimensional units of F-L/L. The first variation of the functional U 0* [ M x, M y ] is defined by

∂ ∂
δU 0* = ( U * ) δM x + ( U * ) δM y (5.56)
∂Mx 0 ∂My 0
Equating eqs. (5.54) and (5.56), and recognizing equality holds for every admissible choice of virtual moments
δM x and δM y , we conclude that

* (M + M t ) – I * (M + M t )
I yy
∂ x x xy y y
( U 0* ) = -------------------------------------------------------------------------
- (5.57)
∂Mx E 0 det ( I * )
and

– I xy* (M + M t ) + I * (M + M t )
∂ x x xx y y
( U 0* ) = ----------------------------------------------------------------------------- (5.58)
∂My E 0 det ( I * )

From eqs. (5.57) and (5.58), the functional U 0* [ M x, M y ] should be quadratic in its arguments M x and M y .
Assume that U 0* [ M x, M y ] is of the form

U o* = A 11 M x2 + A 12 M x M y + A 22 M y2 + A 1 M x + A 2 M y (5.59)

in which A 11, …, A 2 are coefficients to be determined. Taking partial derivative of eq. (5.59), we get


( U * ) = 2 A 11 M x + A 12 M y + A 1 (5.60)
∂Mx 0


( U * ) = A 12 M x + 2 A 22 M y + A 2 (5.61)
∂My 0
Compare eq. (5.57) to eq. (5.60) to identify the coefficients

Thin-Walled Structures 143


Bending Deflections of Beams under Transverse Loads

*
I yy *
I xy * Mt – I* Mt
I yy x xy y
A 11 = --------------------------- A 12 = – -----------------------
- A 1 = -----------------------------------
- (5.62)
2E 0 det ( I * ) E 0 det ( I * ) E 0 det ( I * )
Compare eq. (5.58) to eq. (5.61) to identify the coefficients
*
I xy *
I xx – I xy * Mt + I* Mt
x xx y
A 12 = – -----------------------
- A 22 = --------------------------- A 2 = ----------------------------------------
- (5.63)
E 0 det ( I * ) 2E 0 det ( I * ) E 0 det ( I * )

Note that coefficient A 12 is determined to be the same value from each comparison. If we have a function of the
form of eq. (5.59), then coefficient A 12 can be identified by the mixed partial derivatives in two ways; i.e.,

∂  ∂ ∂  ∂
A 12 = U * or A 12 = U * (5.64)
∂ M x  ∂ M y 0 ∂ M y  ∂ M x 0

The order of differentiation is immaterial if all derivatives of the function U 0* [ M x, M y ] concerned are continu-
ous. Since we started with the partial derivatives given by eqs. (5.57) and (5.58), a necessary check that the func-
tion U 0* [ M x, M y ] exits, is that the mixed partial derivatives obtained from eqs. (5.57) and (5.58) have to be
equal. If the mixed partial derivatives obtained from eqs. (5.57) and (5.58) were not equal, then the complemen-
tary strain energy would not exist. Finally, substitute the coefficients given by eqs. (5.62) and (5.63) into eq.
(5.59) to get

1 1
U o* = ------------------------
- --- ( I * M 2 – 2I xy
* M M + I * M 2 ) + ( I * M t – I * M t )M + ( – I * M t + I * M t )M (5.65)
E 0 det ( I * ) 2 yy x x y xx y yy x xy y x xy x xx y y

Equation eq. (5.65) is the complementary strain energy density in the most complex beam we consider. If the
beam is made of a homogeneous material with modulus of elasticity E, the complementary strain energy density
is

1 1
U o* = --------------------- --- ( I yy M x2 – 2I xy M x M y + I xx M y2 ) + ( I yy M xt – I xy M yt )M x + ( – I xy M xt + I xx M yt )M y (5.66)
Edet ( I ) 2
If, in addition to homogeneity, the beam has a symmetric cross section so that the product area moment vanishes,
then the complementary strain energy is

1 M2 M2 M xt M yt
U o* = -------  -------x + -------y  + ----------M x + ----------M y (5.67)
2E  I xx I yy  EI xx EI yy
Finally, if the beam is homogenous, has symmetric cross section, and there are no thermal strains, the comple-
mentary strain energy density is

1 M2 M2
U o* = -------  -------x + -------y  (5.68)
2E  I xx I yy 

From eqs. (5.51) and (5.52), the internal virtual work can be written as
L
* = *
δW int
∫ δU 0 dz
0

Assuming a linear elastic material, the complementary strain energy density functional was shown to exist and is
given by eq. (5.65). Now interchange the integral operator and variational operator in the previous equation to get

144 Thin-Walled Structures


Beam displacement by Castigliano’s second theorem

* = δU *
δW int ( elastic material only ) (5.69)

where the complementary strain energy of the beam is given by


L

U* *
=
∫ U dz 0 (5.70)
0

5.4 Beam displacement by Castigliano’s second theorem


Consider an elastic beam restrained against rigid body displacements and subjected to generalized forces Q n ,
n = 1, 2, …, N . Distributed loads and thermal strains can act on the beam as well. The generalized forces may
be independently varied since the variations in the reactions at the support points compensate to keep the struc-
ture in equilibrium. The generalized displacements corresponding to the generalized forces are denoted by q n ,
n = 1, 2, …, N . The principle of complementary virtual work is
N

∑ q δQ n n = δU * (5.71)
n=1

where the complementary energy for the beam U * is the integral over the length of the beam of the complemen-
tary strain energy density as expressed by eq. (5.70). The complementary strain energy density is written in terms
of the bending moments M x ( z ) and M y ( z ) as given by eq. (5.65). Following the development in Section 2.10
for the system of axially loaded bars, we use equilibrium conditions to determine the bending moments as func-
tions of z, 0 ≤ z ≤ L , and in terms of the generalized forces Q n . Hence, from eq. (5.65) the complementary strain
energy density is determined as a function of z with the generalized forces appearing as parameters in the func-
tion. Integration of the complementary strain energy density over the length of the beam, refer to eq. (5.70), gives
the complementary energy as a function of the generalized forces; i.e., U * = U * ( Q 1, Q 2, …, Q N ) . The variation
of the complementary strain energy becomes
N

δU * = ∑  ∂ Q U  δQ n
*
n (5.72)
n=1

Substituting eq. (5.72) into eq. (5.71), recognizing that the virtual generalized forces are independent, yields the
equation of the generalized form of Castigliano’s second theorem; i.e.,
∂ *
qn = U (5.73)
∂ Qn
A statement of the theorem is given in Section 2.10.

EXAMPLE 5.3 End rotations of a simply supported beam subject to an end moment

A simply supported, uniform beam of length L is subjected to a moment Q 1 at its left end as is shown in Fig.
5.12. The material is homogeneous and linear elastic, the cross section is symmetric ( I xy = 0 ), and there are no

Thin-Walled Structures 145


Bending Deflections of Beams under Transverse Loads

thermal strains. The bending stiffness is EI. Use Castigliano’s second theorem to determine the rotation at (a) the
left end, and (b) the right end.

EI
Q 1, q 1 q2
z

Fig. 5.12 Simply supported beam subject to an end moment

Solution This beam is statically determinate. Moment equilibrium about


V
point z of the free body diagram of the left section of the beam gives
Q1 M
M = – Q 1 + zR (5.74)
z
R where R is the reaction force at the support.
Free body diagram of left section

Part (a) Moment equilibrium of the free body diagram of the entire beam
gives the support reaction to be

L Q
R = -----1-
L
Q1
Hence the moment in eq. (5.74) becomes
R R
z
(a) Free body diagram of entire beam M = – Q 1  1 – ---
 L
It is important to note that we have determined a moment distribution from equilibrium that contains only the
applied moment Q 1 . That is, the support reaction R is not an independent force, it is a function of Q 1 . The com-
plementary strain energy is obtained from eqs. (5.68) and (5.70) with one of the two moments in eq. (5.68) iden-
tically zero. Castigliano’s second theorem in the case is
L L
∂ M2 M ∂M
-  dz
q1 =
∂ Q 1 2EI∫
--------- dz =
∫ -----
EI  ∂ Q  1
0 0

Substitute for the bending moment in this equation the function determined from equilibrium. We find
L L
1 z z Q1 z 2
q 1 = ------ Q 1  1 – ---  1 – --- dz = ------  1 – --- dz
EI ∫  L   L  EI ∫ L
0 0

Performing the definite integration on the right-hand side of the last equation we get
Q1 L
q 1 = ---------
- q 1 > 0 clockwise
3EI

146 Thin-Walled Structures


Beam displacement by Castigliano’s second theorem

Part (b) To find the rotation at the right end by Castigliano’s second theorem when no moment acts at that point,
we assume an external moment Q 2 acts at the right end, apply the theorem to compute the corresponding rota-
tion q 2 , then set Q 2 = 0 . That is,

L
M ∂M
-  dz
q2 =
∫ -----
EI  ∂ Q 2
(5.75)
0 Q2 = 0

Equation (5.74) for the moment is still valid in this case with two
L
moments applied to the beam. However, the support reaction R is different Q1 Q2
than in part (a). Moment equilibrium of the free body diagram for the entire
beam gives the support reaction as R R
( Q1 + Q2 ) (b) Free body diagram of entire beam
R = -----------------------
-
L
Substitute this result for the support reaction into eq. (5.74) to find the moment as
z
M = – Q 1 + ( Q 1 + Q 2 ) ---
L
Again, it is important to note that we have determined a bending moment distribution from equilibrium that only
contains the applied moment Q 1 and the fictitious external moment Q 2 . Substitute this bending moment into eq.
(5.75) to get
L
1 z z
– Q 1 + ( Q 1 + Q 2 ) ---  --- dz
q 2 = ------
EI ∫ L  L
0 Q2 = 0

Now we can set Q 2 = 0 and obtain


L
1 z z
– Q 1  1 – ---  --- dz
q 2 = ------
EI ∫  L  L
0

Performing the definite integration on the right-hand side of this result we find
Q1 L
q 2 = –  ---------- q 2 > 0 clockwise
 6EI 

Note that a clockwise moment Q 1 > 0 applied at the left end of the beam results in a counter clockwise rotation
at the right end q 2 < 0 .

EXAMPLE 5.4 Tip displacement of a cantilever wing spar a under distributed load

Determine the vertical tip displacement q 1 of a cantilever wing spar subjected to the distributed lift load of inten-
sity p y ( z ) = p 0 ( 1 – z ⁄ L ) , where p 0 is the intensity at the root and L is the span. See Fig. 5.13. The spar mate-

Thin-Walled Structures 147


Bending Deflections of Beams under Transverse Loads

y, v
z
p y ( z ) = p o  1 – ---
 L
q 1, Q 1

z
EI
L

Fig. 5.13 Cantilever wing spar under a distributed lift load

rial is linear elastic and homogeneous, and there are no thermal gradients. The product area moment of the cross
section is zero, and the bending stiffness EI xx = EI is uniform along the span.

Solution Although there is no tip force acting on the beam, to use Castigliano’s second theorem we imagine a
fictitious force Q 1 corresponding to displacement q 1 to act on the spar. We determine the tip displacement via
the theorem, and then set Q 1 = 0 . The complementary energy for this case is given by eq. (5.68) with M y = 0
and eq. (5.70). In mathematical terms the theorem results in
L
M x  ∂M x
q1 =
∫ ------
-  dz
EI  ∂ Q  1
(5.76)
0 Q1 = 0

where EI = EI xx . The bending moment distribution is obtained by solving the differential equations of equilib-
rium, eqs. (5.1) and (5.2), subject to the boundary conditions on the shear force and bending moment at z = L .
Substitute the distributed load intensity into eq. (5.1) to get
dV z
---------y = – p 0  1 – ---
dz  L
Indefinite integration of this equation gives

z2
V y = – p 0  z – ------ + c 1
 2L

The boundary condition on the shear force is V y ( L ) = Q 1 , which allows for the determination of the constant of
integration; i.e.,
p0 L
c 1 = Q 1 + --------
-
2
Hence, the shear force is
p0
V y ( z ) = ------ ( z – L ) 2 + Q 1 (5.77)
2L
Substitute the shear force from eq. (5.77) into the equilibrium differential equation (5.2) for the moment to get

148 Thin-Walled Structures


Beam displacement by Castigliano’s second theorem

dMx p0
----------- = ------ ( z – L ) 2 + Q 1
dz 2L
Introduce a new independent variable by the definition
ζ≡z–L (5.78)

so that the differential equation for the bending moment becomes


d M x dζ p
----------- ------ = -----0-ζ 2 + Q 1
dζ dz 2L

From eq. (5.78) dζ ⁄ dz = 1 . Indefinite integration of the above equation gives


p0
M x = ------ζ 3 + Q 1 ζ + c 2
6L

The boundary condition for the bending moment at z = L is M x = 0 . The value z = L corresponds to ζ = 0 .
This boundary condition determines that the constant of integration c 2 = 0 . Thus, the bending moment distribu-
tion in the equilibrium configuration under the distributed load and the fictitious force Q 1 is

p0
M x = ------ ( z – L ) 3 + Q 1 ( z – L ) (5.79)
6L

Substitute eq. (5.79) for the bending moment in eq. (5.76), recognize that the bending stiffness is indepen-
dent of z, to get
L
1 p
-----0- ( z – L ) 3 + Q 1 ( z – L ) [ z – L ] dz
q 1 = ------
EI ∫ 6L
0 Q1 = 0

Now set Q 1 = 0 , and change independent variables according to eq. (5.78), to write the expression for the tip
displacement as
0 0
p0 p0 ζ 5
q 1 = ------------- ζ 4 dζ = -------------  -----
6EIL ∫ 6EIL  5 
–L
–L

Finally we get the result that

p0 L 4
q 1 = -----------
- (5.80)
30EI

EXAMPLE 5.5 Strut-braced spar

Consider Example 5.4 again, but with a strut attached between the fuselage (simulated by a wall) and tip of the
spar as shown in Fig. 5.14. The strut is a two-force member, or truss bar, aiding to resist the vertical tip displace-
ment of the spar and to reduce the root bending moment in the spar. Determine the vertical displacement of the
tip of the spar using Castigliano’s second theorem.

Thin-Walled Structures 149


Bending Deflections of Beams under Transverse Loads

y, v
z
p y ( z ) = p o  1 – ---
 L
q 1, Q 1

z
α Fig. 5.14 Strut-braced spar
EA, EI

( EA ) b

q 2, Q 2
L

Solution The structural assembly is statically indeterminate. For


y example, the overall free body diagram of the assembly has three
z
p y ( z ) = p o  1 – ---
 L unknown forces of support R y, R z, Q 2 and a moment C x as
Ry
shown in the adjacent sketch. For a coplanar force system there
Rz are three independent equations of equilibrium. Hence, we have
z
one more unknown force than independent equations of equilib-
Cx rium. This structural system is said to have one redundant force.
To proceed in using Castigliano’s second theorem, remove the
strut from the fuselage support and replace the action of the sup-
Q2
port on the strut with force Q 2 . Force Q 2 is selected as the redun-
dant in this approach. In addition to the applied fictitious force
Assembly free body diagram
Q 1 , we regard the redundant force Q 2 as an applied force as well.
In this manner we have converted our structural assembly to be
statically determinate. Next we need to determine the complementary energy of the assembly U * in terms of the
independent forces Q 1 and Q 2 . (Note: support reactions R y, R z, and C x are not independent, since the three
equilibrium equations obtained from the free body diagram of the assembly relate them to forces Q 1 and Q 2 .)
Then, we apply Castigliano’s second theorem to determine the corresponding displacements q 1 and q 2 . How-
ever, the displacement at the end of the strut connected to the fuselage must be zero to correctly model the origi-
nal assembly. Also, the fictitious force Q 1 is set to zero after performing the differentiations in Castigliano’s
theorem. Hence, the mathematical statements of Castigliano’s method to this structural assembly are

∂U *
q 1 = ---------- (5.81)
∂Q 1
Q1 = 0

∂U *
q 2 = ---------- = 0 (5.82)
∂Q 2
Q1 = 0

Complementary energy The complementary energy consists of the sum of the complementary energies stored
in each member of the assembly. The action of the distributed load on the spar is to bend it, so energy is stored in
the spar due to bending deformation. But bending of the spar upward causes its tip to displace upward and stretch

150 Thin-Walled Structures


Beam displacement by Castigliano’s second theorem

the strut. Hence, energy is stored in the strut due to extensional deformation. An axial tensile force in the strut,
results in a compressive force in the spar as a free body diagram of the joint at the spar to strut will show. Thus,
the spar is subject to compressive deformation as well as bending deformation. The energy stored in compression
in the spar must be added to the energy stored in bending. Let N denote the axial force in the spar and N b denote
the axial force in the brace strut, both positive in tension. Then the complementary strain energy is
L
N2
 ---------- M2 N b2 L
U* - + --------x- dz + ------------------  ------------
=
∫  2EA 2EI  2 ( EA ) b cos α 








0









spar strut (5.83)

Equilibrium Free body diagrams of the joints between the strut and
Q1
fuselage and strut and spar tip are shown in the adjacent figure. Equilib-
rium at the joint between the strut and fuselage yields Vy
N b = Q2 (5.84) N α
Mx
Equilibrium at the joint between the spar and strut results in the follow- Nb
ing boundary conditions at z = L in the spar. z
L
– N ( L ) – N b cos α = 0 (5.85)
Nb
– V y ( L ) + Q 1 – N b sin α = 0 (5.86)
Q2
M y( L) = 0 (5.87)
Free body diagrams of the joints
The differential equations of equilibrium for the spar are
dN
------- = 0 0<z<L (5.88)
dz
dV y z
--------- + p 0  1 – --- = 0 0<z<L (5.89)
dz  L

dMy
----------- – V y = 0 0<z<L (5.90)
dz
The solution to eq. (5.88) subject to boundary condition (5.85) is
N ( z ) = – Q 2 cos α (5.91)

in which we used eq. (5.84) to write the axial force in the strut in terms of Q 2 . The solution to eqs. (5.89) and
(5.90) subject to boundary conditions (5.86) and (5.87) is the same procedure explained in Example 5.4. Omit-
ting the details, the solution for the bending moment in the spar is (refer to eq. (5.79))
p0
M x ( z ) = ------ ( z – L ) 3 + ( Q 1 – Q 2 sin α ) ( z – L ) (5.92)
6L
Equations (5.84), (5.91), and (5.92) represent the equilibrium solutions in terms of the known distributed load
and the independent forces Q 1 and Q 2 . This step is crucial to using Castigliano’s theorem.

Thin-Walled Structures 151


Bending Deflections of Beams under Transverse Loads

Evaluation of displacement equations Substitute the complementary strain energy, eq. (5.83), into eq. (5.81)
to get the following expression for displacement q 1 .

L
N ∂N M ∂M N b  L   ∂N b
-------  --------- + -------x  ----------x dz + --------------
q1 =
∫ EA  ∂Q 1 EI  ∂Q 1 
------------ ----------
( EA ) b  cos α  ∂Q 1
0 Q1 = 0

From eq. (5.91) ∂N ⁄ ∂Q 1 = 0 , from eq. (5.92) ∂M x ⁄ ∂Q 1 = z – L , and from eq. (5.84) ∂N b ⁄ ∂Q 1 = 0 . Thus,
the equation for displacement q 1 reduces to
L
1 p
-----0- ( z – L ) 3 + ( – Q 2 sin α ) ( z – L ) [ z – L ] dz
q 1 = ------
EI ∫ 6L
0

in which we used eq. (5.92) again to eliminate M x . Performing the definite integral on the right-hand-side
Q1 = 0
of the previous equation we get

p 0 L 4 ( Q 2 sin α )L 3
q 1 = -----------
- – ----------------------------- (5.93)
30EI 3EI
The first term on the left-hand side of this equation is the tip displacement obtained in Example 5.4 when there
was no strut. The second term on the left-hand-side of eq. (5.93) represents the reduction the tip displacement
due to the presence of the strut.

Next substitute the complementary strain energy, eq. (5.83), into eq. (5.82) to get the following expression
for displacement q 2 .

L
N ∂N M ∂M N b  L   ∂N b
-------  --------- + -------x  ----------x dz + --------------
q2 =
∫ EA ∂Q 2   EI ∂Q 2  
------------ ----------
( EA ) b  cos α  ∂Q 2
0 Q1 = 0

From eq. (5.91) ∂N ⁄ ∂Q 2 = – cos α , from eq. (5.92) ∂M x ⁄ ∂Q 2 = – ( z – L ) sin α , and from eq. (5.84)
∂N b ⁄ ∂Q 2 = 1 . Use these same equations, (5.91), (5.92), and (5.84), evaluated for Q 1 = 0 , to replace N , M x ,
and N b , respectively, in the equation for displacement q 2 . The resulting expression for q 2 is
L
 ( – Q 2 cos α ) 1 p0 
------ ( z – L ) 3 + ( – Q 2 sin α ) ( z – L ) [ – ( z – L ) sin α ] dz +
q2 =
∫  ---------------------------
EA
( – cos α ) + ------
EI 6L 
0

Q2  L 
-------------- ------------ ( 1 )
( EA ) b  cos α

Evaluating some of the terms in this expression we get


L L
Q 2 L cos2 α p o sin α Q 2 sin2 α Q2 L
- – ----------------- ( z – L ) 4 dz + -------------------- ( z – L ) 2 dz + ---------------------------
q 2 = ------------------------
EA 6EIL ∫ EI ∫ ( EA ) b cos α
-
0 0

152 Thin-Walled Structures


Problems

Perform the definite integrations in this equation, rearrange the terms, and recall that q 2 = 0 to get

L cos2 α L 3 sin2 α L p0 L 4
Q 2 ------------------ + -------------------- + ---------------------------- – ------------ sin α = 0 (5.94)
EA 3EI ( EA ) b cos α 30EI

We can solve eq. (5.94) for the redundant force Q 2 in terms of the distributed load intensity p 0 . Then, in
turn substitute the result for Q 2 into eq. (5.93) to find displacement q 1 .

5.5 Problems
1. The uniform cantilever beam shown below is subjected to a uniformly distributed load intensity and is also
restrained by a linear elastic spring at midspan. Neglect the weight of the beam, and determine
v ( z )EI xx
a) The displacement function v ( z ) . Plot the normalized displacement v ( z ) = --------------------
4 )
- versus dimen-
( p 0 z max
sionless coordinate z = z ⁄ z max for 0 ≤ z ≤ 1 .

Mx
b) .Plot the normalized bending moment M x ( z ) = ---------------
2
- for 0 ≤ z ≤ 1 .
p 0 z max

Vy
c) Plot the normalized shear force V y ( z ) = ---------------
- for 0 ≤ z ≤ 1 .
p 0 z max

y, v
p y ( z ) = p 0 = constant

64EI xx
z K = ----------------
3
-
z max
z max ⁄ 2
z max

2. A uniform beam with a rectangular cross section rests on a knife edge at its left end, while the right end is
clamped in rigid disk. The bending stiffness EI xx = EI , the distance between the knife edge and the beam’s
connection to the disk is L and the radius of the disk is R. This disk rotates about a fixed smooth pin through its
center under the action of applied moment Ma as shown. Determine the relation between the applied moment Ma
and rotation angle θ of the disk under the assumption that the angle of rotation θ is small.

Thin-Walled Structures 153


Bending Deflections of Beams under Transverse Loads

Ma, θ
y, v R
θx
EI xx = EI
z

z2 3. A coplanar bar is subjected to an end force Q 1 as


L
shown. The bar is uniform with axial stiffness EA and
bending stiffness EI. Use Castigliano’s second theorem
z1 L to find

q3 a) the end rotation q 2 , and

b) the vertical displacement q 3 at the joint.


q2

Q1

154 Thin-Walled Structures


CHAPTER 6 Shear Flow due to Shear
Forces

6.1 Shear flows and shear stresses due to bending in a rectangular


section beam
Consider bending under transverse loads of a rectangular section beam in the y-z plane as shown in Fig. 6.1. For

y
Vy
x τ yz
h/2 Vy z y
τ zy
Mx σz Mx
Mx z
h/2

Vy
t
Fig. 6.1 Shear and normal stresses in a rectangular section beam under general bending.

a slender beam, it is assumed that the normal stress σ z is determined from the flexure formula, eq. (3.27) on p. 77
, with sufficient accuracy when the beam is subjected to shear. The shear stress component τ zy exists if the shear
force Vy is present. Our purpose is to determine the shear stress in terms of the shear force.

The first subscript on the symbol τ zy for the shear stress refers to the direction of the outward normal to the
face on which it acts, in this case the z-face, and the second subscript refers to the direction of the shear stress, in
this case the y-direction. A positive value for shear stress τ zy is defined as acting in the positive y-direction on a
positive z-face. By the action/reaction law, a positive value for shear stress τ zy means it acts the negative y-direc-
tion on a negative z-face. See the stress element in Fig. 6.1. Moment equilibrium of the infinitesimal stress ele-

Thin-Walled Structures 155


Shear Flow due to Shear Forces

ment about the x-direction requires the shear stress component acting on the y-face in the z-direction, or τ yz , be
non-zero and equal to τ zy ; i.e., τ yz = τ zy . Shear stress component τ yz is sometimes called conjugate of τ zy . The
sign convention for shear stress τ yz is defined in the same manner as for shear stress τ zy . Shear stress compo-
nents τ zy and τ yz are shown acting in their positive senses on the stress element in Fig. 6.1. If the value of the
shear stress is negative, then the senses of the four arrows for the shear stresses in Fig. 6.1 are reversed.

In classical beam theory the shear stress component τ yz is determined by axial force equilibrium of free
body diagrams obtained from selected portions of the beam. Imagine a section of the beam obtained by cutting
the beam with planes perpendicular to the z-axis at z and z + ∆z, ∆z > 0, and a plane perpendicular to the y-axis
h h
at a generic value of y, where – --- ≤ y ≤ --- . The free body diagram of the beam below the y = constant plane is
2 2
shown in Fig. 6.2. The bending normal stress σz acts over a portion of the cross sectional area at z and z + ∆z.

∆F yz
σz ( z ) σ z ( z + ∆z )
y
z F + ∆F
F

h/2

z ∆z

Fig. 6.2 Free body diagram for a selected portion of the beam.

Because of the presence of the shear force, the bending normal stresses at z and z + ∆z are different. The axial
force due to the stress distribution σz acting over a portion of the cross-sectional area is
y t⁄2 y t⁄2 M M y
-------x y dx dy = -------x
F =
∫ ∫
–h ⁄ 2 –t ⁄ 2
σ z dx dy =
∫ ∫
–h ⁄ 2 –t ⁄ 2 I xx I xx ∫ –h ⁄ 2
yt dy (6.1)

where we substituted the flexure formula for the bending normal stress σz. The integral in the last of these equa-
tions for F is the first area moment of the lower portion of the cross-sectional area; i.e.,

y
t h 2
yt dy = --- y 2 –  ---
Qx( y) =
∫ –h ⁄ 2 2  2
(6.2)

Thus, eq. (6.1) becomes


Mx
F = ------- Q x ( y ) (6.3)
I xx

The action of the top portion of the beam on the bottom portion is represented by the force ∆F yz . Notice that the
surface at y = – h ⁄ 2 is stress free, so that no force or shear stress acts on this surface.; i.e., τ yz = 0 at
y = – h ⁄ 2 . As ∆z → 0 , ∆F yz → 0 , since internal forces are only transmitted over finite areas. This force is

156 Thin-Walled Structures


Shear flows and shear stresses due to bending in a rectangular section beam

given by
z + ∆z t⁄2
∆F yz =
∫ z
∫ –t ⁄ 2
τ yz dx dz

The integral of the shear stress across the width is defined as the shear flow and is denoted by q; i.e.,
t⁄2
q =
∫ –t ⁄ 2
τ yz dx (6.4)

Thus, the shear force ∆Fyz is


z + ∆z
∆F yz =
∫ z
q dz (6.5)

Divide this last equation by ∆z and take the limit as ∆z → 0 to get

dF yz
----------
- = q (6.6)
dz
which shows that the shear flow is also interpreted as the shear force per unit length.

The free body diagram of the beam segment shown in Fig. 6.2 is correct with respect to equilibrium in the
axial direction. Equating the forces in the z-direction to zero we get
– F + F + ∆F + ∆F yz = 0

and if we divide this by ∆z and take the limit as ∆z → 0 we get

dF dF yz
------- + ----------- = 0
dz dz
Substitute eq. (6.3) for F and (6.6) for the second term in this equation we get
d M Qx( y)
-----------x -------------
-+q = 0
dz I xx

Moment equilibrium of the beam requires V y = d M x ⁄ dz , so that the shear flow is given by

Vy
q = – ------Q x ( y ) (6.7)
I xx

If we assume that the shear stresses are uniform across the width of the beam, then q = τ yz t , and we get

V yQx( y)
τ yz = – --------------------
- (6.8)
I xx t

Since τ zy = τ yz , eq. (6.8) is also the formula for the shear stress in the cross section at y. It is important to note
that Q x ( y ) is the first area moment of a portion of the cross-sectional area about the centroidal x-axis. For the
rectangular section, I xx = ( h 3 t ) ⁄ 12 and Q x ( y ) is given by eq. (6.2), so that we finally obtain

Thin-Walled Structures 157


Shear Flow due to Shear Forces

3 Vy 2y 2 h h
τ zy = ---  ------ 1 –  ------ – --- ≤ y ≤ ---
2  ht   h 2 2
The shear stress distribution is parabolic in the cross section, with a maximum value of 1.5 times the average
shear stress, where τ ave = V y ⁄ A = V y ⁄ ( ht ) , and the shear stress is zero at the bottom and top of the beam
since τ yz = 0 at y = −
+ h ⁄ 2 . This distribution is sketched in Fig. 6.3. The shear stress distribution is statically

h/2
1.5 τ zy
--------
-
τ ave
h/2

Fig. 6.3 Parabolic shear stress distribution in the rectangular section beam.

equivalent to the shear force. That is,


h⁄2 h⁄2
t⁄2 t⁄2  3  V y 2y 2 
- 1 –  ------ dx dy = V y
Vy =
∫ ∫ –t ⁄ 2
τ zy dx dy =
∫ ∫  --2-  -----
–t ⁄ 2  ht   h 
–h ⁄ 2 –h ⁄ 2

6.2 Shear flows due to transverse shear forces in open section beams
Consider the thin-walled open cross section shown in Fig. 6.4. Let s denote the contour coordinate, which is

t
---
2
Vy Assume
y 0 ≤ τ zn « τ zs
τ zs
q =
∫τ zs dn
t
Vx τ zn – ---
C 2
x
s n
t s q

Fig. 6.4 Shear forces, shear stresses, and shear flow in a general thin-walled open cross section

defined as the arc length of the center line of the wall. The s-direction is tangent to the contour. Let n denote the
thickness coordinate perpendicular to the contour. The thickness coordinate has the range – t ⁄ 2 ≤ n ≤ t ⁄ 2 , where
t denotes the thickness of the wall. Shear forces Vx and Vy are the resultants of the distribution of the shear stress

158 Thin-Walled Structures


Shear flows due to transverse shear forces in open section beams

components τ zs and τ zn , where τ zs is acts tangent to the contour and τ zn acts in the direction normal to the con-
tour, that is, in the thickness direction. The shear stress component conjugate to τ zn is τ nz , and on the stress-free
lateral surfaces where n = ± t ⁄ 2 component τ nz = 0 . Since τ zn = τ nz , τ zn = 0 at n = ± t ⁄ 2 as well. In thin
wall bar theory we neglect the thickness direction component τ zn with respect to the tangential component τ zs ,
because τ zn = 0 at n = ± t ⁄ 2 and the wall is thin so that it appears that the thickness shear stress component is
not much different from zero. The shear flow is defined as the definite integral across the thickness of the shear
stress component tangent to the contour; i.e.,
t⁄2
q =
∫ –t ⁄ 2
τ zs dn (6.9)

The shear flow acts tangent to the contour at s, is positive if it acts in the positive s-direction, and has dimensional
units of F/L. It is a function of the contour coordinate s.

The shear flow at s, or q(s), can be related to the shear flow at s = 0, or q(0), by axial equilibrium. A free
body diagram of a beam segment defined by the portion of the contour from s = 0 to s, and the cross sections at z
and z + ∆z is shown in Fig. 6.5. The axial force due to the bending normal stress σz is

q(s)
F
s
q(s)
q(0) q(0)

F + ∆F
∆z
s = 0

Fig. 6.5 Free body diagram of a segment of a branch in the beam.

s
F =
∫ σ t ds
0
z

where the bending normal stress is evaluated on the contour (n = 0) rather than at a generic point through the
thickness of the wall in the thin-walled beam approximation. That is, we assume the distribution of the normal
stress through the thickness of the wall is small with respect to its value at the contour. Substituting the general
flexure formula for σz, eq. (3.27) on p. 77 , into the above equation we get

I xx M y – I xy M x I yy M x – I xy M y
F = ----------------------------------
2
- Q y ( s ) + ----------------------------------
2
- Qx(s) (6.10)
I xx I yy – I xy I xx I yy – I xy
where the first area moments with respect to the centroidal axes x and y are
s s
Qx(s) =
∫ 0
y ( s )t ds Qy(s) =
∫ x ( s )t ds
0
(6.11)

Since the contour is a curve in the x-y plane, the x and y coordinates of the contour are related to the coordinate s;
i.e., x = x ( s ) and y = y ( s ) on the contour. These coordinate functions x(s) and y(s) are required to compute

Thin-Walled Structures 159


Shear Flow due to Shear Forces

the first area moments defined by eqs. (6.11). Note that the thickness of the wall can be, in general, a function of
contour coordinates as well, as long as t(s) remains small with respect to the overall cross-sectional dimensions.

On the s-faces of the free body diagram in Fig. 6.5, the axial forces are approximated by the shear flow
times the small length ∆z. The shear flows at s = 0 and at s are not the same. Note that the shear flow acting on the
positive z-face is assumed positive in the positive s-direction, so that the positive shear flow on the positive s-face
is in the positive z-direction and a positive shear flow on a negative s-face is in the negative z-direction. Setting
the sum of forces in the axial direction equal to zero in the free body diagram shown in Fig. 6.5 leads to
F + ∆F – F + q ( s )∆z – q ( 0 )∆z = 0

Divide this equation by ∆z and then take the limit as ∆z → 0 to get

dF
------- + q ( s ) – q ( 0 ) = 0 (6.12)
dz
The derivative of eq. (6.10) is

dF I xx V x – I xy V y I yy V y – I xy V x
------- = --------------------------------Q
2 y ( s ) + --------------------------------Q
2 x(s)
dz I xx I yy – I xy I xx I yy – I xy

where we used moment equilibrium of the beam about the x-axis, or V y = d M x ⁄ dz , and about the y-axis, or
V x = d M y ⁄ dz . Substituting this last result into eq. (6.12), we get the formula for the shear flow due to trans-
verse shear forces as
I xx V x – I xy V y I yy V y – I xy V x
q ( s ) = q ( 0 ) – --------------------------------Q
2 y ( s ) – --------------------------------Q
2 x(s) (6.13)
I xx I yy – I xy I xx I yy – I xy

EXAMPLE 6.1 Shear flow distribution in a tee beam

A symmetric tee-section of dimensions h and t with h >> t is shown in Fig. 6.6. The section is subjected to a
5
shear forces V x = 0, V y > 0 . The second area moment I xx = ------ h 3 t and the product area moment I xy = 0 .
24
Determine the shear flow distribution and the maximum shear stress magnitude.

y
h/2 h/2 s,q s,q

x
C
h 3
--- h s,q
4
t, typical

Fig. 6.6 Symmetric tee-section beam.

160 Thin-Walled Structures


Shear flows due to transverse shear forces in open section beams

Solution A separate contour coordinate is selected in the web, left flange, and right flange. Choosing a separate
contour coordinate in each branch, rather than using one contour coordinate for the entire cross section, is often a
convenience for subsequent computations. The direction of positive contour coordinate s and, consequently, the
sense of positive shear flow is as shown in Fig. 6.6. Note that we selected the origin of each branch contour coor-
dinate at a free edge, so that q(0) = 0 in each branch. For this particular section eq. (6.13) reduces to
Vy
q ( s ) = – ------Q x ( s )
I xx

In the web, the cartesian coordinates of the contour are


3
x(s) = 0 y ( s ) = – --- h + s 0≤s≤h
4
The first area moment about the x-axis for the web, refer to the first of eqs. (6.11), is

s s
3 3 s2
Qx(s) =
∫ 0
y ( s )t ds =
∫  – --4- h + s t ds = t  – --4- hs + ---2-
0

Substitute this into the shear flow formula to get


Vy Vy 3 s 18 V y 2s s
q ( s ) = – ------Q x ( s ) = – --------------  – --- h + --- st = ------ ------  1 – --- --- --- 0≤s≤h (6.14)
I xx 5 3  4 2 5 h 3 h h
------ h t
24

6V y
The web shear flow is quadratic in the branch contour coordinate s, is zero at s = 0, and q ( h ) = --- ------ . We set the
5h
derivative of the shear flow with respect to s equal to zero to find the location where the shear flow may have a
maximum magnitude in the web. Hence

 dq 2 s 2s 1 3
------ = 0 →  – ------ --- +  1 – --- --- --- = 0, or s = --- h
 ds   3h h  3 h h 4

3 27 V y
This is the location of the centroid for the entire section, and q  --- h = ------ ------ .
4  20 h

The x and y coordinates of the contour in the left flange are

h h h
x ( s ) = – --- + s y ( s ) = --- 0 ≤ s ≤ ---
2 4 2
The first area moment about the x-axis is
s s
h 1
Qx(s) =
∫ 0
y ( s )t ds =
∫  --4- t ds = --4- hts
0

and hence the shear flow in the left flange is


Vy h 6V y s h
q ( s ) = – -------------- --- ts = – --- ------  --- 0 ≤ s ≤ --- (6.15)
5 3 4 5 h  h 2
------ h t
24

Thin-Walled Structures 161


Shear Flow due to Shear Forces

Thus, the shear flow in the left flange is linear in the contour coordinate, and at the junction with the web
h 3V y
q  --- = – --- ------ .
 2 5h

The x and y coordinates of the contour in the right flange are

h h h
x ( s ) = --- – s y ( s ) = --- 0 ≤ s ≤ ---
2 4 2
The first area moment about the x-axis is
s s
h 1
Qx(s) =
∫ y ( s )t ds = ∫  --4- t ds = --4- hts
0 0

and hence the shear flow in the right flange is


Vy h 6V y s h
q ( s ) = – -------------- --- ts = – --- ------  --- 0 ≤ s ≤ --- (6.16)
5 3 4 5 h  h 2
------ h t
24
Thus, the shear flow in the right flange is linear in the contour coordinate and at the junction with the web
h 3V y
q  --- = – --- ------ .
 2 5h

The shear flow distribution in the cross section is given by eqs. (6.14), (6.15), and (6.16). This distribution is
shown schematically in Fig. 6.7. The shear flows in the two flanges are given by the same equation, and are neg-
ative for Vy > 0. In Fig. 6.7 positive values of the flange shear flows are shown, which means that the arrow indi-
cating the sense of the shear flow in the flanges is drawn in the opposite direction to what was originally assumed
positive. Note that the shear flow into the junction from the web is equal to the sum of the shear flows going out
of the junction into the flanges. The physical basis of this flow relation at the junction is axial equilibrium at the
junction.

3V y
--- ------
5h

6V y 3V y 3V y
--- ------ --- ------ --- ------
Vy q 5h 5h
q 5h

27 V y 6V y
q ------ ------ --- ------
3 20 h 5h
--- h
4 Shear flows at the junction
of the web and flanges

Fig. 6.7 Shear flow distribution in the tee-section.

162 Thin-Walled Structures


Shear center of a thin-walled open section

27 V y
The maximum shear flow magnitude is equal to ------ ------ , and occurs in the web at the centroid of the cross
20 h
section. Since each branch of the cross section has the same thickness, the maximum shear stress magnitude also
occurs at the same location as the maximum shear flow. Thus,

27 V y 27 V y
τ max = ------ ------ = ------ 2  -------- = 2.7τ ave
20 ht 20  2ht
where the average shear stress is defined as the shear force Vy divided by the cross-sectional area 2ht. The maxi-
mum shear stress is 2.7 times larger than the average shear stress value, or 170% larger with respect to the aver-
age!

6.3 Shear center of a thin-walled open section


Consider a thin-walled channel section with a web height denoted by h, with flanges of equal width denoted by b,
and with the web and flanges having the same thickness denoted by t. The section is subjected to a vertical shear
force Vy > 0 and no horizontal shear, or Vx = 0. See Fig. 6.8. The objective is to determine the shear flow distri-

b s,q
y
y
h/2
Vy C C
x x

h/2
t, typical s,q s,q
b
Fig. 6.8 Thin-walled channel section subjected to vertical shear.

bution in the cross section, and then use static equivalence to determine the resultant of the shear flow distribu-
tion.

The x-axis as shown in Fig. 6.8 is an axis of symmetry for the section, so that the product area moment Ixy =
0. The second area moment about the x-axis is

1 h 2 1 1
I xx = ------ h 3 t + 2 0 +  --- bt = ------ h 3 t + --- h 2 bt
12  2 12 2

The positive contour coordinate directions and shear flows in each branch of the cross section are defined as
shown in Fig. 6.8. To determine the shear flows for this problem, eq. (6.13) reduces to
Vy s
q ( s ) = q ( 0 ) – ------Q x ( s )
I xx
where Q x ( s ) =
∫ y ( s )t ds
0

Thin-Walled Structures 163


Shear Flow due to Shear Forces

In the lower flange, s = 0 at a free edge so that q(0) = 0, and y ( s ) = – h ⁄ 2 for 0 ≤ s ≤ b . Thus, the first area
moment of the lower flange segment is Q x ( s ) = – ( hts ) ⁄ 2 , and the shear flow is

V y ht
q ( s ) = ------ ----- s 0≤s≤b (6.17)
I xx 2

The web shear flow at the junction with the lower flange, q(0), is equal to the shear flow of the lower flange
at s = b. This, again, is a consequence of axial equilibrium at the junction. Hence, for the web we have:
q ( 0 ) = ( V y hbt ) ⁄ ( 2I xx ) from eq. (6.17), y ( s ) = – h ⁄ 2 + s for 0 ≤ s ≤ h , and

s
h t
Qx(s) =
∫  – --2- + s t ds = --2- ( – hs + s )
0
2

Combining these results, the shear flow in the web is


V yt
q ( s ) = ---------- ( hb + hs – s 2 ) 0≤s≤h (6.18)
2I xx

The upper flange shear flow at the junction with the web, qweb(0), is equal to the web shear flow at s = h.
Hence, from eq. (6.18) we get q web ( 0 ) = q flange ( b ) = ( V y hbt ) ⁄ ( 2I xx ) . For the upper flange: y ( s ) = h ⁄ 2 for
0 ≤ s ≤ b , so Q x ( s ) = ( hts ) ⁄ 2 . The shear flow in the upper flange is

V yt
q ( s ) = ---------- ( hb – hs ) 0≤s≤b (6.19)
2I xx
Note that the upper flange shear flow at s = b is zero. This is as it should be, since the upper flange at s = b coin-
cides with a free edge. If this were not the case, then an error would have been made in the shear flow computa-
tions, and we would have to go back at this point to find it. The shear flow distributions given by eqs. (6.17) to
(6.19) are shown schematically in Fig. 6.9(a).

Next we determine the resultant of the shear flows in several steps using static equivalence. First, for
straight branches we can integrate the shear flow along the contour to get the branch force. The line of action of
b
the branch force is the contour line. For the lower flange, the branch force is defined by F 1 =
∫ q ( s ) ds . Substi-
0
tute eq. (6.17) for the shear flow to get
b V y ht 
 ----- V y hb 2 t
F1 =
∫ 
- ----- s ds = ----------------
0 I xx 2
 4I xx
- (6.20)

h
For the web, the branch force is defined by F 2 =
∫ q ( s ) ds . Substitute eq. (6.18) for the shear flow to get
0

h V yt
 --------- V yt 2 h3 h3 V 1 h3t
- ( hb + hs – s 2 ) ds = ---------
- h b + ----- – ----- = -----y- --- h 2 bt + -------
F2 =
∫ 
0 2I xx
 2I xx 2 3 I xx 2 12

1 1
but I xx = ------ h 3 t + --- h 2 bt , so that
12 2

164 Thin-Walled Structures


Shear center of a thin-walled open section

V y hbt
---------------
2I xx V y hb 2 t
F 3 = ----------------
-
4I xx

y y

V y hbt  h C C
- 1 + ------
-------------- x ⇒ x
2I xx  4b F2 = V y
static equivalence

V y hb 2 t
F 1 = ----------------
-
V y hbt 4I xx
--------------
-
2I xx

a) Shear flow distribution b) Branch forces


Fig. 6.9 Shear flow distribution due to bending and the statically equivalent branch forces for the
channel section.

F2 = V y (6.21)

b
For the upper flange, the branch force is defined by F 3 =
∫ q ( s ) ds . Substitute eq. (6.19) for the shear flow
0
to get
b V yt
 --------- V y hb 2 t
- ( hb – hs ) ds = ----------------
F3 =
∫ 
0 2I xx
 4I xx
- (6.22)

Note that the branch forces in each flange have the same magnitude, eqs. (6.20) and (6.22), have parallel lines of
action, but they have opposite senses. The branch forces are shown in Fig. 6.9(b), and these branch forces are
statically equivalent to the shear flows. Since the flange branch forces are the same magnitude, we let F1 = F3 =
1 1
Fflange, and use I xx = ------ h 3 t + --- h 2 bt to find
12 2

V y hb 2 t

3b 2 
F flange = ------  ----------------------------------- = V y  --------------------
- (6.23)

4 1 3 1 2   h + 6hb
2
 ------ h t + --- h bt
12 2

The second step in the process of static equivalence is to reduce the coplanar force system consisting of
branch forces to a force and a couple at some convenient point in the cross section. Each time a force is moved to
a parallel line of action a couple is created to maintain static equivalence, as is shown below. If we choose the
center of the web to resolve the force system, we get

Thin-Walled Structures 165


Shear Flow due to Shear Forces

a a a
b b b
F
F
F
F

F
d(F )
d a d a d
a
b b b

R x = F flange – F flange = 0 Ry = V y C z = – hF flange (6.24)

where Rx and Ry are the x- and y-direction components of the resultant force, and Cz is the moment of the couple
about the z-axis due to the force system. The resolution of the coplanar force system at the center of the web is
shown in Fig. 6.10(a). The system in Fig. 6.10(a) is statically equivalent to the planar force system in Fig.
6.9(b).

The third step in the process of static equivalence is to move the force Vy to a parallel position to eliminate
the couple and maintain force equivalence. Force Vy is moved such that the moment of Vy in its new position
about the center of the web is equal to Cz. This single force along a specific line of action is the resultant of the
coplanar force system. That is,

eV y = hF flange

where e denotes the perpendicular distance between the lines of action of Vy in its original position and its final
position. Note that the force Vy must be move to the left of its original position. The result of this final reduction
is shown in Fig. 6.10(b). The resultant is a single force of magnitude Vy in the positive y-direction, whose line of
action is parallel to the web at a distance e to the left of the web. The intersection of the line action of the result-
ant with the x-axis is called the shear center, which is abbreviated as S.C. in Fig. 6.10(b).

Vy

Vy
y y
hF flange
C S.C. C
x ⇒ x
static equivalence hF flange 3b 2
e = -------------------
- = ---------------
Vy h + 6b

a) Force and couple at the center of the web b) Resultant of the planar force system

Fig. 6.10 Reduction of branch forces to a single resultant acting at the shear center

166 Thin-Walled Structures


Shear center of a thin-walled open section

The location of the shear center (S.C.) in the cross section is determined by the shear flows due to bending.
The computations given above for the channel section illustrate the general conclusion that the S.C. location
depends on the pattern of the shear flow distribution, and not on the magnitude of the shear force. Shear forces
dV dV
act in the plane of loading to equilibrate the applied loads. (Recall that ---------y = – p y ( z ) and ---------x = – p x ( z ) .) The
dz dz
line of action of the lateral loads p y ( z ) and p x ( z ) must pass through the shear center if the beam is to bend with-
out twisting about the z-axis. If the line of action of the transverse loads is not through the shear center, then the
loads will cause torsion in addition to bending. For open cross sections with straight branches and one junction,
the shear center is at the junction where all branch forces intersect; i.e., the moment of the branch forces is zero at
the common junction. Three simple open sections illustrating this point are shown in Fig. 6.11.
S.C.

S.C.

S.C.

Fig. 6.11 Shear center locations for open sections with straight branches and one junction

The location of the shear center for a closed section beam cannot be determined using only the steps outlined
above for the open section. In a closed section the shear flow at the origin of the contour in eq. (6.13) is an addi-
tional unknown. To determine the shear flow at the origin of the contour requires consideration of twist of the
section. The procedure to determine the shear center of a closed section is discussed in “Shear center of a closed
section” on page 219 .

EXAMPLE 6.2 Shear center location in an unsymmetrical section

For the thin-walled section shown in Fig. 6.12, determine the location of the shear center. The location of

a s,q

y y free edge
2t
C C
2a x x
a 3a/8
2t t

A
2a
Fig. 6.12 Unsymmetrical thin-walled channel section

the centroid is shown in Fig. 6.12, and the second area moments about the centroidal axes are

Thin-Walled Structures 167


Shear Flow due to Shear Forces

16 53
I xx = ------ a 3 t I yy = ------ a 3 t I xy = – a 3 t
3 24

Solution To find the vertical location of the shear center we can consider shear forces V x > 0 and V y = 0 .
We do not need to compute the shear flows in each branch to find the shear center in this example. A good choice
for the point in the cross section where we would resolve the branch forces can simplify the computations. It is
convenient to use the junction of the lower flange and web, labeled point A, as the point about which to sum
moments of the branch forces, since the only the branch force in the upper flange contributes to the moment
about point A. Then static equivalence of the moment of the branch force and Vx about point A will give the loca-
tion of the line of action for shear force Vx.

First the shear flow in the upper flange is determined using the positive senses for q and the contour coordi-
nate s as shown in Fig. 6.12. For Vy = 0 and q(0) = 0, eq. (6.13) reduces to

( – I xx Q y + I xy Q x )V x
q ( s ) = ------------------------------------------------
2
- (6.25)
I xx I yy – I xy

The cartesian coordinates of the contour coordinate s in the upper flange are
5
x ( s ) = --- a – s y(s) = a 0≤s≤a (6.26)
8
The first area moments of the portion of the contour from s = 0 to s of the upper flange are
s
Qx =
∫ ( a )2t ds = 2ast
0
(6.27)

s
 5--- a – s 2t ds = 2t  5--- as – s---2-
Qy =
∫ 
0 8
 8 2
(6.28)

2 = ( 97a 6 t 2 ) ⁄ 9 . Substituting these results into eq.


The factor in the denominator of eq. (6.25) is I xx I yy – I xy
(6.25) we get

9 Vx 32 5 s2
q ( s ) = ------ -----6- – 2a 4 s – ------ a 3  --- as – ---- 0≤s≤a (6.29)
97 a 3 8 2
The branch force in the upper flange is the integral of this shear flow over the upper flange; i.e.,

a 23V x
F =
∫ q ( s ) ds = – -----------
0 97
- (6.30)

Static equivalence means that the moment of the shear force Vx about point A must equal the moment of the
branch force F about point A, as is illustrated in the sketch below. Let ey denote the moment arm for the shear
force. Static equivalence for the moment about point A gives

2aF = – e y V x (6.31)

Although the branch forces for lower flange and web have not been computed, they do not contribute to the
moment about point A. Substitute eq. (6.30) for F into this moment equivalence relation and solve for ey to get

168 Thin-Walled Structures


Shear center of a thin-walled open section

⇔ Vx
2a
ey
A A

46
e y = ------ a (6.32)
97
This result for ey locates the line of action of the shear force Vx for beam bending. The shear center is on this par-
ticular line of action of Vx, and we can locate the exact point on this line of action by considering the separate
problem of V x = 0 and V y > 0 .

Now consider V x = 0 and V y > 0 . Equation eq. (6.13) becomes

( I xy Q y – I yy Q x )V y
q ( s ) = --------------------------------------------
2
(6.33)
I xx I yy – I xy

Substitute eq. (6.27) for Qx and eq. (6.28) for Qy into this equation to get the shear flow in the upper flange as

9 V y 53 5 s2
q ( s ) = ------ -----6- – ------ a 4 s – 2a 3  --- as – ---- 0≤s≤a (6.34)
97 a 12 8 2
The branch force in the upper flange is
a
45
F =
∫ q ( s ) ds = – --------
0 194
-V y (6.35)

Moment equivalence about point A with the moment arm for shear force Vy denoted as ex, gives

2aF = e x V y (6.36)

See the sketch below. Thus, the location of the line of action of the shear force Vy is

⇔ ex Vy
2a
A A

Thin-Walled Structures 169


Shear Flow due to Shear Forces

45
e x = – ------ a (6.37)
97
The location of the shear center is shown in the sketch below.

S.C. 46
------ a
97

45
------ a
97

6.4 Skin-stringer idealization


The purpose of the numerical example to be presented here is to justify the skin-stringer approximation, or ideal-
ization, to classical engineering beam theory. This approximation is especially well suited to semi-monocoque
(stiffened shell) construction; i.e., reinforced thin-walled structures typically found in vehicles where minimum
weight is important. The cross section of the beam is shown in Fig. 6.13, and we take l = 10 d, d = 2 inches, Mx
= 15,000 in-lb, and Vy = 10,000 lbs for numerical evaluation. To study what happens when the web thickness t
varies we set t = ε d.

d Area A(s)

y
Vy q(s)
l/2
centroid, C
Mx s
x x-axis

l/2 Web shear flow q(s)


t

d
Geometry of cross section Beam resultants

Fig. 6.13 Two-flanged beam with thin web

170 Thin-Walled Structures


Skin-stringer idealization

Mx
The flexure formula for the axial normal stress is σ z = ------- y , in which the second area moment of the cross
I xx
section about the x-axis is

tl 3 d4 l d 2
y 2 dA = ------ + 2 ------ +  --- + --- d 2 = [ 83.33ε + 60.67 ]d 4
I xx =
∫∫ A
12 12  2 2
(6.38)

The average normal stress in the flange is


M x (l + d ) Mx
σ ave = ------- ---------------- = ------- ( 5.5d ) (6.39)
I xx 2 I xx
The portion of the total moment carried by the flanges is

 --l- + d  --l- + d
2  2 
Mx Mx
y  ------- y d dy = 2 ------- d y 2 dy
( M x ) fl = 2
∫ y ( σ dA ) = 2 ∫
z  I xx  I xx ∫
A fl l l
--- ---
2 2

or

( M x ) fl 2d l 3
l 3
--------------- = --- ------  --- + d –  --- = 60.67 ( 83.33ε + 60.67 ) ––1 (6.40)
Mx 3 I xx  2   2

The shear flow in the web is


Vy Vy l d 1 l l
ydA = ------  --- + --- d 2 + ---  --- + s  --- – s t
q ( s ) = ------
I xx
A(s)
∫ I xx  2 2 22  2 
(6.41)
Vy 1 t l 2 Vy ε s 2
q ( s ) = ------ --- ( l + d )d 2 + ---   --- – s 2 = ------ 5.5 + ---  25 –  ---  d 3
I xx 2 2   2  I xx 2   d 
The portion of the total shear force carried by the web is

l l l l
--- --- --- ---
2 2 2 2
Vy  l 2 
q ( s )ds = 2 q ( s )ds = ------  ( l + d )d 2 + t   --- – s 2 ds
( V y ) web =
∫ τ zy tds =
∫ ∫ I xx  ∫   2 

l l 0 0
– --- – ---
2 2

or
( V ) web 1 1 2 l 3 55 + 83.33ε
---------------- = ------ --- ( l + d )ld 2 + --- t  --- = ------------------------------------ (6.42)
Vy I xx 2 3  2 60.67 + 83.33ε

Thin-Walled Structures 171


Shear Flow due to Shear Forces

These results are tabulated below for various values of ε.

Flange Flange Web shear flow q, lb/in, eq. (6.41)


Ixx, in4 σave, psi (Mx)fl/Mx (Vy)web/Vy
ε = t/d eq. (6.38) eq. (6.39) eq. (6.40) s=0 s = ±5 in s = ±10, in eq. (6.42)
1/4 1304.0 126.5 0.744 529.1 481.2 337.4 0.930
1/10 1104.0 149.5 0.879 489.1 466.5 398.6 0.918
1/50 997.33 165.4 0.973 461.2 456.2 441.2 0.909
0 970.67 170.0 1 453.3 453.3 453.3 0.907
A plot of the shear flow in the web for each ε in the table above is shown in Fig. 6.14.

sêH10 inL
1
ε = 0
ε = 1 ⁄ 50
0.5 ε = 1 ⁄ 10
ε = 1⁄4

qêH500 lbêinL
0.7 0.8 0.9 1

-0.5

-1

Fig. 6.14 Web shear flow distributions as a function for decreasing web thickness to flange
width ratio.

The assumptions of the skin-stringer approximation are


• The flanges carry all of the bending moment, but no shear force. Note: a flange is considered a stringer.
• Normal stress is uniform over each flange area. In other words, each flange area is assumed to be concentrated
at its centroid for calculation of total section centroid, Ixx, σz, and the first area moment.

172 Thin-Walled Structures


Skin-stringer idealization

• The web carries all of the shear force, but no bending moment.

Only the transfer term in the parallel axis theorem is needed y


to compute the second area moment about the x- axis.

l+d 2 Af = d2
I xx = 2  ----------- A f = 968in 4
 2 
(l + d)/2
The axial normal stress is uniformly distributed over the x
stringer area, and the shear flow is uniformly distributed along
the web. (l + d)/2

Mx l + d Af
σ z = -------  ± ----------- = ± 170.45 psi
I xx  2 

Vy l + d
q = ------  ----------- A f = 454.55 lb/in
I xx  2 

EXAMPLE 6.3 Shear flows in a stringer-stiffened C-section

The C-section shown Fig. 6.15 is stiffened by four stringers and subjected to a shear force with components
V x = 0 and V y = 40 kN . The stringers are assumed to carry only axial forces and each has the same cross-sec-
tional area A s = 150 mm 2 . Each branch has a thickness t, and dimensions h = 80 mm and b = 20 mm .
Determine the shear flow distributions in each branch for t = 5 mm and t = 0.5 mm and plot them. Also
determine the maximum shear stress for each thickness.

As b As
s 3, q 3

y y
h
Vy ---
2

C x C x
S.C.

h
---
2 t, typical s 2, q 2 branch shear flows

s 1, q 1
As b As

Fig. 6.15 C-section stiffened by four stringers

Thin-Walled Structures 173


Shear Flow due to Shear Forces

Solution From eq. (6.13), the shear flow formula applicable to this problem is
Vy
q ( s ) = q ( 0 ) – ------Q x ( s ) (6.43)
I xx

To use this formula, we should first compute the second area moment I xx of the cross section. Using the thin wall
approximations and the composite body technique, the second area moment about the centroidal x-axis is

h 2 h3t h 2
I xx = 2  --- bt + ------- + 4  --- A s
 2 12  2

Note that the stringers are represented by their cross-sectional areas concen-
ε→0 trated at their centroids, so that only the transfer term in the parallel axis theo-
rem affects the second area moment computation for the entire section. Begin
q 1 ( ε )dz
with determining the shear flow in the first branch shown in Fig. 6.15, or the
lower flange. The shear flow at s 1 = 0 is not zero because of the presence of
the stringer. Axial equilibrium of the free body diagram of the lower right
dN1 dz stringer shown in the adjacent sketch gives

lower right stringer dN


q 1 ( 0 ) = – ---------1- (6.44)
dz
But the axial force in the stringer is N 1 = σ z1 A s , and the derivative of this force is d N 1 ⁄ dz = ( dσ z1 ⁄ dz ) A s .

h h
From the flexure formula σ z1 = M x  – --- ⁄ I xx , and the derivative dσ z1 ⁄ dz = V y  – --- ⁄ I xx . Thus,
 2  2

dN Vy
---------1- = ------Q xs1 (6.45)
dz I xx

where Q xs1 is the first area moment of the lower right stringer about the x-axis given by Q xs1 = ( – h ⁄ 2) A s .
Hence, the shear flow at s 1 = 0 is q 1 ( 0 ) = – ( V y ⁄ I xx )Q xs1 . Substitute this result for q 1 ( 0 ) into eq. (6.43) to
get
Vy
q 1 ( s 1 ) = – ------ [ Q xs1 + Q x1 ( s 1 ) ] (6.46)
I xx
s1

The first area moment of the portion of the branch 1 from s 1 = 0 to s 1 is Q x1 ( s 1 ) =


∫ y ( s )t ds
1 1 1, and the y-
0
coordinate to a generic point on the contour is y 1 ( s 1 ) = – h ⁄ 2 . Hence,

h
Q x1 ( s 1 ) = – ---( ts 1 ) 0 ≤ s1 ≤ b
2
Combining these results, the shear flow in branch 1 becomes
Vy h
q 1 ( s 1 ) = ------  --- [ A s + ts 1 ] 0 ≤ s1 ≤ b (6.47)
I xx  2

174 Thin-Walled Structures


Skin-stringer idealization

The shear flow in branch 2, or the web, is determined from the same gen-
eral formula given in eq. (6.43), and the shear flow at s 2 = 0 is determined q 2 ( 0 )dz
separately from the free body diagram of the lower left stringer shown in the
adjacent sketch. Axial equilibrium of the lower left stringer gives
dN q 1 ( b )dz
q 2 ( 0 ) = q 1 ( b ) – ---------2- (6.48)
dN2
dz
The derivative of the axial force in the stringer is obtained following the same lower left stringer
procedure used for lower right stringer discussed above, which resulted in eq.
(6.45). The result for the lower left stringer is

dN2 Vy
---------- = ------Q xs2 (6.49)
dz I xx

where the first area moment about the x-axis of stringer 2 is Q xs2 = ( – h ⁄ 2 ) A s . Calculating the shear flow at the
end of branch 1 from eq. (6.46), using eq. (6.49) for the derivative the axial force in stringer 2, the shear flow at
the beginning of branch 2 from eq. (6.48) is
Vy
q 2 ( 0 ) = – ------ [ Q xs1 + Q x1 ( b ) + Q xs2 ] (6.50)
I xx
The term in brackets on the right-hand-side of eq. (6.50) represents the first area moment about the centroidal x-
axis of the lower right stringer, the lower flange or branch 1, and the lower left stringer. For the web, or branch 2,
the y-coordinate to the generic point s 2 is y 2 ( s 2 ) = – h ⁄ 2 + s 2 , so the first area moment of branch 2 is
s2
t
= --- ( – hs 2 + s 22 )
Q x2 ( s 2 ) =
∫ y ( s )t ds
2 2 2
2
0

It follows from this result, eqs. (6.43) and (6.50), that the shear flow in branch 2 is
Vy h t
q 2 ( s 2 ) = ------ h A s + --- tb – --- ( – hs 2 + s 22 ) 0 ≤ s2 ≤ h (6.51)
I xx 2 2

For the upper flange, or branch 3, the general shear flow formula is again given by eq. (6.43). It follows from
the procedure used to obtain the beginning shear flow in branch 2 that
Vy
q 3 ( 0 ) = – ------ [ Q xs3 + Q x2 ( h ) + Q xs2 + Q x1 ( b ) + Q xs1 ] (6.52)
I xx

The y-coordinate to a generic point on branch 3 is y 3 ( s 3 ) = h ⁄ 2 . Hence, the first area moment of the portion of
branch 3 is
s3
h
Q x3 ( s 3 ) =
∫ y ( s )t ds
3 3 3 = --- ts 3
2
0

It follows from this result and eqs. (6.43) and (6.52), that the shear flow in branch 3 is

Thin-Walled Structures 175


Shear Flow due to Shear Forces

Vy h
q 3 ( s 3 ) = ------  --- [ A s + tb – ts 3 ] 0 ≤ s3 ≤ b (6.53)
I xx  2

Vy h
q 3 ( b )dz Note that the shear flow at the end of branch 3 is q 3 ( b ) = ------  --- A s . We can
I xx  2 
dz check that this is correct by drawing the free body diagram of the upper right
stringer as shown in the adjacent sketch. Axial equilibrium per unit z-coordinate
dN3 gives
upper right stringer dN
– q 3 ( b ) + ---------3- = 0
dz
Similar to the results for the derivatives of the stringer forces given in eqs. (6.45) and (6.49), we have
dN3 Vy
---------- = ------Q xs3 . Axial equilibrium of the upper right stringer is
dz I xx

Vyh Vyh
– ------ --- A s + ------ --- A s = 0
I xx 2 I xx 2

which, of course, is satisfied. If axial equilibrium of the upper right stringer were not satisfied, we would have to
go back over our previous calculations to search for an error.

Numerical evaluation and plots Given the data h = 80 mm , b = 20 mm , A s = 150 mm 2 , and


3
V y = 40 ×10 N , the shear flows from eqs. (6.47), (6.51), and (6.53) evaluate as

3
( 40 ×10 ) ( 6000 + 4ts 1 )
q 1 ( s 1 ) = -------------------------------------------------------
- 0 ≤ s 1 < 20 mm (6.54)
320000
960000 + ------------------ t
3

t
( 40 ×10 )  12000 + 800t + 40ts 2 – --- s 22
3
 2 
q 2 ( s 2 ) = ----------------------------------------------------------------------------------------------- 0 ≤ s 2 ≤ 80 mm (6.55)
320000
960000 + ------------------ t
3
and
3
( 40 ×10 ) ( 6000 + 800t – 40ts 3 )
q 3 ( s 3 ) = ---------------------------------------------------------------------------
- 0 ≤ s 3 ≤ 20 mm (6.56)
320000
960000 + ------------------ t
3

The dimensional units of the shear flows are N/mm for the thickness t and branch contour coordinates speci-
fied in mm. These shear flows are plotted versus the global section contour coordinate s in Fig. 6.16 for t = 5 mm
(solid lines) and for t = 0.5mm (dashed lines). The global contour coordinate is defined by
s = s1 0 ≤ s 1 ≤ 20 mm
s = s 2 + 20 mm 0 ≤ s 2 ≤ 80 mm
s = s 3 + 100 mm 0 ≤ s 3 ≤ 20 mm

Note that the shear flow distributions exhibit jumps at the stringer locations, and that the shear flow distribution is

176 Thin-Walled Structures


Influence of transverse shear deformations on bending

q, Nêmm

500

400

300

200

100

s, mm
20 40 60 80 100 120

Fig. 6.16Shear flows in the stiffened C-section for t = 5 mm (solid lines) and t =
0.5 mm (dashed lines)

more uniform in each branch for the thinner section. The tendency to a spatially uniform shear flow in the thin-
walled branches between stringers corroborates with the earlier results for the thin web as shown in Fig. 6.14.
From inspection of the Fig. 6.16, the maximum shear flow occurs at the center of the web. Thus,

 535.7 N/mm t = 5 mm
q max = q 2 ( 40 ) = 
 505.3 N/mm t = 0.5 mm

The maximum shear flow is reduced in the thinner section with respect to the thicker section. The maximum
shear stress is estimated from τ max = q max ⁄ t . Hence,

 107 MPa t = 5 mm
τ max = 
 1011 MPa t = 0.5 mm

The maximum shear stress occurs for the thinner section. N.B. A shear stress magnitude of 1011 MPa is very
large. For example, aluminum alloy 2024-T4 has an ultimate shear stress of about 280 MPa. Clearly, an alumi-
num section of this alloy with t = 0.5 mm would fail.

6.5 Influence of transverse shear deformations on bending


Classical beam theory neglects the effects of transverse shear stresses on the bending deformation of slender
beams. That is, classical theory assumes that plane cross sections before deformation remain plane and perpen-
dicular to the neutral axis (z-curve) in the deformed beam. For thin-walled beams the effect of transverse shear-
ing deformation can be significant even for slender beams. An approximate method to account for transverse
shearing deformations is to assumed that cross sections remain plane but not necessarily perpendicular to the
neutral axis in the deformed beam. Again, we will assume the flexure formula for the bending normal stress
obtained from pure bending is sufficiently accurate in the presence of transverse shear, and consequently the

Thin-Walled Structures 177


Shear Flow due to Shear Forces

shear flow formula derived from equilibrium conditions using the flexure formula is also sufficiently accurate.
The beam theory that is developed in this section is sometimes called Timoshenko beam theory to distinguish it
from classical beam theory.

6.5.1 Transverse shear strains, forces, and complementary energy density

To motivate how we will account for transverse shearing deformations, consider the expression we obtained
for the internal complementary virtual work in Section 5.3. The internal complementary virtual work is given by
the right-hand-side of eq. (5.49) on p. 142 . Repeating this result we have
L
* =  du + θ  δV + dθ y δM +  dv + θ  δV + dθ x δM dz
δW int
∫ dz y x
dz y dz x y
dz x (6.57)
0

The factors multiplying the virtual shear forces in this equation, which are zero in the classical theory, represent
beam shear strains. Let ψ x denote the beam shear strain in the x-z plane and let ψ y denote beam shear strain in
the y-z plane. That is, we define

du dv
ψx ≡ + θy ψy ≡ + θx (6.58)
dz dz
These transverse shear strains ψ x and ψ y represent the reduction in the right angles between lines elements in
the deformed beam that were originally parallel to the x-axis and z-axis, and the y-axis and z-axis, respectively, in
the undeformed beam. The displacement gradients du ⁄ dz and dv ⁄ dz , and rotations θ y and θ x are assumed to
be very small in magnitude with respect to unity. The beam shears are depicted in Fig. 6.17. Classical beam the-
ory is characterized by ψ x = 0 and ψ y = 0 for all values of z.

θy θx

x y
du dv
w ------ w ------
dz dz
π π
--- – ψ x v --- – ψ y
u 2
2
z z
(a) x-z plane (b) y-z plane
Fig. 6.17 Transverse shear strains in (a) the x-z plane, and (b) the y-z plane

For an elastic material the integrand of eq. (6.57) is identified as the variation of the complementary strain
energy per unit length of the beam, or the complementary strain energy density U 0* . Using the definitions of
beam shear strains, the variation in the complementary strain energy density is written as

dθ y dθ x
δU 0* = ψ x δV x + δM y + ψ y δV y + δM x (6.59)
dz dz

178 Thin-Walled Structures


Influence of transverse shear deformations on bending

We can identify the terms in this expression for the complementary strain energy density as originating from
transverse shear deformation or bending deformation. Decompose the complementary strain energy density as

U 0* = U 0* s + U 0* b (6.60)

where U 0* s is the complementary strain energy density due to transverse shear deformation and U 0* b is the com-
plementary strain energy density due to bending deformation. Compare the variation of eq. (6.60) to eq. (6.59) to
get

δU 0* s = ψ x δV x + ψ y δV y (6.61)

and

dθ y dθ x
δU 0* b = δM y + δM x (6.62)
dz dz

The derivation of the complementary strain energy density due to bending for a linear elastic, homogeneous
material was obtained in Section 5.3 as eq. (5.66) on p. 144 . We do not need to repeat this derivation here, since
this beam theory, as does classical beam theory, assumes that the bending normal stresses are given by flexure
formula even in the presence of transverse shear deformations. Neglecting the thermal strain terms, the comple-
mentary strain energy density due to bending is obtained as
1
U 0* b = ------------------------ ( I yy M x2 – 2I xy M x M y + I xx M y2 ) (6.63)
2Edet ( I )
2 and E is the modulus of elasticity.
where det ( I ) = I xx I yy – I xy

To obtain an expression for the complementary strain energy density due to shear for a linear elastic mate-
rial, the form of eq. (6.61) suggests that there is a functional of the shear forces, U 0* s [ V x, V y ] , whose variation by
definition is

∂U 0* s ∂U 0* s
δU 0* s [ V x, V y ] = ------------ δV x + ------------ δV y (6.64)
∂V x ∂V y
Comparing eqs. (6.61) and (6.64), the beam shear strains can be identified as

∂U 0* ∂U 0*
ψ x = ------------s ψ y = ------------s (6.65)
∂V x ∂V y
For linear material behavior the beam shear strains are linearly related to the shear forces by expressions of the
form
ψ x = c xx V x + c xy V y
(6.66)
ψ y = c yx V x + c yy V y

where the coefficients c xx , c xy , c yx , and c yy are shear compliances of the beam’s cross section having dimen-
sional units of 1/F. From eqs. (6.65) and (6.66), the complementary strain energy density due to shear is deter-
mined to be
1
U 0* s = --- ( c xx V x2 + 2c xy V x V y + c yy V y2 ) (6.67)
2

Thin-Walled Structures 179


Shear Flow due to Shear Forces

where c xy = c yx if eq. (6.66) is to be derived from eq. (6.67) using the relations in eq. (6.65). The question now
is how do we obtain expression for the shear compliances c xx , c xy , and c yy ? To derive these expressions we
return to the free body diagram of an element of the wall of the beam as was used to derived the shear flow for-
mula in Section 6.2 given as eq. (6.13) on p. 160 .

6.5.2 Complementary energy density obtained from a two-dimensional element of the wall

The complementary strain energy per unit length of the beam is derived again in this subsection, but we start
from a differential element cut from the wall of the beam of dimensions ds × dz × t , where t is the wall thick-
ness. Let n z denote the normal stress resultant defined as the integral of the normal stress over the thickness of
the wall; i.e.,
t⁄2

nz =
∫ σ z dn (6.68)
–t ⁄ 2

where n is the coordinate normal to the contour. If the wall is thin with respect to overall cross-sectional dimen-
sions of the beam, the distribution of the normal stress over the thickness of the wall can be ignored and the nor-
mal stress is represented by a uniform distribution through the thickness equal to its value on the contour. For
thin walls then, the normal stress resultant is approximated by n z ≈ σ z t where σ z is evaluated on the contour at s.
A free body diagram of an element of the wall of the beam is shown in Fig. 6.18. The stress resultant n s acting

s, v t
 q + ∂q
------ ds dz
y n z ds  ∂s 
n
s, v t
qds  q + ∂q
n ds ------ dz ds
x  ∂z 
θ dz
qdz ∂n z 
 n + -------
- dz ds
 z ∂z 
z, w

Fig. 6.18 Free body diagram of an infinitesimal element of the beam wall.

normal to the s-face of the element is assumed to be negligible with respect to the axial resultant n z in beam the-
ory, and so it is not shown in the free body diagram. Sum the forces acting in the z-direction to zero yields
∂n z 
 n + ------- ∂q
- dz ds – n z ds +  q + ------ ds dz – qdz = 0
 z ∂z   ∂s 
In the limit as the element shrinks to zero dimensions we obtain from this equation the differential equation of
equilibrium
∂n z ∂q
-------- + ------ = 0 (6.69)
∂z ∂s

180 Thin-Walled Structures


Influence of transverse shear deformations on bending

Summing forces to zero in the s-direction yields

 q + ∂q
------ dz ds – qds = 0
 ∂z 
In the limit as the element shrinks to zero dimensions, we obtain from s-direction equilibrium that
∂q
------ = 0 (6.70)
∂z

The displacement of the element in the z-direction is denoted w ( z, s ) and the displacement in the s-direc-
tion, or in the direction tangent to the contour is denoted by v t ( z, s ) .The element strains of interest are the axial
normal strain
ε z = ∂w ⁄ ∂z (6.71)

and the shear strain


γ zs = ∂w ⁄ ∂s + ∂v t ⁄ ∂z . (6.72)

The shear strain is depicted in the adjacent sketch. These displacements must be
continuous and single valued so that no gaps or overlaps of material occur in the s, v t π ⁄ 2 – γ zs
deformed configuration. Continuous, single-valued displacements and strains
satisfying the strain-displacement relations (6.71) and (6.72) are said to be com- ∂w
-------
patible. ∂s ∂v t
-------
∂z
Now assume the beam element has displacements and rotations that are
compatible and satisfy eqs. (6.71) and (6.72). From this compatible deformation z, w
state consider a virtual change in the normal stress resultant δn z and a virtual wall element shear strain

change in the shear flow δq , that satisfy equilibrium conditions (6.69) and
(6.70); i.e.,
∂ ( δn z ) ∂ ( δq )
---------------- + -------------- = 0 (6.73)
∂z ∂s
and
∂ ( δq )
-------------- = 0 (6.74)
∂z
Equation (6.74) implies that the virtual shear flow δq is spatially uniform in the z-coordinate. A shear flow that is
uniform along the length of the beam occurs in the case of a cantilevered beam subjected to transverse forces at it
tip with no distributed loads acting on the beam. Hence, we regard the virtual force system acting on the beam to
consist of only virtual transverse forces acting at the tip of a cantilevered beam, so that the virtual shear forces
δV x and δV y are independent of z. Although this may seem like a limiting situation in which to derive the shear
compliances, in the approximate transverse shear theory being developed here the shear compliances obtained
for this case are assumed applicable to other beam loading situations as well. The complementary virtual work
per unit area of the element is denoted by δÛ 0* , and is determined by the displacements acting through the vir-
tual stress resultants. That is, the complementary virtual work of the stress resultants acting on the infinitesimal
element is written as (refer to Fig. 6.18)

Thin-Walled Structures 181


Shear Flow due to Shear Forces

z + dz z + dz s + ds
δÛ 0* dsdz = [ w ( δn z ds ) ] + [ v t ( δqds ) ] + [ w ( δqdz ) ] s
(6.75)
z z

Divide this equation by the differential area dsdz , and take the limit as ds → 0 and dz → 0 to get
∂ ∂
δÛ 0* = [ wδn z + v t δq ] + [ wδq ]
∂z ∂s
Distribute the derivative in this last expression and write it as

∂ ( δn z ) ∂ ( δq ) ∂ ( δq ) ∂w ∂v ∂w
δÛ 0* = w  ---------------
- + -------------- + v t -------------- + ------- δn z +  -------t + ------- δq (6.76)
 ∂z ∂s  ∂z ∂z  ∂z ∂s 
Since the virtual stress resultants satisfy equilibrium conditions (6.73) and (6.74), the first two terms on the right-
hand side of eq. (6.76) vanish. Hence,

δÛ 0* = ε z δn z + γ zs δq (6.77)

in which the axial normal strain and the shear strain are identified from eqs. (6.71) and (6.72). For a Hookean
material the strain-stress relations are
1 q
ε z = -----n z γ zs = ------ (6.78)
Et Gt
where E is the modulus of elasticity, t the wall thickness, and G is the shear modulus. Using this material law in
eq. (6.77) we get that the complementary strain energy per unit area is
1 1
δÛ 0* = -----n z δn z + ------ qδq (6.79)
Et Gt
Equation (6.79) is integrated over the contour of the cross section from s = 0 to s = S , where S is the arc
length of the contour, to obtain the complementary strain energy per unit length of the beam. That is, the varia-
tion of the complementary strain per unit length is determined from the variation of the complementary strain
energy per unit area by the definite integral
S
*
δU 0* =
∫ δÛ ds 0 (6.80)
0

Substitute eq. (6.79) into this result to get


S
1 1
δU 0* =
∫ -----n z δn z + ------ qδq ds
Et Gt
(6.81)
0

The two terms in this complementary strain energy density expression can be identified with bending and shear-
ing deformations. The bending deformation effect is represented by the term
S
nz
δU 0* b =
∫ -----δn
Et
ds z (6.82)
0

Recall that n z = σ z t in the thin wall approximation, and that the bending normal stress σ z is given by the flex-
ure formula, eq. (3.28) on p. 77 . If we substituted the flexure formula for σ z and its variation into eq. (6.82) and
carried out the details to obtain the complementary strain energy density due to bending, then we would obtain

182 Thin-Walled Structures


Influence of transverse shear deformations on bending

eq. (6.63). A result we have already derived. It is the second term in eq. (6.81) that is of interest here. This second
term is the variation of the complementary strain energy density due to transverse shear deformation given by
S
q
δU 0* s =
∫ -----
Gt
- δq ds (6.83)
0

6.5.3 Determination of the transverse shear compliances


Two expressions for the variation of the complementary strain energy density due to transverse shear deforma-
tion have been derived; eqs. (6.61) and (6.83). Equating these results we have
S
q
ψ x δV x + ψ y δV y =
∫ -----
Gt
- δq ds (6.84)
0

From axial equilibrium of an infinitesimal element of the wall of the beam, and using the flexure formula for the
bending normal stress, we derived the shear flow formula which is given in eq. (6.13). We write this formula in a
slightly different form from what is given in eq. (6.13) as
I xx Q y ( s ) – I xy Q x ( s ) I yy Q x ( s ) – I xy Q y ( s )
q ( s ) = q ( 0 ) – ------------------------------------------------ V x – ------------------------------------------------ V y (6.85)
det ( I ) det ( I )
For an open section having a free edge, the origin of the contour coordinate s can be taken at the free edge so that
q ( 0 ) = 0 . If a stringer is located at the contour origin of a branch, then the shear flow q ( 0 ) is obtained in terms
of shear forces V x and/or V y as is shown in Example 6.3. Consequently, we can write eq. (6.85) in the form

q ( s ) = – K x ( s )V x – K y ( s )V y (6.86)

where
I xx Q y ( s ) – I xy Q x ( s )
K x ( s ) = -----------------------------------------------
- (6.87)
det ( I )
and
I yy Q x ( s ) – I xy Q y ( s )
K y ( s ) = -----------------------------------------------
- (6.88)
det ( I )
The variation of the shear flow, or virtual shear flow, is obtained from eq. (6.86) as
δq = – K x ( s )δV x – K y ( s )δV y (6.89)

Substituting eqs. (6.86) and (6.89) into the right-hand side of eq. (6.84) yields
S
( – K x ( s )V x – K y ( s )V y )
ψ x δV x + ψ y δV y =
∫ --------------------------------------------------------
Gt
[ – K ( s )δV
x x – K y ( s )δV y ] ds (6.90)
0

Equation eq. (6.90) must hold for every virtual force δV x and δV y . Recall that we are considering a cantilever
beam subject to virtual forces equal to δV x and δV y applied at its tip. These virtual forces are independent.
Hence, we conclude from eq. (6.90) that

Thin-Walled Structures 183


Shear Flow due to Shear Forces

S
( K x2 ( s )V x + K x ( s )K y ( s )V y )
ψx =
∫ ------------------------------------------------------------------- ds
Gt
(6.91)
0

and
S
( K y ( s )K x ( s )V x + K y2 ( s )V y )
ψy =
∫ ------------------------------------------------------------------- ds
Gt
(6.92)
0

Compare these expressions for the beam shear strains to those given in eq. (6.66), to finally get the expressions
for the shear compliances. These compliance formulas are
S S
K x2 ( s ) K x ( s )K y ( s )
c xx =
∫ -------------- ds
Gt
c xy =
∫ ---------------------------
Gt
- ds (6.93)
0 0

and
S S
K y ( s )K x ( s ) K y2 ( s )
c yx =
∫ ---------------------------- ds
Gt
c yy =
∫ -------------- ds
Gt
(6.94)
0 0

The shear compliances given in the above expressions depend on the shear modulus of the material, and the
geometry of the cross section. Note that the dimensional units of the cross-sectional functions K x ( s ) and K y ( s )
are 1/L. The dimensional units of the shear modulus are F/L2, so that the dimensional units of the shear compli-
ances are 1/F.

EXAMPLE 6.4 Shear compliances of a stiffened blade section

Determine the shear compliances for the blade section stiffened by two
y
stringers as shown in Fig. 6.19. The section is made of an isotropic, homo-
geneous material with a shear modulus G. Also, determine the influence of
As increasing the stringer area 2 A s relative to the web area ht on the shear
h h»t>0 stiffness.
---
2
C x
Solution The second area moments for this thin-walled section are
h
--- t h3t h 2
2 I xx = ------- + 2  --- A s I yy ≈ 0 I xy = 0 (6.95)
12  2
As
Since the second area moment about the y-axis is zero in the thin wall
Fig. 6.19 approximation, this section can only resist bending in the y-z plane. Hence,
we consider beam loading such that V y ≠ 0 and V x = 0 . Let the origin of
the contour coordinate s in the web be located at the lower stringer and take the positive direction of s in the pos-
itive y-direction. The cartesian coordinates of point s are
h
x(s) = 0 y ( s ) = – --- + s 0≤s≤h
2
The first area moments of the portion of the cross section below s are

184 Thin-Walled Structures


Influence of transverse shear deformations on bending

s
h h ht t
Q x ( s ) = – --- A s + y ( s )t ds = – --- A s – ----- s + --- s 2
2 ∫ 2 2 2
0

and
s

Qy(s) =
∫ x ( s )t ds = 0
0

From eq. (6.87) the function K x ( s ) = 0 , and from eq. (6.88) the function K y ( s ) = Q x ( s ) ⁄ I xx . Hence, the for-
mulas for the shear compliances given in eqs. (6.93) and (6.94) become
c xx = c xy = c yx = 0

and
h
1 Qx(s) 2
c yy = ------
Gt ∫ -------------
I xx
- ds
0

The factor Gt is factored out of the integral in the equation for compliance c yy , since the section is homogeneous
and the thickness of the web is uniform. Substitute for the function Q x ( s ) in the expression for the compliance
c yy to get
h
1 h ht t 2
– --- A s – ----- s + --- s 2 ds
c yy = -------------
GtI xx 2 ∫ 2 2 2
0

Performing the definite integral in this equation results in

1  h 3 A s2 h 4 t A s h 5 t 2
c yy = ------------- ------------ + ------------- + ---------
2  4
(6.96)
GtI xx 12 120 

where the second area moment about the x-axis through the centroid I xx given by the first of eqs. (6.95). Equa-
tion (6.96) is the result sought. The only non-zero transverse shear compliance is c yy , so the material law for
shear deformation of the beam in the y-z plane is ψ y = c yy V y . We write the inverse of this material law as
V y = k yy ψ y , where k yy = 1 ⁄ c yy is the transverse shear stiffness.

To study the influence of the stringers on the transverse shear stiffness k yy , we define the ratio of the cross-
sectional area of the stringers to the total cross-sectional area A by
α = ( 2 As ) ⁄ A

where A = ht + 2 A s . From the definition of α , the portion of the cross-sectional area in the web is

( ht ) ⁄ A = 1 – α
For α = 0 the stringers are not present ( A s = 0 ), and the web resists both bending normal stress and shear
stress. In the extreme case of α = 1 the web thickness t = 0 , and the stringers carry all the bending load with

Thin-Walled Structures 185


Shear Flow due to Shear Forces

the web carrying only shear force V y . Introducing the definition of the stringer area and web area in terms of α
and the total area A into the expressions for I xx and k yy we get after some algebra

h2 A
I xx = --------- ( 1 + 2α ) (6.97)
12
and

5 ( 1 – α ) ( 1 + 2α ) 2
k yy = --- GA ----------------------------------------- (6.98)
6 1 + 3α + 3.5α 2

When there are no stringers ( α = 0 ), the shear stiffness of the blade section is 5GA ⁄ 6 . The shear stiffness of
the section decreases, or the section becomes more flexible in shear, as the stringer area increases as a larger por-
tion of the total area ( α increasing from zero, but remaining less than one). A plot of the transverse shear stiff-
ness versus α is shown in Fig. 6.20.

kyy êHGAL
2A
0.8 α = ---------s
A
ht
0.6 ----- = 1 – α
A

0.4

0.2

a
0.2 0.4 0.6 0.8 1

Fig. 6.20 Transverse shear stiffness of the blade section as the area of the stringers increases relative
to the web area.

EXAMPLE 6.5 Deflection of a cantilevered beam due to bending and shear deformation.

Consider a cantilever beam of length L = 50 in. subjected to a vertical tip force F as shown in Fig. 6.22. The
cross section of the beam is the same as in Example 6.5, and the cross-sectional dimensions shown in Fig. 6.19
are h = 10 in. , t = 0.02 in. , and A s = 0.5 in. 2 . The moduli of the material are E = 10 Msi and
G = 3 Msi . Use Castigliano’s second theorem to determine the tip portion of the tip displacement ∆ due to
shear deformation of the web.

Solution Castigliano’s theorem applied to this problem is

∂U *
∆ = ----------
∂F

The complementary strain energy U * is due to bending in the y-z plane and transverse shear in the same plane.
This complementary energy is

186 Thin-Walled Structures


Influence of transverse shear deformations on bending

y y
F, ∆

x
z

Fig. 6.22 A cantilever beam subjected to a tip force

L L
M x2 V y2
U* =
∫ -------------
2EI xx
- dz +
∫ ---------
2k yy
- dz
0 0









bending shear

in which the second area moment I xx is given by eq. (6.97), and the transverse shear stiffness k yy by eq. (6.98),
in Example 6.5. Castigliano’s theorem yields
L L
M x ∂M x V y ∂V y
∫ ∫
∆ = ---------- ---------- dz + ------- --------- dz
EI xx ∂F k yy ∂F
0 0

Equilibrium determines the bending moment in terms of the tip force by the equation M x = – ( L – z )F , and
the shear force in terms of the tip force by V y = F . Substitute these results into Castigliano’s theorem to get

L L
1 1
∆ =  ---------- [ – ( L – z )F ] [ – ( L – z ) ] dz +  ------- [ F ] [ 1 ] dz
 EI xx ∫  k yy ∫
0 0

Performing the integrals in the above expression we obtain


F L3 FL
∆ = -------------- + -------
3EI xx k yy




bending shear

Let ∆ b denote the portion of the tip displacement due to bending, and let ∆ s denote the portion due to shear. The
ratio of the displacement due to shear to the displacement due to bending is
∆ 3EI xx FL
------s =  -------------
- -------
∆b  F L 3  k yy

Substitute into this expression eq. (6.97) for I xx and eq. (6.98) for k yy , and perform some algebra to get

∆ 3 E h 2 ( 1 + 3α + 3.5α 2 )
------s = ------  ----  --- -----------------------------------------
∆b 10  G  L ( 1 – α ) ( 1 + 2α )

For the specified data the factor α is

Thin-Walled Structures 187


Shear Flow due to Shear Forces

2 As 2 × 0.5
α = ------------------------
- = -------------------------------------------------- = 0.8333
( ht + 2 A s ) ( 10 × 0.02 + 2 × 0.5 )

The ratio of tip displacements evaluates as

∆ 3 10Msi 10in. 2 1 + 3 ( 0.8333 ) + 3.5 ( 0.8333 ) 2


------s = ------  ---------------  ------------  ------------------------------------------------------------------------ = 0.5337 (6.99)
∆b 10  3Msi   50in.  ( 1 – 0.8333 ) ( 1 + 2 ( 0.8333 ) ) 
The percentage of the tip displacement due to shear relative to the total tip displacement is given by the expres-
sion
∆s ∆s ⁄ ∆b
----------------- × 100 = ------------------------ × 100 = 34.8%
∆s + ∆b ∆s ⁄ ∆b + 1

Hence, almost 35% of the tip displacement is due to shearing deformation of the web.

6.6 Problems
1. The thickness of each branch in the thin-walled
y
cross section shown below is 3mm and
I xx = 10 5 mm 4 . The shear force V y = 5 kN .

40 a)Determine the shear flow distribution and sketch


x it on the cross section. Indicate on the sketch the
C
A 10 positive sense along the branch.
50 30 30 50 b)Estimate the shear stress due to transverse bend-
ing at point A.
Note: all dimensions in mm
c)Estimate the maximum shear stress due to trans-
verse bending.

2. Consider the thin-walled, Y-section beam of Problem 1 Section 3.6, in


y
which we found I xx = 2.0238l 3 t . Also, I yy = l 3 t . Determine the mag-
nitude, direction, and sense of the shear flow at the centroid if the shear
45° 45°
force components are V x = 0 and V y > 0 .

C x
0.6414l

l l

Note: the length of each branch is l


and the thickness of each branch is t

188 Thin-Walled Structures


Problems

3. Determine the location of the shear center e from point A for


the thin-walled, open section shown. a


S.C. α A axis of symmetry
α

t

4. Determine the coordinates (ex,ey), with respect to point O, of the shear center (S.C.) for the thin-walled open
section shown below. The second area moments with respect to the centroidal axes x and y are

I xx = 0.660155 in. 4 I yy = 0.130305 in. 4 I xy = – 0.127197 in. 4

y
1.5
( x c, y c ) = ( 1.02786, 0.529989 ) y
y S.C.
ey
C x
O x O x
ex
2

t = 0.030 typ.
All dimensions in inches

Thin-Walled Structures 189


Shear Flow due to Shear Forces

190 Thin-Walled Structures


CHAPTER 7 Bars Subjected to
Torsional Loads

7.1 Uniform torsion of a circular tube


Consider uniform tube, or cylindrical shell, of length L subjected to a torque T at each end as shown in Fig. 7.1.

y y
a
b
T z T x

T , θz
L

Fig. 7.1 Circular tube subjected to uniform torsion

The inside radius of the tube is denoted by a and the outside radius by b. Equilibrium requires the internal torque
to be equal to applied torque T at each cross section z = constant, where 0 < z < L . Using the right-hand screw
rule, the sign convention for a positive internal torque is that a positive torque acts in the positive z- direction on
a positive z-face, and by the action/reaction law a positive torque acts in the negative z-direction on a negative z-
face. The tube is assumed to made of a homogeneous, isotropic, linear elastic material.

The geometry of deformation of the tube is based on symmetry of the undeformed configuration, symmetry
of the loading, and on isotropic material behavior. Three symmetry arguments concerning the geometry of the
deformation are
1. The pattern of deformation will not vary along the length of the tube.
Any element cut perpendicular to the z-axis of the tube, say of length ∆z, will have the same original geome-
try and same loading. Thus, we expect each element to have the same deformation.

Thin-Walled Structures 191


Bars Subjected to Torsional Loads

2. Each element has a midplane of symmetry in the sense that the deformed geometry at each end must be the
same.
Consider the element shown in Fig. 7.2(a). If we rotate it and the torques 180 degrees about the x-axis we
get the element shown in Fig. 7.2(b). The elements (a) and (b) are identical in original shapes and loadings.
Thus, the deformed geometry of the left end of (b) must be the same as the left end of (a). But the deformed
geometry of the left end of (b) must be identical to the deformed geometry of the right end of (a), since dia-
gram (b) is a simple rotation of (a).

D A

B C
C B
x
∆z ∆z
180°
A D
(a) (b)

Fig. 7.2 Symmetry about the midplane of an element of the torsion tube.

3. All radial lines in the undeformed cross section must deform into identical curves. This is a result of axial
symmetry (circular cross section symmetric with respect to the z-axis), and isotropy. Note that tube with a
rectangular cross section would not be axially symmetric.

First, we argue that plane and normal sections in the undeformed tube remain plane and normal to the z-axis
in the deformed tube.
On the basis of symmetry argument (3) end surfaces of elements must dish-in or bulge-out as a surface of
revolution, or possibly remain plane as a result of the applied torque. Also symmetry argument (2) implies
both ends do the same. See Fig. 7.3. Further (1) implies each element has the same deformation. But ele-
ments which bulge-out or dish-in cannot fit together in the deformed geometry to form a continuous tube.
Thus, plane sections remain plane.

dish-in bulge-out

∆z ∆z

Fig. 7.3 Potential axial symmetric end surface deformations.

Second, we argue that each cross section rotates without distortion in its plane.
Consider the element of the tube shown in Fig. 7.4(a). It is assumed a diameter deforms in the left-end plane

192 Thin-Walled Structures


Uniform torsion of a circular tube

as shown. Each diameter would have the same shape by symmetry argument (3). By symmetry argument (2)
diameters on the right-end plane should have the same shape. Rotate the element in Fig. 7.4(a) 180 degrees
about the x-axis to get Fig. 7.4(b). The elements in Fig. 7.4(a) and Fig. 7.4(b) have the same original geom-
etry and loading, so the assumed shape of the deformed element should be the same for both. But it is not. If
the family of diameters in the deformed geometry were straight lines, there would be no conflict with sym-
metry. Hence, diameters in the undeformed tube remains straight in the deformed geometry, but they may
rotate in their plane. Due to symmetry argument (3) each diameter in the cross section rotates through the
same angle.

T T

x
∆z ∆z
180°

T T
(a) (b)
Fig. 7.4 A diameter in the end plane of the undeformed body cannot distort from a straight line in
the deformed body due to problem symmetry.

Symmetry of deformation has not ruled out symmetrical


expansion or contraction of the circular cross section, or a y
lengthening or shortening of the tube. It does not seem plau- r∆θ
sible that such dilational deformations would be an important ∆r
part of the deformation. Thus, extensional strains of line seg- ∆z r
ments in the axial direction (z-axis), radial direction (r), and
the circumferential direction ( θ ) are assumed zero. See Fig. θ
7.5. That is extensional strains, ε z = 0 , ε r = 0 , and x
εθ = 0 . z

The only motion of a circular cross section is, then, a Fig. 7.5 Cylindrical coordinates and
differential line elements.
rotation about the z-axis. Thus, the right angle between line
elements originally parallel to the r- and θ -directions does
not change, which means the shear strain γ rθ = 0 . Also, for small strains the right angle between the line ele-
ments originally parallel to the r- and z-directions does not change very much, so shear strain γ rz = 0 . In cylin-
drical coordinates the only non-zero strain is γ θz . To compute this shear strain consider an element of the tube of
length ∆z and radius r, where a ≤ r ≤ b . According to the symmetry arguments presented above, a straight line
on the periphery of the tube parallel to the z-axis will deform into a helix. Line AB is a portion of such a straight
line, and when loads are applied this line moves to become A’B’ as shown in Fig. 7.6. For very small ∆z the seg-
ment A’B’ can be considered a straight line segment forming the angle γ θz with line A’C’ taken parallel to line
AB. This angle is the change of the right angle between originally perpendicular elements ∆z and r∆θ , and is the

Thin-Walled Structures 193


Bars Subjected to Torsional Loads

shear strain. From the diagram in Fig. 7.6, we have the relationship between the shear strain and twist of the tube

γ θz
A’
θz

B’ A
r θ
∆θ z C’ x
B
z

∆z

Fig. 7.6 Reduction in the right angle between line elements originally parallel to the
axial and circumferential directions due to twist of the body.

as
r∆θ z = γ θz ∆z

which in the limit as ∆z → 0 gives


γ θz = r ------ (7.1)
dz
This result shows that the shear strain varies in direct proportion to the radius. Since each element of the tube of
length ∆z deforms in the same way as any other, we conclude dθ z ⁄ dz is a constant along a uniform section of
the tube. The quantity dθ z ⁄ dz is the twist per unit length or simply the unit twist.

We use Hooke’s law to relate the shear stress to the shear strain, since we have assumed the material is linear
elastic and isotropic. Hooke’s law is
τ θz = Gγ θz (7.2)

where G denotes the shear modulus of the material. For an isotropic material recall that
E
G = --------------------
2(1 + ν)
where the modulus of elasticity is denoted by E and Poisson’s ratio is denoted by ν. Eliminating the shear strain
by eq. (7.1), we get
dθ z
τ θz = Gr -------- (7.3)
dz

The stress distribution given by eq. (7.3) leaves the outside and inside cylindrical surfaces of the tube stress
free, as it should. Internally, each differential element of the tube is in equilibrium because τ θz does not change

194 Thin-Walled Structures


Uniform torsion of a circular tube

in the θ -direction (axial symmetry) nor does it change in the z-direction (because of uniformity of deformation
pattern along the length of the tube). The shearing stress is the same on each z- and θ -face. Hence the differential
element is in equilibrium. See Fig. 7.7.

y r
θ
τ θz
τ zθ
τ zθ
z x
θ
z τ zθ = τ θz

T
Fig. 7.7 Shear stresses due to torque

On the end surface of the tube the resultant of the stress distribution must be equal to the applied torque T.
Because of rotational symmetry of the stress distribution, the resultant force due to the shear stresses must be
zero. The only resultant of the stress distribution is the torque

T =
∫ ∫ r(τ
A
zθ ) d A (7.4)

where dA is the element of area and the integral is taken over the cross-sectional area A of the tube. Substituting
eq. (7.3) into eq. (7.4) we get
dθ z
T = GJ -------- (7.5)
dz
where the torsion constant J is defined as
b 2π
π
r 2dA = r 3 ( dθ ) dr = --- ( b 4 – a 4 )
J =
∫∫
A
∫∫
a 0 2
(7.6)

The torsion constant J is the same as the polar area moment for a circular annulus, and has dimensional units of
L4. From eq. (7.5) we find the twist per unit length is
dθ z T
-------- = ------- (7.7)
dz GJ
For tube of length L, the total angle of twist between the ends is found by integrating eq. (7.7); i.e.,
L
T TL
θz ( L ) – θz ( 0 ) =
∫ 0
------- dz = -------
GJ GJ
(7.8)

Now substitute eq. (7.7) into eq. (7.3) to find


Tr
τ zθ = ------ a≤r≤b (7.9)
J

Thin-Walled Structures 195


Bars Subjected to Torsional Loads

Thin-walled approximations Let the mean radius be denoted by r and the wall thickness be denoted by t. The
relationships between the inside and outside radii to the mean radius and wall thickness are
t
a = r – --- 1
2 r = --- ( a + b )
or the inverse 2 (7.10)
t
b = r + --- t = b–a
2
t t 1
A thin-walled tube is quantified by 0 < - « 1 ; e.g. - ≤ ------ . The torsion constant, eq. (7.6), becomes
r r 20

π t 4 t 4 πr 4 t 4 t 4
J = ---  r + --- –  r – --- = --------  1 + ----- –  1 – -----
2  2   2  2  2r   2r

πr 4 t t 2 t 3 t 4
J = --------  1 + 4  ----- + 6  ----- + 4  ----- +  -----  –
2   2r   2r   2r   2r 

t
 1 – 4  ----- t 2 t 3 t 4
+ 6  ----- – 4  ----- +  ----- 
  2r  2r  2r  2r 

π t t 3 t 2
J = --- r 4 8  ----- + 8  ----- = 2πr 3 t 1 +  -----
2  2r  2r  2r

t
Since the quantity 0 <  ----- « 1 , we can neglect this factor squared with respect to one in the above equation to
 2r
get

J = 2πr 3 t thin-walled circular tube (7.11)

The variation of the shear stress, eq. (7.9), over the annular section is small for a thin wall. That is, we can
write eq. (7.9) as

Tr r t r t
τ zθ = ------  -  1 –  ----- ≤ - ≤ 1 +  -----
J  r  2r r  2r

which shows that the range of r ⁄ r is small for a thin-walled section and can be neglected. Thus, the shear stress
is approximated as a uniform distribution across the wall equal to its value at the mean radius.
Tr
τ zθ = ------ thin-walled circular tube (7.12)
J
Since the shear stress is assumed to be uniform across the wall thickness, we define the shear flow due torsion q
as q = τ zθ t . Substituting eq. (7.12) for the shear stress and eq. (7.11) for the torsion constant, the shear flow is
written as
T rt T
q = -------------
3
- = -------- (7.13)
2πr t 2¿

where the area enclosed by the mean radius is ¿ = πr 2 . Equation eq. (7.13) is more general than as presented in
this section for the circular tube. It is applicable to the torsion of thin-walled sections other than the circular tube.
Equation (7.13) is called the Bredt-Batho formula. These torsion results for the thin-walled circular tube are

196 Thin-Walled Structures


Uniform torsion of an open section

depicted in Fig. 7.8., in which we used the contour coordinate s rather than the angle θ to write the shear stress
as τ zθ = τ zs

y T
q = --------
2¿ τ zs = q ⁄ t

r
s = rθ
θ
x
t
T

t
¿
Fig. 7.8 Shear flow and shear stress in a thin-walled circular tube subjected to a torque.

7.2 Uniform torsion of an open section


Consider uniform torsion of the thin-walled rectangular section shown in Fig. 7.9. The correct solution to the

y
T

t
0 < --- « 1 x
b
z
t T
L

b
Fig. 7.9 Uniform torsion of a bar with a rectangular cross section.

problem of uniform torsion of bars with non-circular cross sections was first obtained by Saint-Venant (1855).
The procedure used by Saint-Venant, called the semi-inverse method, was to assume the form of the displace-
ments for the non-circular bar based on the known displacements for the circular section bar. Elasticity theory is
required to fully understand the details of the solution, but elasticity theory is beyond the scope of this text.
Hence, we will only summarize the important aspects of this solution here. The reader is referred to the texts by
Timoshenko and Goodier (1970) and Oden and Ripperger (1981) for in-depth discussions.

The deformed bar is shown in Fig. 7.10. As for the circular section under positive torque, the cross section a
z + dz rotates counter-clockwise relative to the section at z (Fig. 7.10a), but unlike the circular section bar the
cross section does not remain plane. For non-circular cross sections, the cross section displaces out of its plane as
well as rotates about the z-axis. This out-of-plane displacement is a result of the loss of axial symmetry of the
non-circular section relative to the circular section bar. (Refer to the third symmetry argument of Section 7.1.)

Thin-Walled Structures 197


Bars Subjected to Torsional Loads

y T y

x x
z
w

T
w ( x, y ) = – ------- xy
GJ

T
(a) Rotation and warping of the cross sections (b) Warping of the cross section
Fig. 7.10 Deformation of a thin-walled rectangular section bar under uniform torsion

The out-of-plane displacement w(x,y) is called warping of the cross section (Fig. 7.10b). For the thin-walled
rectangular section this warping displacement is approximated by
dθ z
w ( x, y ) = – xy -------- (7.14)
dz
It is assumed in this torsion solution that the cross sections are free to warp as expressed by eq. (7.14). If the bar
cross section is not free to warp as described, then additional normal stresses arise as result of the constraint on
the warping displacement. Constrained warping complicates the state of stress, and is not considered further here
since constrained warping stresses are known to be boundary layer effects in most structural applications. Under
uniform torsion, the twist per unit length is a constant, and from the elasticity solution for the thin-walled section
it is given by
dθ T
--------z = --------------------- (7.15)
dz 1
G  --- bt 3
3 

dθ T
Comparing this result to the standard torsion formula --------z = ------- , we find the torsion constant is
dz GJ

1
J = --- bt 3 (7.16)
3
Shear stress components due torsion are τzx and τzy, as shown in the adjacent sketch. The dom-
τ zy inate stress under free warping of the narrow rectangular section is the shear stress tangent to
t τ zx the long side, τzx, and it is approximated by

dθ z t t
τ zx = – 2G  -------- y – --- ≤ y ≤ --- (7.17)
 dz  2 2
b
The shear stress component τzy, is negligible except near the ends at x = ± --- , where component τzx must vanish
2
b
because the lateral surfaces of the bar at x = ± --- are stress free. The shear stress distribution through the thick-
2
ness of the wall is linear, as is shown by eq. (7.17), with a value of zero along the contour line y = 0, and attaining
198 Thin-Walled Structures
Uniform torsion of an open section

a maximum magnitude at y = ± t ⁄ 2 . Combining eqs. (7.15) and (7.17), the magnitude of the maximum shear
stress is
3T
τ max = ------2- (7.18)
bt

Equations (7.15) to (7.18) are thin-walled approximations. The elasticity solution provides more general for-
mulas based on the width-to-thickness ratio b/t. These relations are
dθ T T
--------z = ---------------------
- τ max = ------------2 (7.19)
dz G ( k 1 bt 3 ) k 2 bt
where the dimensionless parameters k1 and k2 are functions of the ratio of b/t. These parameters are listed for var-
ious b/t ratios in table on the next page. From results listed in the table, eqs. (7.15) and (7.18) are seen to be good
approximations for b/t >10.

If the shear flow is calculated from the shear stress distribution given by eq. (7.17) we get
t⁄2 t⁄2 t⁄2
 – 2G  dθ dθ
--------z y dy = – 2G  --------z
q =
∫ –t ⁄ 2
τ zx dy =
∫ –t ⁄ 2
  dz    dz  ∫ –t ⁄ 2
y dy = 0

That is, the shear flow due to a linear shear stress distribution through the thickness of the wall is zero. The shear
stresses are not zero, but the shear flow is zero for an open section. This is an important difference with respect to
torsion of a closed section. Recall that the shear stress distribution is uniform across the thickness of the wall in a
closed section.

Table of parameters for torsion of a rectangular cross section bara

Width to thickness ratio b/t Unit twist parameter k1 Shear stress parameter k2
1.0 0.1406 0.208
1.2 0.166 0.219
1.5 0.196 0.231
2.0 0.229 0.246
2.5 0.249 0.258
3.0 0.263 0.267
4.0 0.281 0.282
5.0 0.291 0.291
10.0 0.312 0.312
∞ 0.333 0.333

a. Timoshenko and Goodier, 1970, p.312

Now consider torsion of open section bars of more complex shape as are shown in Fig. 7.11. Understanding
the torsional response of these bars with complex, open cross-sectional shapes is facilitated by an analogy to the
response of an initially flat membrane supported on its edges over an opening, where the edges of the opening are
in the same shape as the cross section. The membrane is stretched under a uniform tension, and then subjected to
an internal pressure to cause the membrane to deflect. The deflected shape of this pressurized membrane is anal-
ogous to the torsion problem in that level contours on the surface of the deflected membrane coincide with the
lines of action of the of the shear stresses, and that the slope of the membrane normal to the level contour is pro-
Thin-Walled Structures 199
Bars Subjected to Torsional Loads

b2 b2 b2

t2 t2
b
b1 t b1
t1 t1

1 1 1
J = --- ( b 1 t 13 + 4b 2 t 23 ) J = --- bt 3 J = --- ( b 1 t 13 + 2b 2 t 23 )
3 3 3

Fig. 7.11 Some thin-walled open sections and their torsion constants

portional to the magnitude to the shear stress. Also, the volume between the x-y plane and the deflected surface of
the membrane is proportional to the total torque carried by the section. The following text excerpted from Oden
and Ripperger (1981, p. 46) summarizes this analogy.

This analogy was first discovered by Ludwig Prandtl in 1903 and is known’s as Prandtl’s membrane anal-
ogy. Prandtl took full advantage of the analogy and devised clever experiments with membranes. By measur-
ing the volumes under membranes formed by a soap film subjected to a known pressure, he was able to
evaluate torsional constants. By obtaining the contour lines of the membranes he determined stress distribu-
tions.

Torsional constants and the maximum shearing stress can be found for complex cross sections by using the
results for the thin-walled rectangular section. The membrane analogy shows that the torsional load carrying
capacity of the complex open section is nearly the same as the narrow rectangular section, because the volumes
under the membranes are nearly the same if we neglect the small error introduced at the corners or junctions. In
this way, the membrane analogy implies that the complex open cross section has about the same torsional load
carrying capacity as a thin-walled rectangular section with a length equal to the total arc length of the contour of
the complex section.

Since each branch of the open section is equivalent to a narrow rectangular section with the same developed
length and thickness, we can sum the torques carried by each branch in the following way

dθ z dθ z
T = ∑
branches
Ti = ∑ GJ  -------
i
dz 
- = GJ  --------
 dz 

where the torsion constant for the entire cross section is


1 3
J = ∑
branches
Ji = ∑
branches
--- b i t i
3
(7.20)

Note that the twist per unit length is the same for all branches in the open section, because the cross section is
assumed to be rigid in its own plane. The use of eq. (7.20) for several open sections is shown in Fig. 7.11. Start-
ing from eq. (7.18), the maximum shear stress in the ith branch of the section is given by

200 Thin-Walled Structures


Uniform torsion of an open section

3T 3GJ i dθ z 3G 1 T Tt
- -------- = --------2-  --- b i t i3  ------- = -------i
( τ max ) i = --------2-i = ----------- (7.21)
bi t i bi t i 2 dz bi t i  3   GJ  J

That is, the maximum shear stress in the ith branch of the open section is the total torque divided by the torsion
constant for the entire section times the thickness of the ith branch. Note that the largest shear stress magnitude in
a built-up open section occurs in the thickest branch.

EXAMPLE 7.1 Torsional response of a thin-walled open section and an equivalent closed section

A thin-wall circular tube with contour radius r and wall thickness t is subjected to a torque T. The wall of a
second identical tube is cut parallel to its longitudinal axis along its entire length to make the cross section of this
second tube an open circular arc. See Fig. 7.12. Assume the saw kerf is very small. Compare the unit twist and
maximum shear stress in the closed section to the open section.

small slit

t t
r r

T T

Fig. 7.12 Closed and open thin-walled circular sections

T
- from eq. (7.13), and the torsion constant J = 2πr 3 t
Solution For the closed section the shear flow q = ---------------
2 ( πr 2 )
from eq. (7.11). Hence the maximum shear stress and unit twist are

T dθ z
 ------- T T
τ closed = -------------
- - = ------- = ----------------------
-
2πr 2 t  dz  closed GJ G ( 2πr 3 t )

For the open section the developed length b of the contour is essentially 2πr , since the saw kerf is assumed
to be very small. By the membrane analogy the torsional response is the same as the thin-walled rectangular sec-
tion of length b and thickness t. The maximum shear stress is given by eq. (7.18) and the torsion constant is given
by eq. (7.16). For b = 2πr , we have

3T
τ open = -------------2-  dθ
--------z
T
= ---------------------------
2πrt  dz  open 1
G  --- 2πrt 3
3 

Forming the ratio of the maximum shear stress of the open section to the closed section we find
τ open 3T 2πr 2 t r
- = -------------2- ⋅ -------------- = 3 - » 1
--------------
τ closed 2πrt T t

Thin-Walled Structures 201


Bars Subjected to Torsional Loads

Likewise, the ratio of the unit twists are


( dθ z ⁄ dz ) open 3T G2πr 3 t r 2
- = -----------------3- ⋅ ------------------ = 3  -  » 1
----------------------------------
( dθ z ⁄ dz ) closed G2πrt T  t

Since the ratio of the radius to thickness is greater than ten for a thin-walled section, the above results show that
the shear stress and unit twist of the open circular section are much larger than for the closed section if both sec-
tions are subjected to the same torque.

Hence, if a bar is to resist torsional loading, a closed section is preferable to an equivalent open section bar.
That is, the unit twist is smaller for the closed section bar (it is stiffer), and the maximum shear stress is smaller,
than for the equivalent open section bar subjected to the same torque.

7.3 Non-uniform torsion; governing boundary value problem


In the last two sections the cross section of the bar and the torque were considered to be uniform along the length
of the bar. Assume that an external distributed torque of intensity tz (units of F-L/L, or F) acts on the bar. This
intensity is a mathematical function of z, so we denote it as tz(z). In this case the internal torque in the bar is not
uniform along its length, but is a function of the coordinate z; i.e., T(z). First we determine the differential equa-
tz(z)

z = 0 z = L

z t z ( z • )dz, z < z * < z + dz

T T + dT
z z + dz

Fig. 7.13 Differential element of a bar subjected to a distributed torque

tion of torsional equilibrium for this case, where a free body diagram of a differential element of the bar is shown
in Fig. 7.13. Sum the torques acting on the differential element in the figure to get

T + dT – T + t z ( z * )dz = 0 or dT + t z ( z * )dz = 0

Divide the last equation by dz, take the limit as dz → 0 , and note that z * → z in the limit, to get

dT
------- + t z = 0 0<z<L (7.22)
dz

Equation (7.22) is the differential equation of torsional equilibrium of the bar, and its is applicable in any
interval along the length in which the distributed torque intensity is a continuous function. If an external point
torque acts on the bar, then eq. (7.22) is not applicable at the point of application of this external point torque.
The internal torque T exhibits a jump in value at the point of application of the external torque, and torsional
equilibrium applied to a free body diagram of this point determines the relationship between the value of the
internal torque to the left of this point to the value of the internal torque to the right of this point. The analysis of

202 Thin-Walled Structures


Non-uniform torsion; governing boundary value problem

a bar subjected to a point torque is discussed in Example 7.3. The boundary conditions at the ends of the bar are
that we either prescribe the torque or the angle of twist, but not both. That is,
at z = 0 and z = L prescribe either T or θ z but not both. (7.23)

For a linear elastic material, the constitutive equation of the bar is

dθ z
T = GJ (7.24)
dz
where GJ is the torsional stiffness of the bar. For the thin-walled circular annulus, the torsional constant J is
given by eq. (7.11), and for the thin-walled open sections the torsion constant is given by eq. (7.20). In subse-
quent sections we will obtain the torsion constant J for more complex cross-sectional shapes.

Substitute eq. (7.24) for the torque in the differential equation of equilibrium, eq. (7.22), to get

d  dθ z
GJ = –t z ( z ) 0<z<L
dz dz 
If the torsional stiffness is uniform along the length, then this equation reduces to
2
d θz
GJ 2 = – t z z θz = θz ( z ) 0<z<L (7.25)
dz
This is a second order, linear differential equation for the angle of twist θz and requires two boundary conditions
to determine its solution. These boundary conditions are specified in eq. (7.23), in which the torque is related to
the derivative of the twist angle by eq. (7.24).

EXAMPLE 7.2 A uniform distributed torque acting on a bar with fixed ends.

Consider a uniform bar subjected to a uniform distributed torque of intensity t z ( z ) = t z0 = constant , 0 < z < L .
Each end of the bar is fixed to a rigid support which prevents the twisting rotation of the bar so that θ z ( 0 ) = 0
and θ z ( L ) = 0 . Determine distributions of the angle of twist θz(z) and the internal torque T(z) along the length
of the bar.

Solution The governing differential equation, eq. (7.25), reduces to


2
d θz
GJ 2 = – t z0 0<z<L
dz
divide this equation by the torsional stiffness and integrate twice with respect to z to get
t z0 z 2
θ z = – ------- ---- + c 1 z + c 2
GJ 2
where c1 and c2 are constants of integration. These constants are determined from the boundary conditions. At
the left end we have
t z0 0 2
θ z = – ------- ----- + c 1 0 + c 2 = 0
GJ 2

Thin-Walled Structures 203


Bars Subjected to Torsional Loads

So c2 = 0. Using this fact in the boundary condition at the


qz êqzmax t z0 L 2 right end, we get
1 θ zmax = ------------
8GJ
t z0 L 2
0.8 θ z = – ------- ----- + c 1 L = 0
GJ 2
0.6
0.4 t z0 L
So c 1 = ----------
- . Hence, the solution for the angle of twist is
0.2 2GJ
0.2 0.4 0.6 0.8 1 zêL t z0
θ z = ----------- ( Lz – z 2 ) 0≤z≤L (7.26)
2GJ
TêTmax t z0 L
1 T max = ---------
- The torque is obtained by substituting this result, eq.
2
(7.26), for the twist angle in the material law, eq. (7.24), to
0.5
get
0.2 0.4 0.6 0.8 1 zêL t zo
-0.5 T = ------ ( L – 2z ) 0≤z≤L (7.27)
2
-1 The distributions of the angle of twist, eq. (7.26), and
Fig. 7.14 Distributions of the twist angle torque, eq. (7.27), are plotted in Fig. 7.14.
and torque for the bar with fixed ends
and subjected to a uniform distributed
torque.

EXAMPLE 7.3 A point torque acting on a bar with fixed ends.

Consider a uniform bar of length L subjected to an external point torque,


Q denoted by Q, at z = a. Each end of the bar is fixed to a rigid support so that
θ z ( 0 ) = 0 and θ z ( L ) = 0 . Determine the angle of twist θz and torque T in
z the bar for 0 ≤ z ≤ L. See Fig. 7.15.
a b
L Solution In this case the distributed torque intensity tz in the governing dif-
ferential equation, eq. (7.25), is zero in the open intervals 0 < z < a and a < z
Fig. 7.15 Bar with fixed < L. The governing differential equation is not valid at the point of applica-
ends under a point torque tion of the point torque Q. We will have to derive the torsional equilibrium
condition separately for the point z = a. The differential equation of equilib-
rium reduces to
2
d θz
GJ 2 = 0 for 0 < z < a and a < z < L
dz
The general solution to this equation is

 c1 z + c2 0≤z<a
θz = 
 c3 z + c4 a<z≤L

where c1, c2, c3, and c4 are constants to be determined from the boundary conditions. The fixed end condition at
z = 0 yields c2 = 0, and the fixed end condition at z = L yields c4 = - c3 L. The constants c1 and c3 are determined

204 Thin-Walled Structures


Virtual work and strain energy

from the conditions at z = a. First, the angle of twist at z = a must be continuous for otherwise the bar would be
fractured; i.e., θ z ( a – ) = θ z ( a + ) . Imposing this condition of continuity we have

c1 a = –c3 ( L – a ) = –c3 b

since b = L - a. Second, as shown in the accompanying free body diagram, torsional


equilibrium at point z = a yields Q
T ( a– ) T ( a+ )
T ( a+ ) – T ( a– ) + Q = 0
Using the material law, eq. (7.24), this equilibrium equation yields the second condi- a
tion for constants c1 and c3, which is

Q
GJ ( c 3 ) – GJ ( c 1 ) + Q = 0 or c 1 – c 3 = -------
GJ
Now we have two linear, independent equations to deter-
θz
Q b Q a
mine c1 and c3. We find c 1 = ------- --- and c 3 = – ------- --- . Qab
-----------
GJ L GJ L GJL

Thus, the solution for the angle of twist is


z
Q b
 ------ 0
- --- z 0≤z≤a a
 GJ L L
θz =  (7.28)
T
Q a
 ------
- --- ( L – z ) a≤z≤L
 GJ L
b
The torque is determined from this result and the mate- Q ---
L
rial law, eq. (7.24). We find z
0 a L
 Q --b- 0≤z<a
 L
T =  (7.29)
 – Q --a- a<z≤L a
 L – Q ---
L
Plots of the twist angle and the torque are shown in Fig.
Fig. 7.16 Distributions of the angle of
7.16.
twist and torque in Example 7.3.

7.4 Virtual work and strain energy


Consider a bar in equilibrium under a prescribed external torque Q1 at z = 0, a prescribed external torque Q2 at z
= L, and a prescribed distributed torque of intensity t z ( z ) as shown in Fig. 7.17. From this equilibrium state con-
sider an infinitesimal, virtual rotation of the bar in twist denoted by the function δθ z ( z ) . The external virtual
work is
L


δW ext = Q 2 δθ z ( L ) + Q 1 δθ z ( 0 ) + t z ( z )δθ z ( z ) dz (7.30)
0

The free body diagrams (FBDs) at each end of the bar shown Fig. 7.17 are used to determine the relation of the

Thin-Walled Structures 205


Bars Subjected to Torsional Loads

tz(z)

Q1 Q2 Q1 T (0) T (L) Q2

z z = 0 z = L
L
FBD at each end of the bar

Fig. 7.17 External torques acting on a bar and the FBDs for the end conditions

external torques to the internal torque in the bar. Thus, torsional equilibrium at each end of the bar requires that
the internal torque T(z) satisfy T ( L ) = Q 2 and T ( 0 ) = – Q 1 . Hence, the external virtual work becomes
L


δW ext = T ( L )δθ z ( L ) – T ( 0 )δθ z ( 0 ) + t z ( z )δθ z ( z ) dz
0

This equation can be rewritten in an equivalent form as


L L
d
δW ext =
∫ dz ∫
( Tδθ z ) dz + t z ( z )δθ z ( z ) dz
0 0

Now distribute the derivative of the product in the in the first term to get
L
 dT + t  δθ + T d δθ dz
δW ext =
∫ dz z z
dz z
0

The bar is in equilibrium prior to the consideration of the virtual rotation in twist, so the internal torque and exter-
nal distributed torque intensity satisfy equilibrium equation (7.22). Hence, the last equation reduces to
L
dθ z
Tδ   dz
δW ext =
∫ dz 
0

where we interchanged the variational operator and the derivative as discussed in Chapter 2 with eq. (2.34) on
page 32 as the pertinent result. Since the integrand of this last equation contains variables defined internal to the
bar, we define it as the internal virtual work. That is,
L
dθ z
δW int ≡ Tδ   dz

dz 
(7.31)
0

What the manipulations from eq. (7.30) to (7.31) show is that for a bar in equilibrium, the external virtual
work is equal to the internal virtual work for a kinematically admissible variation in the angle of twist function
θ z ( z ) . This proves the necessary condition that if the body is in equilibrium then the external virtual work equals
the internal virtual work for a kinematically admissible variation in the angle of twist. In problem solving we
assume that equating the external to the internal virtual work for every kinematically admissible twist function is

206 Thin-Walled Structures


Virtual work and strain energy

sufficient for equilibrium of the body. See Section 2.6.2 on page 32 for the general statement of the principle of
virtual work.

The principle of virtual work states that the bar is in torsional equilibrium if
δW ext = δW int for every kinematically admissible δθ z ( z ) (7.32)

where the external virtual work is given by eq. (7.30) and the internal virtual work is given by eq. (7.31) consid-
ering torsion only. The mathematical conditions of kinematic admissibility for the virtual angle of twist function
δθ z ( z ) are that it is continuous, its first derivative is piecewise continuous, and that it vanishes at points where
the angle of twist is prescribed. That is, the virtual twist function, or variation in the twist function, must be a
possible twist rotation of the bar.

7.4.1 Strain energy density


Let’s denote the integrand of the internal virtual work expression in eq. (7.31) as the incremental quantity δU 0 .
That is,

dθ z
δU 0 = Tδ   (7.33)
dz 
For a linear elastic material, we can eliminate the torque in this equation using the material law in eq. (7.24) to
get

dθ z dθ z
δU 0 = GJ   δ   (7.34)
dz  dz 
The form of this incremental quantity suggests there is a functional U0 of the derivative of the angle of twist such
that its variation is given by eq. (7.34). We denote this functional by U 0 [ θ z ′ ] , where the prime means ordinary
derivative with respect to z. The variation of a functional is presented in the discussion leading to eq. (2.46) on
page 37. From the latter result, we have that the variation of U0 is given by

∂U 0
δU 0 = δθ ′ (7.35)
∂ θz ′ z
Comparing eqs. (7.34) and (7.35) we get

∂U 0
= GJ θ z ′ (7.36)
∂ θz ′

Regarding θ z ′ as a simple variable, we can integrate this last equation, neglect the constant of integration since it
is immaterial here, to get
1
U 0 = --- GJ ( θ z ′ ) 2 (7.37)
2
Equation (7.37) is the strain energy per unit length of the bar, or the strain energy density. The total strain energy
stored in the bar due to elastic deformation in twist is the integral of the strain energy density over the length of
the bar; i.e.,

Thin-Walled Structures 207


Bars Subjected to Torsional Loads

L L
1
U 0 dz = --- GJ ( θ z ′ ) 2 dz
U =
∫ 2 ∫ (7.38)
0 0

Notice that the partial derivative of the strain energy density, eq. (7.37), with respect to the derivative of the angle
of twist gives the torque; i.e.,

∂U 0
T = = GJ θ z ′ (7.39)
∂ θz ′

The fact that T = ∂U 0 ⁄ ∂θ z ′ is an important property of the strain energy density. In fact, an elastic material can
be defined as one for which a strain energy density function exists.

7.5 Complementary virtual work and energy


Consider the situation in which the angle of twist at each end of the bar is prescribed and the external distributed
torque function is also prescribed. The angle of twist θz(z) of the bar is a continuous function and its derivative is
at least piecewise continuous; that is, the angle of twist function is compatible. The torque Q1 at z = 0 and torque
Q2 at end z = L are unknown reactions to the prescribed twist rotations at each end of the bar. Now consider a
variation in these end torques, or virtual torques, δQ1 and δQ2. The virtual external torques cause a virtual inter-
nal torque δT(z) in the bar. We assume the virtual torque system satisfies equilibrium; i.e.,
d
( δT ) = 0 0<z<L (7.40)
dz
δT ( 0 ) = – δQ 1 δT ( L ) = δQ 2 (7.41)

The virtual system of torques ( δT ( z ), δQ 1, δQ 2 ) satisfying conditions of equilibrium, eqs. (7.40) and (7.41), is
said to be statically admissible. In this case, eqs. (7.40) and (7.41) are satisfied by
δT ( z ) = constant = – δQ 1 = δQ 2 ,

so that either δQ1 or δQ2 are independent virtual torques, but not both, since δQ1 + δQ2 = 0 for overall equilib-
rium of the bar. The external complementary virtual work is defined by
*
δW ext = δQ 2 θ z ( L ) + δQ 2 θ z ( 0 ) (7.42)

Using the boundary conditions, eq. (7.41), for the internal torque this last equation can be written as
*
δW ext = δT ( L )θ z ( L ) – δT ( 0 )θ z ( 0 )

which is equivalent to
L
* d
δW ext =
∫ d z( δT θ ) dz
z
0

Distribute the differential of the product of two functions in the integrand in this equation to get

208 Thin-Walled Structures


Complementary virtual work and energy

L
* d dθ z
δW ext =
∫ dz
( δT )θ z + δT
dz
dz
0

From the differential equation of equilibrium for the virtual force system, eq. (7.40), the first term in the inte-
grand of this last equation vanishes. Hence, we get
L
dθ z
*
δW ext =
∫ δT  d z  dz
0

Since this last expression involves only quantities defined internal to the bar, we define the right-hand side as the
internal complementary virtual work; i.e.,
L
* ≡ δT 
dθ z
δW int

dz 
dz (7.43)
0

We have shown in the process of going from eq. (7.40) to (7.43) that, for a compatible angle of twist func-
tion, the external complementary virtual work is equal to the internal complementary virtual work for a statically
admissible variation in the torques. That is, we have proved the necessary condition that if the angle of twist
function is compatible, then the external complementary virtual work equals the internal complementary virtual
work for a statically admissible variation in the torques. In problem solving we assume that equating external
complementary virtual to the internal complementary virtual work for every statically admissible system of vir-
tual torques is sufficient for the angle of twist function of the body to be compatible. This is a statement of the
principle of complementary virtual work for the torsion bar. A general statement of the principle of complemen-
tary virtual work (PCVW) is given on page 47. For the bar in torsion, the PCVW means the angle of twist func-
tion θ z ( z ) is compatible if

*
δW ext *
= δW int for every statically admissible virtual torque δT ( z ) (7.44)

where the external complementary virtual work is given by eq. (7.42) and the internal complementary virtual
work is given by eq. (7.43). The principle of complementary virtual work is independent of the material behavior.

7.5.1 Complementary strain energy


Now assume the material of the bar is elastic and linear. Solve the material law given by eq. (7.24) for the deriv-
ative of the angle of twist to get θ z ′ = T ⁄ ( GJ ) , and then substitute this result into the internal complementary
virtual work of eq. (7.43) to get
L
T
*
δW int =
∫  ------
GJ 
- δT dz (7.45)
0

Let the integrand of this equation be denoted by the incremental quantity δU 0* . So

T
δU 0* =  ------- δT (7.46)
 GJ 

This form of the incremental quantity suggests there is a functional of the torque whose variation is given by eq.
(7.46). This functional is called the complementary strain energy density, and is denoted by U 0* [ T ( z ) ] . Based on

Thin-Walled Structures 209


Bars Subjected to Torsional Loads

the definition of the variation of a functional given by eq. (2.45) and eq. (2.46) on page 37, we would find in a
similar manner that the variation of the complementary strain energy density is given by

δU 0* = ( U * )δT (7.47)
∂T 0
Equating eq. (7.46) and eq. (7.47) for every statically admissible virtual torque gives
∂ T
( U 0* ) = ------- (7.48)
∂T GJ
Integrating this expression with respect to T (note that we treat the torque as a simple variable) and taking the
constant of integration to be zero when T = 0, we get the complementary strain energy density as

1 T2
U 0* = --- ------- (7.49)
2 GJ
The important property of the complementary strain energy density is

θz ′ = (U *) (7.50)
∂T 0
which can be seen by differentiating eq. (7.49) and using (7.24). A comparison of eq. (7.50) with eq. (7.39) for
the strain energy density shows the dual attributes that these energies possess: the derivative of the complemen-
tary strain energy density with respect to the torque gives the twist rate, and the derivative of the strain energy
density with respect to the twist rate gives the torque.

Return to the internal complementary virtual work given by eq. (7.45), and use eq. (7.46) to write it as
L
* *
δW int =
∫ δU 0 dz
0

Now interchange the variational operator and the integral operator in accordance with eq. (2.34) on page 32, and
write this equation as
* = δU *
δW int

where the complementary strain energy for the bar is defined as


L L
1 T2
U* =
∫ U 0* dz =
∫ --2-  ------
GJ 
- dz (7.51)
0 0

7.6 Unit twist of a single cell beam due to shear flow


Consider a single cell beam with a contour of arbitrary shape and a wall thickness t ( s ) , where s denotes the con-
tour coordinate, as is shown in Fig. 7.18. The beam is made of an isotropic and homogeneous material that fol-
lows Hooke’s Law. Our purpose is to determine the twist per unit length dθ z ⁄ dz in terms of the shear flow
distribution around the contour. In the derivation presented in this section it must be emphasized that the shear
flow can be cause by the transverse shear forces V x and V y , and the torque T. The relationship of the shear flows
to the shear forces was discussed in Section 6.2 (eq. (6.13) on page 160 is the pertinent result), and for the special

210 Thin-Walled Structures


Unit twist of a single cell beam due to shear flow

z x
Vy
q(s)
t(s) s
Vx
T, θ z

Fig. 7.18 Thin-walled, single cell cross section with an arbitrary shaped contour

case of a circular contour the shear flow was related to the torque via Bredt’s formula, eq. (7.13). In Section 7.7
we will determine how to compute the shear flow due to torsion for a closed contour of arbitrary shape. All we
need to keep in mind in this section is that the shear flow is the superposition of the shear flows due to transverse
shear and torsion. To relate the shear flow to the unit twist of the cross section, we use the material law and the
geometry of the deformation.

7.6.1 Hooke’s law


The shear stress is directly related to the shear strain with the constant of proportionality being the shear modulus
G of the material.
τ zs = Gγ zs (7.52)

Multiply this equation by the wall thickness t to get


q = Gtγ zs (7.53)

The deformation due to the shear flow is depicted in Fig. 7.19.


y

s, v t
C C*
z q
A x
t(s) B A*
z, w
q
B*
π
--- – γ zs
2

Fig. 7.19 Element of the wall deformed in shear.

Thin-Walled Structures 211


Bars Subjected to Torsional Loads

7.6.2 Shear strain-displacement relation


The shear strain γzs is the reduction in the right angle between line elements originally parallel to the s- and z-
directions in the undeformed body as is shown in Fig. 7.19. The shear strain is related to the partial derivatives of
the displacement components of a material point on the contour. Let the vt(s,z) denote the displacement tangent
to the contour, or the s-direction, of the material point at (s,z), and let w(s,z) be the displacement of this point in
the axial direction, or z-direction. The three adjacent points labeled A, B, and C in the undeformed shell displace
to position A*, B*, and C* in the deformed shell as shown in Fig. 7.19 and Fig. 7.20. The displacements of these
adjacent material points are shown in Fig. 7.20. The reduction in the right angle between lines elements origi-

∂w w(s+ds,z)
ds
C* ∂s C*
B* vt(s+ds,z)
B*  1 + ∂v t ds C
 ∂s 
∂v t A*
dz vt(s,z+dz) ds
∂z A* vt(s,z)
 1 + ∂w dz
B dz
 ∂z  A:(s,z)
w(s,z)
w(s,z+dz)

Fig. 7.20 Tangential and axial displacements of three adjacent points in the wall of the shell.

nally parallel to the s- and z-directions is determined from Fig. 7.20 as

∂v t ∂w
dz ds
∂z ∂s
γ zs = --------------------------- + ---------------------------- (7.54)
 1 + ∂w dz  1 + ∂v t ds
 ∂z   ∂s 
Since the displacement derivatives are small in magnitude with respect to unity for infinitesimal deformations,
we get

∂v t ∂w
γ zs = + (7.55)
∂z ∂s
Combining eqs. (7.53) and (7.55) to eliminate the shear strain gives

q ∂w ∂v t
------ = + (7.56)
Gt ∂s ∂z

7.6.3 Tangential displacement of a typical point on the contour


Take the origin of the contour coordinate (s = 0) at some point along the contour with s increasing counterclock-
wise. As s transverses the contour and returns to the origin its value is S, where S is the length of the perimeter of
the contour. See Fig. 7.21. A generic point on the contour is labeled A and has cartesian coordinates

212 Thin-Walled Structures


Unit twist of a single cell beam due to shear flow

[ x ( s ), y ( s ) ] . The directions tangent and normal to the contour at A are denoted by t and n, respectively, as is
shown in Fig. 7.21, and the angle between the positive x-direction and the tangent to the contour is denoted by
θ ( s ) . For a differential length ds along the contour at A we have the trigonometric relations

dx dy
cos θ = sin θ = (7.57)
ds ds
y

t(s)

x t
x(s) ds
dy
θ
s = S Α
s dx
n
s = 0 t
θ(s)
y(s) A
n

Fig. 7.21 Geometry of the contour.

The projection of the cross section in the deformed cylindrical shell onto the x-y plane is assumed not to dis-
tort from its original cross-sectional shape in the undeformed shell. However, the projected cross section can dis-
place in the x-y plane and rotate about the z-direction relative to the cross section in the undeformed shell. The
cross section can warp out of its plane in a manner similar to the warping of the rectangular cross section under
torsion discussed in Section 7.2. This warping displacement of the non-circular contour is in contrast to the circu-
lar contour (Section 7.1) in which plane cross sections remained plane in the deformed shell because of axial
symmetry of the circular section. Again, the loss of axial symmetry results in warping of the contour.

The assumption that the projection of the contour into the x-y plane is in the shape of the undeformed con-
tour results in a kinematic relationship between the tangential displacement of the point A on the contour and the
displacements and rotation of the section. First, we reference the displacement and rotation of the cross section to
an arbitrary point labeled O as shown in Fig. 7.22. The tangential displacement vt of point A consists of a rigid
body translation vt1, in which point O moves to O* and A moves to A1, followed by a rotation θ z about O*. The
tangential displacement of A1 to A* is denoted by vt2. Thus, v t = v t1 + v t2 .

The displacement of point O has a horizontal component denoted by u0(z) and a vertical component denoted
by v0(z). Since the angle between the x-direction and the tangent to the contour at s is denoted by the function
θ ( s ) , the projection of the horizontal component in the tangent direction is u 0 cos θ , and the projection of the
vertical component in the tangent direction is v 0 sin θ . Also, the displacement of A to A1 is the same as the dis-
placement of O to O*. Thus, the tangential displacement of A to A1 due to translation of the section is the sum of
these projections, or

Thin-Walled Structures 213


Bars Subjected to Torsional Loads

x θz + θ
O* A*

θ θz
v0 θ
r β
uo A1 v
t2
O
r –vn
v t1
β θ
vt
A

Fig. 7.22 The tangential displacement of a generic point on the contour is related to the rigid body
displacement and rotation of the section

v t1 = u 0 cos θ + v 0 sin θ (7.58)

The tangential displacement due to the rotation θ z is determined by the projection of line O*A* onto the tangent
direction. First note that the length of lines OA , O * A 1 , and O * A * are the same because of the rigid section
assumption. The angle between the line OA and the normal direction to the contour at point A is denoted by β.
The tangential component of the displacement of A1 to A* is obtained from the geometry shown in Fig. 7.22 as

v t2 = O * A * sin(β + θ z ) – O * A 1 sin β = O * A* [ sin β cos θ z + cos β sin θ z – sin β ]

We assume that the rotation θ z is small such that sin θ z ≈ θ z and cos θ z ≈ 1 . Also, note that OA cos β = r ,
which is the coordinate of A relative to O in the direction normal to the contour at A. Thus,

v t2 = ( O * A * cosβ )θ z = ( OA cos β )θ z = rθ z

The total tangential displacement of A to A* is then


v t ( s, z ) = u 0 ( z ) cos [ θ ( s ) ] + v 0 ( z ) sin [ θ ( s ) ] + r ( s )θ z ( z )


















translation rotation (7.59)

7.6.4 Relation between the shear flow and unit twist


We can take the partial derivative of the tangential displacement, eq. (7.59), with respect to z to get

214 Thin-Walled Structures


Uniform torsion of a thin-walled closed section with a contour of arbitrary shape

∂v t du 0 dv 0 dθ z
= cos θ + sin θ + r (7.60)
∂z dz dz dz
Substitute this result for the derivative of the tangential displacement in eq. (7.56) to get

q ∂w du 0 dv 0 dθ z
------ = + cos θ + sin θ + r
Gt ∂s dz dz dz
Now integrate this last expression completely around the contour to get

q ∂w du 0 dv 0 dθ z
°∫------ ds =
Gt °∫
∂s
ds +
dz °∫
cos θ ds +
dz °∫
sin θ ds + ( rds )
°∫
dz
(7.61)

From eqs. (7.57) we have

°∫ cos θ ds = °∫ dx = x ( S ) – x ( 0 ) = 0
and

°∫ sin θ ds = °∫ dy = y ( S ) – y ( 0 ) = 0
so that the second and third terms on the right-hand side of eq. (7.61) vanish. The first integral on the right-hand
side of eq. (7.61) can be done exactly to give

∂w
°∫ ∂ s ds = [ w ( S, z ) ) – w ( 0, z ) ] = 0 ,
since the axial displacement of the material point at s = 0 and s = S is unique. The fourth integral on the right-
hand side of eq. (7.61) (integral of the quantity rds around the contour) is twice the enclosed area of the contour.
Hence, the final result from eq. (7.61) is

dθ z 1 q
dz 2¿ Gt °∫
= ----------- ------ ds (7.62)

Equation (7.62) is the formula we set out to derive. It relates the shear flow to the twist per unit length of the bar.
The shear flow can be caused by both transverse shear forces and torque. In this sense, eq. (7.62) is more general
than for torsion only.

7.7 Uniform torsion of a thin-walled closed section with a contour of


arbitrary shape
Consider again the thin-walled cylindrical shell with an arbitrary shaped contour subjected to only a torque T and
no transverse shear, as is shown in Fig. 7.23. The cross section and torque are constant along the length of the
shell, and the material is linear elastic and isotropic. However, we permit the shear modulus G of the material to
be a function of the contour coordinate s to model multi-material bars. As in the Section 7.1 and Section 7.2, this
is a uniform torsion problem. Under uniform torsion the twist per unit length dθ z ⁄ dz is constant, since the
torque, the cross section, and material do not change with respect to z. Our purpose is to relate the shear flow q to
the torque T.

Thin-Walled Structures 215


Bars Subjected to Torsional Loads

y
T

z x
t(s)

T, θ z

Fig. 7.23 Uniform torsion of a thin-walled cylindrical shell with an arbitrary shaped contour.

Axial equilibrium requires uniform shear flow around the contour We assume the shear stress τ zs is uni-
formly distributed across the thickness of the wall, based on the analysis of a thin-walled circular tube. The shear
flow is given by q = τ zs t ( s ) , and the shear flow is tangent to the contour. Since there are no axial normal
stresses (no bending nor extension), the shear flow is uniform along the contour. That is,
q = constant with respect to s (7.63)

A free body element from the wall is shown in Fig. 7.24, and axial equilibrium is used to establish the shear flow
is uniform along the contour.

q(s + ds) dz
s
z
s + ds
s q(s) dz
q ( s + ds )dz – q ( s )dz = 0

Fig. 7.24 A free body diagram of an element of the shell for axial equilibrium.

Resultant of the uniform shear flow In general, the resultant of the shear flow resolved at an arbitrary point in
the cross section (again, labeled O) is a force and a couple. Static equivalence gives

Fx =
°∫ ( cos θ )qds Fy =
°∫ ( sin θ )qds T =
°∫ r ( qds ) (7.64)

in which r(s) is the coordinate in the normal direction to the contour at A measured from point O. See Fig. 7.25.
Since the shear flow is independent of the contour coordinate s, it may be brought outside the integral in eqs.
(7.64). We get, using eqs. (7.57) in the process, that

°∫ °∫
F x = q cos θds = q dx = q [ x ( S ) – x ( 0 ) ] = 0 (7.65)

216 Thin-Walled Structures


Uniform torsion of a thin-walled closed section with a contour of arbitrary shape

Fy
1
d¿ = --- rds T
2
O O Fx

r qds
θ
ds
ds dy
A θ
dx

Fig. 7.25 Resultant of the shear flow

°∫ °∫
F Y = q sin θds = q dy = q [ y ( S ) – y ( 0 ) ] = 0 (7.66)

°∫
T = q rds = q 2d¿ = 2¿q
°∫ (7.67)

The last equation is in eqs. (7.67) is called Bredt’s formula, or the Bredt-Batho formula, and it relates the torque
to the shear flow via the area enclosed by the contour. Since Bredt’s formula is the principal result, we repeat it
below as eq. (7.68).
T
q = -------- Bredt's formula (7.68)
2¿

For the shear flow independent of the contour coordinate, we find that the unit twist from eq. (7.62) reduces
to

dθ z q ds
dz 2 ¿ Gt °∫
= --------- ------ (7.69)

Equations (7.68) and (7.69) can be combined by eliminating the shear flow, and we write the result as

dθ z
T = G0 J (7.70)
dz
in which the torsion constant is given by

4¿ 2 (7.71)
J = ---------- single cell section
ds
°∫
-----
t*
and the modulus-weighted thickness is given by
G(s)
t* = -----------t (7.72)
G0
Shear modulus G0 is introduced in the definition of torsional stiffness as a reference shear modulus. It is selected
for convenience in cross sections where the material can vary from branch to branch. If the cross section is com-

Thin-Walled Structures 217


Bars Subjected to Torsional Loads

posed of a single homogeneous material, then we take G0 = G, so that t* is the actual wall thickness t. It is impor-
tant to remember that the formula for the torsion constant J above is only valid for single-cell, closed section.

The torsion constants for circular section and rectangular section made of a single homogeneous material
and uniform wall thickness are shown in Fig. 7.26. Note that the general formula for J, eq. (7.71), reduces to
value of J for the circular tube given in Section 7.1, eq. (7.11).

t ¿ = ab
¿ = πr 2
ds 2(a + b)
r ds 2πr
°∫ ----t- = --------
- a °∫ ----t- = -------------------
t
-
t t
4 ( ab ) 2 2 ( ab ) 2 t
4 ( πr 2 ) 2 J = -------------------- = -------------------
J = ------------------- = 2πr 3 t 2(a + b)
--------------------
a+b
2πr t
---------
t b

Fig. 7.26 Torsion constants for thin-walled, single cell sections with circular and rectangular
contours

EXAMPLE 7.4 Torsion of a circular, bi-material section

Consider the circular section made of two materials as shown in Fig. 7.27. Determine the shear flow, maximum
shear stress, the torsion constant, and the torsional stiffness.
6 6
G = 2.5 ×10 psi G = 5 ×10 psi

10000 lb-in. 2.0 in.

0.250 in. 0.125 in.

Fig. 7.27 Bi-material circular torsion tube

Solution The shear flow is computed from Bredt’s formula, eq. (7.68), as
T 10000 lb-in.
q = -------- = ----------------------------
- = 1591.5 lb/in.
2¿ 2π ( 1 in. ) 2

The shear stress in the left side wall is τ zs = ( 1591.5 lb-in. ) ⁄ ( 0.250 in. ) = 6366.2 psi , and in the right side
wall the shear stress is τ zs = ( 1591.5 lb-in. ) ⁄ ( 0.125 in. ) = 12732.4 psi . Hence, the maximum shear stress is
12.7 ksi, and it occurs in the 0.125-inch-thick section.

218 Thin-Walled Structures


Shear center of a closed section

6
Take the reference shear modulus G 0 = 2.5 ×10 psi. First we compute the denominator of the torsion con-
stant given by eq. (7.71).
ds π ( 1 in. ) π ( 1 in. )
°∫ ----t - = ------------------------------------------------------
*
2.5 ×10 psi
6
- + -------------------------------------------------- = 8π
5 ×10 psi
6
---------------------------- ( 0.250 in. ) ----------------------- ( 0.125 in. )
G0 G0

Note that the modulus-weighted thickness of the right side wall is 2 (0.125 in.) = 0.250 in. From eq. (7.71) the
torsion constant is
4 [ π ( 1 in. ) 2 ] 2 π
J = -------------------------------- = --- in. 4
8π 2

T
the torsional stiffness is defined as ------------------ , which equals G 0 J via eq. (7.70). Thus, the torsional stiffness for
dθ z ⁄ dz
this example is

T π
------------------ = G 0 J = ( 2.5 ×10 psi )  --- in. 4 = 3.93 ×10 lb-in. 2
6 6
dθ z ⁄ dz 2 

7.8 Shear center of a closed section


The unit twist formula given by eq. (7.62) is applicable even if the shear flow is a function of the contour
coordinate s, as is the case when the transverse shear forces are non-zero. Hence, eq. (7.62) can be used to com-
pute the unit twist of the section when the shear forces V x and V y and the torque T act in unison. In the case of
torsion only, we showed in the last section that the shear flow was constant along the contour coordinate based on
axial equilibrium, which led to the simplification given by eq. (7.69). If V x and V y are non-zero, then the com-
putation for the unit twist is a more involved. We illustrate the use of eq. (7.62) for the case of transverse shear
forces without torsion by locating the shear center of a closed section. There are three steps to locating the shear
center of a closed section.
• Determine the shear flow distribution around the section due to bending using eq. (6.13) on page 160.
This step involves selecting a contour origin where s = 0. The shear flow q(0) at the contour origin is
unknown.
• Determine the unknown shear flow q(0) by setting the unit twist, eq. (7.62), to zero.
Setting the unit twist to zero enforces the condition of no torsional deformation for the section.
• Use torque equivalence to locate the shear center.

This procedure is best illustrated by example.

EXAMPLE 7.5 Shear center location of a single-cell, closed section having one axis of symmetry.

Consider the thin-walled closed section shown in Fig. 7.28. The contour of this section is an isosceles triangle
with each branch having the same thickness t and having the same shear modulus G. There is a horizontal axis of
symmetry, which is taken as the x-axis. The second area moment about the x-axis is I xx = 300t cm 4 , where t is
specified in cm. Since the shear center lies on this axis of symmetry, we take the shear force components

Thin-Walled Structures 219


Bars Subjected to Torsional Loads

V x = 0 and V y > 0 . Determine the location ex of the shear center from point O, which is located on the x-axis

y
Vy s 1, q 1
5 13
C O 5
S.C. x s 2, q 2 12
O
5 q0
t s 3, q 3
1
4 ---
3
12
All dimensions in cm
ex

Fig. 7.28 A closed section with an isosceles triangle contour

at the vertex of the two sides of the triangle that are of equal length.

Solution The shear flows are assumed positive in the counter-clockwise direction with the contour origin at
Vy
point O. The shear flow at point O is q0, and eq. (6.13) on page 160 reduces to q ( s ) = q 0 – ------Q x ( s ) . For the
I xx
first branch shown in Fig. 7.28 we have that the shear flow due to bending is given by

Vy s1 Vy 2
5 
 -----
q 1 ( s 1 ) = q 0 – -----------
300t ∫ 0
- s t ds = q 0 – -----------
 13 1 1 1560 1
-(s ) 0 ≤ s 1 ≤ 13 cm (7.73)

The shear flow due to bending in the second branch is


Vy s2
q 2 ( s 2 ) = q 1 ( 13 ) – -----------
300t ∫ 0
( 5 – s 2 )t ds 2 0 ≤ s 2 ≤ 10 cm

From eq. (7.73) q 1 ( 13 ) = q 0 – ( 13V y ) ⁄ 120 . Hence the shear flow in the second branch is

Vy 1
q 2 ( s 2 ) = q 0 – ---------  13 + 2s 2 – --- s 22 0 ≤ s 2 ≤ 10 cm (7.74)
120  5 
The shear flow due to bending in the third branch is
Vy s3
5 
 – 5 + -----
q 3 ( s 3 ) = q 2 ( 10 ) – -----------
300t ∫ 0

- s t ds
13 3 3
0 ≤ s 3 ≤ 13 cm

From eq. (5.44) q 2 ( 10 ) = q 0 – ( 13V y ) ⁄ 120 , which is the same value as the shear flow at the junction of
branches one and two. Hence the shear flow in branch three is
Vy 1
q 3 ( s 3 ) = q 0 – ---------  13 – 2s 3 + ------ s 32 0 ≤ s 3 ≤ 13 cm (7.75)
120  13 

220 Thin-Walled Structures


Shear center of a closed section

Note from this result that q 3 ( 13 ) = q 0 , as it should since the shear flow at point O is unique.

To find the shear flow at point O, we impose the condition of zero unit twist from eq. (7.62). Since the prod-
uct of Gt is the same in each branch, eq. (7.62) reduces to
dθ z 1
-------- = -------------------
dz 2( Gt ) ¿ °∫ qds= 0
which requires the contour integral of the shear flow around the entire section to vanish. Note that the unit twist is
positive counter-clockwise and consequently a counter-clockwise shear flow is positive in eq. (7.62). Substituting
eqs. (7.73) to (7.75) into this integral, we get

13 Vy 2 
 q – -----------
10 Vy 
 q – -------- 1
- 13 + 2s 2 – --- s 22  ds 2 +
°∫qds =
∫ 0
 0
1560 1  1
- ( s ) ds +
∫ 0
 0
120  5 

13 Vy 
 q – -------- 1
- 13 – 2s 3 + ------ s 32  ds 3 = 0
∫ 0
 0
120  13  

After evaluation of the definite integrals, this equation becomes

23 23
36q 0 – ------ V y = 0 Hence q 0 = --------- V y (7.76)
10 360
Substitute q0 from eq. (7.76) into eqs. (7.73) to (7.75) to find that the shear flows due to bending only are

Vy
q 1 ( s 1 ) = ------------ ( 299 – 3s 12 ) (7.77)
4680
Vy
q 2 ( s 2 ) = ------------ ( – 80 – 30s 2 + 3s 22 ) (7.78)
1800
Vy
q 3 ( s 3 ) = ------------ ( – 208 + 78s 3 – 3s 32 ) (7.79)
4680

For each straight branch we can compute the branch force. These branch forces are

13 13 Vy 13
- ( 299 – 3s 12 ) ds 1 = ------ V y
f1 =
∫ 0
q 1 ds 1 =
∫ 0
-----------
4680 36
(7.80)

10 10 Vy 13
- ( – 80 – 30s 2 + 3s 22 ) ds 2 = – ------ V y
f2 =
∫ 0
q 2 ds 2 =
∫ 0
-----------
1800 18
(7.81)

13 Vy 13
- ( – 208 + 78s 3 – 3s 32 ) ds 3 = ------ V y
f3 =
∫ 0
-----------
4680 36
(7.82)

These branch forces are shown in Fig. 7.29. Now torque equivalence is used to determine the location of the
shear center on the axis of symmetry. Torque equivalence gives

13 2
12 cm  ------ V y = e x V y Hence e x = 8 --- cm
 18  3

Thin-Walled Structures 221


Bars Subjected to Torsional Loads

13
------ V y Vy
36

13
------ V y
18
O ⇔ O

13
------ V y ex
36

Fig. 7.29 Static equivalence to find the shear center of the triangular contour

The shear center and the centroid are shown in Fig. 7.30.

S.C. C
O
Fig. 7.30 Shear center and centroid
2
location of the triangular section 7 --- cm
3
2
8 --- cm
3

12cm

7.9 Uniform torsion of multi-cell closed sections


Consider a cylindrical shell-beam of length L whose cross section consists of two closed cells. The z-axis is par-
allel to the length of the cylindrical shell-beam, and the x- and y-axes are in a plane parallel to the cross sections.
The shell-beam is subjected to equal and oppositely directed torques T at each end. Hence, equilibrium requires
that each cross section is subjected to a torque equal to T, and the torque is taken positive if it act counter-clock-
wise on a positive z-face. As is shown in Fig. 7.31, the cell on the left side of Section A-A is designated as cell 1
and the cell to the right is designated 2. The thickness of the exterior branches of cells 1 and 2 are denoted by t1
and t2, respectively. The thickness of the common branch separating the two cells is denoted by t 1 – 2 . All
branches are assumed made of the same homogeneous material with the shear modulus denoted by G. Our pur-
pose is to determined the shear flows and shear stresses due to torsion, and to determine the torsion constant J in
dθ z
the standard formula T = GJ -------- where dθ z ⁄ dz is the twist per unit length.
dz

We assume positive shear flow directions as shown in Fig. 7.32. The shear flows in the exterior branches of
cells 1 and 2 are taken positive counter-clockwise and are denoted by q1 and q2, respectively. The shear flow in
the common branch is denoted by q1-2. The common branch shear flow is related to the shear flows in the exte-
rior branches of cells 1 and 2 by axial equilibrium at the junction of the common branch with the exterior
branches of cells 1 and 2. This junction is labeled O in Fig. 7.32, and the free body diagram at the junction is
shown in Fig. 7.33.

222 Thin-Walled Structures


Uniform torsion of multi-cell closed sections

y
A
T z T

A
L

t1
y
t2

x ¿
Section A-A ¿1 2

t1 – 2

Fig. 7.31 Torsion of a shell-beam with a cross section consisting of two closed cells

y
q1
t1 q2 t2

q1 – 2
x ¿1 ¿2
q1 q2
O t1 – 2

Fig. 7.32 Assumed positive shear flows for the two-cell section

q1 – 2

z
q1
q2 ∆z

q1 O q2

Fig. 7.33 Free body diagram of the junction at point O

Axial equilibrium at the junction point O gives


q1 – 2 = q1 – q2 (7.83)

Axial equilibrium at the upper junction of the common web with the exterior branches of cells 1 and 2 gives the
same result as eq. (7.83). An easy way to determine the axial equilibrium of the junction is to recall that the shear

Thin-Walled Structures 223


Bars Subjected to Torsional Loads

flow into the junction must equal the shear flow out of the junction.

Now we determine the torque about point O due to the shear flows. The shear flow q1-2 does not contribute to
the torque about point O since the contour of the common branch is straight and it passes through point O. Using
Bredt’s formula, the torque about point O is
T = 2 ¿1 q1 + 2 ¿2 q2 (7.84)

where ¿ denotes the area enclosed by a cell. Actually, the same value of the torque is obtained for any point in
the plane of the cross section where the resultant of the shear flow distribution is resolved. The fact that the
torque due to the shear flows is the same about any point in the plane is discussed in Section 7.10.

At this point we have two equations, eqs. (7.83) and (7.84), for the three unknown shear flows assuming that
the torque and geometry of the section are known. An additional equation relating the shear flows is based on the
assumed rigidity of the cross section in its own plane. In torsion, this rigid cross section condition implies the
unit twist of each cell is the same. The unit twist for a single cell is given by eq. (7.62). Since the shear modulus
is the same for all branches in the cross section eq. (7.62) reduces to
dθ z 1 q
-------- = ------------
dz 2 ¿G °∫ --t- ds
This unit twist formula was derived on the basis that a counter-clockwise shear flow tends to produce a counter-
clockwise unit twist. Apply this equation to cell 1 to get

dθ z
 ------- 1 S S
- = ---------------  --- q 1 +  --- q (7.85)
 dz  1 2 ¿ 1G  t  1  t1 – 2 1 – 2

where S1 is the arc length of the exterior branch of cell 1, and S1-2 is the length of the common branch between
cells 1 and 2. For cell 2, the unit twist is

 dθ 1 S S
--------z = ---------------  --- q 2 –  --- q (7.86)
 dz  2 2¿ 2G  t  2  t1 – 2 1 – 2

where S2 is the arc length of the two exterior branches of cell 2. Note that the contribution of the common branch
shear flow is negative in the unit twist formula for cell 2. Relative to an observer in cell 2, a positive value for the
shear flow q1-2 tends to produce a clockwise unit twist, and hence is negative by the convention that counter-
clockwise is positive. Since the unit twist of each cell is the same, we equate eqs. eq. (7.85) and eq. (7.86) to get

1 S S 1 S S
---------------  --- q 1 +  --- q = ---------------  --- q 2 –  --- q (7.87)
2 ¿ 1G  t  1  t1 – 2 1 – 2 2¿ 2G  t  2  t1 – 2 1 – 2

We now have three equations for the three unknown shear flows; eqs. (7.83), (7.84), and (7.87). These equa-
tions are solved for the shear flows in terms of the torque and geometry of the cross section. Once the shear flows
are known, the unit twist for either cell 1, eq. (7.85), or cell 2, eq. (7.86) can be determined in terms of the torque,
geometry, and shear modulus. The unit twist computed for one cell is the unit twist for the entire cross section,
since we equated the unit twist of each cell to one another. Finally, compare the computed unit twist to the stan-
dθ T
dard formula --------z = ------- to identify the value of the torsion constant J. For a multi-cell section, the formula
dz GJ
given for the torsion constant of a single cell section, eq. (7.71), is not applicable.

224 Thin-Walled Structures


Uniform torsion of multi-cell closed sections

EXAMPLE 7.6 Uniform torsion of a two-cell section

All branches of the two cell section shown in Fig. 7.31 are made of the same material, whose shear modulus is
denoted by G. The exterior branches have thickness t and the common branch has a thicknesses 3t. The contour
of the nose cell is semi-circular with radius r, and the contour of the second cell is an isosceles triangle with equal
sides of length πr. Note that the horizontal length of the isosceles triangle is approximately 3r, and we will use
this approximation to simplify the algebra. Assume the torque T is given in addition to the geometry, and deter-
mine
• shear flows q1 and q2 in the exterior branches of cells 1 and 2, and
• the torsion constant J.

πr y
q2
q1 q1 – 2
3t
T
r x ¿1 ¿2

t t q2
∼ 3r

Fig. 7.34 Uniform torsion of a two-cell section

Solution From eq. (7.84) the torque is

πr 2 2r ⋅ 3r
2  -------- q 1 + 2  ---------------- q 2 = T
 2   2 

or

πr 2 q 1 + 6r 2 q 2 = T (7.88)

Using eq. (7.85) for the unit twist for cell 1, and eq. (7.83) for the common web shear flow, we get

 dθ 1 πr 2r
--------z = ----------------------  ------ q 1 +  ----- ( q 1 – q 2 )
 dz  1 πr 2  t  3t 
2  -------- G
 2 

or

 dθ 1 2 2
--------z = -----------------  π + --- q 1 – --- q 2 (7.89)
 dz  1 ( πrt )G  3 3
Using eq. (7.76) for the unit twist for cell 2, and eq. (7.83) for the common web shear flow, we get

 dθ 1
--------z = --------------------
2πr 2r
-  --------- q 2 –  ----- ( q 1 – q 2 )
 dz  2 2 ( 3r )G 2  t   3t 
or

Thin-Walled Structures 225


Bars Subjected to Torsional Loads

dθ z
 ------- 1 2 2
- = ----------------- – --- q 1 +  2π + --- q 2 (7.90)
 dz  2 ( 6rt )G 3  3
Now equate eqs. (7.89) and (7.90) to get

1 2 2 1 2 2
---  π + --- q 1 – --- q 2 = --- – --- q 1 +  2π + --- q 2
π  3 3 6 3  3

which after rearrangement gives

2 1
 1 + ----- π 1 2
- + --- q –  --- + --- + ------ q = 0 or 1.3233177q 1 – 1.3705152q 2 = 0 (7.91)
 3π 9 1  3 9 3π 2

Equations (7.88) and (7.91) are two simultaneous equations for the shear flows q1 and q2 in terms of the torque.
The solution to these simultaneous equations determines the shear flows. Hence,
T T
answer : q 1 = --------------------2- q 2 = --------------------2- (7.92)
8.9350r 9.2536r

To find the torsion constant J, we substitute the solutions for the shear flows into eq. (7.89). The result is

dθ dθ z T
--------z =  -------- = 0.11273713 ----------
-
dz  dz  1 Gr 3 t
The same result is obtained using the unit twist for cell 2; i.e., eq. (7.90). Comparing the unit twist result given
dθ T
above to the standard torsional relation ------ = ------- , we find that the torsion constant J is
dz GJ

1 0.11273713
--- = ---------------------------
- giving the answer J = 8.870r 3 t (7.93)
J r3t

7.10 Resultant of a constant shear flow in a curved branch


In torsion problems it often necessary to find the resultant of a constant shear flow along the contour of a curved
branch. This situation is depicted in Fig. 7.35. The resultant of the shear flow is resolved at point A in this fig-
ure. From the differential element shown in Fig. 7.35(b), the differential force and differential torque resolved at
point A are

dx dy
dF x = ( qds )  ------ = qdx dF y = ( qds )  ------ = qdy dT = r ( qds )
 ds  ds

From the differential geometry shown in Fig. 7.35, the product of r times ds is twice the enclosed area of the tri-
angle with base ds and height r. Integrating the above expressions from point A to point B on the contour we get
B B
Fx =
∫ A
q dx = q ( x B – x A ) Fy =
∫ A
q dy = q ( y B – y A ) T = 2¿q (7.94)

where ¿ denotes is the area enclosed by the contour and the chord connecting the end points of the contour.
From the components of the force given above, the magnitude is determined as

226 Thin-Walled Structures


Resultant of a constant shear flow in a curved branch

y y

yB B
qds
d¿
ds dy
yA q A dx
A r
x x
xA xB

(a) Curved contour (b) Differential element on contour

Fig. 7.35 Constant shear flow along a curved branch.

F = F x2 + F y2 = qL where L = ( xB – x A )2 + ( yB – y A )2

That is, L denotes the length of the chord connecting the end points of the contour. Moreover, the line of action of
the force is parallel to the chord, since F y ⁄ F x = ( y B – y A ) ⁄ ( x B – x A ) . The force and torque resolved at point A
are shown in Fig. 7.36(a). The final reduction is to move the force qL to a parallel line of action from point A to
y y

yB B B

qL
¿ qL
2¿q
L A
yA
A
2¿ ⁄ L
x x
xA xB
(a) Force and torque at A (b) Resultant

Fig. 7.36 Steps to resolving the constant shear to a single force along a specific line of action.

eliminate the torque, and this is shown in Fig. 7.36(b). Thus, the resultant of a constant shear flow in a curved
branch is a force of magnitude equal to the shear flow times the length of the chord connecting the end points of
the contour. The line of action of the resultant force qL is parallel to the chord, offset from the chord by a distance
equal to twice the area enclosed by the contour and the chord divided by the length of the chord, measured per-
pendicular to the chord.

Thin-Walled Structures 227


Bars Subjected to Torsional Loads

From the above discussion, a constant shear flow in a closed contour is statically equivalent to a zero force,
since the beginning and end points of the closed contour coincide. Refer to eqs. eq. (7.94). However, a constant
shear flow in a closed contour is statically equivalent to a torque, and this torque is the same for any point in the
plane about which moments are computed. The fact that the torque is a “free vector” is depicted in Fig. 7.37, in
which it is shown that some of the enclosed area used in Bredt’s formula can add as a negative quantity if the
torque produced by the constant shear flow is clockwise over a segment of the branch. About point O in this fig-
ure, the torque produced by the shear flow from point B to A in the right half of the contour is counter-clockwise,
and the torque produced by the shear flow from A to B in the left half is clockwise. Hence, the total torque is the
sum of these two torques with due respect to the sign. This summation shows that the total torque is proportional
to the area enclosed by the contour.

¿a – ¿b = ¿
q q
A A ¿
¿a ¿b
q

B B
2¿q
O 2¿ a q O O
2¿ b q

Fig. 7.37 The torque about an arbitrary point O of a constant shear flow in a closed contour is twice the
enclosed area of the circuit times the shear flow.

Finally consider torsion of a multi-cell section. Each branch composing the cross section has a constant
shear flow, but the shear flow from branch-to-branch is, in general, a different value. Axial equilibrium requires
that the shear flows into a junction must equal the shear flow out of the junction. If the shear flows satisfy this
condition at all junctions, then the shear flow distribution is statically equivalent to a torque and no force. Hence,
if the resultant is only a torque, then the point in the cross section about which we sum moments is immaterial
because the torque is the same about any point. A two cell section is shown in Fig. 7.38. Since the shear flow is
constant in each branch, the resultant for each branch can be determined by considerations of the single curved
branch given above. The resultant of the exterior branch in cell 1 is a downward force of magnitude q 1 S 1 – 2 ,
where S 1 – 2 is the length of the chord connecting the end points of the contour of the exterior branch. This chord
length, as can be seen in the figure, is the same as the length of the common branch. The line of action of this
resultant is parallel to the chord with the distance from the chord equal to twice the enclosed area of cell 1
divided by the chord length. Similarly, the resultant of the exterior branches of cell 2 is an upward force of mag-
nitude q 2 S 1 – 2 with a line of action parallel to the common branch. The common branch resultant is an upward
force of magnitude ( q 1 – q 2 )S 1 – 2 whose line of action is the contour of the common branch. From Fig. 7.38,
the force and torque about any point on the contour of the common branch are
F = – q 1 S 1 – 2 + ( q 1 – q 2 )S 1 – 2 + q 2 S 1 – 2 = 0

2 ¿1 2 ¿2
T = ----------- ( q 1 S 1 – 2 ) + ----------- ( q 2 S 1 – 2 ) = 2 ¿ 1 q 1 + 2 ¿ 2 q 2
S1 – 2 S1 – 2

These resultants confirm eq. (7.84).

228 Thin-Walled Structures


Resultant of a constant shear flow in a curved branch

( q 1 – q 2 )S 1 – 2 F
q1 – q2 T
q2 S1 – 2
¿1 ¿2 q1 S1 – 2

⇔ S1 – 2 ⇔
q1
q2

2 ¿1 2 ¿2
------------ ------------
S1 – 2 S1 – 2

Fig. 7.38 Branch-by-branch statical equivalence of a two-cell section

EXAMPLE 7.7 Torsion of a five cell closed section; circuit shear flow

Consider the five cell section shown in Fig. 7.39. All branches have the same thickness t and shear modulus G.
For an applied torque T, determine
• shear flows q1, q2, q3, q4, and q5, and
• the torsion constant J
a a a

q5 q4 q3 a
q4 – q1 q2 – q4
T

q2 q1 – q2
q1 a
Balance of shear flows
at the center junction

3a/2 3a/2

Fig. 7.39 Uniform torsion of a five cell section

Solution It is convenient to define circuit shear flows in each cell. Circuit shear flows are assumed to be positive
counter-clockwise in each cell and are equal to the actual shear flows in the exterior branches of the cell, if there
are any exterior branches. However, the shear flow in a common branch between cells is the difference in the cir-
cuit shear flows sharing the common branch. This procedure automatically satisfies shear flow continuity at the
junctions;. i.e., axial equilibrium at the junctions. See the sketch of the center junction in Fig. 7.39.

Thin-Walled Structures 229


Bars Subjected to Torsional Loads

Using Bredt’s formula to compute the torque for each cell, then summing these individual torques, assuming
counter-clockwise as positive, results in

3 3
2  --- a 2 q 1 + 2  --- a 2 q 1 + 2 ( a 2 )q 3 + 2 ( a 2 )q 4 + 2 ( a 2 )q 5 = T (7.95)
2  2 

The twist per unit length for each cell, with positive shear flows taken counter-clockwise within a cell, is

dθ z
 ------- 1 5 a
- = ------------------------  --- a q 1 + a ( q 1 – q 2 ) +  --- ( q 1 – q 4 ) + a ( q 1 – q 5 ) (7.96)
 dz  1 3  2   2
2  --- a 2 Gt
2 

 dθ 1 5 a
--------z = ------------------------  --- a q 2 + a ( q 2 – q 3 ) +  --- ( q 2 – q 4 ) + a ( q 2 – q 1 ) (7.97)
 dz  2 3 2   2
2  --- a 2 Gt
2 

dθ z
 ------- 1
- = -------------------
- [ ( 2a )q 3 + a ( q 3 – q 4 ) + a ( q 3 – q 2 ) ] (7.98)
 dz  3 2 ( a 2 )Gt

 dθ 1
--------z = -------------------
a a
- aq 4 + a ( q 4 – q 5 ) +  --- ( q 4 – q 1 ) +  --- ( q 4 – q 2 ) (7.99)
 dz  4 2 ( a 2 )Gt  2  2

 dθ 1
--------z = -------------------
- [ ( 2a )q 5 + a ( q 5 – q 1 ) + a ( q 5 – q 4 ) ] (7.100)
 dz  5 2 ( a 2 )Gt

Since the cross section is assumed to be rigid in its own plane, the unit twist of each cell must be the same. This
kinematic condition can be written as

dθ z
 ------- dθ dθ z dθ dθ z dθ dθ z dθ
- =  --------z  -------
- =  --------z  -------
- =  --------z  -------
- =  --------z (7.101)
 dz  1  dz  2  dz  2  dz  3  dz  3  dz  4  dz  4  dz  5
Substitute eqs. (7.96) to (7.100) into eqs. (7.101) to get four equations for the five shear flows. Then add eq.
(7.95) to get the fifth equation. These five equations in matrix form are

1 1
2 –2--- 0 – ---
( dθ z ⁄ dz ) 1 = ( dθ z ⁄ dz ) 2 : 3 3
q1 0
1 13 7 1
( dθ z ⁄ dz ) 2 = ( dθ ⁄ dz ) 3 : – --- ------ – --- --- 0 q2
3 6 3 3 0
( dθ z ⁄ dz ) 3 = ( dθ z ⁄ dz ) 4 : 1 1 1 1 q3 = 0 (7.102)
--- – --- – 2 --- ---
( dθ z ⁄ dz ) 4 = ( dθ z ⁄ dz ) 5 : 4 4 2 2 q 0
4
1 1 5 q T
eq. (5.65): --- – --- 0 2 – --- 5
4 4 2
3a 3a 2 2a 2
2 2a 2 2a 2
Solving eq. (7.102) for the shear flows, we get
23 T 22 T 55 T 22 T
q 1 = q 2 = --------- ----2- q 3 = --------- ----2- q 4 = --------- ----2- q 5 = --------- ----2-
281 a 281 a 562 a 281 a
The unit twist of the section can be determined by substituting the above shear flows into any one of the eqs.

230 Thin-Walled Structures


Torsion of hybrid sections

(7.96) to (7.100). Using eq. (7.96), the unit twist is


dθ z 75 T
-------- = ------------ -----------
-
dz 1124 a 3 tG

dθ T
Compare this to the standard formula --------z = ------- to find
dz GJ

1 75 1 1124
--- = ------------ ------
- Hence J = ------------ a 3 t = 14.9867a 3 t
J 1124 a 3 t 75

If all the common branches were removed to make the section shown in Fig. 7.38 a single cell of rectangular
section 2a by 3a, then from eq. (7.71) the torsion constant is J = 14.40a 3 t . For this example, subdividing the
single cell section into five cell section shown in Fig. 7.38 increases the torsional stiffness by only 4.07% with
respect to the single cell section, while the weight of the five cell section increases by 60% with respect to the
weight of the single cell section. However, a multi-cell section may be required for improved damage tolerance;
i.e., if we modeled damage as a fracture, or cut, of an exterior branch, then the loss of torsional stiffness of the
single cell would be substantial since it becomes an open section. Damage to an exterior branch of a multi-cell
section on the other hand results in less of a reduction in torsional load carrying capability since some closed
cells remain intact to carry the torsional load.

7.11 Torsion of hybrid sections


Consider a hybrid section composed a single closed cell
and open branches, or fins, as shown in Fig. 7.40 The total
torque carried by the section is the sum of the torques car- ti bi
ried by the closed cell and open branches. For n open t 2
branches, we have
n
¿ i =1
T = T closed + ∑ Ti (7.103) T
i=1
n-1
with

dθ z dθ z i=n
T closed = ( GJ ) closed  -------- T i = ( GJ ) i  -------- (7.104)
 dz   dz 
Fig. 7.40 Torsion of a hybrid section
The torsional stiffness for the closed cell is

4¿ 2
( GJ ) closed = ---------- (7.105)
ds
°∫
------
Gt
and for each open branch the torsional stiffness is

1
( GJ ) i = G i  --- b i t i3 (7.106)
3 

Combining eqs. (7.103) and (7.104), we have

Thin-Walled Structures 231


Bars Subjected to Torsional Loads

n
dθ z
T = ( GJ ) closed + ∑ ( GJ ) i --------
dz
i=1

dθ z
Comparing this result to the standard torsional formula T = ( GJ ) eff  -------- , the effective torsional stiffness for
 dz 
the entire section is
n

( GJ ) eff = ( GJ ) closed + ∑ ( GJ ) i (7.107)


i=1

where the closed and open parts of the torsional stiffness are given by eqs. (7.105) and (7.106).

The shear stress in the closed cell is τ closed = T closed ⁄ ( 2¿t ) , where the portion of the torque carried by the
closed cell is T closed = ( GJ ) closed ( dθ z ⁄ dz ) . The unit twist is given by ( dθ z ⁄ dz ) = T ⁄ ( GJ ) eff . Combining
these results, the shear stress in the closed cell is
( GJ ) closed T
τ closed = ------------------------- ⋅ ---------- (7.108)
( GJ ) eff 2¿t
For the open branches, we use the formula for the shear stress given by the first of eqs. (7.21). Repeating this
equation, using the second of eqs. (7.104) for the torque carried by the branch, and then using eq. (7.106) for the
torsional stiffness of the branch, we can formulate the shear stress as

3T 3 1 dθ z 3 1 T
τ i = --------2-i = --------2- ⋅  G i --- b i t i3 -------- = --------2- ⋅  G i --- b i t i3 ⋅ ------------------
bi t i bi t i  3  dz bi t i  3  ( GJ ) eff

The unit twist was eliminated in this equation using the overall section formula for it. After simplification of the
above result we get
Gi t i
τ i = ------------------ T (7.109)
( GJ ) eff
If the shear modulus is the same in all branches, then eqs. (7.108) and (7.109) reduce to
J closed T Tt
τ closed = --------------- ⋅ ---------- τ i = -------i (7.110)
J 2 ¿t J

7.12 References
Oden, J. T., and Ripperger, E. A., 1981, Mechanics of Elastic Structures, Second Edition, Hemisphere Publish-
ing Corporation, New York, pp.51-56.

Timoshenko, S.P., and Goodier, J.N., 1970, Theory of Elasticity, Third Edition, McGraw-Hill Book Company,
New York, pp. 291-313.

232 Thin-Walled Structures


Problems

7.13 Problems
1. Determine torsion constant J and the magnitude of the maximum shear stress 3b ⁄ 2
τ max for the section shown.
t
2. The wall thickness for each single cell section shown below is 1.5 mm, and t
the maximum shear stress for the material is 2.5 MPa. Determine the maximum
torque that can be applied to section (a) and section (b).
b

50 mm 50 mm
b 2t
10mm
20mm

50 mm 1.5mm 50 mm
1.5mm

10mm 20mm
(a) (b)

Figure for problem 2

3. The uniform, thin-walled, box beam shown below is clamped at the wall and free at the tip. It is subjected to
a uniform distributed load of intensity 40 N/mm along the front web. The shear modulus of the top and bottom
skins is 18 GPa, and the shear modulus of the webs is 26GPa. Determine the distribution of the twist; i.e, function
θ z ( z ) , 0 ≤ z ≤ 2500mm .

y 40 N/mm 40N/mm

x 1.2mm G = 18GPa
G = 26GPa y
2.1mm 1.2mm x 250mm
2.1mm

1000mm
θz 2500mm
cross section
z
Figure for problem 3

4. The thin-walled, closed section shown below consists of two straight branches inclined at 45 degrees from
the horizontal axis and a circular branch of radius r. The x-axis is an axis of symmetry and the second area
moment about the x-axis is I xx = 0.618731r 3 t . All branches have the same thickness t and same shear modulus
G.
a) Determine the shear flow distributions q 1 ( s ), q 2 ( s ), and q 3 ( s ) in each branch due to bending in terms

Thin-Walled Structures 233


Bars Subjected to Torsional Loads

of shear force V y , the shear flow q 0 at point O, and radius r. Take positive shear flow counter-clockwise
around the section as shown with the contour origin at point O. Note: in the circular branch 2 take
s 2 = rθ , 0 ≤ θ ≤ π ⁄ 2 , and hence the y-coordinate to point s2 is y ( s 2 ) = r cos ( π ⁄ 4 + θ )

b) Determine the shear flow q 0 at point O in terms of V y and r by enforcing zero unit twist.(answer:
q 0 = 0.495416V y ⁄ r )

c) Determine the location of the shear center e from point O in terms of r by torque equivalence.

y
Vy s 2, q 2

s 1, q 1
r
S.C. 45° O x
axis of symmetry
45°
t s 3, q 3

Figure for problem 4


e

5. The two cell closed section bar is subjected to uniform torsion with a torque T as shown. Each branch has the
same thickness t and same shear modulus G.
a) Determine the shear flows q1 and q2 in the exterior branches.
b) Determine the torsion constant J.

q1 q2

a cell 1 cell 2 t, typ.


T

a 2a

Figure for problem 5

234 Thin-Walled Structures


Problems

6. A small slit is cut in the lower exterior branch in cell 2 of problem 5.

a small slit t, typ.


T

a 2a
Figure for Problem 6

a) Determine the torsion constant J.


b) Determine the maximum shear stress.

7. For the homogeneous cross section shown below, take dimensions b = 30 mm and t = 3 mm .
a) Determine the torsion constant J.
b) For a torque T = 600 Nm , determine shear stress in the closed circular portion and the maximum
shearing stress in the open portion.

8. Determine the location e y of the shear center in terms of dimension b for the section shown below.

b
--- the thickness of all branches is t
2
y c = 0.8487b

2 π
C x I yy =  --- + --- b 3 t
b SC  3 8
yc
ey

b b

Figure for Problems 7 and 8

Thin-Walled Structures 235


Bars Subjected to Torsional Loads

236 Thin-Walled Structures


CHAPTER 8 Criteria for Initial
Yielding

The yield stress of a material is determined from the tensile test as discussed in “The tensile test” on page 19. The
yield stress is an important material property in the design of structures made of ductile materials, so ductility is
discussed in this chapter. Also, the tensile test is conducted under axial loading such that the state of stress in the
gage length is simple tension. So the questions arises as to how to use the yield stress determined in simple labo-
ratory specimens in actual structural components subjected to multi-axial states of stress. Criteria are discussed
in this chapter for the prediction of the initiation of material yield under combined stress states.

8.1 Ductile and brittle behavior


A ductile material, like many metals, is one for which the plastic deformation before fracture is much larger than
the elastic deformation. A brittle material, like ceramics and glasses, exhibits little deformation before fracture;
i.e., fracture occurs without very much plastic deformation. See Fig. 8.1. A measure of ductility is the engineer-

σ σ

fracture
fracture

0 ε 0 ε
ductile material brittle material
Fig. 8.1 Ductile and brittle material responses

ing strain at fracture, which is usually report in percent: i.e, 100ε f . A second, widely reported measure of ductil-
ity is the percent reduction in area, denoted as %RA, that is defined by

Thin-Walled Structures 237


Criteria for Initial Yielding

Ai – A f d 2f – d i2
%RA = 100  ----------------- = 100  -----------------
 Ai   d i2 

where A i is the initial cross-sectional area, A f is the minimum cross-sectional area at fracture (occurring in the
necked-down region), and d i and d f are the corresponding diameters for round specimens. Example values of
the percent reduction in area (Dowling,1992, p. 157) are
• 35%RA for 2024-T4 aluminum,
• 33%RA for 7075-T6 aluminum,
• 61%RA for AISI 1020 steel, and
• 6%RA for AISI 4142 steel as quenched.
The abbreviation AISI stands for the American Iron and Steel Institute. Some high strength alloys (like spring
steel for instance) have ductilities as low as 2%, but even this is enough to insure that the material yields before it
fractures and that fracture, when it occurs, is of tough ductile type (Ashby, 1992, p. 14).

An important attribute in design with ductile materials is their capacity to accommodate stress concentra-
tions through plastic deformation and hence to redistribute the stresses more evenly. Stress concentrations occur
at stress raisers which are either geometric discontinuities (e.g., holes, sharp corners, cracks, fillets, etc.) and/or
material discontinuities. Geometric discontinuities are commonly referred to as notches. The capacity to redis-
tribute stresses at stress riser makes a ductile material “tough”, giving the material a defense mechanism against
stress concentrations. Brittle materials, on the other hand, do not possess such a defense against stress concentra-
tion so that even small scratches and cracks as naturally occur in their fabrication can lead to brittle fracture. For
this reason, brittle materials have to be used with extreme caution in tension structures. Ductile engineering
materials are those for which static strength in engineering applications is limited by yielding and not fracture.

8.2 Criteria for initial yielding of ductile materials


The yield stress σ y is determined from the tensile test data, but the tensile test is designed to produce a uniaxial
state of stress. However, we would like to know what governs yielding under combined states of stress that occur
in structural components under service loads. That is, for a three-dimensional state of stress as shown in Fig. 8.2,
what is condition for initiation of yield?

There is no theoretical way to correlate yielding in a three-dimensional stress state with yielding in the
uniaxial tensile test. Two empirical equations have been proposed which are reasonably simple and are descrip-
tive of the data. These criteria are:
• the Mises yield criterion (also known as the distortion energy criterion or the octahedral shear stress crite-
rion), and the
• Maximum shear-stress criterion (also known as the Tresca criterion).
Each criterion is based on two observations, which are:
• The state of stress can be completely described by the magnitude and direction of principal stresses. For an
isotropic material, the orientation of the principal stresses is unimportant, so only the magnitudes enter the
criteria.
• Experiments show that a hydrostatic state of stress does not affect yielding (Dowling, 1993, pp. 106-108).
In the remainder of this section, principal stresses and the hydrostatic stress are defined and discussed. Mises cri-

238 Thin-Walled Structures


Criteria for initial yielding of ductile materials

n
σn
τ ns
τ nz
τ sn

τ zn σs
σ z τ zs τ zn
σz σ = τ sz σ s τ sn
s
P τ nz τ ns σ n
τ sz

τ zs

z
Fig. 8.2 Three dimensional state of stress at a point in the material.

terion and the maximum shear-stress criterion are presented in Section 8.6 and Section 8.7, respectively.

Consider an infinitesimal rectangular parallelepiped at a material point, denoted by P, in a thin-walled beam.


The element is referenced to the three mutually perpendicular directions z, s, and n, where the z-direction is par-
allel to the beam longitudinal axis, the s-direction is tangent to the contour, and the n-direction is parallel to the
thickness of the wall measured normal to the contour. In the most general state of stress there are three stress
components acting on each of the three faces normal to the coordinate directions. These nine stress components
can be written in the matrix form

σ z τ zs τ zn
σ = τ sz σ s τ sn (8.1)

τ nz τ ns σ n

in which the first row contains stress components acting on the z-face, the second row contains components act-
ing on the s-face, and the third row contains stresses acting on the n-face. Column one contains stress compo-
nents acting in the z-direction, the second column contains stress components acting in the s-direction and the
third column contains stress components acting in the n-direction. At point P, it is possible to find three mutually
perpendicular planes through the point on which only normal stresses act, and the shear stresses vanish. This par-
ticular triple of normal stresses are called principal stresses, and principal stresses are an equivalent description
of the state of stress at point P. See Fig. 8.3. The principal stresses are denoted by σ 1, σ 2, and σ 3 . The determi-
nation of the principal stresses is discussed Section 8.4. The important point is to realized that although the stress
matrix is fully populated in, say, the z, s, n coordinate system, it is possible to find three mutually perpendicular
directions at the point, labeled x 1, x 2, and x 3 , in which the stress matrix is a diagonal matrix. An important part
of the proof of this statement, which is not presented here, is that the stress matrix is symmetric for any three
mutually perpendicular directions at the point. That is, moment equilibrium about the n-axis at point P leads to
τ zs = τ sz . Likewise, moment equilibrium about the z-axis and the s-axis at P leads to τ sn = τ ns and τ nz = τ zn ,
respectively. In matrix notation moment equilibrium at a point implies symmetry of the stress matrix, which is
written as

Thin-Walled Structures 239


Criteria for Initial Yielding

x2
n
σn
τ ns
τ nz σ2
τ sn

τ zn σs

σz P
s
P x1
τ sz
σ1
τ zs
σ3
z x3

σ z τ zs τ zn σ1 0 0
σ = τ sz σ s τ sn [σp] = 0 σ2 0
τ nz τ ns σ n 0 0 σ3

Fig. 8.3 Equivalent descriptions of the state of stress at a material point in the body.

σ z τ zs τ zn σ z τ sz τ nz
[ σ ] = [ σ ] T or τ sz σ s τ sn = τ zs σ s τ ns (8.2)

τ nz τ ns σ n τ zn τ sn σ n

where the superscript T denotes matrix transpose. Thus, there only six independent stress components at a point.

Hydrostatic stress, denoted by σ h , is defined by

1 1
σ h = --- ( σ x + σ y + σ z ) = --- Tr [ σ ] (8.3)
3 3
where Tr [ σ ] denotes the trace of the stress matrix. The trace of a square matrix is defined as the sum of its diag-
onal elements. We will show in Section 8.3, eq. (8.15), that the trace of the stress matrix is invariant with respect
to a rotation of the three mutually perpendicular axes through the point. Thus, we also have
1
σ h = --- ( σ 1 + σ 2 + σ 3 ) (8.4)
3
In the case of fluids at rest, the pressure, denoted by p, is the hydrostatic stress. The stress matrix for any three
mutually perpendicular axes at a point in the fluid is

–p 0 0
[σ] = 0 –p 0 (8.5)

0 0 –p

240 Thin-Walled Structures


Stress transformation equations for generalized plane stress

That is, the normal stress on all faces through the point is the negative of the pressure, and all directions at the
point are principal stress directions. (See Example 8.2). Hence, by the definition given in eq. (8.3), the hydro-
static stress σ h = – p .

8.3 Stress transformation equations for generalized plane stress


For thin-walled structures we neglect the shear stress components in the thickness direction, τ zn and τ sn with
respect to the shear stress component tangent to the contour τ zs . The stress matrix in the z, s, n system reduces to
what is called generalized plane stress. That is,

σ z τ zs 0
σ = τ sz σ s 0 (8.6)

0 0 σn

For this generalized state of plane stress, we want to determine the principal stresses and the maximum shear
stress at a material point, again labeled P, in the wall. To determine principal stresses we use equilibrium to find
the normal stress and shear stress on face through point P whose normal is in an arbitrary direction in the z-s
plane. Hence, consider a second orthogonal coordinate system z′, s′, n′ which is rotated about the n-axis through
the angle θ , defined positive counter-clockwise looking down the positive n-axis . The n′-axis and n-axis coin-
cide in this transformation of coordinates. The face normal to the z′ -axis is the oblique face of an infinitesimal
wedge at point P. Looking down the positive n-axis, the wedge appears as a triangle as is shown in Fig. 8.4. You

σz s′
σz d Az τ sz d A s

τ zs θ
P s′ s′
τ sz s
θ θ θ
σs τ zs ′ τ zs d A z σs d As
θ θ
σz ′
θ z′ z′
z′ z-face components s-face components
z

Fig. 8.4 Wedge element as seen looking down the positive n-axis

can think of the wedge as having a unit depth in the n-direction. The normal and shear stress components on this
oblique face are denoted by σ z ′, τ zs ′ , respectively. Also, the positive stress components acting on the negative z-
and negative s-faces are shown on the wedge element in the figure. If the area of the wedge face is denoted by dA,
then from the geometry of Fig. 8.4, the area of the z-face d A z = dA cos θ , and the area on the s-face
d A s = dA sin θ . The stresses are converted to differential forces acting on the wedge by multiplying them by the
area of the face on which they act. Force equilibrium in the z′-direction gives

Thin-Walled Structures 241


Criteria for Initial Yielding

σ z ′dA – ( σ z d A z ) cos θ – ( τ zs d A z ) sin θ – ( σ s d A s ) sin θ – ( τ sz d A s ) cos θ = 0

Divide this equation by dA, take the limit as the wedge shrinks down to point P, and recognize that
d A z ⁄ dA = cos θ , d A s ⁄ dA = sin θ , and τ zs = τ sz , to get

σ z ′ = σ z cos2 θ + σ s sin2 θ + τ zs 2 sin θ cos θ (8.7)

Force equilibrium in the s′ – direction on the wedge element gives

τ zs ′dA + ( σ z d A z ) sin θ – ( τ zs d A z ) cos θ – ( σ s d A s ) cos θ + ( τ sz d A s ) sin θ = 0

Again divide by dA and let the wedge element shrink to point P in the limit to get

τ zs ′ = – σ z cos θ sin θ + σ s sin θ cos θ + τ zs ( cos2 θ – sin2 θ ) (8.8)

The wedge element used to determine the stress components on the s′-face is shown in Fig. 8.5. For the element

σs ′ s′
τ sz
θ
σs
τ sz ′ θ
P s Fig. 8.5 Wedge element with an oblique face
τ zs normal to the s’-axis
σz

θ
z′
z

in Fig. 8.5, the areas are related by d A z = dA sin θ and d A s = dA cos θ . Following procedures similar to those
resulting in eq. (8.7), force equilibrium of this wedge element in the s′-direction leads to

σ s ′ = σ z sin2 θ + σ s cos2 θ – τ zs 2 sin θ cos θ (8.9)

It is also possible to obtain eq. (8.9) from eq. (8.7) by letting θ → θ + 90° and σ z ′ → σ s ′ , since this would rotate
the z′-axis to the original direction of the s′-axis . Equilibrium of the wedge element in Fig. 8.5 in the
z′-direction leads to

τ sz ′ = – σ z sin θ cos θ + σ s cos θ sin θ + τ zs ( cos2 θ – sin2 θ ) (8.10)

which when compared to eq. (8.8) shows that τ zs ′ = τ sz ′ , as would be expected for moment equilibrium about
the n-axis at point P. The stress transformation eqs. (8.7) to (8.10) are written in matrix form as

σ z ′ τ zs ′ 0 cos θ sin θ 0 σ z τ zs 0 cos θ – sin θ 0


τ sz ′ σ s ′ 0 = – sin θ cos θ 0 τ sz σ s 0 sin θ cos θ 0 (8.11)

0 0 σn ′ 0 0 1 0 0 σn 0 0 1

or in the more compact form as


T
σ′ = T σ T (8.12)

242 Thin-Walled Structures


Principal stresses and maximum shear stress

in which [ T ] is the matrix of direction cosines as given in the table below. Since the n′-axis and the n-axis

z s n
z′ cos θ sin θ 0

s′ – sin θ cos θ 0

n′ 0 0 1

coincide in the rotation of ( z′, s′, n′ ) relative to ( z, s, n ) , the normal stress σ n ′ = σ n . The trace of stress matrix
in the coordinates ( z′, s′, n′ ) is given by

σ z ′ τ zs ′ 0
Tr τ sz ′ σ s ′ 0 = σz ′ + σs ′ + σn ′ (8.13)

0 0 σn ′

Using eqs. (8.7), (8.9), and σ n ′ = σ n , we find

σz ′ + σs ′ + σn ′ = σz + σs + σn (8.14)

In matrix notation this is written as

σ z ′ τ zs ′ 0 σ z τ zs 0
Tr τ sz ′ σ s ′ 0 = Tr τ sz σ s 0 (8.15)

0 0 σn ′ 0 0 σn

In other words, the trace of the stress matrix is invariant with respect to a coordinate rotation.

8.4 Principal stresses and maximum shear stress


Mohr’s circle The graphical representation of the stress transformation equations is called the Mohr’s circle.
Mohr’s circle is drawn on a plane with the normal stress plotted on the abscissa the shear stress plotted on the
ordinate. To show that the stress transformation equations can be graphed as a circle on this plane, we first use the
trigonometric identities
1 1
cos2 θ = --- ( 1 + cos 2θ ) sin2 θ = --- ( 1 – cos 2θ ) 2 sin θ cos θ = sin 2θ
2 2
to write eqs. (8.7) to (8.9) as
1 1
σ z ′ = --- ( σ z + σ s ) + --- ( σ z – σ s ) cos 2θ + τ zs sin 2θ (8.16)
2 2

1
τ zs ′ = – --- ( σ z – σ s ) sin 2θ + τ zs cos 2θ (8.17)
2

Thin-Walled Structures 243


Criteria for Initial Yielding

1 1
σ s ′ = --- ( σ z + σ s ) – --- ( σ z – σ s ) cos 2θ – τ sz sin 2θ (8.18)
2 2
Define the average stress σ ave and radius R by

2
1 1 2
σ ave = --- ( σ z + σ s ) R = --- ( σ z – σ s ) + τ zs (8.19)
2 2
Using these definitions we write eqs. (8.16) and (8.17) as
1
--- ( σ z – σ s ) τ zs
2
σ z ′ = σ ave + R ------------------------- cos 2θ + ------ sin 2θ
R R

1
--- ( σ z – σ s ) τ zs
2
τ zs ′ = R – ------------------------
- sin 2θ + -----
- cos 2θ
R R

Define an angle ψ via the trigonometric relations 1


--- ( σ z – σ s )
2
1
--- ( σ z – σ s ) τ zs
2 ψ
cos ψ = ------------------------- sin ψ = -----
- (8.20)
R R
τ zs
R
which is depicted in the triangle at right. Using this definition of ψ we get

σ z ′ = σ ave + R [ [ cos ψ cos 2θ + sin ψ sin 2θ ] ]

τ zs ′ = R [ – cos ψ sin 2θ + sin ψ cos 2θ ]

Using the trigonometric identities


cos ( ψ – 2θ ) = cos ψ cos 2θ + sin ψ sin 2θ

sin ( ψ – 2θ ) = sin ψ cos 2θ – cos ψ sin 2θ


we finally get the form
σ z ′ = σ ave + R cos ( ψ – 2θ ) (8.21)

τ zs ′ = R sin ( ψ – 2θ ) (8.22)

Similar manipulations applied to eq. (8.18) results in


σ s ′ = σ ave – R cos ( ψ – 2θ ) (8.23)

The center of the circle is at ( σ, τ ) = ( σ ave, 0 ) and its radius is R. A positive shear stress on the positive z-face,
or τ zs , is defined positive downward on the ordinate. A positive shear stress on a positive s-face, or τ sz , is
defined positive upward on the ordinate. The circle is shown in Fig. 8.6. The locus of points on the circle repre-
sent all possible combinations of normal and shear stresses on the oblique faces through point P. In particular, the
locations of the z-face stresses and s-face stresses are shown on the circle of Fig. 8.6 assuming the stresses are all
positive in value. (Set θ = 0 in eqs. (8.21) and (8.22) to get the z-face stresses.) The diametrically opposite
points of the circle on the normal stress axis represent the oblique faces on which the shear stresses vanish.

244 Thin-Walled Structures


Principal stresses and maximum shear stress

τ sz
s – face
τ sz

s′-face
τ sz ′ σ ave
σ2 σ1
0 σz σ z'
σ
σs ′ σs
'
τ zs ψ
z ' – face
R 2θ
τ zs
z – face

τ zs
Fig. 8.6 Mohr’s circle

Hence, these diametrically opposite points are represent principal stresses σ 1 and σ 2 at point P. Principal stress
σ 3 = σ n since no shear stresses act on the n-faces . Hence, the principal stresses are given by

σ 1 = σ ave + R σ 2 = σ ave – R (8.24)

The rotation about the n-axis to the principal stress directions occurs when 2θ = ψ . From eqs. (8.20) we get

τ zs
tan 2θ p = ------------------------
- (8.25)
1
--- ( σ z – σ s )
2
where the rotation angle to the principal direction is denoted θ p . The definition for plotting positive shear in the
Mohr’s circle plane results in the sense of the rotation on the circle to be the same sense of the rotation about the
n-axis in the physical plane. However, the rotation angle on Mohr’s circle is twice the rotation in the physical
plane. If the angle θ p is substituted for θ in the matrix transformation eq. (8.11) we get

σ z ′ τ zs ′ 0 σ1 0 0
τ sz ′ σ s ′ 0 = 0 σ2 0 θ → θp (8.26)

0 0 σn ′ 0 0 σ3

As stated above, the third principal stress σ 3 = σ n ′ = σ n .

Thin-Walled Structures 245


Criteria for Initial Yielding

EXAMPLE 8.1 Maximum shear stress in tensile test

Draw Mohr’s circle and determine the maximum shear stress in the uniaxial tension test; i.e, the stress matrix is
given by

σ z τ zs 0 σz 0 0
[ σ ] = τ sz σ s 0 = 0 00 (8.27)

0 0 σn 0 00

Solution First note that the principal stresses are σ 1 = σ z , σ 2 = 0 , and σ 3 = 0 .The average stress
σ ave = σ z ⁄ 2 and radius R = σ z ⁄ 2 . Mohr’s circle is shown in Fig. 8.7 The z-, s-, and n-directions are principal

s σz ⁄ 2 s′
τ sz s′-face σz ⁄ 2
σz ⁄ 2 σz ⁄ 2
σz σz
0 σz ⁄ 2 σ P
45° σz ⁄ 2
σz z
P
90°
σz ⁄ 2
principal stress element z′
τ zs z′-face
maximum shear stress element

Fig. 8.7 Mohr’s circle and stress elements in the tension test

stress directions. The maximum shear stress is τ max = σ z ⁄ 2 , and the maximum shear stress element is rotated
45° clockwise about the n-axis as shown in Fig. 8.7. Note that the normal stresses on the planes where the max-
imum shear stress occurs are non-zero and equal to σ z ⁄ 2 as well.

EXAMPLE 8.2 Mohr’s circle for hydrostatic stress state

Draw Mohr’s circle for the hydrostatic stress state given by

σ z τ zs 0 –p 0 0
[ σ ] = τ sz σ s 0 = 0 –p 0
0 0 σn 0 0 –p

where p is the pressure.


τ sz
Solution The average stress σ ave = – p and the radius R = 0 . Hence,
Mohr’s circle reduces to a point on the negative normal stress axis as shown –p σ
0
in Fig. 8.8. The shear stress is zero and the normal stress is compressive
with magnitude p on every plane through point P. The principal stresses are
σ1 = σ2 = σ3 = – p
Fig. 8.8 τ zs

246 Thin-Walled Structures


Principal stresses and maximum shear stress

EXAMPLE 8.3 Principal stresses and maximum shear stress

Determine the principal stresses and the maximum shear stress for the state of given by the matrix

σ z τ zs 0 60 – 12 0
[ σ ] = τ sz σ s 0 = – 12 15 0 MPa
0 0 σn 0 0 0

Sketch the principal stress element and the maximum shear stress element.

Solution The average stress is given by σ ave = ( 60 + 15 ) ⁄ 2 = 37.5MPa , and the radius is

2
1
R = --- ( 60 – 15 ) + 12 2 = 25.5MPa
2

Mohr’s circle is shown in Fig. 8.9. The principal stresses are σ 1 = 63MPa , σ 2 = 12MPa , and σ 3 = 0 . The

stresses in MPa s x2
15 12
14.04°
z-face 12
– 12
R σ
0 60
15 37.5 60 12 63
z n, x 3
n
x1
R = 25.5
τ zs principal stress element

Fig. 8.9 Mohr’s circle and the principal stress element

rotation to the principal stress element is


– 12
tan 2θ p = -------------------------- = – 0.53333333 θ p = – 14.04°
1
--- ( 60 – 15 )
2
The principal stress element is sketched in Fig. 8.9.

The maximum shear stress on the oblique plane of a wedge element obtained by a rotation about the n-axis,
or x3-axis, is 25.5MPa. However, this is not the maximum shear stress at the material point in question. To see
this, begin with the principal stress element and consider a second Mohr’s circle for a rotation about the x 2 -axis ,
which would have principal stresses σ 1 and σ 3 defining its diameter. This second Mohr’s circle is centered at
( σ 1 + σ 3 ) ⁄ 2 with a radius ( σ 1 – σ 3 ) ⁄ 2 . The locus of points on this second Mohr’s circle represents all possible
combinations of the normal and shear stresses on wedge elements with oblique faces normal to the x 1 ′-axis ,
where the x 1 ′-axis is rotated from the x 1 -axis counter-clockwise as view down the positive x 2 -axis . See Fig.
8.10. The maximum shear stress on this second Mohr’s circle is simply its radius, and it occurs for counter-clock-

Thin-Walled Structures 247


Criteria for Initial Yielding

stresses in MPa x2
12
τ x 1 ′-face 14.04°
principal stress element

z-face 63
n, x 3
x 3 -face x 2 -face 90° x 1 -face x1
0 63 σ
12 31.5
x1 ′

63 σ 1 ′ = 31.5
45°

x 2, x 2 ′ 12 τ 13 ′ = 31.5

τ x1
45° maximum shear stress
on a wedge element
x3 x3 ′

Fig. 8.10 Mohr’s circle associated with the maximum shear stress

wise rotation of 90° from the x 1 face stress location on the circle. This rotation corresponds to a counter-clock-
wise rotation of 45° in the physical plane as shown in Fig. 8.10. Equilibrium of the wedge element with an
oblique face normal to the x 1 ′-axis determines the shear and normal stresses on the wedge face, which from
Mohr’s circle we know to both have a magnitude of 31.5MPa. It is informative to verify this by summing forces
to zero on the wedge element shown in Fig. 8.10. For convenience, take the area of the oblique face to be unity.
Then the area on the negative x 1 -face is 1 × cos 45° . Note that the negative x 3 face on the wedge element is
stress free. Summing forces in the x 1 ′-direction to zero we get

σ 1 ′ × 1 – [ 63 × 1 × cos 45° ] × cos 45° = 0

Equating forces to zero in the x 3 ′-direction we get

τ 13 ′ × 1 – [ 63 × 1 × cos 45° ] × sin 45° = 0

Hence, the stresses on the x 1 ′-face are

σ 1 ′ = 31.5MPa τ 13 ′ = 31.5MPa (8.28)

The maximum shear stress for the stress state is 31.5MPa.

248 Thin-Walled Structures


Octahedral shear stress

8.5 Octahedral shear stress


Since the hydrostatic stress σ h is observed not to effect yielding, we find the plane at the material point for which
the normal stress is σ h , and then assume that the shear stress acting on this plane, denoted by τ h , is the stress
component responsible for initiating yielding. We start the analysis to find the shear stress τ h with the principal
stress element at point P, and consider equilibrium of an infinitesimal tetrahedron at P. The direction of the unit
normal to the oblique face of the tetrahedron, denoted by n̂ , is unknown initially, but we can determine it from
the condition that the normal stress acting on the oblique face is the hydrostatic stress. Let S ( n ) denote the stress
vector acting on the oblique face, and let dA denote the area of the oblique face of the tetrahedron. See Fig.
8.11. The stress vectors acting on the three triangular faces normal to the principal stress directions are
x2
x2

σ 3 ( – î 3 )d A 3

σ 1 ( – î 1 )d A 1
d x2
– d x 3 î 3 + d x 2 î 2
P dx
1
d x3 P
x1
– d x 3 î 3 + d x 1 î 1 x1

σ 2 ( – î 2 )d A 2
x3 n̂
x3 2dAn̂
S ( n ) dA

Fig. 8.11 Tetrahedron element at point P.

σ 1 d A 1 ( – î 1 ) , σ 2 d A 2 ( – î 2 ) , and σ 3 d A 3 ( – î 3 ) , where d A 1 , d A 2 , and d A 3 are the areas of the triangular faces of


the tetrahedron normal to the unit vectors î 1 , î 2 , and î 3 along the principal directions, respectively. Force equi-
librium of the tetrahedron gives the vector equation

S ( n ) dA – σ 1 d A 1 î 1 – σ 2 d A 2 î 2 – σ 3 d A 3 î 3 = 0 (8.29)

Divide this equation by the area of the oblique face dA and shrink the tetrahedron to point P in the limit to get
d A1 d A2 d A3
S ( n ) = σ 1 ---------î 1 + σ 2 ---------î 2 + σ 3 ---------î 3 (8.30)
dA dA dA

The ratio of the areas of the faces normal to the principal directions to the area of the oblique face is obtained
from geometry of the tetrahedron. Consider two of the edge vectors of the oblique face given by

( – d x 3 î 3 + d x 1 î 1 ) and ( – d x 3 î 3 + d x 2 i 2 ) (8.31)

as are shown in Fig. 8.11. The geometric interpretation of the vector cross product of two position vectors is a
vector normal to the plane of the two vectors with magnitude equal to the area of a parallelogram formed by the

Thin-Walled Structures 249


Criteria for Initial Yielding

two position vectors. Using the above two edge vectors we have by this interpretation

2dAn̂ = ( – d x 3 î 3 + d x 1 î 1 ) × ( – d x 3 î 3 + d x 2 i 2 ) (8.32)

Expanding the cross product in this equation gives

2dAn̂ = ( – d x 3 d x 2 ) ( î 3 × î 2 ) – d x 1 d x 3 ( î 1 × î 3 ) + d x 1 d x 2 ( î 1 × î 2 )

2dAn̂ = d x 2 d x 3 î 1 + d x 1 d x 3 î 2 + d x 1 d x 2 î 3

The areas of the triangular faces of the tetrahedron normal to the principal directions are d A 1 = ( d x 2 d x 3 ) ⁄ 2 ,
d A 2 = ( d x 1 d x 3 ) ⁄ 2 , and d A 3 = ( d x 1 d x 2 ) ⁄ 2 . Hence, the above equation reduces to

d A1 d A2 d A3
n̂ = ---------î 1 + ---------î 2 + ---------î 3 (8.33)
dA dA dA
In general, a unit vector is represented as

n̂ = n 1 î 1 + n 2 î 2 + n 3 î 3 where n 12 + n 22 + n 32 = 1 (8.34)

Component n 1 is the cosine of the angle between the normal and the positive x 1 -axis with similar interpreta-
tions for components n 2 and n 3 . Comparing eqs. (8.33) and (8.34), it is clear that the direction cosines of the unit
normal also relate the areas of the faces of the tetrahedron as
dA dA dA
n 1 = ---------1 n 2 = ---------2 n 3 = ---------3 (8.35)
dA dA dA
Substitute the direction cosines of the unit normal from eq. (8.35) into eq. (8.30). Thus, we find the stress vector
on the oblique face of the tetrahedron is related to the principal stresses by the vector relation

S ( n ) = σ 1 n 1 î 1 + σ 2 n 2 î 2 + σ 3 n 3 î 3 (8.36)

The normal stress component on the oblique face of the tetrahedron is given by the scalar product
σ nn = n̂ • S ( n ) . Using eqs. (8.34) and (8.36) this scalar product is

σ nn = σ 1 n 12 + σ 2 n 22 + σ 3 n 32 (8.37)

We need to find the direction of the oblique face, i.e., to find direction cosines n 1, n 2, and n 3 , such that the nor-
mal stress on this face is equal to the hydrostatic stress, σ h . From eq. (8.4) the hydrostatic stress is

1 1 1
σ h = σ 1 --- + σ 2 --- + σ 3 --- (8.38)
3 3 3
Setting σ nn = σ h and substituting eqs. (8.38) into (8.37), determines the square of the direction cosines as

1 1 1
n 12 = --- n 22 = --- n 32 = --- (8.39)
3 3 3
The result for the direction cosines in this equation implies that there are actually eight oblique faces at the point

250 Thin-Walled Structures


Octahedral shear stress

P on which the hydrostatic acts. Unit normal vectors to these eight oblique faces are given the table below. These

Components
Unit normal
vectors n1 n2 n3
n̂ 1 1⁄ 3 1⁄ 3 1⁄ 3
n̂ 2 –1 ⁄ 3 1⁄ 3 1⁄ 3
n̂ 3 –1 ⁄ 3 –1 ⁄ 3 1⁄ 3
n̂ 4 1⁄ 3 –1 ⁄ 3 1⁄ 3
n̂ 5 1⁄ 3 1⁄ 3 –1 ⁄ 3
n̂ 6 –1 ⁄ 3 1⁄ 3 –1 ⁄ 3
n̂ 7 –1 ⁄ 3 –1 ⁄ 3 –1 ⁄ 3
n̂ 8 1⁄ 3 1⁄ 3 –1 ⁄ 3

eight particular oblique planes at point P are called octahedral planes. The normal stress on the octahedral planes
is the hydrostatic stress.

Now we need to find the shear stress on the octahedral plane. The shear stress acting on an octahedral plane
is called the octahedral shear stress, and it will have the same value on all eight of the octahedral planes at point
P. To find its value, consider the octahedral plane in the first octant of the cartesian space defined by the principal
directions as shown in Fig. 8.12 The stress vector can be written as

x3
1
σ h n̂ n̂ = n̂ 1 = ------- ( î 1 + î 2 + î 3 )
3

S(n)
P x2

τ h t̂
x1
Fig. 8.12 Normal stress and shear stress components acting on an octahedral plane.

S ( n ) = σ h n̂ + τ h t̂ (8.40)

where t̂ is a unit vector tangent to the octahedral plane and τ h is the octahedral shear stress. The scalar product
of the stress vector with itself yields the square of its magnitude. Using eq. (8.36) we have

S ( n ) • S ( n ) = ( σ1 n1 ) 2 + ( σ2 n2 ) 2 + ( σ3 n3 ) 2 (8.41)

While from eq. (8.40) we have

Thin-Walled Structures 251


Criteria for Initial Yielding

S ( n ) • S ( n ) = σ h2 + τ h2 (8.42)

Equations (8.41) and (8.42) are equated to one another, eq. (8.39) is substituted for the square of the direction
cosines, and then the resulting equation is solved for the octahedral shear stress to get
1 1 1
τ h2 = σ 12 --- + σ 22 --- + σ 32 --- – σ h2 (8.43)
3 3 3
Now substitute eq. (8.38) for the hydrostatic stress in this equation to get

1 1 1 1 1 1 2
τ h2 = σ 12 --- + σ 22 --- + σ 32 --- –  σ 1 --- + σ 2 --- + σ 3 --- , or
3 3 3  3 3 3

2
τ h2 = --- ( σ 12 + σ 22 + σ 32 – σ 1 σ 2 – σ 1 σ 3 – σ 2 σ 3 )
9
This last expression can be written as
1
τ h2 = --- [ ( σ 1 – σ 2 ) 2 + ( σ 2 – σ 3 ) 2 + ( σ 3 – σ 1 ) 2 ]
9
Hence the octahedral shear stress magnitude is
1
τ h = --- [ ( σ 1 – σ 2 ) 2 + ( σ 2 – σ 3 ) 2 + ( σ 3 – σ 1 ) 2 ] (8.44)
3

8.6 Mises criterion for initiation of yielding


Mises Criterion
Yielding begins in a three-dimensional stress state when the octahedral shear stress τ h is
equal to its value at yield initiation in the uniaxial tension test

In the uniaxial tension test the principal stresses at the initiation of yielding are σ 1 = σ y , σ 2 = 0 σ 3 = 0 .
(Refer to Example 8.1). Substitute these principal stress values into eq. (8.44) to get

2
τ h = ------- σ y (8.45)
3
For the three-dimensional state of stress, we substitute eq. (8.45) for the octahedral shear stress in eq. (8.44) to
get

1
--- [ ( σ 1 – σ 2 ) 2 + ( σ 2 – σ 3 ) 2 + ( σ 3 – σ 1 ) 2 ] = σ y (8.46)
2
Equation (8.44) is the mathematical statement of Mises criterion for the initiation of yielding.

Mises criterion can be visualized in principal stress space, which has orthogonal axis σ 1 , σ 2 , and σ 3 as is
shown in Fig. 8.13 In this space the yield surface given by eq. (8.46) is a right-circular cylinder of radius
2 ⁄ 3 σ y whose axis makes equal angles with respect to the σ 1 , σ 2 , and σ 3 coordinate axes. Points within the
cylinder correspond to stress states which have not initiated yielding, while points outside the cylinder corre-

252 Thin-Walled Structures


Mises criterion for initiation of yielding

axis of cylinder
σ 1 s1
----- 0
σy 1
2

2
s3σ 3 --- σ y radius
----- 3
σy 1

0
1
σ
-----2
σy 2
s2

Fig. 8.13 Mises yield surface in principal stress space

spond to stress states that are beyond yielding. Notice that for any stress state on the axis of the cylinder, for
which σ 1 = σ 2 = σ 3 , no yield is predicted no matter the magnitude of the stresses! This reflects the assumption
that hydrostatic stress does not initiate yielding.

In design, it is convenient to define the Mises stress, denoted σ M , by the definition

1
σM = --- [ ( σ 1 – σ 2 ) 2 + ( σ 2 – σ 3 ) 2 + ( σ 3 – σ 1 ) 2 ] (8.47)
2
Then, the constraint that yielding not occur is
σM < σy (8.48)

Also, it is convenient to define the factor of safety, denoted as FS, with respect to yielding as
σy
FS = ------- > 1 for no yielding (8.49)
σM

Thin-Walled Structures 253


Criteria for Initial Yielding

For the most general state of stress given by

σ z τ zs τ zn
σ = τ sz σ s τ sn
τ nz τ ns σ n

It can be shown after much algebra (which was done using Mathematica) that the Mises stress can be written in
the equivalent form

σM = σ z2 + σ s2 + σ n2 – σ z σ s – σ z σ n – σ s σ n + 3τ zs
2 + 3τ 2 + 3τ 2
zn sn (8.50)

This equation represents a simplification in computing the Mises stress, since we can omit the step of computing
the principal stresses. For the case of generalized plane stress where τ zn = τ sn = 0 , eq. (8.50) reduces to

σM = σ z2 + σ s2 + σ n2 – σ z σ s – σ z σ n – σ s σ n + 3τ zs
2 (8.51)

8.7 Maximum shear-stress criterion


This second empirical criterion assumes that yielding occurs whenever the maximum shear stress attains the
value it has when yielding begins in the tension test. The maximum shear stress is one half of the difference of
the maximum and minimum principal stresses; i.e.,
τ max = ( σ max – σ min ) ⁄ 2 (8.52)

where
σ max = Max ( σ 1, σ 2, σ 3 ) σ min = Min ( σ 1, σ 2, σ 3 ) (8.53)

From Example 8.1, the maximum shearing stress in the tensile test at the initiation of yield is
τ max = σ y ⁄ 2 (8.54)

For a three-dimensional state of stress, we substitute eq. (8.52) for the maximum shear stress in eq. (8.50) to get
σ max – σ min = σ y (8.55)

Equation (8.53) is the mathematical statement for the initiation of yielding by the maximum-shear stress crite-
rion.

A plot of the maximum shear-stress criterion and the Mises criterion for σ 3 = 0 is shown in Fig. 8.14. Cal-
culating the maximum shear stress depends on the numerical order of the principal stresses. In the following
paragraphs this procedure is discussed in terms of plotting the maximum shear-stress criterion shown in Fig.
8.14.

In the first quadrant of Fig. 8.14, both σ 1 and σ 2 are positive. If σ 1 > σ 2 , then
Mohr’s circles appear as shown in the figure to the right, and the maximum shear
σ3 σ2 σ1
stress is σ 1 ⁄ 2 . Therefore, maximum shear-stress criterion plots as a vertical line
0
with σ 1 = σ y for 0 < σ 2 < σ y . (Note that stresses normalized by σ y are plotted in

254 Thin-Walled Structures


Maximum shear-stress criterion

σ2 ⁄ σ y

1
Mises criterion

0.5
Maximum
shear-stress criterion

σ1
-----
-1 -0.5 0.5 1 σy

-0.5

σ3 = 0

-1

Fig. 8.14 Criteria for yield initiation in the σ 1 – σ 2 principal stress plane.

Fig. 8.14.) If σ 2 > σ 1 > 0 , then the role of σ 1 and σ 2 interchange on the Mohr’s circles, so the maximum shear
stress is now σ 2 ⁄ 2 . Therefore, maximum shear-stress criterion plots as the horizontal line σ 2 = σ y for
0 < σ1 < σ y .

In the second quadrant of Fig. 8.14 σ 1 < 0 and σ 2 > 0 , so that Mohr’s circles
appear as shown in the figure to the right. The maximum shear stress is
σ1 σ3 σ2
( σ 2 – σ 1 ) ⁄ 2 . Therefore, the maximum shear-stress criterion, eq. (8.55), gives
0
σ 2 = σ y + σ 1 in the second quadrant. This is a straight line with a one-to-one slope
intersecting the σ 2 -axis at σ y . Plotting the maximum shear-stress criterion in quad-
rants three and four in Fig. 8.14 proceed in a similar manner.

The maximum shear-stress envelope in Fig. 8.14 is contained within the Mises envelope. Hence, the maxi-
mum shear stress criterion is conservative from a design perspective, with the largest differences between the
predictions being about 15%. However, in analysis the Mises criterion is easier to implement than the maximum
shear stress criterion. Mises criterion is a single equation, see eq. (8.50), but the maximum shear stress criterion
requires that we compute the principal stresses and find their numerical order. Also note that the maximum shear
stress acts on four planes at the material point, refer to Fig. 8.10, while the octahedral shear stress acts on eight
planes at the material point. Laboratory tests on thin-walled tubes subject to an axial force, torque, and internal
pressure are often used to study yielding under combined stress states. The experimental data for ductile metal
tubes fall between the maximum shear-stress criterion and the Mises criterion on a plot such as Fig. 8.14, with
the data closer to the Mises prediction (Dowling, 1993, pp. 251 and 252).

In design, the limit state for no yielding by maximum shear-stress criterion is simply

Thin-Walled Structures 255


Criteria for Initial Yielding

τ max < σ y ⁄ 2 (8.56)

and we can define a factor of safety against yielding as


σy ⁄ 2
FS = ------------ > 1 for no yielding (8.57)
τ max

EXAMPLE 8.4 Factor of safety against initial yielding

Compute the factor of safety against the initiation of yield by Mises criterion and the maximum shear-stress
criterion for 2024-T6 aluminum alloy that has a yield stress of 325 MPa.

Solution From Example 8.3, the principal stresses are σ 1 = 63MPa , σ 2 = 12MPa , and σ 3 = 0 , and the
maximum shear stress is 31.5MPa . If we calculate the Mises stress from eq. (8.47), we get

1
σM = --- ( ( 63 – 12 ) 2 + ( 12 – 0 ) 2 + ( 0 – 63 ) 2 ) = 57.94MPa
2
If we calculate the Mises stress via eq. (8.51), we get

σM = 60 2 + 15 2 – 60 × 15 – 0 – 0 + 3 × ( – 12 ) 2 = 57.94MPa

which is the same value as obtained from eq. (8.47). Hence the factor of safety against yield, eq. (8.49), is
325
FS = ------------------- = 5.61
57.9396

The factor of safety against yield using the maximum shear stress criterion, eq. (8.57), is
325 ⁄ 2
FS = ---------------- = 5.16
31.5
In linear structural analysis where the stresses are proportional to the load, the factor of safety means that the load
can be increased by 5.61, in the case of Mises criterion, before the material initiates yielding. In the case of the
maximum shear-stress criterion, the load can be increased by 5.16 before the material begins to yield. The lower
factor of safety predicted by the maximum shear stress criterion illustrates it is slightly conservative with respect
Mises prediction (in this case by about 8%).

256 Thin-Walled Structures


Maximum shear-stress criterion

EXAMPLE 8.5 Stress responses of a stringer-stiffened, single cell beam.

The thin-walled, prismatic beam of length L shown in Fig. 8.15 is clamped at z = 0. It is subjected to a linearly
distributed load p y ( z ) = p 0 ( 1 – z ⁄ L ) acting through the locus of shear centers, and a torque T applied at z = L.
The cross-sectional contour is an isosceles triangle with branches one and three having the same length b and
the same thickness t 1 . The vertical branch, or second branch, has length h and thickness t 2 . The beam is stiff-
py
T
py( z) As t1 y

h s1
T --- s2
2 S.C. C
z x
h t2
---
2 t1
s3
L b
As

Cross-section

Fig. 8.15 A stringer-stiffened, cantilevered beam. The contour in the cross section is an isosceles
triangle.

ened by two longitudinal stringers of cross-sectional area A s . The overall dimensions are specified as L = 1800
mm, h = 100 mm, and b = 130 mm. The material is 2024-T4 aluminum alloy whose properties are listed in the
table below.
Aluminum alloy 2024-T4

Property Value Units


E, modulus of elasticity 73 GPa
ν, Poisson’s ratio 0.33 none
G, Shear modulus 27 GPa
ρ, density 2800 kg/m3
σyield, yield strength in tension 325 MPa

Take p 0 = 3.0 N/mm , T = 750 N-m , t 1 = t 2 = 1.0 mm , and


A s = 45 mm 2 . For the cross section at z = 0, τ zs s

d. plot the axial normal stress σ z in MPa versus the contour coordinate s in σz s = 0

mm, and
e. the shear stress τ zs in MPa versus the contour coordinate s in mm.
The contour coordinate is related to the branch contour coordinates by

Thin-Walled Structures 257


Criteria for Initial Yielding

 s1 0 ≤ s1 ≤ b

s =  s2 + b 0 ≤ s2 ≤ h

 s3 + b + h 0 ≤ s3 ≤ b

Equations for the stress analysis y


Geometry
As q1
h s1 s1
y 1 ( s 1 ) = --- ---- 0 ≤ s1 ≤ b C
2b s2 x
h
h q2 q3
y 2 ( s 2 ) = --- – s 2 0 ≤ s2 ≤ h contour origin s
2 s3
As
h h s3
y 3 ( s 3 ) = – --- + --- ---- 0 ≤ s3 ≤ b
2 2b c
( a)

b h b
h 2 h 2
y 12 ( s 1 )t 1 ds 1 + y 22 ( s 2 )t 2 ds 2 + y 32 ( s 3 )t 1 ds 3 +  --- A s +  – --- A s
I xx =
∫ ∫ ∫  2  2
( b)
0 0 0

Axial normal stress due to bending

M x(z)
σ z ( z, s ) = --------------- y ( s ) ( c)
I xx

Shear stress tangent to the contour

q ( z, s )
τ zs ( z, s ) = --------------- q ( z, s ) = q b ( z, s ) + q t ( z ) ( d)
t(s)
Shear flow due to torque

T 1
q t = -------- ¿ = enclosed area of the cell = --- hc ( e)
2¿ 2
Shear flow due to transverse the shear force

s1
V y(z)
I xx ∫
q b1 ( z, s 1 ) = q 0 – ------------- y 1 ( s 1 )t 1 ds 1 ( f)
0

s2
V y(z) h V y(z)
q b2 ( z, s 2 ) = q b1 ( b ) – -------------  --- A s – ------------- y 2 ( s 2 )t 2 ds 2
I xx  2 I xx ∫




0
upper stringer ( g)

258 Thin-Walled Structures


Maximum shear-stress criterion

s3
V y ( z )  h V y(z)
q b3 ( z, s 3 ) = q b2 ( h ) – ------------- – --- A s
I xx  2 I xx ∫
– ------------- y 3 ( s 3 )t 1 ds 3








0
lower stringer ( h)

To determine the shear flow at the contour origin, q 0 , due to the transverse shear force V y acting through the
shear center, we invoke the condition of no twist. That is,

dθ z q ( s ) ds q(s)
dz °∫
Gt ( s ) t(s) °∫
= ---------------- = 0 → ---------- ds = 0 ( i)

or

b h b
q b1 ( s 1 ) q b2 ( s 2 ) q b3 ( s 3 )
∫ t1 ∫
----------------- ds 1 + ----------------
t2 ∫
- ds 2 + ----------------
t1
- ds 3 = 0 ( j)
0 0 0

Equilibrium py
Vy
L
 dV
---------y = – p y ( z ) and V y ( L ) = 0 → V y ( z ) =
 dz  ∫ p ( z ) dz
y ( k)
Mx
z

L
 d----------
Mx
- = V y ( z ) and M x ( L ) = 0 → M x ( z ) = – V y ( z ) dz
 dz  ∫ ( l)
z

To locate the centroid (the point labeled C) y


As
A = area = 2bt 1 + ht 2 + 2 A s ( m)
s1
b b s2 C
h x
xc
Qy =
∫ ∫
x 1 ( s 1 )t 1 ds 1 + x 3 ( s 3 )t 1 ds 3 xc = Q y ⁄ A ( n)
s3
0 0
As
x1 ( s1 ) = c ( 1 – s1 ⁄ b ) x3 ( s3 ) = c ( s3 ⁄ b ) ( o) c
To locate the shear center (the point labeled S.C.)

From the solution of equations (f) to (j), we can find the shear flow in branch two due to the shear force V y
only. Now we use torque equivalence to locate the S.C. Sum torques about the apex to get


– c q b2 ( s 2 ) ds 2 = ( c – x sc )V y ( p)
0

Thin-Walled Structures 259


Criteria for Initial Yielding

Equation (p) yields a relation to find x sc , or the location of the shear center. The location of the shear center is
independent of the magnitude of the shear force.
y y
Vy
s2
S.C.
h
q b2 ( s 2 )
x ⇔ x
x sc

c – x sc
c c

Results The shear force and bending moment attain maximum magnitudes at the root. A the root
6
V y ( 0 ) = 2700 N and M x ( 0 ) = ( – 1.62 ×10 ) N-mm . The torque at the root section is
3
T ( 0 ) = 750 ×10 N-mm . Note that 1 N/mm2 equals 1 MPa
a) The axial normal stress is plotted as a function of the contour coordinate in graph below.

sz ,Nêmm2
150

100

50

s,mm
50 100 150 200 250 300 350
-50

-100

-150

b) The shear stress is plotted as a function of the contour coordinate in the following graph.

260 Thin-Walled Structures


Maximum shear-stress criterion

tzs ,Nêmm2
80
70
60
50
40
30
20
10
s,mm
50 100 150 200 250 300 350

EXAMPLE 8.6 Minimum weight design of the beam in Example 8.5 subject to a constraint on initial
yielding

Consider the design for minimum weight of the aluminum alloy beam in Example 8.5, which is shown in Fig.
8.15. The design is constrained by material yielding with the factor of safety specified as 1.5. Use the material
data for 2024-T4 aluminium as listed in the Example 8.5 problem statement, Mises yield criterion, and take the
value of the local acceleration due to gravity as 9.81 m/s2. The specified dimensions of the beam are
L = 1800 mm , h = 100 mm , and b = 130 mm . Take the value of the applied loads as p 0 = 3 N/mm and
3
T = – 250 ×10 N-mm . (Note the value of the torque is changed with respect to its value in Example 8.5.)

The objective is to minimize the weight subject to no yielding given the loads p 0 and T . That is, what are
the thicknesses t 1 and t 2 and the stringers’ cross-sectional area A s for minimum weight? Parameters t 1 , t 2 and
A s are called design variables. This is a problem in constrained optimization, which is stated mathematically as

minimize W ( t 1, t 2, A s )
such that g i ( t 1, t 2, A s ) > 0

where W ( t 1, t 2, A s ) is the objective function, or weight in this problem, and g i ( t 1, t 2, A s ) are constraint func-
tions. For design against yielding the constraint functions are defined as
σ all – ( σ M )
g i = ----------------------------i
σ all

Thin-Walled Structures 261


Criteria for Initial Yielding

point 2, s 1 = b where the allowable stress, σ all , is defined as

point 3, s 2 = 0 point 1, s 1 = 0 σ all = σ yield ⁄ F.S. and the Mises stress is σ M . These par-
y ticular constraint functions are called static margins, and
point 4, s 2 = h ⁄ 2 x positive values indicate the degree of safety against exceed-
C
ing the allowable stress. Due to symmetry about the x-axis
you only need to calculate the margin of safety for yielding
at four points in the cross section at the root as indicated in
Fig. 8.16 Critical points for yield evaluation Fig. 8.16. That is, compute the margins of safety at the
in the cross section at the root points labeled 1, 2, 3, and 4 in Fig. 8.16. In addition, the
shear force and bending moment attain their largest magni-
tude simultaneously at z = 0, so the four constraints are evaluated at z = 0.

The intent of this exercise is to study the influence of the stringers on the design for minimum weight. For
each stringer area given in the table below, determine the values of thicknesses t1 and t2 for minimum weight.
List these values along with the weight, and the four margins of safety in the table.

Influence of the stringer area on the minimum weight designs


Stringer Beam Thicknesses in mm Margins of safety, dimensionless
area As in weight in
mm2 N t1 t2 point 1 point 2 point 2 point 4

50
60
70
80
90
100

To calibrate the computations, the beam weight is 57.87 N and the margins of safety at points 1 to 4 are
50.3716, 1.92285, 1.92539, and 8.56734, respectively, for the design variable values of t1 = 2.84 mm, t2 = 3.42
mm, and As = 45 mm2.

262 Thin-Walled Structures


Maximum shear-stress criterion

Results s

0.8

t 2 , mm
Minimum static margin = 0
0.7

constant weight lines


0.6
feasible designs

0.5 least weight design

18 N
0.4

infeasible designs

0.3 16 N
14 N

t 1 , mm
0.2 0.3 0.4 0.5

Design plane for A s = 100 mm 2

As Wt. t1 t2 M1 M2 M3 M4
50 14.3955 0.437253 0.774719 3.2 9.4 µ 10 -10 0.0029 1.1
60 13.5688 0.342651 0.653471 2.1 -1.7 µ 10-12 0.0008 0.82
70 13.3053 0.279341 0.564789 1.4 0.001 -5.8 µ 10 -8 0.57
80 13.4942 0.239829 0.505728 0.97 0.0024 -2.4 µ 10 -11 0.41
90 13.9704 0.214902 0.466853 0.72 0.0034 3.1 µ 10 -12 0.3
100 14.6189 0.198474 0.440723 0.57 0.0039 -1. µ 10 -10 0.23

Thin-Walled Structures 263


Criteria for Initial Yielding

8.8 References
Ashby, M.F., 1992, Materials Selection in Mechanical Design, Butterworth-Heinemann, Ltd., Oxford.

Dowling, N.E., 1993, Mechanical Behavior of Materials, Prentice-Hall, Inc., Englewood Cliffs, New Jersey.

264 Thin-Walled Structures


CHAPTER 9 Buckling

Buckling of a structure means


• failure due to excessive displacements (loss of structural stiffness), and/or
• loss of stability of an equilibrium configuration of the structure

Stability of equilibrium means that the response of the structure due to a small disturbance from its equilib-
rium configuration remains small; the smaller the disturbance the smaller the resulting magnitude of the displace-
ment in the response. If a small disturbance causes large displacement, perhaps even theoretically infinite, then
the equilibrium state is unstable. Practical structures are stable at no load. Now consider increasing the load
slowly. We are interested in the value of the load, called the critical load, at which buckling occurs. That is, we
are interested in when a a sequence of equilibrium stable states as a function of the load, one state for each value
of the load, ceases to be stable.

If buckling occurs before the elastic limit of the material, which is roughly the yield stress of the material,
then it is called elastic buckling. If buckling occurs beyond the elastic limit, it is called inelastic buckling, or plas-
tic buckling if the material exhibits plasticity during buckling (mainly metals). Most thin-walled structural com-
ponents buckle in compression below the elastic limit. Therefore, buckling determines the limit state in
compression rather than material yielding. In fact, about 50% of an airplane structure is designed based on buck-
ling constraints.

9.1 One-degree of freedom model


To illustrate the physical nature of buckling as a stability problem and failure by excessive displacements, it is
instructive to analyze the response of a simple structural model to a compressive force. This model is shown in
Fig. 9.1 and has one coordinate θ, – π < θ < π , to describe the configuration of the model under the deadweight
load P. The model consists of a rigid rod of length l, connected by smooth hinge to a rigid base. The rod can
rotate about the hinge but it is restrained by a linear elastic torsional spring of stiffness K (dimensional units of F-
L/ radian). The spring is unstretched at θ = 0. Neglect the weight of the rod with respect to the applied load P.

Thin-Walled Structures 251


Buckling

P P

θ θ
l l
l
FBD
K Kθ
Ox
initial deflected Oy

Fig. 9.1 One degree of freedom structural model

From the free body diagram of the rod shown in Fig. 9.1, the equation of motion for rotation about the fixed
hinge is
2

Pl sin θ – Kθ = I 0 θ = θ(t ) t>0 (9.1)
dt2
where I0 is the moment of inertia of the rod about the fixed point and t is time.

9.1.1 Static equilibrium

Consider equilibrium states under the static, downward load P which are characterized by the angle θ being
independent of time t. Hence, the inertia term in eq. (9.1) vanishes and we have
Pl sin θ – Kθ = 0 θ <π (9.2)

The solutions to eq. (9.2) are


P 1 : θ = 0 for any P (9.3)

and

K θ
P 2 : P =  ---- ----------- (9.4)
 l  sin θ

P Recall from the calculus using l’Hôpital’s rule that the limit of the
---------- indeterminate form θ ⁄ ( sin θ ) as θ → 0 is one. The two equilib-
K⁄l
rium paths are plotted in the load-deflection diagram shown in Fig.
2
9.2. Equilibrium path P1 coincides with the load axis in the plot and
P1 is called the primary equilibrium path, or the trivial equilibrium
P2 P2 path. Equilibrium path P2 is called the secondary path and we note
1 it is symmetric about θ = 0. The two equilibrium paths intersect at
(θ,P) = (0,K/l). This intersection of the two paths is called a bifurca-
P1
tion point. At no load the rod is vertical and this corresponds to the
origin in the load-deflection diagram. As the load P is slowly
–π 0 π θ increased from zero the rod remains vertical (θ = 0), and at P = K/l
Fig. 9.2 Equilibrium paths. adjacent equilibrium states exists on the secondary path. The exist-
ence of adjacent equilibrium states in the vicinity of the primary

252 Thin-Walled Structures


One-degree of freedom model

equilibrium path has been noted by investigators of structural stability as the onset of buckling. Hence, buckling
is characterized by the bifurcation point on the load-deflection diagram. For this reason, the term bifurcation
buckling is used to describe this condition. As we will show later, the rod will not remain vertical for loads P >
K/l if there are infinitesimal disturbances present (there always are), but will rotate either to the left or right
depending on type of infinitesimal disturbance. We note that the magnitude of the angle θ becomes large as the
load is increased from K/l on the secondary path. The load at the bifurcation point is called the critical load and is
denoted as P cr . Thus,

P cr = K ⁄ l (9.5)

Small θ analysis (9.6)

Consider the small angles of rotation such that sin θ ≈ θ for θ measured in radians. Equilibrium eq. (9.2)
becomes
Plθ – Kθ = 0 (9.7)

And the solutions of this equation are


P 1 ′: θ = 0 for any P , and (9.8)

P 2 ′: P = K ⁄ l for any small θ (9.9)

These solutions are shown in the load-deflection plane in Fig. 9.3. The equilibrium P
----------
path P 1 ′ coincides with path P 1 , but path P 2′ is not a good approximation to path K⁄l
P2 ′
P 2 unless θ is very small. However, the bifurcation point is the same as obtained in
1
the large θ-analysis. Hence, the critical load from the small θ-analysis is the same as
obtained in eq. (9.5) from the large θ-analysis. P1 ′

0 θ
9.1.2 Stability analysis Fig. 9.3 Small θ analysis

Let the rotation angle


θ ( t ) = θ0 + ϕ ( t ) (9.10)

where θ 0 is independent of time and satisfies the equilibrium eq. (9.2); i.e.,

Pl sin θ 0 – K θ 0 = 0 (9.11)

Consider the additional rotation angle ϕ ( t ) to be small in magnitude but a function P


of time. Thus, we are considering small oscillations about an equilibrium state
θ(t )
( P, θ 0 ) as shown in Fig. 9.4. Substitute eq. (9.10) for q in the equation of motion, ϕ(t )
eq. (9.1), to get
θ0
I 0 ϕ̇˙ + K ( θ 0 + ϕ ) – Pl sin ( θ 0 + ϕ ) = 0 (9.12)
l
2

where the dots denote derivatives with respect to time; e.g., ϕ̇˙ = . Using the
dt2
trigonometric identity for the sine of the sum of two angles and performing some Fig. 9.4 Rotations in
minor rearrangements the last equation becomes the stability analysis

Thin-Walled Structures 253


Buckling

I 0 ϕ̇˙ + K θ 0 + Kϕ – Pl [ sin θ 0 cos ϕ + cos θ 0 sin ϕ ] = 0 (9.13)

Now expand the trigonometric functions of angle ϕ in a Taylor Series about ϕ = 0 to get

1 1
I 0 ϕ̇˙ + K θ 0 + Kϕ – Pl sin θ 0 1 – --- ϕ 2 + O ( ϕ 4 ) – Pl cos θ 0 ϕ – --- ϕ 3 + O ( ϕ 5 ) = 0 (9.14)
2 6

in which O ( ϕ n ) means terms of order ϕ n and higher. Arrange eq. (9.14) in powers of ϕ to get

Pl Pl
I 0 ϕ̇˙ + ( K θ 0 – Pl sin θ 0 ) + ( K – Pl cos θ 0 )ϕ +  ----- sin θ 0 ϕ 2 +  ----- cos θ 0 ϕ 3 + O ( ϕ 4 ) = 0
2  6 






= 0 (9.15)

Note that "coefficient" of the term ϕ 0 vanishes because of the equilibrium condition given by eq. (9.11).

For very small additional rotation angles ϕ ( t ) about the equilibrium configuration, eq. (9.15) is approxi-
mated by

I 0 ϕ̇˙ + ( K – Pl cos θ 0 )ϕ = 0 or ϕ̇˙ + ω 2 ϕ = 0 (9.16)

where

ω 2 = ( K – Pl cos θ 0 ) ⁄ I 0 (9.17)

The solution of the second order differential equation, eq. (9.16), for ω 2 > 0 is

ϕ ( t ) = A 1 sin ( ωt ) + A 2 cos ( ωt ) ω2 > 0 (9.18)

in which constants A 1 and A 2 are determined by initial conditions for ϕ ( 0 ) and ϕ̇ ( 0 ) . The solution given by
eq. (9.18) is a harmonic oscillation about the equilibrium configuration and ω is interpreted as the natural fre-
quency in radians per second. Initial conditions ϕ ( 0 ) and ϕ̇ ( 0 ) are considered to be very small but arbitrary to
simulate and arbitrary small initial disturbance. The smaller the initial disturbance, the smaller the maximum
amplitude of the oscillation in ϕ . Thus, ω 2 > 0 is a condition for a stable equilibrium configuration with respect
to infinitesimal disturbances.

The solution of the second order differential equation, eq. (9.16), for ω 2 < 0 is

ϕ ( t ) = A 1 e ωt + A 2 e –ωt ω 2 = – ω 2 = – ( K – Pl cos θ 0 ) ⁄ I 0 (9.19)

For arbitrary initial conditions, the term with the positive exponent in the dominates the solution. This corre-
sponds to large values of the ϕ no matter how small the initial disturbance. Hence, ω 2 < 0 is a condition of
unstable equilibrium configuration with respect to infinitesimal disturbances. The dynamic criterion for structural
stability is

Dynamic criterion for stability of an equilibrium state


The equilibrium state is stable if ω2 > 0
The equilibrium state is critical if ω2 = 0
The equilibrium state is unstable if ω2 < 0

254 Thin-Walled Structures


Perfect Columns

On the primary equilibrium path P 1 given by eq. (9.3), we have from eq. (9.17) that

ω 2 = ( K – Pl ) ⁄ I 0 on P 1 (9.20)

Thus, equilibrium configurations are stable if P < K ⁄ l , critical if P = K ⁄ l , and unstable if P > K ⁄ l . The pri-
mary equilibrium path ceases to be stable at P = P cr , and P cr is the buckling load.

9.2 Perfect Columns


Consider a perfectly straight, uniform column of length L and cross-sectional area A subjected to a centric end
load P as shown in Fig. 9.5. The column is long relative to its largest cross-sectional dimension, and the column
consists of a homogeneous, linear elastic material whose modulus of elasticity is denoted by E. The equilibrium
configuration of this column is pure compression. Let N(z) denote the internal axial force. From equilibrium of

y,v

P
z,w

Fig. 9.5 A straight column subjected to a centric, compressive axial force.

a differential element shown below we have dN ⁄ dz = 0 , and from Hooke’s law N = EAε z where ε z is axial
normal strain. The axial normal strain is related to the axial displacement by ε z = dw ⁄ dz as is shown in Fig.
9.6. The boundary conditions for the column are w ( 0 ) = 0 and N ( L ) = – P . Hence, the internal axial load is

dw
y dz + w ( z + dx ) – w ( z ) ≈  1 +  dz
 dz 

z N N + dN
z
dz w ( z + dz )
w(z)

Fig. 9.6 An element of the column in the pre-buckling equilibrium state

uniform and compressive along the length of the column and equal in magnitude to the applied load P. Summa-
rizing the equilibrium solution we have
Pz
N = –P w ( z ) = – ------- v(z) = 0 0≤z≤L (9.21)
EA
Thin-Walled Structures 255
Buckling

The end shortening under the compressive load is – w ( L ) , and this is plotted on the load-end shortening plot
shown in Fig. 9.7. The equilibrium configuration of pure compression of the perfect column is called the trivial

Fig. 9.7 Load-end shortening EA


plot in pre-buckling -------
L
1

0 –w ( L )

equilibrium state. Note that in the trivial equilibrium state the lateral displacement of the column, v ( z ) is zero
for all values of the compressive load P. Researchers in structural stability recognized from experience that buck-
ling of the column is associated with the appearance of second, non-trivial, equilibrium configuration at the buck-
ling load. This observation is the basis of the adjacent equilibrium method of stability analysis. The question
characterizing the method of adjacent equilibrium is
What is the value of the load for which the perfect system admits non-trivial equilibrium configura-
tions?

To answer this question we consider equilibrium of a slightly deflected element of the column at the same
value of the external load P. The free body diagram of this element is shown in Fig. 9.8. The displacement due to

 1 + dw
------- dz
 dz  P V y1 + dV y1
θ x1 dv 1
– --------dz
y dz M x1 P M x1 + d M x1
v1 dv 1 V y1
v 1 + -------- ( dz )
dz
z

Fig. 9.8 Free body diagram of an element of the column in the buckled state.

buckling is denoted by v1(z), and all quantities due to buckling are labeled with the subscript 1. Vertical force
equilibrium gives
dV y1
------------ = 0 (9.22)
dz
where V y1 is the y-direction shear force due to buckling. Moment equilibrium about the x-axis gives

d M x1  dw dv
- – 1 + ------- V y1 +  – -------1- P = 0
------------
dz  dz   dz 

where M x1 is the bending moment due to buckling. In general, the axial strain in the equilibrium configuration
dw ⁄ dz is very small in magnitude compared to unity and is then neglected with respect to unity in this equation.

256 Thin-Walled Structures


Perfect Columns

Then, the moment equation becomes

d M x1 dv
- – V y1 +  – -------1- P = 0
------------ (9.23)
dz  dz 

Neglecting dw ⁄ dz with respect to unity, and assuming the rotation θ x1 due to buckling is small, implies that the
rotation is given by θ x1 ( z ) = – ( dv 1 ⁄ dz ) . A very important term in eq. (9.23) is the contribution of the axial
compressive load P through the buckling displacement v1 to moment equilibrium. This term that couples the
axial compressive load in equilibrium state to the buckling displacement arises only because we took equilibrium
on the slightly deflected column element. Hooke’s law for the bending moment is
2
 dv
M x1 = EI xx  – 21 (9.24)
 dz 
where Ixx is the second area moment of the cross section about the x-axis. (We assume the cross section is sym-
metric about either the x-axis or y-axis, or that these axes are principal axes of the cross section if no symmetry is
present. In design we use the minimum second are moment of the cross section). If we take the derivative of eq.
(9.23), use eq. (9.22), and then substitute eq. (9.24) for the bending moment due to buckling we get
2 2
2
d  d v 1  d v 1
 – EI xx 2  +  – 2  P = 0 (9.25)
dz2  dz   dz 

This is the governing fourth order, ordinary differential equation for the buckling displacement v1(z). For conve-
nience in writing, we will drop the subscripts on the second area moment in the following developments. Also
note that in the buckling theory the vertical shear force is determined in terms of the buckling displacement v1 by
substituting eq. (9.24) into eq. (9.23) to get
2
d  d v 1 dv 1
V y1 = –  EI 2  – -------- P (9.26)
d z  d z  dz

To determine the buckling displacement v1 we need boundary conditions at z = 0 and z =L in addition to the
o.d.e. given by eq. (9.25). There are four standard boundary conditions. These are

A. Pinned-pinned
v1 ( 0 ) = 0 v1 ( L ) = 0 P
M x1 ( 0 ) = 0 M x1 ( L ) = 0 z
L

B. Clamped-free
P
v1 ( 0 ) = 0 M x1 ( L ) = 0
θ x1 ( 0 ) = 0 V y1 ( L ) = 0 z
L

Thin-Walled Structures 257


Buckling

C. Clamped-clamped
v1 ( 0 ) = 0 v1 ( L ) = 0 P
θ x1 ( 0 ) = 0 θ x1 ( L ) = 0 z
L

D. Clamped-pinned
v1 ( 0 ) = 0 v1 ( L ) = 0 P
θ x1 ( 0 ) = 0 M x1 ( L ) = 0 z
L

One solution to the o.d.e., eq. (9.25), subject to boundary conditions A-D is v 1 ( z ) = 0 for all values of the
load P. This is the trivial solution. Are there any other solutions? Can we get them? The answer is yes to both
questions if EI = constant. For EI = constant, eq. (9.25) becomes
4 2
d v1 d v1
EI 4 + P 2 = 0
dz dz
or
4 2
d v1 d v1
4
+ k2 2 = 0 0<z<L (9.27)
dz dz
where
P
k 2 = ------ (9.28)
EI
The general solution of eq. (9.27) for k 2 > 0 is

v 1 ( z ) = A 1 sin ( kz ) + A 2 cos ( kz ) + A 3 z + A 4 (9.29)

where A1, A2, A3, and A4 are arbitrary constants to be determined by boundary conditions.

EXAMPLE 9.1 Critical load for clamped-free boundary conditions (B)

Consider the clamped-free boundary conditions; i.e. b.c.’s (B) above. Determine the buckling load P cr for
which the perfect column ceases to be stable.

Solution The boundary conditions in this case become

v1 ( 0 ) = 0 v1 ′ ( 0 ) = 0 EI v 1 ″ ( L ) = 0 [ v 1 ′′′ + k 2 v 1 ′ ] z = L = 0
where the primes denote derivatives with respect to z. Taking derivatives of eq. (9.29) we have

258 Thin-Walled Structures


Perfect Columns

v 1 = A 1 sin ( kz ) + A 2 cos ( kz ) + A 3 z + A 4
v 1 ′ = A 1 k cos ( kz ) – A 2 k sin ( kz ) + A 3
v 1 ″ = – A 1 k 2 sin ( kz ) – A 2 k 2 cos ( kz )
v 1 ′′′ = – A 1 k 3 cos ( kz ) + A 2 k 3 sin ( kz )

Substitute these solutions into the four boundary conditions to get

0 1 0 1 A1
k 0 1 0 A2
= 0 (9.30)
– k 2 sin ( kL ) – k 2 cos ( kL ) 0 0 A3
0 0 k2 0 A4
A non-trivial solution for A1 to A4 requires the determinate of coefficients to vanish

0 1 0 1
k 0 1 0
det = 0 (9.31)
2 2
– k sin ( kL ) – k cos ( kL ) 0 0
0 0 k2 0
After expanding this determinate we get

– k 5 cos ( kL ) = 0 (9.32)

which gives “n-values”, n = 1, 2, 3,..., of kL; or

( 2n – 1 ) π P
k n = -------------------- --- = -----n- n = 1, 2, 3, … (9.33)
L 2 EI

For k n L = ( 2n – 1 ) ( π ⁄ 2 ) , the fourth row of matrix eq. (9.30) gives A 3 = 0 ; using this result in the sec-
ond row of matrix eq. (9.30) gives A 1 = 0 ; the first row of matrix eq. (9.30) gives A 2 = – A 4 . Note that the
third row of matrix eq. (9.30) is identically satisfied. So eq. (9.33) implies

π 2 EI
P n = ( 2n – 1 ) --- -----2- n = 1, 2, 3, … , (9.34)
2 L
where Pn are the buckling loads. Equation (9.32) is called the characteristic equation, and the roots of this equa-
tion determine the buckling loads. For each value of n we have an associated buckling mode (A1 = A3 = 0, A2 =
– A4)

v 1n ( z ) = A 4 [ 1 – cos ( k n z ) ] (9.35)

where coefficient A4 is arbitrary. The first three buckling modes are shown in Fig. 9.9. Note that the amplitude
of the buckling mode is not known. However, we can plot its shape. The critical load, denoted by P cr , is the low-
est buckling load. That is

π 2 EI
P cr = P 1 = ----- -----2-
4L

Thin-Walled Structures 259


Buckling

Remember that in design we use the minimum EI for the cross section.

v1/A4
π 2 EI
1 P 1 = ----- -----2-
0.8 4L
n=1 0.6
0.4
0.2
z/L
0.2 0.4 0.6 0.8 1
v1/A4
2
1.5 9π 2 EI
n=2 1 P 2 = --------- -----2-
0.5 4 L
z/L
0.2 0.4 0.6 0.8 1
v1/A4
2
n=3 1.5 25π 2 EI
1 P 3 = ------------ -----2-
4 L
0.5
z/L
0.2 0.4 0.6 0.8 1

Fig. 9.9 First three buckling modes for the clamped-free column.

260 Thin-Walled Structures


Imperfect columns

The critical loads for boundary conditions A through D and for EI = constant are given in Fig. 9.10 .

EI
A. Pinned-pinned P cr = π 2 -----2-
L
L

π 2 EI
P cr = ----- -----2-
4L
B. Clamped-free

EI
C. Clamped-clamped P cr = 4π 2 -----2-
L
L

EI EI
D. Clamped-pinned P cr = 2.046π 2 -----2- = ( 4.49 ) 2 -----2-
L L
L

Fig. 9.10 Buckling loads for the standard boundary conditions A to D.

9.3 Imperfect columns

9.3.1 Eccentric load


Consider a uniform (EI = constant), pinned-pinned column subjected to an eccentric axial load P. Let e denote
the perpendicular distance between the line of action of load P and the z-axis. This situation is statically equiva-
lent to a centric axial load P and a moment of magnitude eP applied to the ends of the column. See Fig. 9.11.
Hence, the eccentric axial load will simultaneously subject the column to compression and bending in the equi-
librium state. The analysis for the equilibrium response of the column in compression (axial force N and axial
displacement w(z)) is identical to the case of the perfect column, since the z-axis passes through the centroid of
each cross section (decoupling the axial compression from bending in the material law). The equilibrium
response of the column in bending, which includes the influence of axial compression on bending, is determined
by the same analysis that led to eqs. (9.27) to (9.29), except that we drop the subscript “1” on the lateral displace-
ment, since in the eccentric load case the lateral displacement refers to an equilibrium state and not to a buckling
mode. The differential equation and boundary conditions for equilibrium displacement v(z) are

Thin-Walled Structures 261


Buckling

4 2
dv dv
+ k2 2 = 0 0<z<L (9.36)
dz4 dz
2 2
dv dv
v(0) = 0 – EI ( 0 ) = eP v( L) = 0 – EI ( L ) = eP (9.37)
dz2 dz2
where k2 is as given in (9.28). Note that the boundary conditions, eqs. (9.37), are inhomogeneous. Thus, eqs.
(9.36) and (9.37) do not have the trivial solution v(z) = 0 for all values of the load P. Using the general solution

y,v

EI P P
z,w

e eP eP
e

P L P

Fig. 9.11 Column subjected to an eccentric axial load

form given by (9.29) for the solution of differential equation (9.36), subject to the boundary conditions (9.37), we
find

kL
v ( z ) = e – 1 + cos ( kz ) + tan  ------ sin kz (9.38)
 2

kL 1 – cos ( kL )
By a trigonometric identity, tan  ------ = ----------------------------- . Let δ denote the midspan displacement; i.e., δ = v(L/2).
 2 sin ( kL )
From eq. (9.38) we get

kL kL kL 1
δ = e – 1 + cos  ------ + tan  ------ sin  ------ = e --------------------- – 1
 2  2  2 kL
cos  ------
 2

The factor kL/2 can be written in terms of the eccentric axial load P and the critical load Pcr for the pinned-
pinned uniform column subjected to centric load as

kL 1 P  π 2 EI  π P
------ =  --- ------ L  ------------- = --- ------- .
2  2 EI   L 2 P cr  2 P cr

Thus, the center deflection of the eccentrically loaded column becomes

262 Thin-Walled Structures


Imperfect columns

1
δ = e ------------------------------ – 1 (9.39)
π P
cos  --- -------
 2 P cr

The load-displacement response is shown in Fig. 9.12. Note that δ → ∞ as P → P cr for e ≠ 0 . That is, no

P/Pcr
e=0
1

EI
increasing e P cr = π 2 -----2-
L

0 δ

Fig. 9.12 Load-deflection curves for an eccentrically loaded column

matter the magnitude of the eccentricity as P → P cr the value of the center deflection δ gets very large.

9.3.2 Geometric imperfection


Consider a uniform, pinned-pinned column that is slightly crooked under no load. The initial shape under no load
is described by the function v 0 ( z ) . The column is subjected to a centric, axial compressive load P. The lateral
displacement of the column is denoted by v ( z ) , so that v ( z ) = v 0 ( z ) when P = 0. Also, the bending moment
in the column is zero under no load. Thus, we write the material law for bending as

2 2
d v d v 
M x = – EI  2 – 20 (9.40)
dz dz 
Vertical force equilibrium of the deflected column leads to a differential equation similar to eq. (9.22) except that
we drop the “1” subscript since it is the configuration of the column is one of equilibrium and not a buckling
mode. Similarly, moment equilibrium of the imperfect column leads to a differential equation similar to eq.
(9.23) with the subscript “1” dropped. Combining these differential equations of vertical force equilibrium and
moment equilibrium via elimination of the vertical shear force gives
2 2
d M x  d v
+  – 2 P = 0 (9.41)
dz2  dz 
The pinned-pinned boundary conditions are
v(0) = 0 M x(0) = 0 v( L) = 0 M x(L) = 0 (9.42)

Thin-Walled Structures 263


Buckling

πz
Consider the imperfection shape v 0 ( z ) = a 1 sin  ----- , where a 1 denotes the amplitude at midspan of the
 L
slightly crooked column. Substitute eq. (9.40) into eq. (9.41) to eliminate the moment to get
4 2 2
dv dv d v0 π 2 πz
+ k2 2 = = a 1  --- sin  ----- (9.43)
dz 4 dz dz 2  L  L

where k 2 is given by eq. (9.28). The boundary conditions, eqs. (9.42), for this imperfection shape lead to
2
dv
v = = 0 at z = 0 and z = L (9.44)
dz2
The solution of the differential equation (9.43) subject to boundary conditions (9.44) is
a1 πz
v ( z ) = ----------------------2- sin  ----- 0≤z≤L (9.45)
kL  L
1 –  ------
 π

It is convenient to measure the deflection of the imperfect column under load with respect to its original unloaded
state. That is, let δ define the additional displacement at midspan by δ = v ( L ⁄ 2 ) – v 0 ( L ⁄ 2 ) . Hence,

P
 -------
 P cr
δ = a 1 ---------------------- (9.46)
P
1 –  -------
 P cr

The load-displacement response is sketched in Fig. 9.13.


Note that δ → ∞ as P → P cr for a 1 ≠ 0 . That is, for a non- P ⁄ P cr EI
P cr = π 2 -----2-
zero value of the imperfection amplitude, the displacement L
gets very large as the axial force approaches the buckling 1.0
load of the perfect column. Also, the imperfect column
deflects in the direction of imperfection; e.g., if a 1 > 0 , then
δ>0. increasing a 1

Collectively the eccentric load and the geometric shape


imperfection are called imperfections. All real columns are 0 δ
imperfect. Even for a well manufactured column whose geo-
Fig. 9.13 Load-deflection response plots
metric imperfections are small and with the load eccentricity for geometrically imperfect columns
small, the displacements become excessive as the axial com-
pressive force P approaches the critical load P cr of the per-
fect column. Hence, the critical load determined from the
analysis of the perfect column is meaningful in practice.

264 Thin-Walled Structures


Column Design Curve

9.4 Column Design Curve


2
Consider the pinned-pinned uniform column whose critical load is given by P cr = π 2 ( EI ⁄ L ) . Let A denote the
cross-sectional area of the column. At the onset of buckling the critical stress is defined as

σ cr = P cr ⁄ A = ( π 2 EI ) ⁄ ( AL 2 ) (9.47)

We write the second area moment as I = r 2 A , where r denotes the minimum radius of
h
gyration of the cross section. For the rectangular section shown in the adjacent sketch,
I min = ( bh 3 ) ⁄ 12 and A = bh , so that r = h ⁄ 12 , where 0 < h < b . Thus, the critical b
stress becomes
E
σ cr = π 2 ----------------2- (9.48)
(L ⁄ r)
and L ⁄ r is called the slenderness ratio. The slenderness ratio is the column length divided by a cross-sectional
dimension significant to bending.

For any set of boundary conditions we define the effective length KL by the formula

EI
P cr = π 2 --------------2 (9.49)
( KL )
The effective lengths for the four standard boundary conditions are as follows:

A pinned-pinned EI EI KL = L K = 1
P cr = π 2 -----2- = π 2 --------------2
L ( KL )
B clamped-free π 2 EI EI KL = 2L K = 2
P cr = ----- -----2- = π 2 --------------2
4L ( KL )
C clamped-clamped EI EI KL = L ⁄ 2 K = 1⁄2
P cr = 4π 2 -----2- = π 2 --------------2
L ( KL )
D clamped-pinned EI EI KL = 0.699L K = 0.7
P cr = 20.2 -----2- = π 2 --------------2
L ( KL )
The definition of effective length uses case A boundary conditions as a reference. the concept of effective length
accounts for boundary conditions other than simple support, or pinned-pinned end conditions.

The column curve is a plot of the critical stress versus the effective slenderness ratio; i.e., σ cr versus KL ⁄ r .
For elastic column buckling under all boundary conditions

π2E
σ cr = ---------------2 (9.50)
 KL
-------
 r 

which is a hyperbola that depends only on the modulus of elasticity E of the material. This equation governing
elastic buckling is called the Euler curve, and columns that buckle in the elastic range are called long columns.
See Fig. 9.14

Thin-Walled Structures 265


Buckling

σ cr

σp
Euler curve, depends only on E

Fig. 9.14 Column curve for elastic


buckling

long columns
KL
0 -------
r
KL E
------- = π ------
r σp

9.4.1 Inelastic buckling


The column curve equation, eq. (9.50), is valid up to the proportional limit of the material, denoted by σ p . The
proportional limit is defined as the stress where the compressive stress-strain curve of the material deviates from
a straight line. If the stress at the onset of buckling is greater than the proportional limit, then the column is said
to be of intermediate length, and the Euler formula, eq. (9.50), cannot be used. The proportional limit is difficult
to measure from test data because its definition is based on the deviation from linearity. In particular, the com-
pressive stress-strain curves for aluminum alloys typically used in aircraft construction do not exhibit a very pro-
nounced linear range. For aluminum alloys a material law developed by Ramberg and Osgood (1943) is often
used to describe the nonlinear compressive stress-strain curve. The Ramberg-Osgood equation is a three parame-
ter fit to the compressive stress-strain curves of aluminum alloys. From the experimental compressive stress-
strain curve we measure the slope near the origin, which is the modulus of elasticity E, the stress σ 0.85 where the
secant line drawn from the origin with slope 0.85 E intersects the stress-strain curve, and the stress σ 0.7 where a
second secant line drawn from the origin with slope 0.7E intersects the stress-strain curve. These data are
depicted in Fig. 9.15. Note that the compressive normal strain corresponding to the stress σ 0.7 is usually about
σ

σ 0.7 experimental compressive


stress-strain curve
σ 0.85
Fig. 9.15 Data used to fit the compression 0.7E
stress-strain curve of aluminum alloys. 1
0.85E
1
E
1 ε
0 ≈ 0.002
the 0.2% offset yield strain for the material. Hence, stress σ 0, 7 is close to the 0.2% offset yield stress of the alu-
minum alloy. The Ramberg-Osgood equation is

σ 3 σ n–1
ε = --- 1 + ---  --------- (9.51)
E 7  σ 0.7

where the shape parameter n is given by


266 Thin-Walled Structures
Column Design Curve

17
ln  ------
 7
n = 1 + ---------------------- (9.52)
σ 0.7
ln  -----------
 σ 0.85

We can re-write eq. (9.51) as

Eε σ 3 σ n
--------- = --------- + ---  --------- (9.53)
σ 0.7 σ 0.7 7  σ 0.7

and plot σ ⁄ σ 0.7 versus ( Eε ) ⁄ σ 0.7 for various values of the shape parameter n., and this plot is shown in Fig.

n = 2
1.2 n = 5
n = 10
1 50 n = 20
0.8 5 n = 50
σ
---------
σ 0.7 0.6 n = 2

0.4

0.2

0.5 1 1.5 2

---------
σ 0.7
Fig. 9.16 A normalized plot of the Ramberg-Osgood material law for various values of
the shape parameter n.

9.16. Some approximate values for common aluminum alloys are given in the table below.

AL E in106psi σ 0.7 in 103psi n

2014-T6 10.6 60 20

2024-T4 10.6 48 10

6061-T6 10.0 40 30

7075-T6 10.4 73 20

From the Ramberg-Osgood equation, eq. (9.51), we can determine the local slope of the compressive stress-
strain curve as a function of the stress. This slope of the compressive stress-strain curve is called the tangent

modulus; i.e., ------ = E t where Et is the tangent modulus. Differentiating eq. (9.51) we get

Thin-Walled Structures 267


Buckling

dσ 3 nσ n – 1 dε 1 1 3n σ n–1
dε = ------ + --- ---------------
-dσ ------ = ----- = --- + --- ---  --------- (9.54)
E 7 Eσ 0.7 n–1 dσ Et E 7 E  σ 0.7

or
E
E t = --------------------------------------- (9.55)
3 σ n–1
1 + --- n  ---------
7 σ 0.7

For intermediate length columns it has been demonstrated by extensive testing that the critical stress is rea-
sonably well predicted using the Euler curve, eq. (9.50), with the modulus of elasticity replaced by the tangent
modulus. This inelastic buckling analysis is called the tangent modulus theory. That is,
Et
σ cr = π 2 ---------------2 (9.56)
 KL
-------
 r 

Now substitute eq. (9.55) for the tangent modulus in this equation, noting that σ = σ cr , to get

π2 E
σ cr = ---------------2 ---------------------------------------
KL 3
 ------- 1 + --- n  -------- σ cr  n – 1
-
 r  7  σ 0.7

After division by σ 0.7 , this equation can be written as

σ cr 3  σ cr  n 1
- + --- n --------- = --------------------------------2
-------- (9.57)
σ 0.7 7  σ 0.7 ( KL ⁄ r )
------------------------
π E ⁄ σ 0.7
A plot of the column curve given by eq. (9.57) is shown in Fig. 9.17.

268 Thin-Walled Structures


Bending of thin plates

1.2

n = ∞
1 n = 50
n = 20
0.8
σ cr
--------
-
σ 0.7 0.6
n = 10
0.4 n = 5

0.2 n = 2

0.5 1 1.5 2 2.5 3 3.5


( KL ) ⁄ r
------------------------
π E ⁄ σ 0.7

Fig. 9.17 Column curves for a Ramberg-Osgood material law with different shape factors.

9.5 Bending of thin plates


Recall that bars and beams are structural elements characterized by having two orthogonal dimensions, say the
thickness and width, that are small compared to the third orthogonal dimension, the length. Thin plates, both flat
and curved, are common structural elements in flight vehicle structures, and they are characterized by one dimen-
sion being small, say the thickness, with respect to the other two orthogonal dimensions, say the width and
length. A flat plate with rectangular planform is shown in Fig. 9.18 referenced to cartesian axes x, y, and z,
where the x-direction is parallel to the length, the y-direction is parallel to the width, and the z-axis is parallel to
the thickness of the plate. We denote the length of the plate by a, the width by b, and the thickness by t. Trans-
σz z

τ zy
τ zx
τ xz τ yz
x y
t
σx τ xy τ σy
yx

3-D stress state


τ xy τ a
σx yx σy b a»t>0
b»t>0
primary stresses in plate theory

Fig. 9.18 Illustration of the nomenclature and primary stresses for a flat, rectangular plate

verse loads, or lateral loads, acting in the z-direction applied to the plate are primarily carried by the in-plane

Thin-Walled Structures 269


Buckling

stress components σx, σy, and τxy. Transverse shear stresses τxz and τyz are necessary for force equilibrium in the
z-direction under transverse loads, but are smaller in magnitude with respect to the in-plane stresses. In plate the-
ory, the transverse normal stress σz is small with respect to the in-plane normal stresses and, hence, is neglected.
In contrast a bar carries the transverse load primarily carried by the longitudinal normal stress σx, and the so-
called lateral stresses σy,σz, τyx, and τyz are assumed to be negligible. Thus, a plate resists transverse loads
through stress components σx, σy, and τxy while a beam resists transverse loads with only the longitudinal stress
σx.

Now consider the deformation, or strains, caused by the normal stresses. Hooke’s law for the normal stresses
and strains in a three-dimensional state of stress is
1
ε x = --- ( σ x – νσ y – νσ z )
E
1
ε y = --- ( – νσ x + σ y – νσ z ) (9.58)
E
1
ε z = --- ( – νσ x – νσ y + σ z )
E
where E is the modulus of elasticity and ν is Poisson’s ratio. For both the plate and beam, the thickness normal
stress σz is assumed negligible and is set to zero in Hooke’s law. We neglect the third of eqs. (9.58), so that these
equations reduce to
1
ε x = --- ( σ x – νσ y )
E
(9.59)
1
ε y = --- ( – νσ x + σ y )
E

Consider pure bending of the plate or beam subjected to moment M. We consider two cases. In the first case
the cross section is compact with dimension b nearly equal to thickness t, and in the second case with dimension
b is much larger that thickness t. In the first case the structure is a beam and the second case it is a plate. In pure

z ρ
A ---
ν

A
x ρ
M M Section A-A
z

Fig. 9.19 Pure bending of a beam in the x-z plane and the associated anticlastic curvature of its
cross section.

bending the neutral axis of the beam deforms into an arc of a circle with radius ρ , and the normal strain in the x-

270 Thin-Walled Structures


Bending of thin plates

z
direction is ε x = --- . Note that we assumed that the z-axis coincided with the neutral axis in the undeformed
ρ
beam. Hence, longitudinal line elements above the neutral axis, z > 0, are stretched, and line elements below the
neutral axis, z < 0, are compressed. In the case of a beam, the normal stress in the y-direction, σ y , is also very
small and is neglected with respect to the longitudinal normal stress σ x . That is, the beam carries the applied
bending moment by the longitudinal normal stress σ x . Since σ y = 0 , we get from Hooke’s law, eq. (9.59), that

σ x = Eε x
ν z
ε y = – νε x = – --- z = – --------- (9.60)
ρ ρ ---
 ν

Hence, the longitudinal normal stress is the modulus of elasticity times the longitudinal normal strain, and the
normal strain in the y-direction is just Poisson’s ratio times the longitudinal normal strain. The form of the last
expression for εy in eq. (9.60) shows that the line elements in the cross section parallel to the y-axis before defor-
mation bend into circular arcs. The transverse line element at z = 0 in the undeformed beam has a radius of curva-
ρ
ture of --- . This transverse curvature is called anticlastic curvature, and is illustrated in Fig. 9.19.
ν

Now consider pure bending of a plate under the same moment M, where now the dimension b is much larger
than thickness t. In this case experiments show that the transverse line elements remain straight over the central
section of the plate, so that the anticlastic curvature is suppressed. In this central section of the plate the trans-
verse normal stress σy is non-zero. However, the transverse normal stress must vanish at the free edges at y = 0
and y = b, so that anticlastic curvature develops only in narrow zones near the free edges to adjust to vanishing of
the normal stress σ y there. In the central portion of the plate, the associated normal strain is zero. The suppres-
sion of anticlastic curvature is characterized by the vanishing of the transverse normal strain εy. Hence from
Hooke’s law, eq. (9.59), for εy = 0 we get

E
σ x = -------------2-ε x σ y = νσ x (9.61)
1–ν
Since denominator in the expression for σx is positive but less than unity, the plate is stiffer than the beam owing
to the presence of the non-zero transverse normal stress σ y to help in carrying the applied moment. Compare
eqs. (9.60) and (9.61) for the normal stress σx. The quantity E/(1 - ν2) is an effective modulus of the plate.

Thin-Walled Structures 271


Buckling

9.6 Compression buckling of thin plates

272 Thin-Walled Structures


Compression buckling of thin plates

Thin-Walled Structures 273


Buckling

274 Thin-Walled Structures


Compression buckling of thin plates

Thin-Walled Structures 275

Potrebbero piacerti anche