Beams 3 D

Elastic Beams in Three Dimensions
Lars Andersen and Sren R.K. Nielsen
ISSN 1901-7286
DCE Lecture Notes No. 23 Department of Civil Engineering
Aalborg University
Department of Civil Engineering
Structural Mechanics
DCE Lecture Notes No. 23
by
Lars Andersen and Sren R.K. Nielsen
August 2008
c Aalborg University

Scientific Publications at the Department of Civil Engineering
Technical Reports are published for timely dissemination of research results and scientific work
carried out at the Department of Civil Engineering (DCE) at Aalborg University. This medium
allows publication of more detailed explanations and results than typically allowed in scientific
journals.
Technical Memoranda are produced to enable the preliminary dissemination of scientific work
by the personnel of the DCE where such release is deemed to be appropriate. Documents of
this kind may be incomplete or temporary versions of papersor part of continuing work. This
should be kept in mind when references are given to publications of this kind.
Contract Reports are produced to report scientific work carried out under contract. Publications
of this kind contain confidential matter and are reserved for the sponsors and the DCE. Therefore,
Contract Reports are generally not available for public circulation.
Lecture Notes contain material produced by the lecturers at the DCE for educational purposes.
This may be scientific notes, lecture books, example problems or manuals for laboratory work,
or computer programs developed at the DCE.
Theses are monograms or collections of papers published to report the scientific work carried
out at the DCE to obtain a degree as either PhD or Doctor of Technology. The thesis is publicly
available after the defence of the degree.
Latest News is published to enable rapid communication of information about scientific work
carried out at the DCE. This includes the status of research projects, developments in the labora-
tories, information about collaborative work and recent research results.
Published 2008 by
Aalborg University
Department of Civil Engineering
Sohngaardsholmsvej 57,
DK-9000 Aalborg, Denmark
Printed in Denmark at Aalborg University
ISSN 1901-7286 DCE Lecture Notes No. 23

Preface
This textbook has been written for the course Statics IV on spatial elastic beam structures
given at the 5th semester of the undergraduate programme in Civil Engineering at Aalborg Uni-
versity. The book provides a theoretical basis for the understanding of the structural behaviour
of beams in three-dimensional structures. In the course, the text is supplemented with labora-
tory work and hands-on exercises in commercial structural finite-element programs as well as
M ATLAB. The course presumes basic knowledge of ordinary differential equations and struc-
tural mechanics. A prior knowledge about plane frame structures is an advantage though not
mandatory. The authors would like to thank Mrs. Solveig Hesselvang for typing the manuscript.
Aalborg, August 2008 Lars Andersen and Sren R.K. Nielsen
i
ii

Contents
1 Beams in three dimensions 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Equations of equilibrium for spatial beams . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Section forces and stresses in a beam . . . . . . . . . . . . . . . . . . . 3
1.2.2 Kinematics and deformations of a beam . . . . . . . . . . . . . . . . . . 5
1.2.3 Constitutive relations for an elastic beam . . . . . . . . . . . . . . . . . 10
1.3 Differential equations of equilibrium for beams . . . . . . . . . . . . . . . . . . 12
1.3.1 Governing equations for a Timoshenko beam . . . . . . . . . . . . . . . 13
1.3.2 Governing equations for a Bernoulli-Euler beam . . . . . . . . . . . . . 14
1.4 Uncoupling of axial and bending deformations . . . . . . . . . . . . . . . . . . . 15
1.4.1 Determination of the bending centre . . . . . . . . . . . . . . . . . . . . 15
1.4.2 Determination of the principal axes . . . . . . . . . . . . . . . . . . . . 21
1.4.3 Equations of equilibrium in principal axes coordinates . . . . . . . . . . 25
1.5 Normal stresses in beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.6 The principle of virtual forces . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.7 Elastic beam elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.7.1 A plane Timoshenko beam element . . . . . . . . . . . . . . . . . . . . 34
1.7.2 A three-dimensional Timoshenko beam element . . . . . . . . . . . . . . 41
1.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2 Shear stresses in beams due to torsion and bending 45

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2 Homogeneous torsion (St. Venant torsion) . . . . . . . . . . . . . . . . . . . . . 46
2.2.1 Basic assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2.2 Solution of the homogeneous torsion problem . . . . . . . . . . . . . . . 48
2.2.3 Homogeneous torsion of open thin-walled cross-sections . . . . . . . . . 57
2.2.4 Homogeneous torsion of closed thin-walled cross-sections . . . . . . . . 59
2.3 Shear stresses from bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.3.1 Shear stresses in open thin-walled cross-sections . . . . . . . . . . . . . 69
2.3.2 Determination of the shear centre . . . . . . . . . . . . . . . . . . . . . 75
2.3.3 Shear stresses in closed thin-walled sections . . . . . . . . . . . . . . . . 82
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
References 93
iii
iv Contents

C HAPTER 1
Beams in three dimensions
This chapter gives an introduction is given to elastic beams in three dimensions. Firstly, the
equations of equilibrium are presented and then the classical beam theories based on Bernoulli-
Euler and Timoshenko beam kinematics are derived. The focus of the chapter is the flexural de-
formations of three-dimensional beams and their coupling with axial deformations. Only a short
introduction is given to torsional deformations, or twist, of beams in three dimensions. A full de-
scription of torsion and shear stresses is given in the next chapters. At the end of this chapter, a
stiffness matrix is formulated for a three-dimensional Timosheko beam element. This element can
be used for finite-element analysis of elastic spatial frame structures.
1.1 Introduction
In what follows, the theory of three-dimensional beams is outlined.
1.2 Equations of equilibrium for spatial beams

An initially straight beam is considered. When the beam is free of external loads, the beam
occupies a so-called referential state. In the referential state the beam is cylindrical with the
length l, i.e. the cross-sections are everywhere identical. The displacement and rotation of the
beam is described in a referential (x, y, z)-coordinate system with base unit vectors {i, j, k}, the
origin O placed on the left end-section, and the x-axis parallel with the cylinder and orientated
into the beam, see Fig. 11. For the time being, the position of O and the orientation of the y-
and z-axes may be chosen freely.
The beam is loaded by a distributed load per unit length of the referential scale defined by
the vector field q = q(x) and a distributed moment load vector per unit length m = m(x). A
differential beam element of the length dx is then loaded by the external force vector qdx and
external moment vector mdx as shown in Fig. 11. The length of the differential beam element
may change during deformations due to axial strains. However, this does not affect the indicated
load vectors which have been defined per unit length of the referential state. Measured in the
(x, y, z)-coordinate system, q and m have the components

qx mx
q = qy , m = my . (11)
qz mz
1
2 Chapter 1 Beams in three dimensions
qdx mdx
F + dF
M + dM
idx
M
y F
x
i
k
dx
z
l
Figure 11 Beam in referential state.
As a consequence of the external loads, the beam is deformed into the so-called current state
where the external loads are balanced by an internal section force vector F = F(x) and an
internal section moment vector M = M(x). These vectors act on the cross-section with the base
unit vector i of the x-axis as outward directed normal vector. With reference to Fig. 12, the
components of F and M in the (x, y, z)-coordinate system are:

N Mx
F = Qy , M = My (12)
Qz Mz
Here, N = N (x) is the axial force, whereas the components Qy = Qy (x) and Qz = Qz (x) sig-
nify the shear force components in the y- and z-directions. The axial component Mx = Mx (x) of
the section moment vector is denoted the torsional moment. The components My = My (x) and
Mz = Mz (x) in the y- and z-directions represent the bending moments. The torsional moment
is not included in two-dimensional beam theory. However, in the design of three-dimensional
frame structures, a good understanding of the torsional behaviour of beams is crucial.
Assuming that the displacements remain small, the equation of static equilibrium can be
established in the referential state. With reference to Fig. 11, the left end-section of the element
is loaded with the section force vector F and the section moment vector M. At the right
end-section, these vectors are changed differentially into F + dF and M + dM, respectively.
Force equilibrium and moment equilibrium formulated at the point of attack of the section force
vector F at the left end-section then provides the following equations of force and moment

1.2 Equations of equilibrium for spatial beams 3
Qy
y
My
x
Mx N
z Mz
Qz
Figure 12 Components of the section force vector and the section moment vector.
equilibrium of the differential beam element:
F + F + dF + qdx = 0
dF
+q=0 (13a)
dx
M + M + dM + idx (F + dF) + mdx = 0
dM
+iF+m =0 (13b)
dx
From Eqs. (11) and (12) follows that Eqs. (13a) and (13b) are equivalent to the following
component relations:
dN dQy dQz
+ qx = 0, + qy = 0, + qz = 0, (14a)
dx dx dx
dMx dMy dMz
+ mx = 0, Qz + my = 0, + Qy + mz = 0. (14b)
dx dx dx
At the derivation of Eq. (14b), it has been utilised that
i F = i (N i + Qy j + Qz k) = N i i + Qy i jQz i k = 0i Qz j + Qy k. (15)
Hence, i F has the components {0, Qz , Qy }. It is noted that a non-zero normal-force compo-
nent is achieved when the moment equilibrium equations are formulated in the deformed state.
This may lead to coupled lateral-flexural instability as discussed in a later chapter.
1.2.1 Section forces and stresses in a beam

On the cross-section with the outward directed unit vector co-directional to the x-axis, the normal
stress xx and the shear stresses xy and xz act as shown in Fig. 13. These stresses must be

statically equivalent to the components of the force vector F and the section moment vector M
as indicated by the following relations:
Z Z Z
N= xx dA, Qy = xy dA, Qz = xz dA, (16a)
A A A
Z Z Z
Mx = (xz y xy z)dA, My = zxx dA, Mz = yxx dA. (16b)
A A A
y Qy
My
x
xy
Mx N
dA
xx
z Mz xz
Qz
Figure 13 Stresses and stress resultant on a cross-section of the beam.
yy
yx
yz
xy dy
zy xx
xz
x
zx
zz
dz
dx
z
Figure 14 Components of the stress tensor.
On sections orthogonal to the y- and z-axes, the stresses {yy , yx , yz } and {zz , zx , zy }
act as shown in Fig. 14. The first index indicates the coordinate axis co-directional to the
outward normal vector of the section, whereas the second index specifies the direction of action
of the stress component. The stresses shown in Fig.14 form the components of the stress tensor

in the (x, y, z)-coordinate system given as

xx yx zx
= xy yy zy . (17)
xz yz zz
Moment equilibrium of the cube shown in Fig. 14 requires that
xy = yx , xz = zx , yz = zy . (18)
Hence, is a symmetric tensor.
1.2.2 Kinematics and deformations of a beam

The basic assumption in the classical beam theory is that a cross-section orthogonal to the x-axis
at the coordinate x remains plane and keeps its shape during deformation. In other words, the
cross-section translates and rotates as a rigid body. Especially, this means that Poisson contrac-
tions in the transverse direction due to axial strains are ignored. Hence, the deformed position
of the cross-section is uniquely described by a position vector w = w(x) and a rotation vector
= (x) with the following components in the (x, y, z)-coordinate system:

wx x
w = wy , = y . (19)
wz z
Further, only linear beam theory will be considered. This means that the displacement compo-
nents wx , wy and wz in Eq. (19) all small compared to the beam length l. Further the rotation
components x , y and z are all small. Especially, this means that
sin tan , (110)
where represents any of the indicated rotation components measured in radians. The various
displacement and rotation components have been illustrated in Fig. 15. The rotation component
around the x-axis is known as the twist of the beam.
Now, a material point on the cross-section with the coordinates (x, y, z) in the referential
state achieves a displacement vector u = u(x, y, z) with the components {ux , uy , uz } in the
(x, y, z)-coordinate system given as (see Fig. 15):
ux (x, y, z) = wx (x) + zy (x) yz (x), (111a)
uy (x, y, z) = wy (x) zx (x), (111b)

uz (x, y, z) = wz (x) + yx (x). (111c)
It follows that the displacement of any material point is determined if only the 6 components of
w(x) and (x) are known at the beam coordinate x. Hence, the indicated kinematic constraint
reduces the determination of the continuous displacement field u = u(x, y, z) to the determina-
tion of the 6 deformation components wx = wx (x), wy = wy (x), wz = wz (x), x = x (x),
y = y (x) and z = z (x) of a single spatial coordinate along the beam axis.

z y
y z
dwy dwz
dx dx
wy wz
x x
z y
wx wx
y
z
x
Figure 15 Deformation components in beam theory.
The strains conjugated to xx , xy and xz are the axial strain xx and the angular strains
xy = 2xy and xz = 2xz . They are related to displacement components as follows:
ux dwx dy dz
xx = = +z y , (112a)
x dx dx dx
ux uy dwy dx
xy = + = z z (x), (112b)
y x dx dx
ux uz dwy dx
xz = + = +y + y (x). (112c)
z x dx dx
From Eq. (112) follows that xy = xy (x, z) is independent of y as a consequence of the
presumed plane deformation of the cross-section. Then, the shear stress xy = xy (x, z) must
also be constant over the cross-section. Especially, xy 6= 0 at the upper and lower edge of the
cross-section as illustrated in Fig. 16a. However, if the cylindrical surface is free of surface
shear tractions, then yx = 0 at the edge. Hence, xy 6= yx in contradiction to Eq. (18).
In reality xy = 0 at the edges, corresponding to xy = 0. This means that the deformed
cross-section forms a right angle to the cylindrical surface as shown in Fig. 16b.
The displacement fields Eq. (111) are only correct for beams with cross-sections which are
circular symmetric around the x-axis. In all other cross-sections, the torsional moment Mx will

y yx y
xy
xy dwy xy dwy
dx dx
wy wy
x x
z z
wx wx
(a) (b)
Figure 16 Shear stresses on deformed beam section: (a) Deformation of cross-section in beam theory and (b) real
deformation of cross-section.
induce an additional non-planar displacement in the x-axis, which generally can be written in the
form ux (x, y, z) = (y, z)dx /dx. This is illustrated in Fig. 17. Hence, the final expression
for the axial displacement reads
dx
ux (x, y, z) = wx (x) + zy (x) yz (x) + (y, z) . (113)
dx
The expressions for uy and uz in Eq. (111) remain unchanged, and (y, z) is called the warp-
ing function. Whereas y and z in Eq. (113) may be considered as shape functions for the
deformations caused by the rotations z (x) and y (x), the warping function is a shape function
defining the axial deformation of the cross-section from the rotation component. The definition
and determination of the warping function is considered in a subsequent section.
Section AA
A Undeformed state B
Deformed top flange
Deformed bottom flange

A B Section BB
Figure 17 Warping deformations in an I-beam induced by homogeneous torsion. The cross sections AA and BB are
shown with the top flange on the left and the bottom flange on the right.
As a consequence of the inclusion of the warping, the strain components in Eq. (112) are
modified as follows:
ux dwx dy dz d2 x
xx = = +z y + 2 , (114a)
x dx dx dx dx

ux uy dwy dx
xy = + = z + z , (114b)
y x dx y dx

ux uz dxz dx
xz = + = + y + +y . (114c)
z x dx z dx

z y
y z
dwy dwz
dx dx
wy wz
x x
z y
wx wx
Figure 18 Kinematics of Bernoulli-Euler beam theory.
Bernoulli-Euler beam kinematics presumes that the rotated cross-section is always orthogo-
nal to the deformed beam axis. This involves the following additional kinematical constraints on
the deformation of the cross-section (see Fig. 18):
dwz dwy
y = , z = . (115)
dx dx
Assuming temporarily that x 0 in bending deformations, i.e. disregarding the twist of the
beam, Eqs. (114) and (115) then provide:
xy = xz = 0. (116)
Equation (116) implies that the shear stresses are xy = xz = 0, and in turn that the shear
forces become Qy = Qz = 0, cf. Eq. (16). However, non-zero shear forces are indeed present
in bending of Bernoulli-Euler beams. The apparent paradox is dissolved by noting that the shear
forces in Bernoulli-Euler beam theory cannot be derived from the kinematic condition, but has
to be determined from the static equations.
The development of the classical beam theory is associated with names like Galilei (15641642),
Mariotte (16201684), Leibner (16461716), Jacob Bernoulli (16541705), Euler (17071783),
Coulomb (17361806) and Navier (17851836), leading to the mentioned Bernoulli-Euler beam
based on the indicated kinematic constraint. The inclusion of transverse shear deformation was
proposed in 1859 by Bresse (18221883) and extended to dynamics in 1921 by Timoshenko (1878
1972). Due to this contribution, the resulting beam theory based on the strain relations Eq. (112),
is referred to as Timoshenko beam theory (Timoshenko 1921).
The first correct analysis of torsion in beams was given by St. Venant (1855). The underlying
assumption was that dx /dx in Eq. (113) was constant, so the warping in all cross-sections
become identical. Then, the axial strain xx from torsion vanishes and the distribution of the
shear strains xy and xz are identical in all sections. Because of this, St. Venant torsion is also
referred to as homogeneous torsion.

Whenever the twist or the warping is prevented at one or more cross-sections, dx /dx is no
longer constant as a function of x. Hence, axial strains occur and, as a consequence of this,
axial stresses arise and the shear strains and shear stresses are varying along the beam. These
phenomena were systematically analysed by Vlasov (1961) for thin-walled beams, for which
reason the resulting theory is referred to as Vlasov torsion or non-homogeneous torsion. Notice
that the shear stresses from Vlasov torsion have not been included in the present formulation.
These will be considered in a subsequent chapter.
Seen from an engineering point of view, the primary advantage of Vlasov torsion theory is that it
explains a basic feature of beams, namely that prevention of warping leads to a much stiffer struc-
tural elements than achieved in the case of homogeneous warping, i.e. a given torsional moment
will induce a smaller twist. Warping of the cross-section may, for example, be counteracted by the
inclusion of a thick plate orthogonal to the beam axis and welded to the flanges and the web. The
prevention of torsion in this manner is particularly useful in the case of slender beams with open
thin-walled cross-sections that are prone to coupled flexuraltorsional buckling. Obviously, Vlasov
torsion theory must be applied for the analysis of such problems as discussed later in the book.
Next, the deformation of the cross-section may be decomposed into bending and shear com-
ponents. The bending components are caused by the bending moments My and Mz and deform
as a Bernoulli-Euler beam. Hence, the bending components are causing the rotations y and z
of the cross-section. The shear components are caused by the shear forces Qy and Qz . These
cause the angular shear strains xy and xz without rotating the cross-section. Further, the dis-
placement of the beam axis in shear takes place without curvature. Hence, the curvature of the
beam axis is strictly related to the bending components, see Fig. 19.
With reference to Fig. 110, the radii of curvatures ry and rz are related to the rotation
increments dz and dy of the end-sections in the bending deformations of a differential beam
element of the length dx as follows

ry dz = dx y = 1/rz = dy /dx
(117)
rz dy = dx z = 1/ry = dz /dx

Here, y and z denote the components of the curvature vector of the x-axis. Especially, for a
Bernoulli-Euler beam the curvature components become, cf. Eq. (115),
d2 wz d2 wy
y = 2
, z = . (118)
dx dx2
From Eqs. (114) and (117) follows that the axial strain may be written as
d2 x
xx (x, y, z) = (x) + zy (x) yz (x) + (y, z) , (119)
dx2
where (x) denotes the axial strain of the beam along the x-axis given as
dwx
(x) = . (120)
dx
Here, (x), y (x) and y (x) define the axial strain and curvatures of the beam axis, i.e. the
(x)-axis.

x
z
y =
Mz
x
z
y
+ xy
Qy
x
z
xy
Figure 19 Decomposition of cross-section deformation into bending and shear components.
y z
dz dy
ry rz
wy ds dx wz ds dx
x x
z y
Figure 110 Definition of curvature.
1.2.3 Constitutive relations for an elastic beam

In what follows we shall refer to N (x), Qy (x), Qz (x), Mx (x), My (x) and Mz (x) as generalised
stresses. These are stored in the column matrix

N (x)
Qy (x)

Qz (x)
(x) = Mx (x) .
(121)

My (x)
Mz (x)

The internal virtual work of these quantities per unit length of the beam is given as
= N + Qy xy + Qz xz + Mx x + My y + Mz z = T , (122)
where

(x)

xy (x)

xz (x)
(x) = . (123)

x (x)

y (x)
z (x)
The components of (x) are referred to as the generalised strains. The components of (x) and
(x) are said to be virtual work conjugated because these quantities define the internal virtual
work per unit length of the beam.
Let E and G denote the elasticity modulus and the shear modulus. Then, the normal stress
xx and the shear stresses xy and xz may be calculated from Eq. (114) as follows:
d2 x

dwx dy dz
xx = Exx = E +z y + (y, z) 2 , (124a)
dx dx dx dx

dwy dx
xy = Gxy = G z + z , (124b)
dx y dx

dw2 dx
xz = Gxz = G + y + +y . (124c)
dx z dx
By integration over the cross-sectional area, it then follows that
d2 x

dwx dy d2
N =E A + Sy Sz + S 2 , (125a)
dx dx dx dx

dwy dx
Qy = G Ay z + Ry , (125b)
dx dx

dwz dx
Qz = G Az + y + Rz , (125c)
dx dx

dwz dwy dx
M x = G Sz + y S y z + K , (125d)
dx dx dx

dwx dy dz dx
M y = E Sy + Iyy Iyz + Iz , (125e)
dx dx dx dx

dwx dy dz dx
Mz = E Sz Iyz + Izz Iy , (125f)
dx dx dx dx
where Ay , Az , Ry , Rz , Sy , Sz , S , K, Iyy , Izz , Iyz = Izy , Iy and Iz are cross-sectional (or
geometrical) constants identified as:
Z
A= dA, Ay = y A, Az = z A, (126a)
A

Z Z

Ry = z dA, Rz = + y dA, (126b)
A y A z
Z Z Z
Sy = zdA, Sz = ydA, S = dA, (126c)
A A A
Z Z
Iyy = z 2 dA, Izz = y 2 dA, (126d)
A A
Z Z Z
Iyz = yzdA, Iy = ydA, Iz = zdA, (126e)
A A A
Z
2 2
K= y +z +y z dA. (126f)
A z y
Here, A is the cross-sectional area, whereas Ay and Az signify the so-called shear areas. Beam
theory presumes a constant variation of the shear stresses in bending, whereas the actual variation
is at least quadratic. The constant variation results in an overestimation of the stiffness against
shear deformations, which is compensated by the indicated shear reduction factors y and z . If
the actual distribution of the shear stresses is parabolic, these factors become y = z = 5/6.
For an I-profile, the shear area is approximately equal to the web area.
For GAy we have xy = Qy /(GAy ) = 0. Bernoulli-Euler beam theory is charac-
terised by xy = 0. Hence, Timoshenko theory must converge towards Bernoulli-Euler theory
for the shear areas passing towards infinity. The magnitude of the shear deformations in propor-
tion to the bending deformations depends on the quantity (h/l)2 , where h is the height and l is
the length of the beam. This relation is illustrated in Example 1-3.
Ry and Rz are section constants which depend on the warping mode shape (y, z) as well
as the bending modes via y and z. Further, the section constants Sy and Sz are denoted the static
moments around the y- and z-axes. S specifies a corresponding static moment of the warping
shape function.
Iyy and Izz signify the bending moments of inertia around the y- and z-axes, respectively.
Iyz is denoted the centrifugal moment of inertia, whereas Iy and Iz are the corresponding
centrifugal moments of the warping shape function and the bending mode shapes.
K is the so-called torsion constant. This defines merely the torsional stiffness in St. Venant
torsion. As mentioned above, the additional contribution to Mx from Vlasov torsion will be
considered in a subsequent section.
1.3 Differential equations of equilibrium for beams

In what follows, the governing differential equations for Timoshenko and Bernoulli-Euler beams
are derived. At this stage, the twist x and the torsional moment Mx are ignored. With no further
assumptions and simplifications, Eq. (125) reduces to

N EA ESy ESz 0 0 dwx /dx
My ESy EIyy EIyz 0 0 dy /dx

Mz = ESz EIyz EIzz 0 0 dz /dx . (127)

Qy 0 0 0 GAy 0 dwy /dx z
Qz 0 0 0 0 GAz dwz /dx + y

1.3 Differential equations of equilibrium for beams 13
The coefficient matrix of Eq. (127) is symmetric. When formulated in a similar matrix format,
the corresponding matrix in Eq. (125) is not symmetric. This is a consequence of the ignorance
of the Vlasov torsion in Mx .
1.3.1 Governing equations for a Timoshenko beam

Next, Eq. (127) is inserted into the equilibrium equations (14a) and (14b), which results in
the following system of coupled ordinary differential equations for the determination of wx , wy ,
wz , y and z :

dN/dx 0 qx
dMy /dx Qz my

dMz /dx = Qy mz

dQy /dx 0 qy
dQz /dx 0 qz

EA ESy ESz 0 0 dwx /dx
ESy EIyy EIyz 0 0 dy /dx
d
ESz EIyz EIzz 0 0 dz /dx
dx

0 0 0 GAy 0

dwy /dx z
0 0 0 0 GAz dwz /dx + y

0 0 0 0 0 dwx /dx qx
0 0 0 0 GAz dy /dx my

0 0 0 GAy 0 dz /dx
= mz . (128)

0 0 0 0 0 dz /dx z qy
0 0 0 0 0 dwz /dx + y qz
Equation (128) specifies the differential equations for Timoshenko beam theory. These should
be solved with proper boundary condition at the end-sections of the beam. Let x0 denote the
abscissa of any of the two end-sections, i.e. x0 = 0 or x0 = l, where l is the length of the beam.
At x = x0 either kinematical or mechanical boundary conditions may be prescribed.
Kinematical boundary conditions mean that values of wx , wy , wz , y and z are prescribed,

wx (x0 ) = wx,0
wy (x0 ) = wy,0

wz (x0 ) = wz,0 , x0 = 0, l, (129)
y (x0 ) = y,0

z (x0 ) = z,0

whereas mechanical boundary conditions imply the prescription of N , Qy , Qz , My and Mz ,

N (x0 ) = N0
Qy (x0 ) = Qy,0

Qz (x0 ) = Qz,0 , x0 = 0, l. (130)
My (x0 ) = My,0

Mz (x0 ) = Mz,0


In Eq. (130), the left-hand sides are expressed in kinematical quantities by means of Eq. (127).
Of the 10 possible boundary conditions at x = x0 specified by Eqs. (135) and (130), only 5
can be specified. The 5 boundary conditions at x0 = 0 and x0 = l can be selected independently
from Eq. (135) and Eq. (130).
With given boundary conditions Eq. (128) can be solved uniquely for the 5 kinematic quan-
tities wx , wy , wz , y , z , which make up the degrees of freedom of the cross-section. Although
an analytical solution may be cumbersome, a numerical integration is always within reach.
1.3.2 Governing equations for a Bernoulli-Euler beam

Next, similar differential equations are specified for a Bernoulli-Euler beam. At first the shear
forces Qy and Qz in the equations of equilibrium for My and Mz in Eq. (14b) are eliminated
by means of the 2nd and 3rd equations in Eq. (14a):
d2 My /dx2 dQz /dx + dmy /dx = 0

2
d My /dx2 + qz + dmy /dx = 0

(131)
d2 Mz /dx2 + dQy /dx + dmz /dx = 0 d2 Mz /dx2 qy + dmz /dx = 0.

Using the Bernoulli-Euler kinematical constraint Eq. (115), the constitutive equations for the
resulting section forces may be written as

N EA ESy ESz dwx /dx
My = ESy EIyy EIyz d2 wz /dx2 . (132)
Mz ESzz EIyz EIzz d2 wy /dx2
Then, the equations of equilibrium Eq. (14a) and Eq. (131) may be recasted as the following
system of coupled ordinary differential equations
d2 wz d2 wy

d dwx
EA ESy ESz + qx = 0, (133a)
dx dx dx2 dx2
d2 d2 wz d2 wy

dwx dmy
ESy EIyy EI yz + qz + = 0, (133b)
dx2 dx dx2 dx2 dx
d2 d2 wz d2 wy

dwx dmz
ES z + EI yz + EI zz + qy + = 0. (133c)
dx2 dx dx2 dx2 dx
The governing equations (133) should be solved with 5 of the same boundary conditions as
indicated by Eqs. (135) and (130). The difference is that y (x0 ), z (x0 ), Qy (x0 ) and Qz (x0 )
are represented as, cf. Eqs. (14b) and (115),
dwz (x0 ) dwz (x0 )

= y,0 , = z,0 , (134a)
dx dx
dMz (x0 ) dMy (x0 )

mz (x0 ) = Qy,0 , + my (x0 ) = Qz,0 . (134b)
dx dx

1.4 Uncoupling of axial and bending deformations 15
With this in mind, the kinematic boundary conditions for Bernoulli-Euler beams are given in the
form

wx (x0 ) = wx,0
wy (x0 ) = wy,0

wz (x0 ) = wz,0 , x0 = 0, l, (135)
dwz (x0 )/dx = y,0

dwy (x0 )/dx = z,0

whereas the mechanical boundary conditions defined in Eq. (130) are still valid.
1.4 Uncoupling of axial and bending deformations

Up to now the position of the origin O and the orientation of the y- and z-axes in the cross-
section have been chosen arbitrarily. As a consequence of this, the deformations from the axial
force and the deformation from the bending moments My and Mz will generally be coupled.
This means that the axial force N referred to the origin O will not merely induce a uniform
displacement wx of the cross-section, but also non-zero displacements wy and wz of O as well
as rotations y and z . Similarly, the bending moment My will not merely cause a displacement
wy and a rotation y of the cross-section, but also a non-zero displacement wy and a rotation z
in the orthogonal direction in addition to an axial displacement wy of the origin. The indicated
mechanical couplings are the reason for the couplings in the differential equations (128) and
(133). The couplings may have a significant impact on the structural behaviour and stability of
an engineering structure and the position of the origin for a given beam element as well as the
orientation of the coordinate axes must be implemented correctly in a computational model.
In this section, two coordinate transformations will be indicated, in which the axial force re-
ferred to the new origin B, called the bending centre, only induces a uniform axial displacement
over the cross-section. Similarly, the bending moments My and Mz around the new rotated y-
and z-axes, referred to as the principal axes, will only induce the non-zero deformation compo-
nents (wz , y ) and (wy , z ), respectively. Especially, the moments will induce the displacement
wx = 0 of the bending centre, B.
1.4.1 Determination of the bending centre

The position of the bending centre B is given by the position vector rB with the components
{0, yB , zB } in the (x, y, z)-coordinate system. In order to determine the components yB and zB ,
a translation of the (x, y, z)-coordinate system to a new (x , y , z )-coordinate system with origin
in the yet unknown bending centre is performed (see Fig. 111). The relations between the new
and the old coordinates read
x = x , y = y + yB , z = z + zB , (136)
In the new coordinate system, the displacement of B (the new origin) in the x -direction (the
new beam axis) becomes (see Fig. 111):
wx = wx + zB y yB z . (137)

z , M z z , Mz
z z
B y y , My
zB
N
rB
O y y , M y
N yB
Figure 111 Translation of coordinate system.
The axial strain of fibres placed on the new beam axis becomes
dw dw
(x ) = = = + z B y y B z , (138)
dx dx
where Eq. (117), Eq. (120) and Eq. (137) have been used. The components of the rotation
vector of the cross-section are identical, i.e.
x = x , y = y , z = z . (139)
In turn, this means that the components of the curvature vector in the two coordinate systems
are identical as well
dy dy dz dz
y = = = y , z = = = z . (140)
dx dx dx dx
Further, the components of the section force vector F in the two coordinate systems become
identical, i.e.
N = N, Qy = Qy Qz = Qz . (141)
As a consequence of referring the axial force N = N to the new origin B, the components
of the section vector in the (x , y , z )-coordinate system are related to the components in the
(x, y, z)-coordinate as follows:
Mx = Mx , My = My zB N, Mz = Mz + yB N. (142)
Equations (138) and (140) provide the following relation for in terms of , y and z :
= zB y + yB z = zB y + yB z . (143)

Then the relation between {N, My , Mz } and {N , My , Mz } and {, y , z } and { , y , z }

may be specified in the following matrix formulation:
= AT , (144a)
= A , (144b)
where

N
= My , = y , (145a)
Mz z

N
= My , = y , (145b)
Mz y

1 zB yB
A = 0 1 0 . (145c)
0 0 1
The components {N, My , Mz } and {, y , z } of and may be interpreted as work conjugated
generalised stresses and strains.
With reference to Eq. (132), the constitutive relation between and is given as
= C, (146)
where C denotes the constitutive matrix,

A Sy Sz
C = E Sy Iyy Iyz . (147)
Sz Iyz Izz
Likewise, the constitutive relation in the (x , y , z )-coordinate system reads
= C (148)
where the constitutive matrix has the form

A Sy Sz
C = E S y Iyy Iy z . (149)
Sz Iyz Izz
Obviously, as given by Eqs. (147) and (149), the cross-sectional area A is invariant to a rotation
of the cross-section about the x-axis and a translation in the y- and z-directions.
From Eqs. (144a), (144b) and (146) follows that
= AT C = AT CA

1 0 0 A Sy Sz 1 zB yB
C = AT CA = E zB 1 0 Sy Iyy Iyz 0 1 0
yB 0 1 Sz Iyz Izz 0 0 1


A Sy zB A (Sz yB A)
2
C = E Sy zB A Iyy 2zB Sy +zB A Iyz + yB Sy +zB (Sz yB A) .
2
(Sz yB A) Iyz + yB Sy +zB (Sz yB A) Izz 2yB Sz +yB A
(150)
The idea is now to use the translational coordinate transformation to uncouple the axial defor-
mations from the bending deformations. This requires that Sy = Sz = 0. Upon comparison
of Eq. (149) and Eq. (150), this provides the following relations for the deformation of the
coordinates of the bending centre:
Sz Sy
yB = , zB = . (151)
A A
With yB and zB given by Eq. (151), the bending moments of inertia, Iyy and Izz , and the
centrifugal moment of inertia, Iyz , in the new coordinate system can be expressed in terms of
the corresponding quantities in the old coordinate system as follows:
2 2
Iyy = Iyy 2zB (AzB ) + zB A = Iyy zB A, (152a)
2 2
Izz = Izz 2yB (AyB ) + yB A = Izz yB A, (152b)
Iyz = Iyz yB (AzB ) zB (AyB ) + yB zB A = Iyz yB zB A. (152c)

The final results in Eq. (152) are known as Knigs theorem.
O
z
x
yB
Figure 112 Single-symmetric cross-section.
If the cross-section is symmetric around a single line, and the y-axis is placed so that it
coincides with this line of symmetry, then the static moment Sy vanishes, i.e.
Z
Sy = zdA = 0 (153)
A
As a result of this, the bending centre B will always be located on the line of symmetry in a
single-symmetric cross-section, see Fig. 112. Obviously, if the cross-section is double symmet-
ric, then the position of B is found at the intersection of the two lines of symmetry.

Example 1.1 Determination of bending and centrifugal moments of inertia of non-symmetric

thin-walled cross-section
The position of the bending centre of the cross-section shown in Fig. A is determined along with the
bending moments of inertia Iy y and Iz z and the centrifugal moment of inertia Iy z .
2a
x O
z
2t
2a
t
t
a
y
Figure A Thin-walled cross-section.
The (x, y, z)-coordinate system is placed as shown in Fig. A. Then, the following cross-sectional con-
stants are calculated:
A = 2a 2t + 2a t + a t = 7at, (a)
t a 1
Sy = 2a 2t a + 2a t + a t = (2t + 9a) + ta, (b)
2 2 2

t 1
Sz = 2a 2t t + 2a t (2t + a) + a t 2t + 2a + = (21t + 8a)ta, (c)
2 2
1 1 1 1
Iyy = 2t (2a)3 + 2a t3 + t a3 = 2t2 + 17a2 ta,

(d)
3 3 3 3
2
1 1 1 t
Izz = 2a 2a (2t)3 + (2a)3 t + 2a t (2t + a)2 + a t3 + a t 2t + 2a +
3 12 12 2
1
59t2 + 54ta + 20a2 ta,

= (e)
3

1 t a 1
8t2 + 25ta + 4a2 ta. (f)

Iyz = 2a 2t t a + 2a t (2t + a) + a t 2t + 2a + =
2 2 2 4
Here, use has been made of Knigs theorem at the calculation of contributions to Iyy , Izz and Iyz from
the three rectangles forming the cross-section.
The coordinates of the bending centre follow from Eq. (151) and Eqs. (a) to (c). Thus,
Sz 3 4 Sy 1 9
yB = = t + a, zB = = t+ a. (g)
A 2 7 A 7 14
(continued)

3
2
t + 47 a
1 9
7
t + 14
a
B z
x
y
Figure B Position of bending centre in the thin-walled cross-section.
Subsequently, the moments of inertia around the axes of the (x , y , z )-coordinate system follow from
Eq. (152), Eq. (153) and Eqs. (d) to (f):
2
1 1 9 1
Iy y = (2t2 + 17a2 )ta t+ a 7ta = (44t2 108ta + 233a2 )ta, (h)
3 7 14 84
2
1 2 2 3 4 1
I z z = (59t + 54ta + 20a )ta t + a 7ta = (329t2 + 504ta + 368a2 )ta, (i)
3 2 7 84

1 1 9 3 4 1
Iy z = (8t2 + 25ta + 4a2 )ta t+ a) t + a 7ta = (7t2 15ta 22a2 )ta. (j)
4 7 14 2 7 14
Now, for a thin-walled cross-section the thickness of the flanges and the web is much smaller than the
widths of the flanges and the height of the web. In the present case this means that t a. With this in
mind, Eqs. (g) to (j) reduce to
4 9
yB a, zB a (k)
7 14
and
233 3 92 3 11 3
Iy y ta , Iz z ta , Iy z ta . (l)
84 21 7
It is noted that the error on Iz z estimated by Eq. (l) increases rapidly with increasing values of t/a.
Thus, for t/a = 0.1, the error is about 13%. The errors related to the estimated values of Iy y and Iy z
are somewhat smaller, i.e. about 5% and 7%, respectively.
From now on, the origin of the (x, y, z)-coordinate system is placed at the bending centre.
Then, the constitutive matrix given by Eq. (147) takes the form

A 0 0
C=E 0 Iyy Iyz . (154)
0 Iyz Izz

As a result of this, an axial force N no longer induces deformations in the y- and z-directions,
and the bending moments My and Mz do not induce axial displacements. However, the bending
moment My will still induce displacements in the y-direction in addition to the expected dis-
placements in the z-direction. Similarly, the bending moment Mz induces displacements in both
the y- and z-directions.
1.4.2 Determination of the principal axes

In order to uncouple the bending deformations, so that My will only induce deformations in the
z-direction, and Mz only deformations in the y-direction, a new (x , y , z )-coordinate system is
introduced with origin in B and rotated the angle around the x-axis as shown in Fig. 113.
z , Mz
z , Mz
y , My
z z
y
y y , My
B
N = N
Figure 113 Rotation of coordinate system.
Let {N, My , Mz } and {N , My , Mz } denote the components of the generalised stresses in

the (x, y, z)-coordinate system and the (x , y , z )-coordinate system, respectively. The two sets
of generalised stresses are related as
= B , (155a)
where

N N 1 0 0
= My , = My , B = 0 cos sin . (155b)
Mz Mz 0 sin cos
Likewise, the components of the generalised strains in the two coordinate systems are denoted
as { , y , z } and {, y , z }, respectively. These are related as
= B , (156a)

where

1 0 0
= y , = y , B = 0 cos sin . (156b)
z z 0 sin cos
The constitutive relation in the (x , y , z )-coordinate system reads

A 0 0
= C , C = E 0 Iyy Iyz . (157)
0 Iyz Izz
The corresponding constitutive relation in the (x, y, z)-coordinate system is given by Eq. (146)
with C given by Eq. (154). Use of Eqs. (155a) and (156a) in Eq. (146) provides
B = C = CB = BT CB , (158)
where it has been utilised that B1 = BT . Comparison of Eq. (157) and Eq. (158) leads the
following relation between the constitutive matrices
C = BT CB

1 0 0 A 0 0 1 0 0
=E 0 cos sin 0 Iyy Iyz 0 cos sin
0 sin cos 0 Iyz Izz 0 sin cos

A 0 0

= E 0 C22 C23 . (159a)

0 C32 C33
where

C22 = cos2 Iyy 2 cos sin Iyz + sin2 Izz , (159b)

C23
= C32 = sin cos (Iyy Izz ) (cos2 sin2 )Iyz , (159c)

C33 = cos2 Izz + 2 cos sin Iyz + sin2 Izz . (159d)
From Eq. (157) and Eq. (159) follows that
1 1
Iyy = (Iyy + Izz ) + (Iyy Izz ) cos(2) Iyz sin(2), (160a)
2 2
1
Iyz = sin(2)(Iyy Izz ) cos(2)Iyz , (160b)
2
1 1
Izz = (Iyy + Izz ) (Iyy Izz ) cos(2) + Iyz sin(2), (160c)
2 2
where use has been made of the relations
sin(2) = 2 sin cos , cos(2) = cos2 sin2 ,

1 1
cos2 = (1 + cos 2), sin2 = (1 cos2 ).
2 2

Uncoupling of bending deformations in the (x , y , z )-coordinate system requires that Iyz = 0.

This provides the following relation for the determination of the rotation angle :
2Iyz
1 tan(2) = Izz I yy
for Iyy 6= Izz ,
sin(2)(Iyy Izz ) cos(2)Iyz = 0 (161)
2
cos(2) = 0 for Iyy = Izz .
Note that cos(2) = 0 implies that sin(2) = 1. The sign of sin(2) is chosen as follows:
= 14 for Iyz < 0

sin(2) = 1
. (162)
sin(2) = 1 = 34 for Iyz > 0

Then, Eq. (160) provides the following solutions for Iyy and Izz :
1 1
Iyy = (Iyy + Izz )+ | Iyz |, Izz = (Iyy + Izz ) | Iyz | . (163)
2 2
For Iyy 6= Izz the solution for tan(2) is fulfilled for the following two alternative solutions
for sin(2) and cos(2):
2Iyz Iyy Izz

sin(2) = , cos(2) = , (164a)
J J
2Iyz Iyy Izz
sin(2) = , cos(2) = , (164b)
J J
where
q
J = (Iyy Izz )2 + 4Iyz
2 . (165)
The sign definition in Eq. (164a) is chosen. This implies that
2 [0, ] for Iyz < 0 and 2 [, 2] for Iyz > 0. (166)
Insertion of the solution for sin(2) and cos(2) into Eq. (160) provides the following
results for Iyy and Izz :
1 1 (Iyy Izz )2 + 4Iyz

2
Iyy = (Iyy + Izz ) + , (167a)
2 2 J
1 1 (Iyy Izz )2 + 4Iyz

2
Izz = (Iyy + Izz ) , (167b)
2 2 J
or, by insertion of J, cf. Eq. (165),
1 1q
Iyy = (Iyy + Izz ) + (Iyy Izz )2 + 4Iyz
2 , (168a)
2 2
1 1q
Izz = (Iyy + Izz ) (Iyy Izz )2 + 4Iyz
2 . (168b)
2 2

z z
z y
y
x = x
B y y
B
x = x
(a) (b)
Figure 114 Position of principal axes: (a) Iyz < 0 and (b) Iyz > 0.
The coordinate axes y and z are known as the principal axes of the cross-section, whereas Iyy
and Izz are called the principal moments of inertia. It follows from Eqs. (163) and (167) that
the choices of signs for sin(2) implies that Iyy becomes the larger of the principal moments
of inertia and Izz is the smaller principal moment of inertia. It is emphasised that this choice
is performed merely to have a unique determination of . Three other choices of are possible
obtained by additional rotations of the magnitudes 2 , and 32 relative to the indicated.
If the cross-section has a symmetry line, and the y-axis is placed along this line, then Iyz = 0.
Hence, a symmetry line is always a principal axis. Since the principal axes are orthogonal, the
z-axis is also a principal axiseven if the cross-section is not symmetric around the axis.
Example 1.2 Determination of principal axes coordinate system

The cross-section analysed in Example 1.1 is reconsidered. The thin-wall approximation is used, so the
moments of inertia are given by Eq. (l) in Example 1.1 and repeated here (without the primes):
233 3 92 3 11 3
Iyy ta 2.7738 ta3 , Izz ta 4.3810 ta3 , Iyz ta 1.5714 ta3 .(a)
84 21 7
The position of the bending centre relatively to the top-left corner of the cross-section is provided in
Fig. A. From Eq. (a) and Eq. (165) follows that
s 2 2
233 3 92 3 11 9769
J= ta ta + 4 ta3 = ta3 , (b)
84 21 7 28
which by insertion into Eq. (164) provides:
2 11 ta3 28

7 88
sin(2) = = 0.8903, (c)
ta3 9769 9769
233
ta3 92 ta3 28

84 21 3780
cos(2) = = 0.4553. (d)
3
ta 9769 84 9769
(continued)

yB 4
a z
7
2a
9 2t
zB 14
a
B z
x= x
2a
t y
t
a
y
Figure A Position of principal axes coordinated system for the thin-walled cross-section.
From Eqs. (c) and (d) it is found that = 1.0217 radians corresponding to = 58.5418 . Hence,
[0, 2 ] in agreement with Iyz = 117
ta3 < 0.
Finally, the moments of inertia in the principal axes coordinate system follow from Eq. (167), i.e.
3

Iy y 5.3423 ta ,
s 2 2
1 233 92 1 233 92 11 3
= + +4 ta = (e)
2 84 21 2 84 21 7
Iz z 1.8124 ta3 .

Clearly, Iy y is greater than any of Iyy or Izz , whereas Iy y is smaller than the bending moments of
inertia defined with respect to the original y- and z-axes.
1.4.3 Equations of equilibrium in principal axes coordinates

From now on it will be assumed that the (x, y, z)-coordinate system forms a principal axes coor-
dinate system with origin at the bending centre. In this case, the system of differential equations
(128) for a Timoshenko beam uncouples into three differential subsystems. Thus, the axial
deformation is governed by the equation

d dwx
EA + qx = 0, (169)
dx dx
whereas bending deformation in the y-direction is defined by the coupled equations

d dz dwy
EIz + GAz z + mz = 0, (170a)
dx dx dx

d dwy
GAy z + qy = 0, (170b)
dx dx
where the double index yy on the bending moment of inertia has been replaces by a single index
y in order to indicate that the principal-axes coordinates are utilised.

Similarly, the flexural deformations in the z-directions are determined by

d dy dwz
EIy GAz + y + my = 0, (171a)
dx dx dx

d dwz
GAz + y + qz = 0, (171b)
dx dx
where again the double index on the bending moment of inertia has been replaced by a single
index. As seen from Eqs. (170) and (171), {wy , z } and {wz , y } are still determined by
pairwise coupled ordinary differential equations of the second order.
For a Bernoulli-Euler beam, the system of ordinary differential equations (133) uncouples
completely into the following differential equations for the determination of wx , wy and wz :

d dwx
EA + qx = 0, (172a)
dx dx
d2 d2 wy

dmz
EIz qy + = 0, (172b)
dx2 dx2 dx
d2 d2 wz

dmy
2
EI y 2
qz = 0. (172c)
dx dx dx
Example 1.3 Plane, fixed Timoshenko beam with constant load per unit length
Figure A shows a plane Timoshenko beam of the length l with constant bending stiffness EIz and shear
stiffness GAy . The beam is fixed at both end-sections and is loaded with a constant load qy and a constant
moment load mz . The displacement wy (x), the rotation z (x), the shear force Qy (x) and the bending
moment are to be determined.
qy mz
z x
EIzz , GAy
y
Figure A Fixed beam with constant load per unit length.
The differential equations for determination of wy (x) and z (x) follow from Eq. (170). Thus,
d 2 z

dwy d dwy
EIz + GA y z + m z = 0, GA y z + qy = 0. (a)
dx2 dx dx dx
According to Eq. (135), the boundary conditions are:
wy (0) = wy (l) = 0, z (0) = z (l) = 0. (b)
(continued)

Integration of the second equation in Eq. (a) provides:

dwy
Qy = GAy z = qy x + c1 (c)
dx
Then, the following solution is obtained for z (x) from the first equation in Eq. (a):
d 2 z 1 1
EIz = qy x (c1 + mz ) EIz z (x) = qy x3 (c1 + mz )x2 + c2 x + c3 . (d)
dx2 6 2
Further, the boundary conditions z (0) = z (l) = 0 provide
1 1
c3 = 0, c2 = qy l2 + (c1 + mz )l. (e)
6 2
Hence, the following reduced form is obtained for z (x):
1
qy (x3 xl2 ) 3(c1 + mz )(x2 xl) .

z (x) = (f)
6EIz
Next, Eq. (f) is inserted into Eq. (c) which is subsequently integrated with respect to x, leading to the
following solution for wy (x):
dwy GAy
qy (x3 xl2 ) 3(c1 + mz )x2 xl)

GAy = qy x + c1 +
dx 6EIz

1 GAy 1 1
GAy wy (x) = qy x2 + c1 x + c4 + qy (x4 2x2 l2 ) (c1 + mz )(2x3 3x2 l) . (g)
2 6EIz 4 2
The boundary conditions wy (0) = wy (l) = 0 provide the integration constants

1 1
c4 = 0, c1 = qy l mz , (h)
2 y + 1
where
EIz
y = 12 (i)
GAy l2
Then, Eq. (a) and Eq. (f) provide the following solutions:
qy qy mz x mz y
wy (x) = (l x)x + (l x)2 x2 (2x3 3x2 l)
2GAy 24EIz GAy y + 1 12EIz y + 1
qy mz y
(l x)2 x2 + y l2 (l x)x (2x3 3x2 l + xl2 ),

wy (x) = (j)
24EIz 12EIz y + 1
qy mz y
z (x) = (2x3 3x2 l + xl2 ) + (l x)x. (k)
12EIz 2EIz y + 1
The non-dimensional parameter y is a measure of the influence of the shear deformations. For a
1 2
rectangular cross-section with the height h we have Iz = 12 h A and Ay = 56 A. Then y becomes
72 h2 E
y = . (l)
5 l2 G
Hence, shear deformations are primarily of importance for short and high beams. On the other hand, for
long beams with a small height of the cross-section, shear deformations are of little importance, i.e. only
the bending deformation is significant. (continued)

For Bernoulli-Euler beams we have xy = 0, corresponding to GAy = , cf. Eqs. (116) and (127).
Then, y = 0 and Eqs. (j) and (k) reduce to
qy
wy (x) = (l x)2 x2 , (m)
24EIz
qy
z (x) = (2x3 3x3 l + xl). (n)
12EIz
It is remarkable that the distributed moment load mz does not induce any displacements or rotations in
the considered beam with Bernoulli-Euler kinematics.
The shear force Qy (x) and bending moment Mz (x) follow from Eq. (c), Eq. (h) and Eq. (k), respec-
tively, i.e.
1 1
Qy (x) = qy x + c1 = qy (l 2x) mz , (o)
2 y + 1
dz qy mz y
Mz (x) = EIz = (6x2 6xl + l2 ) + (l 2x). (p)
dx 12 2 y + 1
For a Bernoulli-Euler beam these results reduce to
1
Qy (x) = qy (l 2x) mz , (q)
2
qy
Mz (x) = (6x2 6xl + l2 ). (r)
12
The constant moment load mz only induces a constant shear force of magnitude mz , whereas Mz (x)
is not affected by this load. Especially, for x = 0 and x = l, Eq. (o) and Eq. (p) provide
1 1 1 y
Qy (0) = qy l mz , My (0) = qy l + mz l, (s)
2 12 2 y + 1
1 1 1 y
Qy ( l) = qy l mz , My ( l) = qy l2 mz l. (t)
2 12 2 y + 1
The displacement at the midpoint x = l/2 follows from Eq. (k):
qy l4
wy (l/2) = (1 + 4y ). (u)
384EIz
The first and second terms within the parenthesis specify the contributions from bending and shear, re-
spectively. Again the parameter y reveals itself as a measure of the relative contribution from shear
deformations.
1.5 Normal stresses in beams

For at beam without warping, the normal stress xx (x, y, z) in terms of the generalised strains
follows from Eqs. (117), (120) and (124):
xx = E( z y + y z). (173)
In the principal axes coordinate system, where Sy = Sz = Iyz = 0, the generalised strains
{, y , z } are related to the conjugated generalised stresses {N, My , Mz } as determined from

1.6 The principle of virtual forces 29
Eqs. (145a), (145b), (146) and (147) for Sy = Sz = Iyz = 0, i.e.
N My Mz
= , y = , z = . (174)
EA EIy EIz
Insertion of Eq. (174) into Eqs. (119) and (124) provides the result for the axial stress in
terms of the generalised stresses,
N Mz My
xx = y+ z. (175)
A Iz Iy
Equation (175) is due to Navier, and is therefore referred to as Naviers formula. It should be
noticed that Eq. (175) presumes that the stresses are formulated in a principal axes coordinate
system, so Iy and Iz indicate the principal moments of inertia. The relation is valid for both
Timoshenko and Bernoulli-Euler beams. This is so because only the relation (115), but not
the relation (118) has been utilised. Hence, Eq. (175) is based on the assumption that plane
cross-sections remain plane, but not that they remain orthogonal to the beam axis.
The so-called zero line specifies the line in the (y, z)-plane on which x = 0. The analytical
expression for the zero line becomes
N Mz My
y+ z = 0. (176)
A Iz Iy
It is finally noted that warping introduces displacements in the axial direction in addition to
those provided by bending. However, if the torsion is homogeneous, these displacements will
not introduce any normal strains and therefore no normal stresses. Hence, Naviers formula is
also valid in the case of St. Venant torsion, but in the case of Vlasov torsion, or inhomogeneous
torsion, additional terms must be included in Eq. (175).
1.6 The principle of virtual forces

In this section the principle of virtual forces is derived for a plane Timoshenko beam of the length
l. The deformation of the beam is taking place in the (x, y)-plane. In the referential state, the
left end-section is placed at the origin of the coordinate system and the x-axis is placed along the
bending centres of the cross-sections, see Fig. 115.
The principle of virtual forces is the dual to the principle of virtual displacements. In the
principle of virtual displacements the actual sectional forces and sectional moments are assumed
to be in equilibrium with the loads and the reaction forces applied at the end sections. The virtual
displacements and rotations are considered as arbitrary increments to the actual displacements
and they only need to fulfil homogeneous kinematic boundary conditions, so that the combined
field made up by the actual and the virtual fields always fulfils the actual non-homogeneous
boundary conditions as given by Eq. (135). Further, the generalised virtual strains defining the
internal virtual work must be derived from the virtual displacement and rotation fields.
In contrast, the principle of virtual forces presumes that the displacements and rotations of the
beam are fulfilling the kinematic boundary conditions, and that the generalised internal strains
are compatible to these fields. The actual loads on the beam are superimposed with the virtual
incremental loads per unit length qx and qy , the virtual moment load per unit length mz ,

qx qy
z mz
Mz,1
N N2
N1 x
Mz Mz,2
Qy Qy,2
Qy,1
Figure 115 Virtual internal and external forces.
the virtual reaction forces Nj and Qy,j along the x- and y-directions, and the virtual reaction
moments Mz in the z-directions, where j = 1 and j = 2 indicate the left and right end-sections,
respectively. Due to the load increments, the internal section forces and section moment achieve
increments N , Qy and Mz , see Fig. 115. These variational fields are assumed to be in
equilibrium with the variational load fields qx , qy and mz , and to comply with the variations
Nj , Qy,j and Mz,j of the reaction forces and reaction moments. In what follows, N , Qy
and Mz will be referred to as the virtual internal forces, whereas qx , qy , mz , Nj , Qy,j
and Mz,j are called the virtual external forces.
The starting point is taken in the kinematical conditions provided by Eqs. (112) and (117)
and rewritten in the form
dwx dwy dz
= 0, z xz = 0, z = 0. (177)
dx dx dx
The virtual internal forces are related to the virtual external loads per unit length qx , qy and
mz via the following equations of equilibrium, cf. Eqs. (14a) and (14b):
d(N ) d(Qy ) d(Mz )

+ qx = 0, + qy = 0, + Qy + mz = 0. (178)
dx dx dx
The first equation in Eq. (178) is multiplied with N (x), the second equation is multiplied
with Qy (x), and the third equation with M (z). Next, the equations are integrated from x = 0
to x = l, and the three resulting equations are added, leading to the identity
Z l
dwx dwy dz
N + Qy z xy + Mz z dx = 0 (179)
0 dx dx dx
Integration by parts is carried out on the first terms within the innermost parentheses, leading to
l Z l
d(N ) d(Qy ) d(Mz )
N wx + Qy wy + Mz Qz wx + wy + z Qy z dx
0 0 dx dx dx
Z l
= N + Qy xy + Mz z dx. (180)
0

1.6 The principle of virtual forces 31
Upon utilisation of Eq. (178), this is reduced to
h il Z l
N wx + Qy wy + Mz z + qx wx + qy wy + mz z dx
0 0
Z l
= N + Qy xy + Mz z dx. (181)
0
The generalised strains on the right-hand side of Eq. (181) are now expressed in mechanical
quantities by means of Eq. (174). Further, N , Qy and Mz fulfil the following boundary
conditions at x = 0 and x = l, cf. Fig. 115,
N (0) = N1 , Qy (0) = Qy,1 , Mz (0) = Mz,1 , (182a)
N ( l) = N2 , Qy ( l) = Qy,2 , Mz ( l) = Mz,2 . (182b)
Equation (181) then obtains the following final form:
2
X Z l
Nj wx,j + Qy,j wy,j + Mz,j z,j + qx wx + qy wy + mz z dx
j=1 0
Z l
N N Qy Qy Mz Mz
= + + dx, (183)
0 EA GAy EIz
where wx,j , wy,j and z,j denote the displacements in the x- and y-directions and the rotation
in the z-direction at the end-sections, respectively. Equation (183) represents the principle of
virtual forces. The left- and right-hand sides represent the external and internal virtual work,
respectively.
The use of Eq. (183) in determining the displacements and rotations of a Timoshenko beam
is demonstrated in Examples 1.4 and 1.5 below. Furthermore, the principle of virtual forces may
be used to derive a stiffness matrix for a Timoshenko beam element as shown later.
Example 1.4 End-displacement of cantilevered beam loaded with a force at the free end
Figure A shows a plane Timoshenko beam of the length l with constant axial stiffness EA, shear stiffness
GAy and bending stiffness EIz . The beam is fixed at the left end-section and free at the right end-section,
where it is loaded with a concentrated force Qy,2 in the y-direction. The displacement wy,2 at the free
end is searched.
The principle of virtual forces Eq. (183) is applied with the following external virtual loads: qx =
qy = mz = 0, N1 = Qy,1 = Mz,1 = N2 = Mz,2 = 0 and Qy,2 = 1. Further N (x) = 0.
Then, Eq. (183) reduces to
Z l
Qy Qy Mz Mz
1 wy,2 = + dx. (a)
0 GAy EIz
(continued)

Qy,2 Qy,2 = 1
EA, GAy , EIz
z x z x
l l
y y
Mz Mz
Qy,2 l
Qy Qy
Qy,2 1
Figure A Fixed plane Timoshenko beam loaded with a concentrated force at the free end: Actual force and section
forces (left) and virtual force and section forces (right).
The variation of the bending moment Mz (x) and the shear force Qy (x) from the actual load Qy,2 has
been shown in Fig. A on the left. The corresponding variational moment field Mz (x) and shear force
Qy (x) from Qy,2 = 1 are shown in Fig. A on the right. Insertion of these distributions in Eq. (b)
provides the solution
Qy,2 l 1 Qy,2 l3 1 Qy,2 l3

wy,2 = + = (4 + y ) , (b)
GAy 3 EIz 12 EIz
where y is given by Eq. (i) in Example 1.3. The deformation contributions from shear and bending
are additive. This is a consequence of the additive nature of the flexibilities indicated by Eq. (183)
in contribution to the fact that the beam is statically determinate, which provides the fields Mz (x) and
Qy (x) as well as Mz (x) and Qy (x) directly.
Example 1.5 End-deformations of fixed beam loaded with a moment at the free end
The beam described in Example 1.4 is considered again. However, now the free end is loaded with a
concentrated moment Mz,2 . The displacement wy,2 and the rotation z,2 of the end-section is to be
found.
At the determination of wy,2 from Mz,2 , the principle of virtual forces given by Eq. (183) is again
applied with Qy,2 = 1 and all other external variational loads equal to zero, leading to Eq. (a) in
Example 1.4. However, Mz (x) and Qy (x) are now caused by Mz,2 , and are given as shown in Fig. A,
whereas Mz (x) and Qy (x) are as shown in Fig. A of Example 1.4. Then, wy,2 becomes
Z l
Mz Mz 1 My,2 l2
wy,2 = dx = . (a)
0 EIz 2 EIz
(continued)

1.7 Elastic beam elements 33
EA, GAy , EIz

z x z x
Mz,2 Mz,2 = 1
l l
y y
Mz Mz
Mz,2 1
Qy = 0 Qy = 0
Figure A Fixed plane Timoshenko beam loaded with a moment at the free end: Actual moment and section forces
(left) and virtual moment and section forces (right).
At the determination of z,2 the principle of virtual forces Eq. (183) is applied with the following
external virtual loads qx = qy = mz = 0, N1 = Qy,1 = Mz,1 = N2 = Qy,2 = 0 and
Mz,2 = 1. Then, Eq. (183) reduces to
Z l
Qy Qy Mz Mz
1 z,2 = + dx. (b)
0 GAy EIz
The variation of Qy (x) and Mz (x) from My,2 has been shown in Fig. A on the left, and the variation of
Qy (x) and Mz (x) from Mz,2 = 1 is shown in Fig. A on the right. Then z,2 becomes
Z l
1 Mz,2 Mz,2 l
z,2 = dx = . (c)
0 EI z EIz
In the present load case, the shear force is given as Qy (x) = 0. Consequently it will not induce any
contributions in Eq. (a) and Eq. (c).
1.7 Elastic beam elements

When frame structures consisting of multiple beams are to be analysed, the establishment of
analytical solutions is not straightforward and instead a numerical solution must be carried out.
For this purpose, a discretization of the frame structure into a number of so-called beam ele-
ments is necessary, eventually leading to a finite-element model. The aim of the present section
is not to provide a full introduction to the finite-element method for the analysis of frame struc-
tures, e.g. tower blocks with a steel frame as the load-carrying structure. However, a formula-
tion is given for a single beam element to be applied in such analyses. Both the Timoshenko
and Bernoulli-Euler beam theories are discussed in this context, and plane as well as three-
dimensional beams are touched upon.

1.7.1 A plane Timoshenko beam element

Firstly, the stiffness matrix and element load vector is derived for a plane Timoshenko beam ele-
ment with constant axial stiffness EA, shear stiffness GAy and bending stiffness EIz , cf. Fig. 1
16. The stiffness relation is described in an (x, y)-coordinate system with origin at the left end-
section and the x-axis along the bending centres.
z
EA, GAy , EIz
Mz,1 , z,1
N2 , wx,2
1 2 x
N1 , wx,1
Mz,2 , z,2
Qy,1 , wy,1 Qy,2 , wy,2
Figure 116 Plane Timoshenko beam element with definition of degrees of freedom and nodal reaction forces.
At the end-nodes, nodal reaction forces Nj and Qy,j are acting along the x- and y-directions,
respectively, and reaction moments Mz,j are applied around the z-axis. Here, j = 1 and j = 2
stand for the left-end and right-end nodes of the beam element, respectively, and the reaction
forces and moments are in equilibrium with the remaining external loads on the element for
arbitrary deformations of the beam.
The element has 6 degrees of freedom defining the displacements and rotations of the end-
sections, cf. Fig. 116. These are organised in the column vector

we1 T
we = = wx,1 wy,1 z,1 wx,2 wy,2 z,2 (184)
we 2
The sub-vector wej defines the degrees of freedom related to element node j.
Similarly, the reaction forces Nj , Qy,j and Mz,j , j = 1, 2, at the end-sections, work conju-
gated to wx,j , wy,j and z,j , are stored in the column vector

re 1 T
re = = N1 Qy,1 Mz,1 N2 Qy,2 Mz,2 (185)
re2
py
px mz
Mz,1
N2
x
N1
z Mz,2
Qy,1 Qy,2
y
Figure 117 External loads and reaction forces from external loads on a plane beam element.
The equilibrium of the beam element relating the nodal reaction forces to the degrees of
freedom of the element may be derived by the principle of virtual displacements as demonstrated

in a subsequent paper. The resulting equilibrium equations on matrix form may be written on the
form
re = Ke we + fe . (186)
The vector fe in Eq. (186) represents the nodal reaction forces from the external element loads
when we = 0, i.e. when the beam is fixed at both ends as shown in Fig. 117. We shall merely
consider constant element loads qx and qy per unit length in the x- and y-directions, and a con-
stant moment load per unit length mz in the z-direction, see Fig. 117. The reaction forces and
reaction moments follow from Eqs. (s) and (t) in Example 1.3:
21 qx l

21 qy l + mz
1 q l2 1 y m l

12 y 2 y +1 z
fe = . (187)
21 qx l
1
2 qy l mz

1 2 1 y
q
12 y l m
2 y +1 z l
The matrix Ke in Eq. (186) denotes the stiffness matrix in the local (x, y, z)-coordinate system.
Let wi denote the ith component of we . Then, the ith column in Ke represents the nodal reaction
forces for fe = 0, and with wi = 1 and wj = 0, j 6= i. These forces are obtained following
the derivations in Example 1.3 from Eq. (a) to Eq. (t) with qy = mz = 0 and with the boundary
condition in Eq. (b) replaced by the indicated conditions. Because of the symmetry of the prob-
lem, only two such analyses need to be performed. Still, this is a rather tedious approach. Partly
because of this, and partly in order to demonstrate an alternative approach, the stiffness matrix
will be derived based on the principle of virtual forces.
Undeformed state Undeformed state
x x
z
wy
wx
y y
(a) (b)
Figure 118 Rigid-body modes of a plane beam element: (a) Translation and (b) rotation.
The beam element has 6 degrees of freedom, by which a total of 6 linear independent modes
of deformation may be defined. These consist of 3 linear independent rigid body modes and
3 linear independent elastic modes. The rigid modes may be chosen as a translation in the
x-direction, a translation in the y-direction and a rotation around the z-direction as shown in
Fig. 118. Any rigid body motion of the beam element may be obtained as a linear combination
of these component modes of deformation. Obviously, the rigid body motions do not introduce
stresses in the beam. Hence, the axial force N , the shear force Qy and the bending moment Mz
are all zero during such motions.
Since, axial elongations are uncoupled from bending deformations, the elastic elongation
mode is uniquely defined as shown in Fig. 119a. The two bending deformation modes may be

N0 N0
x
u0 u0
2 2
y N
N0
(a)
2Ma
l
Ms Ms Ma
a
x x
s s a
2Ma Ma
l
y y Qy
Qy = 0 2Ma
l
Mz Ma
Ms Ma
Mz
(b) (c)
Figure 119 Elastic modes and related section forces in a plane beam element: (a) Axial elongation; (b) symmetric
bending and (c) antisymmetric bending.
chosen in arbitrarily many ways. Typically, these are chosen by prescribing an angle of rotation
at the other end-section. Following an idea by Krenk (2001), a more convenient formulation may
be obtained by choice of two bending modes symmetric and anti-symmetric around the mid-point
of the beam element as shown in Figs. 119b and 119c. It should be noticed that these modes
also apply if the material properties of the beam are not symmetrical around the mid-point.
The axial elongation and conjugated axial force related to the axial elongation mode are de-
noted u0 and N0 , respectively. The symmetric and anti-symmetric bending modes are described
by the end-section rotations s and a defined in Figs. 119b and 119c, respectively. The con-
jugated moments are denoted Ms and Ma , respectively. The related distributions of the shear
force Qy (x) and the bending moment Mz (x) are shown in Figs. 119b and 119c.
The shear force is equal to Qy = 0 in symmetric bending, because the bending moment is
constant. Then, no shear deformations are related to this mode. In contrast, a constant shear
force appears in the anti-symmetric bending mode. Hence, the deformations occurring in this
mode are affected by bending as well as shear contributions.
At first the constitutive relations between the deformation measures and the conjugated gen-
eralised strains for the indicated elastic modes are found by means of the principle of virtual
forces Eq. (183). In all cases, the beam element is unloaded, so qx = qy = mz = 0. For the
axial elongation mode N = N0 , Qy = Mz = 0, and N = 1, N1 = 1, N2 = 1. Further,

wx,1 = 12 u0 , wx,2 = 12 u0 , wy,j = z,j = 0. Then, Eq. (183) reduces to

Z l Z l
1 1 1 N0 dx l
(1) ( u0 ) + 1 u0 = dx u0 = N0 = N0 . (188)
2 2 0 EA 0 EA EA
The last statement holds for a beam element with constant axial stiffness EA. If EA varies, the
integral in the middlemost statement must be evaluated analytically or numerically.
For the symmetric bending mode N = Qz = 0, Mz = Ms , Mz = 1, Mz,1 = 1, and
Mz,2 = 1. Further, z,1 = s , z,2 = s and wx,j = wy,j = 0. Then, Eq. (183) provides
Z l Z l
(1)(Ms ) dx l
1 s + (1) (s ) = dx 2s = Ms = Ms . (189)
0 EIz 0 EIz EI z
Again, the last statement only applies for a homogeneous beam, whereas the middlemost state-
ment applies for any variation of the bending stiffness EIz along the beam.
For the anti-symmetric bending mode N = 0, Qy = 2 Ml a , Mz (x) = (1 + 2x/l)Ma , and
Qy = 2Mz (x)/l = (1 + 2x/l). Further, z,1 = a , z,2 = a , wx,j = wy,j = 0. Then,
Eq. (183) provides
Z l 2 2Ma !
l l (1 + 2 xl )(1 + 2 xl )Ma
1 a + 1 a = + dx
0 GAy EIz
!
4 l dx (1 + 2 xl )2
Z Z l
2a = Ma 2 + dx
l 0 GAy 0 EIz

1 1 l 1 l EIz
2a = Ma 4l + = (1 + y )Ma , y = 12 . (190)
GAy l2 3 EIy 3 EIy GAy l2
As discussed in Example 1.3, the non-dimensional parameter y defines the contribution of shear
flexibility relatively to the bending flexibility.
The flexibility relations provided by Eqs. (188), (189) and (190) may be written in the
following equivalent stiffness matrix formulation:
r0 = K0 w0 , (191)
where
EA
0 0

N0 u0 l
EIz
r0 = Ms , w0 = 2s , K0 = 0 l 0 . (192)
3 EIz
Ma 2a 0 0 1+y l
The nodal reaction forces re and r0 in Eq. (185) and Eq. (192) are related via the transfor-
mation
re = Sr0 = SK0 w0 , (193)
where

1 0 0

0 0 2/l

0 1 1
S= . (194)

1 0 0

0 0 2/l
0 1 1

Similarly, the elastic deformation measures stored in w0 can be expressed by the degrees of
freedom of the element stored in we as follows (see Fig. 120):
u0 = wx,2 wx,1 , (195)
z,1 = a + s + 1l (wy,2 wy,1 ) 2a = z,1 + z,2 2l (wy,2 wy,1 )

(196)
z,2 = a s + 1l (wy,2 wy,1 ) 2s = z,1 z,2 .

wy,1
wy,2
a + s
z,1
z,2
a s
Figure 120 Connection between elastic deformation measures and element degrees of freedom.
Equations (195) and (196) may be rewritten in the common matrix form
w0 = ST we , (197)
where S is given by Eq. (194). Insertion of Eq. (197) into Eq. (193) provides upon compari-
son with Eq. (186):
re = SK0 ST we Ke = SK0 ST . (198)
Insertion of K0 and S as given by Eq. (192) and Eq. (194) provides the following explicit
solution for Ke :
Al2 0 0 Al2 0 0

12 6 12 6
0 1+y Iz 1+y Iz l 0 1+ y
Iz 1+y Iz l

4+y 2y

6 2 6 2
E 0 1+y Iz l 1+y Iz l 0 1+y Iz l 1+y Iz l

Ke = 3 . (199)

l Al2 0 0 Al2 0 0
12 6 12 6
0 1+ Iz 1+ Iz l 0 1+y Iz 1+y Iz l

y y
6 2y 2 6 4+y 2
0 1+y Iz l 1+y Iz l 0 1+ y
Iz l 1+y Iz l
The corresponding result for a plane Bernoulli-Euler beam element is obtained simply by
setting y = 0. The equivalent element relations for a three-dimensional beam formulated in a
(x, y, x) principal axes coordinate system are given in the next section.

Example 1.6 Deformations of a plane Timoshenko beam structure

Figure A shows a plane beam structure ABC consisting of two Timoshenko beam elements AB and
BC of the lengths l and l/2, respectively. The loads on the structure and the resulting displacements are
described in the indicated (x, y)-coordinate system. The shear stiffness GAy and the bending stiffness
EIz are constant and the same in both beam elements. The structure is fixed at point A, free at point C,
and simply supported at point B. Both beam elements are loaded with a constant load per unit length qy .
Additionally, beam BC is loaded with a concentrated load Py = qy l at the free end C.
qy Py = q y l
z x
A 1 B 2 C
GAy , EIz
l l/2
y
Figure A Plane Timoshenko beam structure consisting of two beam elements.
We want to determine the displacement of point C in the y-direction, and the reaction forces and
moments at the support points A and B. The calculations are performed with the shear stiffness given as
EIz
GAy = 120 . (a)
l2
We shall refer to the beam elements AB and BC with the index e = 1 and e = 2, respectively. With
reference to Eq. (i) in Example 1.3, the shear flexibility parameters for the two beam elements become:
12EIz 12EIz
y1 = = 0.1, y2 = = 0.4. (b)
GAy l2 GAy (l/2)2
1 , M 1 2 2 3
1 2
w1 w2 w2 w3
Q1 Q2
Figure B Global degrees of freedom and reaction forces in the plane Timoshenko beam structure.
Since, the axial deformations are disregarded, each element has 4 degrees of freedom out of which two
are common, namely the displacement and rotation at point B. The related global degrees of freedom
have been defined in Fig. B. The element stiffness matrices follow from Eq. (199) and Eq. (b):

12 6l 12 6l 12 3l 12 3l
2 2 2 2
EIz 6l 4.1l 6l 1.9l , K2 = 8EIz 3l
1.1l 3l 0.4l
K1 = . (c)
1.1l3 12 6l 12 6l 1.4l3 12 3l 12 3l
6l 1.9l2 6l 4.1l2 3l 0.4l2 3l 1.1l2
(continued)

The corresponding element loads become, cf. Eq. (187):
12 qy l 21 qy 2l

6 12
1 2 1 qy ( l )2
qy l
12
1 l
12 2 = 1 qy l
l
1 qy l = 12 qy l 6 , f2 = 1 qy l + Py
f1 =
60 .
(d)
2 2 2
48
1
q l2
12 y
l 1
q ( l )2
12 y 2
l
The global equilibrium equation, made up of contributions from both elements, has the structure
r = Kw + f , (e)
where

p q Q1 w1 p q 24
| r1 | M1 1 | f1 | 4l

| p | q = Q2
w2 | p | q 1
= qy l 36 ,

r=
x |
, w= , f = (f)
y |

0

2

x |
y | 48 3l

| r2 | 0 w3 | f2 | 60
x y 0 3 x y l

p q
| K1 |

| p | q
K=

x | y |
| K2 |
x y
12 6 12 6

1.1 1.1
l 1.1 1.1
l 0 0
6 4.1 2 6 1.9 2

1.1
l 1.1
l 1.1 l 1.1
l 0 0
12 6 12 96 6 24 96 24

EIz 1.1 1.1 l 1.1
+ 1.4
1.1 + 1.4 l 1.4 1.4
l
= 2 6 1.9 2 6 24 4.1 8.8
2 24 3.2 2
. (g)
l
1.1
l 1.1
l 1.1 + 1.4 l 1.1
+ 1.4
l 1.4 l 1.4
l

96 24 96 24
0 0 1.4 1.4 l 1.4
1.4 l
24 3.2 2 24 8.8 2
0 0 1.4
l 1.4
l 1.4 l 1.4
l
The details in the derivation of the matrix equation Eq. (e), including the formation of the column vectors
r and f and the global stiffness matrix K by adding contribution from element components, will be
explained in a later chapter.
The displacement degrees of freedom w1 and w2 at the nodes A and C are both equal to zero. Similarly,
the rotation 1 at the fixed support at A is zero. When these values are introduced in w0 , Eq. (e) provides
the values of the reaction components Q1 , M1 and Q2 if the remaining unconstrained degrees of freedom
2 , w3 and 3 are inserted. These are determined from the corresponding equations in Eq. (e) using
w1 = 1 = w2 = 0, leading to
+ 8.8 )l2 1.4 l2

4.1 24
l 3.2

( 2 3l 0
EI 1.1 241.4 96
1.4
24 1
1.4 l 1.4
1.4 l w3 qy l 60 = 0
l3 3.2 2 24 8.8 2 48
1.4
l 1.4
l 1.4
l 3 l 0

2 l 4 0.1453
w3 = qy l 0.1274 . (h)
EIz
3 l 0.2912
Insertion of Eq. (h) along with w1 = 1 = w2 = 0 into the remaining equation (e) provides the
following solution for the reaction components Q1 , M1 and Q2 : (continued)

6

Q1 l 0 0 4 0.1453/l 24
M1 = EI z
1.1
1.9 2 qy l 1

1.1
l 0 0 0.1274 qy l 4l
l3 6 24 96 24 EIz 48
Q2 ( 1.1 + 1.4 )l 1.4 1.4
l 0.2912/l 36

Q1 l 0.2927
M1 = qy l2 0.1677 . (i)
O2 l 2.7927
In case of Bernoulli-Euler kinematics, corresponding to y1 = y2 = 0, the corresponding solutions

become:

1 l 4 0.1354
w3 = qy l 0.1172 (j)
EI
2 l 0.2812

Q1 l 0.3125
M1 = qy l2 0.1875 (k)
Q2 l 2.8125
Comparison of Eqs. (h) and (j) reveals that the displacement w3 as well as the rotations 1 and 3
are increased by the shear flexibility. This is so because bending and shear deformations in general are
coupled for statical indeterminate structures. At the same time, a comparison of Eqs. (i) and (k) shows
that Bernoulli-Euler beam kinematics lead to higher stresses than Timoshenko beam theory, which is due
to the fact that the shear stiffness in the Bernoulli-Euler beam is infinite.
1.7.2 A three-dimensional Timoshenko beam element

The formulation of the beam-element stiffness matrix is now extended to three dimensions. This
involves flexural displacements in two directions, axial displacements and, in addition to this,
twist of the beam. It is assumed that the beam element is straight, of the length l and with
constant cross-section. The element relation is described in a principal axes (x, y, z)-coordinate
system with origin at the left end-section and the x-axis placed along the bending centres of the
cross-sections. Only St. Venant torsion is taken into consideration. The axial stiffness EA, the
shear stiffnesses GAy and GAz in the y- and z-directions, respectively, the torsional stiffness
GK, and the principal inertial bending stiffnesses EIy and EIz around the y- and z-axes are all
constant along the beam element.
The degrees of freedom of the element are made up by the 6 components wx,j , wy,j and wz,j ,
j = 1, 2, providing the displacements of the bending centres, and the 6 components x,j , y,j
and z,j defining the rotation of the end-sections. Again, j = 1 and j = 2 refer to the left and
right end-sections, respectively. The nodal reaction forces conjugated to the indicated degrees
of freedom consist of the axial forces Qy,j and Qz,j in the y- and z-directions, respectively,
the torsional moments Mx,j and the bending moment components My,j and Mz,j in the y- and
z-directions. Finally, the element loadings consist of constant loads per unit length {qx , qy , qz }
and constant moment loads per unit length {mx , my , mz } in the x-, y- and z-directions. No
concentrated element forces or moments are considered. The loads qy and qz as well as the
shear forces Qy,j and Qz,j are assumed to act through the so-called shear centre, leading to an
uncoupling of the flexural and torsional deformations. The definition of the shear centre and
further details about uncoupling of torsional and flexural displacements are given in Chapter 2.

A, GA
qx
mx
Mx,1 N1 N2 Mx,2
x
x,1 wx,1 wx,2 x,2
EA, GK
qy
mz
z
z,1 , Mz,1 z,2 , Mz,2 x

GAy , EIz
wy,1 , Qy,1 wy,2 , Qy,2

y
qz
my
y
y,1 , My,1 y,2 , My,2 x
GAz , EIy
wz,1 , Qz,1 wz,2 , Qz,2
z
Figure 121 Three-dimensional Timoshenko beam element with definition of degrees of freedom, nodal reaction forces,
element loads and sectional properties.
The element equilibrium equations may be expressed on the matrix form, cf. Eq. (186),
re = Ke we + fe (1100)
Here, re and we are 12-dimensional column vectors storing the reaction forces and the element
degrees of freedom, respectively, cf. Eqs. (184) and (185),

N1 wx,1
Qy,1 wy,1

Qz,1 wz,1

Mx,1 x,1

My,1 y,1

re 1 Mz,1 we 1 z,1
re = =
,
we = = . (1101)
re2 N2 we2 wx,2

Qy,2 wy,2

Qz,2 wz,2

Mx,2 x,2

My,2 y,2
Mz,2 z,2
Likewise, fe is 12-dimensional column vector storing the contributions to the reaction forces

from the element loads, given as, cf. Eq. (187),
12 qx l

1
2 qy l + mz
21 qz l my

12 mx l

1 2 1 z
12 qz l 2 1+z my l

y

1
12 qy l2 12 1+ mz l

fe 1

fe = = y
, (1102)

fe2 12 qx l
21 qy l mz

21 qz l + my

12 mx l

z
12 qz l 12 1+
1 2
my l

z

1 2 1 y
q
12 y l 2 1+y mz l
where y and z are given as, cf. Eq. (i) in Example 1.3,
EIz EIy
y = 12 , z = 12 . (1103)
GAy l2 GAz l2
Finally, Ke specifies the element stiffness given as, cf. Eq. (199),
EA
0 EA

l 0 0 0 0 l 0 0 0 0 0
0 z z z z
k11 0 0 0 k12 0 k13 0 0 0 k14

0 y y y y
0 k11 0 k12 0 0 0 k13 0 k14 0
GK
0 GK

0 0 0 0 0 0 0 0 0
l l
y y y y

0 0 k12 0 k22 0 0 0 k23 0 k24 0
z z z z

0 k12 0 0 0 k22 0 k23 0 0 0 k24
Ke =
EA EA
, (1104)
l 0 0 0 0 0 l 0 0 0 0 0

0 z z z z
k13 0 0 0 k23 0 k33 0 0 0 k34
y y y y

0 0 k13 0 k23 0 0 0 k33 0 k34 0

0 0 0 GK
l 0 0 0 0 0 GK
l 0 0
y y y y

0 0 k14 0 k24 0 0 0 k34 0 k44 0
z z z z
0 k14 0 0 0 k24 0 k34 0 0 0 k44
where
z z z z

k11 k12 k13 k14 12 6l 12 6l
z z z
k22 k23 k24 EIz (4 + y )l2 6l (2 y )l2
z =
z
, (1105)
k33 k34 (1 + y )l2 12 6l
z 2
k44 (4 + y )l
y y y y
k11 k12 k13 k14 12 6l 12 6l
y y y
k22 k23 k24 EIy (4 + z )l2 6l (2 z )l2
y =
y
. (1106)
k33 k34 (1 + z )l2 12 6l
y 2
k44 (4 + z )l

The element equilibrium relation provided by Eq. (1100) presumes that only St. Venant
torsion is taken into consideration, corresponding to the torsional equilibrium equations

Mx,1 GK 1 1 x,1 1 1
= mx l . (1107)
Mx,2 l 1 1 x,2 2 1
The torsional constant K is determined in the next chapter. It will be shown that the inclusion
dx,1 dx,2
of Vlasov torsion requires the introduction of two extra degrees of freedom dx and dx .
The conjugated generalised stresses are the so-called bimoments. Hence, a Timoshenko beam
element, where both St. Venant and Vlasov torsion are taken into consideration, is described by
a total of 14 degrees of freedom.
1.8 Summary
In this chapter, the basic theory of Timoshenko and Bernoulli-Euler beams in three-dimensional
space has been presented. Some of the main topics covered are summarised below.
Beams are one-dimensional structures that may carry loads in three dimensions including axial
forces, shear forces in two orthogonal directions and moments around three directions.
Bernoulli-Euler beam kinematics assume that cross-sections remain orthogonal to the beam
axis during deformation. Hence, no shear deformation occurs.
Timoshenko beam kinematics include shear flexibility, but still a cross-section remains plane
during deformation. Hence, shear strains and stresses are homogeneous over the beam height.
The bending centre of a beam cross-section is defined as the point of attack of an axial force
not producing a bending moment.
The principal axes of a beam cross-section are defined as the axes around which a bending
moment will neither produce an axial force nor flexural displacements in the other direction.
The principle of virtual forces can by applied to the analysis of deformations in a beam. In the
case of Timoshenko beam theory, both shear and bending deformation occurs, whereas only
bending deformation is present in a Bernoulli-Euler beam.
Plane beam elements have six degrees of freedom with three at either end, i.e. two displace-
ments and one in-plane rotation. In the general case, the rotations and the axial displacements
are coupled, but in a principle-axes description, they become uncoupled.
Spatial beam elements have 12 degrees of freedom, that is three displacements and three rota-
tions at either end. Generally, the displacements and rotations are coupled, but an uncoupling
can be achieved by a proper choice of coordinate system.
Thus, a detailed description has been given of the lateral and flexural deformations in a beam.
However, in this chapter only a brief introduction has been given to twist and torsion of a beam.
A thorough explanation and analysis of these phenomena will be the focus of the next chapter.

C HAPTER 2
Shear stresses in beams due to
torsion and bending
In this chapter, a theoretical explanation is given for the shear forces in beams stemming from
bending as well as torsion. In this regard, the coupling of torsion and bending is discussed, and
the so-called shear centre is introduced. The derivations are confined to homogeneous torsion, or
St. Venant torsion. Later, the strains and stresses provided in the case of non-homogeneous torsion,
or Vlasov torsion, will be dealt with.
2.1 Introduction
When the beam is exposed to the loads per unit length qy and qz , the beam will generally deform
with bending deformations {uy , z } and {uz , y }, respectively. Depending on the line of action
of these loads, the bending deformations will be combined with a torsional rotation x around
the x-axis as illustrated in Fig. 21a and Fig. 21c. Under certain conditions, the bending of the
beam is not associated with a torsional deformation. This happens if the loads per unit length
qy and qz , the reaction forces at the ends of the beam as well as the shear forces Qy and Qz are
acting through a special point S, known as the shear centre, as illustrated in Fig. 21b.
When this is the case, torsion is caused solely by the moment load mx per unit length, which
in part includes contributions from the translation of qy and qz to S. These torsional deformations
take place without bending deformation as illustrated in Fig. 21d. Hence, bending and torsion
can be analysed independently.
The position of the shear centre depends on the geometry of the cross-section and is generally
different from the position of the bending centre B. However, for double-symmetric cross-
sections, the positions of the bending and torsion centres will coincide.
The shear forces Qy and Qz as well as the torsional moment Mx bring about shear stresses
xy and xz on the beam section. In what follows these will be determined independently for
the two deformation mechanisms. Hence, qy and qz are presumed to be referred to the shear
centre. The shear stresses caused by the torsional moment Mx are statically equivalent to the
shear forces Qy = Qz = 0. The position of the shear centre has no influence on the distribution
of shear stresses in this case.
Likewise, in the decoupled bending problem, in which the cross-section is exposed to the
shear forces Qy and Qz , the position of the shear centre is determined from the requirement
that the resulting shear stresses are statically equivalent to Qy and Qz , and produce the torsional
moment Mx = 0 around S.
45
46 Chapter 2 Shear stresses in beams due to torsion and bending
Qy , qy Qy , qy
B B
z z
S x S x
(a) y (b) y
Mx , mx
Qy , qy
B B
z z
S x S x
(c) y (d) y
Figure 21 Coupled and uncoupled bending and torsion. Coupling exists in cases (a) and (c), whereas cases (b) and (d)
involve no coupling.
2.2 Homogeneous torsion (St. Venant torsion)

It is assumed that the torsional moment Mx and the incremental twist per unit length dx /dx and
the warping of the cross-sections remain unchanged along the beam. Then all cross-sections of
the beam are exposed to the same distribution of the shear stresses xy and xz . For this reason,
this case is referred to as homogeneous torsion. Since the solution of the problem was given
by St. Venant (ref.), the case is also called St. Venant torsion. Inhomogeneous torsion refers to
the case, where either Mx or the material properties vary along the beam. Then dx /dx or the
warping will vary as well.
Figure 22a shows a cross-section of a cylindrical beam of the length l. The cross-sectional
area is A. The curve along the outer periphery is denoted 0 . The cross-section may have a
number N of holes determined by the boundary curves j , j = 1, 2, . . . , N . At the boundary
curves arc-length coordinates s0 , s1 , . . . , sN are defined. The arc-length coordinate s0 along
0 is orientated in the anti clock-wise direction, whereas s1 , s2 , . . . , sN , related to the interior
boundaries 1 , 2 , . . . , N , are orientated in the clock-wise direction. The outward directed unit

2.2 Homogeneous torsion (St. Venant torsion) 47
n0
s0 n0
0 s0
N
s0
B zS
z sj
j Mx nj
yS nj
S
j
sj sj
(a) y (b) 0
Figure 22 Cross-section with holes: (a) Interior and exterior edges; (b) definition of local (x, nj , sj )-coordinate
systems.
vector on a point of the exterior or interior boundaries j is denoted nj , j = 0, 1, . . . , N . The

unit tangential vector to a boundary curve is denoted sj and is co-directional to the arc-length
coordinate sj , see Fig. 22b. Thus, a local (x, nj , sj )-coordinate system may be defined with
the base unit vectors {i, nj , sj }. The indicated orientation of the exterior and interior arc-length
coordinates sj , j = 0, 1, . . . , N , insures that the related (x, nj , sj )-coordinate system forms a
right-hand coordinate system.
The beam material is assumed to be homogeneous, isotropic linear elastic with the shear
modulus G. In homogeneous torsion, only shear stresses are present for which reason G is the
only needed elasticity constant.
2.2.1 Basic assumptions

For convenience the index x is omitted on the twist x (the rotation angle around the x-axis),
i.e. x . Figure 23 shows a differential beam element of the length dx. Both end-sections
of the element are exposed to the torsional moment Mx , so the element is automatically in
equilibrium. On the left and right end-sections the twists are and + d, respectively. The
increment d may be written as
d
d = dx. (21)
dx
Since Mx and the material properties are the same in all cross-sections, d/dx must be constant
along the beam. Further, the warping must be the same in all cross-sections, i.e. ux = ux (y, z).
This implies that the warping in homogeneous torsion does not induce normal strains, i.e.
ux
xx = = 0. (22)
x
In turn this means that the normal stress becomes xx = Exx = 0. Hence, only the shear
stresses xy and xz are present on a cross-section in homogeneous torsion.

Mx
y l
z
Mx
dx
Mx
+ d x
dx Mx
Figure 23 Beam and differential beam element subjected to homogeneous torsion.
The only deformation measure of the problem is the twist gradient d/dx. Then, due to the
linearity assumptions, the torsional moment Mx must depend linearly on d/dx. Further, xy
and xz (and hence Mx ) depend linearly on G. This implies the following relation:
d
Mx = GK . (23)
dx
The proportionality constant K with the dimension [unit of length]4 is denoted the torsional
constant. The determination of this constant is a part of the solution of the torsion problem.
2.2.2 Solution of the homogeneous torsion problem

As for all beam theories, the shape of the cross-section is assumed to be preserved during the
deformation. Then, the displacements in the (y, z)-plane are caused merely by the rotation
around the shear centre S. The warping displacements ux must also be linearly dependent on the
strain measure d/dx, corresponding to the last term in Eq. (113). This implies the displacement
components
d
ux = ux (y, z) = (y, z) , uy = (z zs ), uz = (y ys ). (24)
dx
Here (y, z) is the so-called warping function as discussed in Section 1.2.2, and its determination
constitutes the basic part of the solution of the homogeneous torsion problem.
Similarly to Eq. (114), it follows from Eq. (24) that the components of the strain tensor
become:
ux uy uz
xx = = 0, yy = = 0, zz = = 0, (25a)
x y z


1 uy uz
1
yz = + =
( + ) = 0, (25b)
2 z y2

1 ux uy 1 d
xy = + = (z zs ) , (25c)
2 y x 2 y dx

1 ux uz 1 d
xz = + = + (y ys ) . (25d)
2 z x 2 z dx
As seen only xy and xz are non-vanishing. Correspondingly, all components of the Cauchy
stress tensor become equal to zero, save the shear stresses xy and xz . These are given as

d
xy = xy (y, z) = 2Gxy = G (z zs ) , (26a)
y dx

d
xz = xz (y, z) = 2Gxz = G + (y ys ) . (26b)
z dx
Ignoring the volume loads, the equilibrium equations read
xx xy xz
+ + = 0, (27a)
x y z
xy yy yz
+ + = 0, (27b)
x y z
xz yz zz
+ + = 0. (27c)
x y z
With xx = yy = zz = yz = 0, and xy and xz only dependent on y and z, the two last
equations are identically fulfilled, and the first equation reduces to
xy xz
+ = 0. (28)
y z
Equation (28) may be formulated at a point on the boundary curve j , j = 0, 1, . . . , N , in
the related local (x, nj , sj )-coordinate system. Ignoring the index j, the non-trivial equilibrium
equation then reads
xn xs
+ = 0, (29)
n s
where xn and xs denote the shear stresses along the local n- and s-axes. According to Cauchys
boundary condition (ref.), xn can be expressed in terms of the shear stress components xy and
xz as
xn = xy ny + xz nz . (210)
Here ny and nz denote the components of the unit normal vector n along the y- and z-axes.
The symmetry of the stress tensor implies that xn = nx . Further, since the exterior and all
interior surfaces are free of surface traditions, corresponding to nx = 0, it follows that xn = 0
(see Fig. 24). Then, Eq. (210) reduces to
xy ny + xz nz = 0. (211)

xn nx = 0
n
x
Figure 24 Shear stress in the normal direction at an exterior or interior boundary.
Finally, insertion of Eq. (26) into Eq. (211) provides the following boundary condition formu-
lated in the warping function

ny + ny (z zs )ny + (y ys )nz = 0
y z
= (z zs )ny (y ys )nz , (212)

n
where /n denotes the partial derivative of in the direction of the outward directed unit
normal.
Equation (212) must be fulfilled at the exterior and all interior boundaries. The partial
differential for to be fulfilled in the interior A of the profile follows from insertion of Eq. (26)
into the equilibrium equation (28), leading to
2 2
+ = 0. (213)
y 2 z 2
If a solution to Eq. (213) with the boundary conditions (212) is obtained, the shear stresses are
subsequently determined from Eq. (210). Equation (213) is Laplaces differential equation,
and the boundary conditions Eq. (213) are classified as the so-called Neumann boundary con-
ditions. Notice that the solution to Eqs. (212) and (213) is not unique. Actually, if (y, z) is a
solution, then (y, z) + 0 will be a solution as well, where 0 is an arbitrary constant. Since the
shear stresses are determined by partial differentiation of the of the warping function, all these
solutions lead to the same stresses. The boundary value problem for the warping function has
been summarised in Box 2.1.

Box 2.1 Boundary value problem for the warping function

The differential equation, representing the non-trivial equation of equilibrium, reads:
2 2
2
+ = 0, (y, z) A. (214a)
y z 2
The Neumann boundary condition, representing the relevant Cauchy boundary condition, reads:

= (z zs )ny (y ys )nz , (y, z) 0 1 N . (214b)
n
An alternative formulation for the solution of the problem can be obtained by the introduction
of the so-called Prandtls stress function S with the defining properties
S S
xy = , xz = . (215)
z y
Upon insertion of Eq. (215) into Eq. (29), the equilibrium equation is seen to be identical
fulfilled, i.e.
xy xz 2S 2S
+ = 0. (216)
y z yz zy
From Eq. (26) follows
2 2
xy xz d d d
= 1 G +1 G = 2G . (217)
z y zy dx yz dx dx
Then, the differential equation for S is obtained by insertion of Eq. (215) on the left-hand side
of Eq. (217), leading to
2S 2S d
+ = 2G . (218)
y 2 z 2 dx
Equation (218) is a compatibility condition for S in order that the kinematical conditions (26)
are fulfilled.
The boundary condition for S follow upon insertion of Eq. (215) into Eq. (211), i.e.
S S
ny nz = 0 (219)
z y
The tangential unit vector is given as sT = [sy , sz ] = [nz , ny ], cf. Fig. 22b, where {ny , nz }
denotes the Cartesian components of the outward directed unit normal vector at any of the bound-
ary curves j , j = 0, 1, . . . , N , cf. Fig. 22b. Then, Eq. (219) may be written as
S S S
sy + sz = = 0. (220)
y z s
where S/s denotes the directional derivative of S in the direction of the tangential unit vector
s. Equation (220) implies that S is constant along the exterior and the interior boundary curves,
i.e.
S = Sj , j = 0, 1, . . . , N. (221)

Equation (218) is a Poisson differential equation (inhomogeneous Laplace equation), and the
boundary conditions Eq. (221) are classified as the so-called Dirichlet boundary conditions.
In principle, the solution to the indicated boundary value problem is unique. The problem is
that the constant values Sj of the stress function along the boundary curves are unknown. The
determination of these is a part of the problem. For profiles with interior holes this is only
possible by the introduction of additional geometric conditions. The boundary value problem
for the Prandtl stress function has been summarised in Box 2.2. The formulation in terms of
Prandtls stress function is especially useful in relation to homogeneous torsion of thin-walled
profiles and will be utilised in a number of examples below.
Box 2.2 Boundary value problem for the Prandtl stress function
The equilibrium equation is automatically fulfilled. The compatibility condition is represented by the
following differential equation:
2S 2S d
2
+ = 2G , (x, y) A. (222a)
y z 2 dx
The Dirichlet boundary condition, representing the Cauchy boundary condition, reads:
S = Sj , j = 0, 1, . . . , N, (x, y) 0 1 N . (222b)
The shear stresses xy and xz must be statically equivalent to the shear forces Qy = Qz =
0 and the torsional moment Mx . Application of Gausss theorem on the vector field vT =
[vy , vz ] = [0, S] provides,
Z N I N
0 S
Z X X I
Qy = xy dA = + dA = (0 dz S dy) = Sj dy, (223)
A A y z j=0 j j=0 j
where the circulation is taken anticlockwise along 0 and clockwise along j , j = 1, , N .

Further it has been used that Sj is constant along
H the boundary curve, and hence may be trans-
ferred outside the circulation integral. Now, j dy = 0. Hence, it follows that any solution
to the boundary value problem defined by Eqs. (222a) and (222b) automatically provides a
solution fulfilling Qy = 0. Using vT = [S, 0] it can in the same way be shown that
Z
S 0
Z
Qz = xz dA = + dA
A A y z
N I
X N
X I
Qz = (S dz 0 dy) = Sj dz = 0. (224)
j=0 j j=0 j
The torsional moment Mx must be statically equivalent to the moment of the shear stresses
around S, i.e.
Z Z
Mx = ((y yS )xz (z zS )xy ) dA = (yxz zxy )dA, (225)
A A

z B zS
z
y Mx
xz
dA yS
xy S
Figure 25 Static equivalence of torsional moment to shear stresses.
R R
where it has been used that A
xy dA = xz dA = 0. Next, insertion of Eq. (215) provides
Z Z
S S
Z
Mx = y +z dA = (yS) + (zS) dA + 2 SdA. (226)
A y z A y z A
The divergence theorem with vT = [yS, zS] provides

Z N I N
X X I
(yS) + (zS) dA = (ySdz zSdy) = Sj (ydz zdy). (227)
A y z j=0 j j=0 j
Let A0 and Aj denote the area inside the boundary curves 0 and j . Then, use of Eq. (14a) in
Eq. (227) provides
Z N
X
(yS) + (zS) dA = 2A0 S0 2Aj Sj . (228)
A y z j=1
The negative sign of the last term is because the circulation on the interior boundaries is taken
clockwise, cf. the discussion subsequent to Eq. (14a). Insertion of Eq. (228) gives the follow-
ing final result:
Z N
X
Mx = 2A0 S0 + 2 SdA + 2 Aj Sj . (229)
A j=1
The shear stresses remain unchanged if an arbitrary constant is added to S. Then, without re-
striction one can choose S0 = 0, which is assumed in what follows.
Let the domain of definition for S(y, z) be extended to the interior of the holes, where S(y, z)
is given the same value Sj as the boundary value along the holes. With S0 = 0, Eq. (229)
determines Mx as twice the volume below S(y, z). Further, with S(x, y) determined along with

0 , S0 = 0
B
z
1
y
S(0, z)
S1
z
Figure 26 Variation of Prandtls stress function over a cross-section with a hole.
the boundary values Sj , j = 1, 2, . . . , N , the torsional constant K can next be determined upon
comparison of Eqs. (23) and (229). This is illustrated by examples below.
It is remarkable that no reference is made to the position of the shear centre, neither in the
boundary value problem (222), for the Prandtl stress function, nor in the expression (225) for
the torsional moment. In contrast, the coordinates of the shear centre enter the boundary value
problem (214) for the warping function.
Defining the same homogeneous torsion problem, the warping function and Prandtls stress
function cannot be independent function. The relation follows from a comparison of Eqs. (26)
and (215):

S d
= (z zs ) G , (230a)
z y dx

S d
= + (y ys ) G , (230b)
y z dx
where it is recalled that x .
Example 2.1 Homogeneous torsion of infinitely long rectangular cross-section

Figure A shows an infinitely long rectangular cross-section with the thickness t exposed to a torsional
moment Mx . The torsional moment is carried by shear stresses uniformly distributed in the y-direction.
Then, S = S(z) is independent of y, and the boundary value problem (222) reduces to
d2 S d
= 2G , S(t/2) = S(t/2) = 0 (a)
dz 2 dx
with the solution
1 d
S(z) = (4z 2 t2 )G . (b)
4 dx
(continued)

The shear stresses follow from Eq. (215):

S d
xy = = 2zG , xz = 0. (c)
z dx
The shear stresses are linearly distributed in the thickness direction, and has been illustrated in Fig. A.
xs = xy
d
= G dx t dMx
dy z dy z
y y
Figure A Torsion of a infinitely long rectangular cross-section: Distribution of shear stresses (left); torsional moment
on differential cross-sectional segment (right).
Due to the independence of the shear stresses on y, the torsional problem can be analysed by merely
considering a differential cross-sectional segment of the length dy exposed to the torsional moment dM x,
see Fig. A. The increment dM x is related to the stress function by Eq. (228), i.e.
Z t/2 Z d/2
1 d 1 d
dMx = 2 S(z)dzdy = dyG , (4z 2 t2 )dz = t3 dyG . (d)
t/2 2 dx t/2 3 dx
At the same time dMx = GdKd/dx, where dK denotes the torsional constant related to the differential
segment. As seen from Eq. (d), this is given as
1 3
dK = d dy. (e)
3
The value dK of the torsional constant for the differential segment will be applied below.
Example 2.2 Homogeneous torsion of a solid ellipsoidal cross-section

Figure A shows an ellipsoidal cross-section without holes with semi-axes a and b, exposed to a torsional
moment Mx . At first, it is verified that the Prandtls stress-function of this problem is given as
a 2 b2
2
z2

y d
S(y, z) = 2 + 1 G . (a)
a + b2 a 2 b2 dx
(continued)

b
z
Mx a
y
Figure A Ellipsoidal cross-section.
The boundary curve 0 is described by the ellipsis
y2 z2
2
+ 2 = 1. (b)
a b
Hence, S(x, y) = 0 for (y, z) 0 , so the boundary condition (222b) is fulfilled by Eq. (a). The
Laplacian of s(y, z) becomes
2S 2S a 2 b2

2 2 d d
+ = 2 + 2 G = 2G . (c)
y 2 z 2 a + b2 a 2 b dx dx
Then, also the differential equation (222a) is fulfilled, from which is concluded that Eq. (a) is indeed the
solution to the homogeneous torsion problem.
From Eq. (229) follows that
a 2 b2
Z 2
z2 a 3 b3
Z
d y d
Mx = 2 SdA = 2 2 2
G 2
+ 2
1 dA = 2 G , (d)
A a +b dx A a b a + b2 dx
where the following result has been utilised:

Z 2
z2

y
2
+ 2 1 dA = ab. (e)
A a b 2
From Eq. (23) and Eq. (d) follows that the torsional constant for an ellipsoidal cross-section becomes:
a 3 b3
K= . (f)
a 2 + b2
Gd/dx follows from Eq. (d), i.e.
d 1 a 2 + b2
G = Mx . (g)
dx a 3 b3
Then, the shear stresses become, cf. Eqs. (215) and (a):
S 2z a2 b2 1 a 2 + b2 2 Mx
xy = = 2 2 Mx = z, (h)
z b a + b a 3 b3
2 ab3
S 2y a2 b2 1 a 2 + b2 2 Mx
xz = = Mx = y. (i)
y a 2 a 2 + b2 a 3 b3 a3 b
(continued)

The warping of the cross-section is given by Eq. (24). Due to the symmetry of the cross-section, it is
observed that ys = zs = 0, i.e. the shear centre coincides with the bending centre. Then, the warping
function is determined from, cf. Eq. (230):
1 S 2a2 b2 a 2
= +z = 2 2
z+z = 2 z, (j)
y Gd/dx z a +b b + a2
1 S 2b2 b2 a 2
= y = 2 2
yy = 2 y. (k)
z Gd/dx y a +b b + a2
The solution to Eqs. (j) and (k) is given as
b2 a 2
(y, z) = yz. (l)
b2 + a 2
It is left as an exercise to prove that Eq. (l) fulfils the boundary value problem (214).
The warping follows from Eq. (24), Eq. (g) and Eq. (l)
d b2 a 2 1 a2 + b2 Mx 1 b2 a2 Mx
ux (y, z) = (y, z) = 2 yz yz . (m)
dx a + b2 a 3 b3 G a 3 b3 G
The solution (l) has been chosen so that the warping from torsion provided by Eq. (m) is zero at the
bending centre. This will generally be presumed in what follows. Then, the displacement of the bending
centre in the x-direction is caused entirely by the axial force N . Finally, it is noted that, especially for a
circular profile with a = b, the warping vanishes everywhere on the profile.
2.2.3 Homogeneous torsion of open thin-walled cross-sections

Figure 27 shows an open cross-section of a cylindrical beam. An arc-length coordinate s is
defined along the midpoints of the profile wall, where s = 0 is chosen at one of the free ends,
and s = L at the other free. Further, L specifies the total length of the profile wall, and the wall
thickness at arc-length coordinate s is denoted t(s).
s
s=L
xs
Mx , mx
ds
n
S
t(s)
s s=0
Figure 27 Open thin-walled cross-section of cylindrical beam exposed to homogeneous torsion.

For thin-walled cross-sections it is assumed that t(s) L. Then, the profile can be consid-
ered as built-up of differential rectangles of the length ds, similar to those considered in Exam-
ple 2.2. Each has the torsional constant dK = 13 t3 (s)ds, cf. Eq. (e) in Example 2.2. Hence, the
torsional constant for the whole profile is given as
1
Z
K= t3 (s)ds, (233)
3 L
where the index L indicates that the line integral is extended over the whole length of the profile
measured along the profile wall.
The shear stresses are specified in a local (x, n, s)-coordinate system as shown in Fig. 27.
These coordinates follow from Eq. (c) upon replacing y with s and z with w. Then,
d
xs = 2nG , xn = 0. (234)
dx
Finally, using Gd/dx = Mx /K, the maximum shear stresses for n = t(s)/2 becomes
d Mx
=G t(s) = t(s). (235)
dx K
The computation of the shear stress in a U -profile is considered in the example below.
Example 2.3 Homogeneous torsion of a U -profile

Figure A shows a U -profile exposed to homogeneous torsion from the torsional moment Mx . With the
thin-wall approximation t a, the position of the bending centre is as shown in the figure. Further, the
cross-sectional area and the bending moments of inertia around the y- and z-axes become
64 3 26 3
A = 10at, Iy = a t, Iz = a t, (a)
15 3
where the single indices y and z indicate that the corresponding axes are principal axes. The cross-
sectional constants given in Eq. (a) have been calculated for later use. With reference to Eq. (233), the
torsional constant becomes:
2 1 34 3
K= 2a(2t)3 + 2at3 = a t. (b)
3 3 5
(continued)

2a
4 2
5
a
2t
a
Mx Mx
B
z
t
a
2t
y 2
2
Figure A Homogeneous torsion of U -profile: Dimensions (left) and shear stresses (right).
The distribution of shear stresses follows from Eq. (235). The maximum shear becomes and 2 ,
respectively, where is given as
Mx 3 Mx
= t= . (c)
K 34 at2
Hence, in the present case, the shear stresses in the flanges are higher than those in the web.
2.2.4 Homogeneous torsion of closed thin-walled cross-sections
The boundary value problem for the warping function (y, z), defined by Eq. (214), has a
unique solution (save an arbitrary additive constant) no matter if the profile has an interior hole
or not. Although an analytical solution is seldom obtainable, a numerical solution can always
be achieved by a discretization of the Laplace operator by a finite-difference or a finite-element
approach.
In contrast to this, the boundary value problem defined by Eq. (222) for Prandtls stress
function cannot immediately be solved, because the boundary values Sj , j = 1, 2, . . . , N , are
not known. The determination of these values requires the formulation of additional geomet-
rical conditions which express that the warping function (y, z) shall be continuous along the

boundary curves j . This may be formulated as

I

I
d = dy + dz = 0, j = 1, 2, . . . , N. (237)
j j y z
Equation (237) can be expressed in the stress function by use of Eq. (230):
I
S S d
I
dy dz + G ((z zs )dy (y ys )dz) = 0. (238)
j z y dx j
It can be shown that the integrand of the first integral can be rewritten in terms of the coordinates
s and n by the substitution
S S S
dy dz = ds. (239)
z y n
The second integral in Eq. (238) may be recast as
I I
((z zs )dy (y ys )dz) = (ydz zdy) = 2Aj . (240)
j j
The change ofH sign in Eq. (240)

H is because Hall circulations are taken clockwise. Further, it has
utilised that j ys dz = ys j dz = 0 and j zs dy = 0. Insertion of Eqs. (239) and (240)
into Eq. (238) provides the following form for the geometrical conditions:
S d
I
ds = 2Aj G . (241)
j n dx
Only thin-walled cross-sections are considered. At first a cross-section with a single cell
is considered as shown in Fig. 28. An arc-length coordinate is introduced, orientated in the
clockwise direction. Correspondingly, a local (x, n, s)-coordinate system is defined at each point
of the boundary curve with the n-axis orientated inward into the cavity, whereas the s-axis is
tangential to the boundary curve and unidirectional to the arc-length coordinate. Then, the shear
stresses along the n- and s-directions become, cf. Eq. (215),
S S
xn = , xs = . (242)
s n
Here, S is constant along the exterior and interior boundary curve of the wall, given as S = S0 =
0 and S = S1 , respectively. Hence, xn = 0 along these boundaries. If the thickness t(s) of the
wall is small compared to a characteristic diameter of the profile, it then follows from continuity
that xn is ignorable in the interior of the wall, i.e.
xn 0. (243)
The stress function decreases from S = S1 at the inner side of the wall to S = S0 = 0 at
the outer side. If t(s) is small, the variation of S(n, s) must vary approximately linearly over the
wall thickness t(s), see Fig. 28. This implies the following approximation
S S1 S0 S1
xs = = . (244)
n t(s) t(s)

xn xs
s
s
n
s
n
0 B
z
Mx
S
y
S(0, z)
S1
z
Figure 28 Closed thin-walled cross section of a cylindrical beam exposed to homogeneous torsion.
Then, the geometrical condition Eq. (241) may be written as
ds d
I
S1 = 2A1 G , (245)
1 t(s) dx
where A1 is area of the cavity.

The torsional moment is given by Eqs. (23) and (229):
d
Mx = GK = 2A1 S1 (246)
dx
Combining Eq. (245) and Eq. (246) provides the following result for the torsional constant:
S1 4A21 ds
I
K = 2A1 = , J= . (247)
Gd/dx J 1 t(s)
Equation (247) is known as Bredts formula.

The shear stresses follow from Eqs. (244) and (246):
S1 Mx
xs (s) = = . (248)
t(s) 2A1 t(s)
Notice, that the shear stress xs (s) is uniformly distributed over the wall-thickness as shown in
Fig. 28. This is in contrast to an open section, where a linear variation was obtained, as given
by Eq. (234) .

Example 2.4 Homogeneous torsion of a closed thin-walled cross-section

Figure A shows the cross-section of a beam with a single cell exposed to homogeneous torsion. The
thin-wall approximation t a is assumed to be valid.
1
t 6

s
a t n
1
6

t
1
Mx Mx 12

1
6

2t 2t 1
a
12
2t
1
12

a a
Figure A Homogeneous torsion of a closed thin-walled cross-section: Geometry (left) and distribution of shear
stresses with = Mx /(a2 t) (right).
The area of the cavity and the line integral entering Eqs. (245) and (247) become
I
ds 4a 4a 6a
A1 = 3a2 , = + = . (a)
1 t(s) t 2t t
Then, the torsional constant becomes, cf. Eq. (247),
4(3a2 )2
K= = 6a3 t. (b)
6a/t
The shear stresses follow from Eq. (248):
Mx
xs = . (c)
6a2 t(s)
The distribution has been shown in Fig. A along with the direction of action. Note that = Mx /(a2 t)
has been introduced as a normalisation quantity.
Example 2.5 Comparison of homogeneous torsion of open and closed profiles

Figure A shows two cylindrical beams, both with a cylindrical thin-walled cross-section with the side
length a and the thickness t. In one case, the cross-section is closed, whereas the cross-section in the
other case has been made open by a cut along a developer. The rotation gradient d/dx and the shear
stresses in two beams due to homogeneous torsion are compared below. (continued)

a
Mx,o , o
c a
Mx,c , c
Figure A Homogeneous torsion of open and closed cross-sections.
According to Eqs. (233) and (247), the torsional constants Ko and Kc for the beams with open and
closed cross-sections are given as
Z
1 4
Ko = t3 (s)ds = at3 , (a)
3 3
4A2 4a4
Kc = H ds = = a3 t. (b)
t
4a/t
If the two beams are exposed to the same torsional moment Mx , it follows from Eq. (23) that the rotation
gradients do /dx and dc /dx for the open and closed sections are related as
do /dx Kc 4 a2
= = . (c)
dc /dx Ko 3 t2
Next, assume that both cross-sections are fully stressed, i.e. the maximum shear stress o = c is in
both cases equal to the yield shear stress y . The corresponding torsional moments that can be carried
by the profiles are denoted Mx,o and Mx,c , respectively. With reference to Eqs. (235) and (248), these
torsional moment are related as
Mx,o Ko y /t 2a
= = . (d)
Mx,c 2a2 ty 3t
Hence, whereas the deformations of the two beams depend on the fraction a2 /t2 , the torsional moment
which can be carried depends on a/t. In conclusion, open sections are extremely ill-conditioned to carry
torsional moments in homogeneous torsion compared to closed sections. However, in many cases another
mechanics is active, which involves inhomogeneous torsion. This substantially increases the torsional
properties of open sections.

As illustrated by Example 2.5, closed thin-walled cross-sections, e.g. tubes and box gird-
ers, have a much higher torsional strength and stiffness than open thin-walled cross-section.
A mechanical explanation for the highly increased stiffness obtained for a closed cross-section
compared with the open cross-section is the fact that warping is hindered by the fact that the
displacements in the axial direction must be continuous along the s-direction, i.e. along the wall.
Especially, for a circular tube no warping is achieved in homogeneous torsion, making this profile
particularly useful for structural elements with the primary function of carrying torsional loads.
Next, consider a thin-walled cellular cross-section with a total of N cavities as illustrated in
Fig. 29. Each cavity has the cross-sectional area Aj , j = 1, 2, . . . , N , and the boundary of the
cell is denoted j . In each cell, a local (x, nj , sj )-coordinate system is defined with the nj -axis
oriented into the cavity as illustrated in Fig. 29. The local arc-length coordinate sj is defined
along the cell boundary in the clockwise direction, and the wall-thickness is t(sj ).
SN
xs (s) = t(s)
0
N
s j
sN x sj
B j
z
n
s1 1 Mx
n sk n s
k
s
s Sj Sk
xs (s) = t(s)
Figure 29 Cellular thin-walled cross section of a cylindrical beam exposed to homogeneous torsion.
As discussed in Subsection 2.2.2 and illustrated in Fig. 26, Prandtls stress function S is
constant along each interior boundary. Hence, on the cell boundary j , S has the constant value
Sj , j = 1, 2, . . . , N . However, different values of Prandtls stress function are generally present
on either side of a common wall between two adjacent cells, j and k. Because of the thin-wall
assumption the variation of S over the wall thickness must be approximately linear. Hence, with
reference to Eq. (242),
S Sj Sk
xs = . (251)
nj t(sj )
Thus, xs is uniformly distributed over the wall-thickness as illustrated in Fig. 29. The shear
stress component xn vanishes along j as well as k . It then follows from continuity that xn
is ignorable in the interior of the wall as well, i.e.
xn 0. (252)

Insertion of Eq. (251) into Eq. (241) provides:

dsj d
X Z
(Sj Sk ) = 2Aj G , j = 1, 2, . . . , N (253)
k jk t(sj ) dx
where the summation on the left-hand side is extended over all cavities adjacent to cell j. Notice
that Sk = 0 at the part of cell j adjacent to the outer periphery. Then, Eq. (253) represents N
coupled linear equations for the determination of Sj , j = 1, 2, . . . , N .
Subsequently the shear stress xs in all interior and exterior walls is determined from Eq. (2
51). According to Eq. (229), the torsional moment is given as
N
X
Mx = 2 Aj Sj , (254)
j=1
R
where the contribution A S dA can be ignored due to the thin-wall approximation. Then, the
torsional constant follows from Eq. (23):
N
Mx X Sj
K= =2 Aj . (255)
Gd/dx j=1
Gd/dx
Since Sj turns out to be proportional to Gd/dx, the right-hand side of Eq. (255) will be
independent of this quantity.
Example 2.6 Homogeneous torsion of a rectangular thin-wall profile with two cells
Figure A shows a rectangular thin-walled cross-section with two cells exposed to homogeneous torsion
from the torsional moment Mx . The wall-thickness is everywhere t a.
8 1 9
sx = 52
sx = 52
sx = 52

a t 1 Mx 2
n n
s1 s s2 s
a 2a
Figure A Homogeneous torsion of a rectangular thin-walled cross-section with two cells. Definition of local coordi-
nate systems and distribution of shear stresses with = Mx /(a2 t). (continued)

For the two cells, Eq. (253) takes the form:

a a a a d
(S1 0) + (S1 0) + (S1 0) + (S1 S2 ) = 2a2 G , (a)
t t t t dx
2a a a 2a d
(S2 0) + (S2 S1 ) + (S2 0) + (S2 0) = 2 2a2 G . (b)
t t t t dx
d d
4S1 S2 = 2atG , S1 + 6S2 = 4atG . (c)
dx dx
The solution of Eq. (c) reads
16 d 18 d
S1 = atG , S2 = atG , (d)
23 dx 23 dx
and by Eq. (254), the torsional moment is derived as
104 3 d d 23 Mx
Mx = 2a2 S1 + 22a2 S2 = a tG G = . (e)
23 dx dx 104 a3 t
Then, S1 and S2 may be written as
8 Mx 9 Mx
S1 = , S2 = . (f)
52 a2 52 a2
The shear stresses xs follow from Eq. (251) and Eq. (f). The distribution has been shown in Fig. A
with their direction of action. The quantity = Mx /(a2 t) represents a normalisation quantity for the
shear stresses.
The torsional constant follows from Eq. (255) and Eq. (d)
16 18 104 3
K = 2 a2 at + 2 2a2 at = a t. (g)
23 23 23
If the interior wall is skipped, the corresponding quantities become:
1 Mx 1 Mx 9 3
S1 = , sx = , K= a t. (h)
6 a2 6 a2 t 2
Hence, the interior wall increases the shear stress in the exterior cell walls of the right cell from 16 to
9
52
(1.3%). The torsional constant is increased merely from K = 92 a3 t to K = 104 23
a3 t (2.2%).
As indicated by Example 2.6, only a small increase of the torsional stiffness is achieved by
the inclusion of internal walls in a cross-section. Hence, from an engineering point of view,
the benefits of applying structural members with a cellular cross-sections are insignificant in
relation to torsion. Since the advantages regarding flexural deformations are also limited, and
profiles with two or more cavities are not easily manufactured, it may be concluded that open
cross-sections are generally preferred for beams loaded in unidirectional bending. Likewise,
closed cross-sections with a single cavity (pipes or box girders) are preferable when the beam is
primarily subjected to torsion and/or bending in two directions.
Finally it is noted that the solution method based on Prandtls stress-function will only be
used in relation to hand calculation for thin-walled cross-sections with at most two or three cells.
Otherwise, for thick-walled sections or multi-cell problems, the homogeneous torsion problem
will be solved numerically by a finite-element approach or another spatial discretization method
based on the Neumann boundary problem specified in Box 2.1.

2.3 Shear stresses from bending 67
2.3 Shear stresses from bending

In the previous section, the shear stresses occurring due to homogeneous torsion of a beam
have been analysed. A second contribution to the shear stresses stem from bending or flexural
deformations of a beam and with a proper choice of coordinates, the two contributions decouple.
Figure 210 shows a cross-section exposed to bending without torsion. Correspondingly, the
loads per unit length qy and qz as well as the shear forces Qy and Qz have been referred to the
shear centre S, the position of which is discussed in this section.
s0
N 0
n
sN
n xz
xx
x s
zS z Mz
xy
1 B
s
s1
n
S Qz qz
s yS
y Qy
qy
My
Figure 210 Cross-section exposed to bending.
The cross-section has the external boundary 0 and may have a number of cavities, N ,
bounded by the interior boundary curves j , j = 1, 2, . . . , N . Again, a local arc-length co-
ordinate sj is defined along each boundary, orientated clockwise for all interior boundaries,
j = 1, 2, . . . , N , whereas the arc-length coordinate s0 along the outer boundary curve is orien-
tated anti clockwise. At any point along the outer and inner boundary curves, local right-handed
(x, nj , sj )-coordinate systems are defined as shown in Fig. 210. The (x, y, z)-coordinate sys-
tem with origin at the bending centre B is assumed to be a principal-axes coordinate system.
On the cross-section, the normal stress xx as well as the shear stresses xy and xz are
acting. With reference to Eq. (27), these stresses fulfil the equilibrium equation
xx xy xz
+ + = 0, (257)
x y z
where xx is determined from Naviers formula (175), i.e.
N (x) My (x) Mz (x)
xx (x, y, z) = + z y. (258)
A Iy Iz
It then follows that
xx dN 1 dMy z dMz y qx y z
= + = + (Qy + mz ) + (Qz my ) , (259)
x dx A dx Iy dx Iz A Iz Iy

where the following equilibrium equations have been used in the statement:
dN dMy dMz
+ qx = 0, Qz + my = 0, + Qy + mz = 0. (260)
dx dx dx
A stress function T = T (x, y, z) is introduced with the defining properties
T T
xy = , xz = . (261)
y z
Hence, T is defined differently from the somewhat similar Prandtls stress function, cf. Eq. (2
15). Insertion of Eqs. (259) and (261) into Eq. (257) provides the following Poisson partial
differential equation for T :
2T 2T qx y z
2
+ 2
= (Qy + mz ) (Qz my ) . (262)
y z A Iz Iy
With reference to Eqs. (210) and (211), the boundary conditions read
xn = xx nx + xy ny + xz nz = xy ny + xz nz = 0, (263)
where nx = 0 because the beam is cylindrical. Insertion of Eq. (261) into this equation provides
the following homogeneous Neumann boundary conditions to be fulfilled on all exterior and
interior boundary curves:
T T T
= ny + nz = 0. (264)
n y z
The boundary value problems defined by Eqs. (214) and (222) for the warping function
and Prandtls stress function, respectively, are independent of x, i.e. the solution applies for all
cross-sections of the beam. In contrast, the corresponding boundary value problem defined by
Eqs. (262) and (264) for T = T (x, y, z) must be solved at each cross-section defined by x,
where the shear stresses are determined. This is so, because qx , my , mz , Qy and Qz entering
the right-hand side of Eq. (262) may vary along the beam. The solution to Eq. (262) with
boundary conditions given by Eq. (264) is unique save an arbitrary function T0 = T0 (x) which
has no influence on the shear stresses. The method can be applied to thick-walled or thin-walled
cross-sections with or without interior cavities. Normally the boundary value problem can only
be solved numerically based on a discretization of the Laplace operator entering Eq. (262). The
boundary value problem for the stress function T has been summarised in Box 2.3.
Box 2.3 Boundary value problem for the stress function determining shear stresses in bending
For a given cross-section determined by the abscissa the stress-function T = T (x, y, z) is obtained from
the Poisson partial differential equation
2T 2T qx (x) y z
2
+ = (Qy (x) + mz (x)) (Qz (x) my (x)) , (y, z) A (265a)
y z 2 A Iz Iy
with the homogeneous Neumann boundary conditions
T
= 0, (y, z) 0 1 N . (265b)
n

The indicated method is not applicable for hand-calculations. For this purpose a method
will be devised in the following sub-sections, which solves the problem for thin-walled open
cross-sections and closed thin-walled sections with few cells.
2.3.1 Shear stresses in open thin-walled cross-sections

Figure 212 shows the same open thin-walled cross-section as shown in Fig. 27, when ex-
posed to homogeneous torsion. Now the shear stresses xs and xn defined in local (x, n, s)-
coordinates are requested, caused by the shear forces Qy and Qz acting through the shear centre
S. We shall return to the definition and determination of the shear centre later in this chapter.
xs s
s=L
S Qz
x
z Mz
B
Qy
y
t(s) (s)
My
s
s=0
The shear stress component xn still vanishes at the surfaces n = 12 t(s), cf. Eq. (234).
For continuity reasons, Eq. (243) remains valid in the interior of the wall, i.e.
xn (x, n, s) 0. (266)
Then, with reference to Eq. (257), the equilibrium equation of stress components in the (x, n, s)-
coordinate system reduces to
xx xs
+ 0. (267)
x s
From Eq. (258) follows that xx varies linearly over the cross-section. Thus it must be almost
constant over the thin wall and can be replaced by its value at the midst of the wall. From
Eq. (267) it then follows that xs must also be approximately constant in the n-direction, i.e.
xs (x, n, s) xs (x, s). (268)
The constancy of the shear stress component xs in the thickness direction has been illustrated
in Fig. 212. This variation should be compared to Eq. (234) for homogeneous torsion of an
open cross-section, where a linear variation with n is obtained as illustrated in Fig. 27.

ds
xx
xs
H + dH, sx + dsx
t(s)
s
dx
H, sx
xs + dxs
xx + dxx
x
Figure 212 Differential element of a thin-walled cross-section.
A differential element with the side lengths dx and ds is cut free from the wall at the axial
coordinate x and at the arc-length coordinate s, see Fig. 212. On the sections with the arc-length
coordinates s and s + ds, the shear forces per unit length H and H + dH are acting, where
Z t/2
H(x, s) = sx (x, n, s)dn xs (x, s)t(s). (269)
t/2
Here it has been exploited that sx = xs , and Eq. (268) has been utilised. On the sections
with the axial coordinates x and x + dx, the normal stresses xx and xx + dxx are acting (see
Fig. 212). Equilibrium in the x-direction then provides
(xx + dxx )t(s)ds xx t(s)ds + (H + dH)dx Hdx = 0

H xx
+ t(s) = 0, (270)
s x
where it has been utilised that xx has an ignorable variation in the thickness direction. It has
therefore been replaced with a constant value equal to the value present at the midst of the wall.
Alternatively, Eq. (270) may be obtained simply by multiplication of Eq. (267) with t(s) and
use of Eq. (269).
With H0 (x) representing an integration constant, integration of Eq. (270) provides the so-
lution
Z s
xx
H(x, s) = H0 (x) t(s)ds. (271)
0 x
Especially, in the open thin-walled section the boundary condition sx = 0 applies at the ends of
the profile, corresponding to the arc-length coordinates s = 0 and s = L, i.e.
H(x, 0) = H(x, L) = 0. (272)
Thus, for an open thin-walled cross-section, the integration constant in Eq. (271) is H0 (x) = 0.
The partial derivative xx /x is given by Eq. (259). In what follows it is for ease assumed
that qx = my = 0. Equation (259) then reduces to
xx (x, s) Qy (x) Qz (x)
= y(s) + z(s), (273)
x Iz Iy

where (y(s), z(s)) denotes the principal-axes coordinate of a position on the wall determined by
the arc-length coordinate s.
Then, insertion in Eq. (271) and use of Eq. (269) provide the following solution for xs :
1 Qy (x) Qz
xs (x, s) = H(x, s) = Sz (s) Sy (s), (274a)
t(s) t(s)Iz t(s)Iy
where it has been utilised that H0 (x) = 0 for the open section, and
Z s Z Z s Z
Sy (s) = z(s)t(s)ds = zdA, Sz (s) = y(s)t(s)ds = ydA. (274b)
0 0
The quantities Sy (s) and Sz (s) denote the statical moment around the y- and z-axes of the
area segment (s), shown with a dark grey signature in Fig. 212 and defined as
Z s Z
(s) = t(s)ds = dA. (275)
0
Equation (274) is known as Grashofs formula. The formula is valid for Timoshenko as well
as Bernoulli-Euler beams with a thin-walled open cross-section. In this context it is noted that
Bernoulli-Euler beam theory is based on the kinematic constraint that plane cross-sections or-
thogonal to the beam axis in the referential state remain plane and orthogonal to the deformed
beam axis. In turn this implies that the angular strains xy and xz vanish, and hence that
xy = xy = 0. Hence, the shear stresses in bending cannot be determined from the beam
theory itself. Instead, these are determined from Eqs. (266) and (274) which are derived from
static equations alone and, hence, are independent of any kinematic constraints.
The shear strain caused by the shear stress xs is given as
1 1
xs = xs = H(x, s). (276)
2G 2Gt(s)
The shear strain xs implies a warping ux0 (x, 0) of the cross-section, additional to the dis-
placement in the x-direction caused by the axial force and the bending moments. The latter
is described by the kinematic conditions defined by Eqs. (111a) and (115) for Bernoulli-Euler
beam theory. Hence, the displacements of the cross-section relative to the principal-axes coordi-
nate system can be written as
dwy (x) dwz (x)

ux (x, s) = wx (x) y(s) z(s) + u0 (x, s), (277a)
dx dx
uy (x, s) = wy (x), (277b)
uz (x, s) = wz (x). (277c)
The component us (x, s) of the displacement vector u(x, s) in the tangential s-direction in the
local (x, n, s)-coordinate system shown in Fig. 212 is given as
" #T
dy(s)
T ds ux (x, s) dy(s) dz(s)
us (x, s) = s u(x, s) = dz(s) = ux (x, s) + uy (x, s), (278)
uy (x, s) ds ds
ds

where sT = sT (s) = [dy/ds, dz/ds] signifies the unit tangential vector. Then, the strain
xs (x, S) follows from Eqs. (277) and (278):

1 us ux 1 dy dwy dz(s) dwz dy dwy dz dwz ux0
xs = + = + +
2 x s 2 ds dx ds dx ds dx ds dx s
1 ux0
xs = . (279)
2 s
The warping ux0 (x, s) is next determined by integration of Eq. (279):
Z s
ux0 (x, s) = 2 xs (x, s)ds + u0 (x). (280)
0
The arbitrary function u0 (x) is adjusted, so that ux0 = 0 at the bending centre B. When xs is
determined from Eq. (276), the warping of the cross-section additional to displacements in the
x-direction predicted by Bernoulli-Euler beam theory can be determined by Eq. (280).
Example 2.7 Shear stresses and warping due to bending of a rectangular cross-section
Figure A shows a rectangular beam of height h and thickness t exposed to a shear force Qy . The bending
moment of inertia, the area segment (s) and static moment Sz (s) around the z-axis become

1 3 h s
Iz = h t, (s) = st, Sz (s) = st . (a)
12 2 2
h
where the arc-length parameter is defined from the upper edge of the profile, i.e. s = 2
+ y. Then,
xs = xy is determined from Eq. (274), that is
s2 Q y y 2 Qy

Qy s 3
xy = Sz (s) = 6 2 = 14 2 . (b)
tIz h h ht 2 h ht
Equation (b) specifies a parabolic distribution over the cross-section, where xy = 0 at the edges y = h2
Q
in agreement with the boundary conditions, and the maximum value 32 hty is achieved at the bending
centre B at y = 0.
Next, the warping is determined from Eqs. (276) and (280) together with Eq. (b):
1 s
Z s
s2
2
s3
Z
Qy s Qy s 1
ux0 (x, s) = u0 + xs ds = u0 + 6 2 ds = 3 2 2 3 , (c)
G 0 Ght 0 h h Gt h h 2
h h
where u0 is adjusted, so ux0 = 0 at the bending centre, i.e. at s = 2
. With s = 2
+ y, the warping
displacements become
3 !
Qy 2y 2y
ux0 (y) = 3 . (d)
4Gt h h
Equation (d) specifies a cubic polynomial variation, leading to an S-shape of the warping function. The
distribution of xy (y) and ux0 (y) has been shown in Fig. A. (continued)

1 Qy
t 2G t
s
(s)
h
2 3 Qy
x z 2 th z
z x x
B=S ux0
h Qy
2 xy
1 Qy
2G t
y y
Figure A Rectangular cross-section subjected to bending: Area segment (left); shear stresses (centre); warping
(right).
Example 2.8 Shear stresses due to bending of a double-symmetric I-profile

Figure A (left) shows a double-symmetric cross-section exposed to the shear forces Qy and Qz . The wall
thickness is everywhere t a, i.e. the thin-wall assumption applies. The cross-sectional area and the
principal bending moments of inertia become
1 3 1 a 2 1 3 7 3
A = 3at, Iy = 2 a t = a3 t, Iz = 2 at + a t= a t. (a)
12 6 2 12 12
Due to the symmetry, the bending and shear centres are coinciding.
n
t s1 s s2
s
a n s5
2 t
B=S n
Qz z s
x
a Qy
2 n
t s3 s4
a a s s
2 2
n
y
Figure A Double-symmetric I-profile: Geometry (left); definition of arc-length coordinates and local coordinate
systems (right).
At first the shear stresses caused by Qz are considered. For each of the four branches and the web of
the profile, arc-length coordinates sj and local right-handed (x, nj , sj )-coordinate systems are defined
as shown in Fig. A (right). At the free edge, the shear stresses are xs = 0. Hence, the shear force per
unit length H(x, sj ) vanishes at this point, and Eq. (271) becomes valid for each branch. (continued)

1 2 3 Qy
Sy (s1 ) 8
a t xs (s1 ) 4 at
Sy (s2 ) xs (s2 )
B=S B=S
Qz Qz
Sy (s3 ) 1 2 Sy (s4 ) xs (s3 ) 3 Qy xs (s4 )

8
a t 4 at
Figure B Double-symmetric I-profile: Distribution of statical moment of the area segment (sj ) around the y-axis
(left); distribution of the shear stresses xs (sj ) from Qz (right).
On the five branches, the area segments become (sj ) = sj t. These have been shown with a dark grey
signature in Fig. A (right). For the flanges, the statical moments of (sj ) around the y-axis become
1 1
Sy (s1 ) = ts1 (s1 a), Sy (s2 ) = ts2 (s2 a), (b)
2 2
1 1
Sy (s3 ) = ts3 (s3 a), Sy (s4 ) = ts4 (s4 a). (c)
2 2
Further, Sy = 0 along the web. The distribution has been shown in Fig. B (left).
1 2 3 Qy
4
a t 7 at
Sz (s1 ) Sz (s2 ) xs (s1 ) xs (s2 )
B=S 1 2 B=S 3 Qy
8
a t 14 at
Qy Qy
1 2 6 Qy
2
a t 7 at
Sz (s3 ) Sz (s4 ) xs (s3 ) xs (s4 )

1 2 3 Qy
4
a t 7 at
Figure C Double-symmetric I-profile: Distribution of statical moment of the area segment (sj ) around the z-axis
(left); distribution of the shear stresses xs (sj ) from Qy (right). (continued)

The shear stresses xs (sj ) follow from Eq. (274):

Qz Qz Qz
xs (s1 ) = Sy (s1 ) = 3 3 s1 (a s1 ), xs (s2 ) = 3 s2 (a s2 ), (d)
tIy a t a3 t
Qz Qz
xs (s3 ) = 3 s3 (a s3 ), xs (s4 ) = 3 3 s4 (a s4 ), xs (s5 ) = 0. (e)
a3 t a t
The distribution has been shown in Fig. B (right). The sign refers to the local x, nj , sj )-coordinate
system. Hence, a negative sign implies that the shear stress is acting in the negative sj -direction and,
hence, is co-directional to Qz . Then, the shear stresses in the flanges are distributed parabolically in the
same way as stresses in the rectangular cross-section considered in Example 2.7. Each flange carries the
shear force Qy /2, and no shear force is carried by the web in the present case.
Next, the shear stresses caused by Qy are analysed. The statical moment of the area segment (sj )
around the z-axis becomes:
1 1
Sz (s1 ) = ats1 , Sz (s2 ) = ats2 , (f)
2 2
1 1 a 1
Sz (s3 ) = ats3 , Sz (s4 ) = ats4 , Sz (s5 ) = at ts5 (a s5 ), (g)
2 2 2 2
and by Eq. (274) the shear stresses xs (sj ) are determined as
Qy 6 Qy 6 Qy
xs (s1 ) = Sz (s1 ) = s1 , xs (s2 ) = s2 , (h)
tIz 7 a2 t 7 a2 t
6 Qy 6 Qy 6 Qy
xs (s3 ) = s3 , xs (s4 ) = 2 s4 , xs (s5 ) = (a2 + s5 a s25 ) . (i)
7 a2 t 7a t 7 at
The distribution of (sj ) and xs (sj ) has been shown in Fig. C.
2.3.2 Determination of the shear centre

So far in this section we have only considered double-symmetric cross-sections for which the
shear centre S coincides with the bending centre B. However, in the general case, the shear
centre will be different from the bending centre as suggested by Fig. 21. In any case, the shear
forces Qy and Qz must be statically equivalent to the shear stresses integrated over the cross-
section, i.e.
Z Z
Qy = xy dA, Qz = xz dA. (283)
A A
In the present case, we are concerned with bending uncoupled from torsion. Hence, the torsional
moment Mx stemming from the shear stresses on the cross-section should vanish. According to
Eq. (225), this implies the identity
Z
((y ys )xz (z zs )xy ) dA = 0. (284)
A
Combining Eqs. (283) and (284) provides the relation for the coordinates of the shear centre:
Z
Qy zs + Qz ys = (zxy + yxz )dA. (285)
A

For a thin-walled cross-section, this identity may be recast in terms of the local (x, s, n)-coordinates,
providing
Z
Qy zs + Qz ys = h(s)xs (x, s)t(s)ds, (286)
L
where h(s) is the moment arm of the differential shear force xs (x, s)t(s)ds,
dz(s) dy(s)
h(s) = y(s) z(s) . (287)
ds ds
It is noted that, with the given definition, h(s) may have positive as well as negative values.
Insertion of Eq. (274) into Eq. (286) provides the identity
Qz Qy
Z Z
Qy zs + Qz ys = Sy (s)h(s)ds Sz h(s)ds (288)
Iy L Iz L
which holds for arbitrary Qy and Qz . This implies the following solutions for the coordinates of
the shear centre:
1 1
Z Z
ys = Sy (s)h(s)ds, zs = Sz (s)h(s)ds. (289)
Iy L Iz L
In particular, for a thin-walled section without branching, Sy (s) and Sz (s) become
Z s Z s
Sy (s) = z(s)t(s)ds, Sz (s) = y(s)t(s)ds. (290)
0 0
Hence, Eq. (289) involves a double integration with respect to the arc-length parameter over the
length L. We shall arrange the calculation in a way such that only a single line integral needs to
be evaluated.
(s)
s=0
ds (s)
h(s)
s
S Qz
z
x B
Qy
y
t(s)
s=L

At first the so-called sector coordinate (s) with pole B is introduced. The sector coordinate
(s) is equal to the area shown in Fig. 213, delimited by the bending centre B and the area
segment (s). Thus, (s) is given as
Z s
(s) = h(s)ds. (291)
0
d
It follows that (s) may be considered as an integral of h(s) with respect to s, i.e. ds (s) = h(s).
Hence, from Eq. (290) follows that the derivation of Sy (s) and Sz (s) with respect to s
become
d d
Sy (s) = z(s)t(s), Sz (s) = y(s)t(s). (292)
ds ds
Then, integration by parts of Eq. (289) and use of Eq. (292) provides
!
1 h iL Z L
ys = Sy (s)(s) z(s)(s)t(s)ds (293)
Iy 0 0
Now, Sy (L) = Sy = 0, because the (x, y, z)-coordinate system has origin in the bending
centre. Further, both Sy (0) = 0 and (0) = 0, so the first term within the parentheses vanishes
in both limits and, hence,
Z L
Iz
Z
ys = , Iz = (s)z(s)t(s)ds = zdA. (294a)
Iy 0 A
Similarly, it can be shown that

L
Iy
Z Z
zs = , Iy = (s)y(s)t(s)ds = ydA. (294b)
Iz 0 A
The quantities Iy and Iz are denoted sector centrifugal moments. An arbitrary constant 0 can
R added to (s)Rin Eq. (294) without
be R changing the value of Iz and Iy . This is so because

A 0
ydA = 0 A ydA = 0 and
A 0
zdA = 0. Hence, the sector coordinate is determined
within an arbitrary constant.
As a third method, the coordinates of the shear centre may be determined from Eq. (286), if
the shear stress xs (x, s) is calculated from Qy and Qz separately. Thus,
1
Z
ys = xs (x, s)h(s)t(0)ds for Qy = 0, (295a)
Qz L
1
Z
zs = xs (x, s)h(s)t(s)ds for Qz = 0. (295b)
Qy L
Obviously, if the profile has a line of symmetry, then the shear centre is placed on this line.
Finally, a note is made regarding the notation in this chapter. Previously, (x, y) has denoted
the normalised warping function or, simply, the warping function. However, we shall later see
that for open thin-walled sections the warping function is identical to the sector coordinate, which
motivates the naming of the latter quantity.

Example 2.9 Determination of the shear centre for an I-profile with a single line of symmetry
The thicknesses of the flanges and the web of the profile shown in Fig. A (left) are all t. Due to the
symmetry around the y-axis, the y-axis as well as the z-axis become principal axes. The bending centre
is placed as shown in the figure and the principal moments of inertia become:
91 3 1044 3
Iy = a t, Iz = a t. (a)
12 13
The profile is exposed to a horizontal shear force Qz , and the position of the shear centre will be deter-
mined both by Eqs. (289), (294) and (295). Due to the symmetry, the shear centre is placed on the
y-axis, i.e. zs = 0.
2a2 t
Sy (s1 ) Sy (s2 )
4a n2
t
s2
36
a n1 s 1
13
Qz
S z
6a B x
t 42
a
13 n4
y t s3 s4
3a n3
Sy (s3 ) 9 2
Sy (s4 )
8
a t
Figure A Single-symmetric I-profile: Geometry (left); definition of arc-length coordinates and distribution of static
moments (right).
For each of the four branches indicated in Fig. A (right), an arc-lengthR parameter sj and a local
(x, nj , sj ) coordinate system are defined. The statical moments Sy (sj ) = z(sj )t(sj )dsj become:
Z s1
1 1
Sy (s1 ) = (2a s)tds1 = (4a s1 )s1 t, Sy (s2 ) = (4a s2 )s2 t, (b)
0 2 2
Z s3
3 1 1
Sy (s3 ) = a s3 tds3 = (3a s3 )s3 t, Sy (s4 ) = (3a s4 )sy t. (c)
0 2 2 2
The shear stresses xs (x, sj ) are positive, when acting in the direction of the arc-length parameter sj .
These follow from Grashofs formula (274):
Qz 12 Qz
xs (sj ) = Sy (sj ) = Sy (sj ). (d)
tIy 91 a3 t2
The distribution of shear stresses has been shown in Fig. B (left). (continued)

72 2
48 a
Q1 182
13
n2
s2
36 36
13
a 13
a
Qz 1
(s2 )
2
S z z
B x B x
42 42
13
a 13
a
y y
Q2 27
182

63 2
13
a
Figure B Single-symmetric I-profile: Shear stresses from the shear force Qz with = Qz /(at) (left); distribution
of the sector coordinate (s) (right).
The shear stress resultants Q1 and Q2 in the top and bottom flanges become:
2 48 Qz 64 2 27 Qz 27
Q1 = 4at = Qz , Q2 = 3at = Qz . (e)
3 182 at 91 3 182 at 91
The shear centre then follows from Eq. (295):
36 42 90
Qz ys = Q1 a + Qz a ys = a. (f)
13 13 91
Next, employing another approach, the moment arm h(sj ) defined by Eq. (287) is negative on the
branches described by arc length parameters s1 and s2 , and positive along the arc-length parameters s2
and s3 . Then, from Eq. (b) it follows that
Z
Sy (s)h(s)ds
L
Z 2a Z 2a
1 36 1 36
= (4a s1 )s1 t a ds1 + (4a s2 )s2 t + a ds2
0 2 13 0 2 13
Z 3a/2 Z 3a/2
1 42 1 42
+ (3a s3 )s3 t + a ds3 + (3a s4 )s4 t a ds4
0 2 13 0 2 13
96 4 96 4 189 4 189 4 15 4
= s t+ a t a t a t= a t. (g)
13 13 52 52 2
Then, from Eqs. (289) and (a) it follows that
1 15 4 90
ys = 91 a t = a. (h)
12
3
a t 2 91
This result is identical to the result achieved in Eq. (f). However, it is noted that the result in Eq. (h) has
been achieved without the determination of the shear stresses. Hence, the second approach may appear
to be simpler than the first approach, but it does not provide any information about the stress distribution.
(continued)

Finally, the same Rcalculation is performed by means of Eq. (294). The distribution of the sector
s
coordinate (sj ) = 0 j h(s)ds with the pole B for each of the four branches becomes:
Z s1
36 36 36
(s1 ) = a ds1 = as1 , (s2 ) = as2 , (i)
13 0 13 13
Z s4
42 42 42
(s4 ) = a ds4 = as4 , (s3 ) = as4 . (j)
13 0 13 13
R R
The sector centrifugal moment Iz = A zdA = L (s)z(s)t(s)ds becomes:
Z 2a Z 2a
36 3 42 15
Iz = 2 (2a s1 ) as1 2tds1 + 2 a + s4 as4 2tds4 = a4 t, (k)
0 13 0 2 13 2
where the symmetry of the integrand (s)z(s) has been exploited. Then, from Eqs. (294) and (a) the
shear centre is determined as
1 15 4 90
ys = 91 3t 2
a t = a. (l)
12
a 91
Hence, the three different approaches lead to the same result.
Example 2.10 Determination of the shear centre for a symmetric U -profile

The bending centre of the U -profile has the position shown in Fig. A. Since the profile is symmetric, the
indicated (x, y, z)-coordinate system is a principal axis system. The moments of inertia with respect to
the principal axes become:
64 3 26 3
Iy = a t, Iz = a t. (a)
15 3
2a
4 4
a a
5 2t 5
a Q2
B B
S Qz S
z z
x x
t Q1
a
2t Q2
Qy Qy
112
y 65
a y
Figure A Symmetric U -profile: Geometry (left); position of the shear centre (right).
Local arc length coordinates s1 , s2 and s3 are introduced as indicated in Fig. B. The distributions of
the static moments Sy (sj ) and Sz (sj ) are shown in the figure. The analytic expressions are given as
(continued)

Z s1
6 12
Sy (s1 ) = a s1 2tds1 = as1 s21 t, (b)
0 5 5
Z s2
4 4 4
Sy (s2 ) = a2 t + a tds2 = at(a s2 ), (c)
5 0 5 5
Z s3
4 4 4 8
Sy (s3 ) = a2 t + a + s3 2tds3 = a2 t as3 t + s23 t, (d)
5 0 5 5 5
Z s1
Sz (s1 ) = (a)2tds1 = 2as1 t, (e)
0
Z s2
1
Sz (s2 ) = 4a2 t + (a + s2 )tds2 = 4a2 t as2 t + s22 t, (f)
0 2
Z s3
Sz (s3 ) = 4a2 t + a 2tds3 = 4a2 t + 2as3 t. (g)
0
Sy (s1 )
4 2
a t 4a2 t Sz (s1 )
26 2
5
5
a t
s1Sz (s2 ) s1
s2 s2
B B
z z
x 1 2
a t x
Sy (s2 ) 2
s3 y s3 y
26 2
4 2
a t 5
a t 4a2 t
5 Sz (s3 )
Sy (s3 )
Figure B Symmetric U -profile: Distribution of the static moments Sy (sj ) (left) and Sz (sj ) (right).
xs (s1 )
3
6
xs (s1 )
32 z 27 26 y

160 z
xs (s2 )
B B
z z
x 3
x
xs (s2 ) 52 y
y y
27
3

160 z
6

32 z 26 y xs (s3 )
xs (s3 )
Figure C Symmetric U -profile: Distribution of the shear stresses from the shear forces Qz with z = Qz /(at)
(left) and Qy with y = Qy /(at) (right). (continued)

The shear stresses follow from Grashofs formula (274). Figure B (left) shows the distribution of
shear stresses from Qz , and Fig. B (right) shows the distribution from Qy . The position of the shear
centre is given by Eq. (295). Since the shear stresses xs from Qz are distributed symmetrically around
the z-axis it follows that ys = 0, reflecting the fact that the z-axis is a line of symmetry.
The resulting shear forces Q1 and Q2 in Fig. A then become

1 6 6 2 3 6
Q1 = 2t 2a y = Qy , Q2 = t + 2ay = Qy . (h)
2 26 13 3 52 13
Finally, from Eq. (295) follows that

1 4 112
zs = Q1 a + a Q2 + Q1 a = a. (i)
Qy 5 65
Hence, for a U -profile the shear centre lies at a considerable distance outside the profile.
2.3.3 Shear stresses in closed thin-walled sections

Figure 214 shows a closed single-cell section of a thin-walled beam. The shear force H(x, s)
acting within the wall per unit length of the beam (see Fig. 212) in a corresponding open section
is uniquely determined by Eq. (271) due to the condition H(x, 0) = H0 (x) = 0 at the boundary
s = 0. However, in a closed section H(x, s) is not a priori known in one or more points of the
periphery 1 , for which reason the shear stresses cannot be determined from static equations
alone. Hence, a geometrical condition must be introduced, from which the initial value H0 (x) at
s = 0 can be determined. The derivation of this geometric condition will be considered at first.
xs
s
n
s
n
B
z
(s) S Qz
1
Qy
s=L
t(s) H
0
s=0
y
Figure 214 Closed single-cell section exposed to bending.

The arc-length coordinate s is orientated clock-wise with an arbitrarily selected origin O.

Then, the local (x, n, s)-coordinate system is orientated with the n-axis orientated toward the
interior of the cell. The shear stress xs (x, s) follows from Eqs. (271) and (273):
1 H0 (x) Qy (x) Qz (x)
xs (x, s) = H(x, s) = Sz (x, s) Sy (x, s). (298a)
t(s) t(s) t(s)Iz t(s)Iy
where the static moments of the area segment (s) are again given as
Z s Z Z s Z
Sy (s) = z(s)t(s)ds = zdA, Sz (s) = y(s)t(s)ds = ydA. (298b)
0 0
Equation (298a) is the equivalence of Grashofs formula (274) for a closed thin-walled section.
Unlike the case of the open section, H0 (x) is generally different from zero and the determination
of H0 (x) is part of the problem.
The shear strain xs in the local (x, n, s)-coordinate system is given by Eq. (276) which is
valid for open as well as closed thin-walled sections. This implies the warping u0 (x, s) defined
by Eq. (280), where it is recalled that u0 (x, s) = 0 in the bending centre B. This is so, because
wx (x) has been defined as the total displacement in the x-direction of B. Now, for the closed
section, the displacement ux (x, s) must be continuous as a function of s along the periphery 1 .
The axial and bending contributions, i.e. the first three contributions on the right-hand side of
Eq. (277), are always continuous. A possible discontinuity then stems from the warping. If the
profile is open, as illustrated in Fig. 212, the ends at s = 0 and s = L can move freely relatively
to each other. Hence, a warping discontinuity develops between these two endpoints. However, in
a closed section, the warping of these points must be identical. Hence, the geometrical condition
can be formulated as
ux0 (x, 0) = u(x, L) = u0 (x). (299)
Insertion of Eqs. (276) and (299) into Eq. (280) provides

L L
1 H(x, s)
Z Z
u0 (x) = u0 (x) + 2 xs (x, s)ds = u0 (x) + ds
0 G 0 t(s)
L
H(s) H(s)
Z I
ds = ds = 0. (2100)
0 t(s) 1 t(s)
Insertion of Eq. (298a) gives the following formulation of the geometrical condition from which
the initial condition H0 can be determined:
ds Qz Sy Qy Sz
I I I
H0 (x) ds ds = 0. (2101)
1 t(s) Iy 1 t(s) Iz 1 t(s)
With H0 (x) determined, H(x, s) and the shear stress xs (x, s) can next be determined from
Eq. (298a).
Example 2.11 Shear stresses in a symmetric thin-walled single-cell section

Figure A shows a double-symmetric thin-walled single-cell profile. The thickness is everywhere t. We
want to determine the shear stresses from a shear force Qy acting at the shear centre. (continued)

The cross-sectional area and moment of inertia around the z-axis become
5 3
A = 8at, Iz = a t. (a)
3
The origin of the arc length coordinate
Rs is chosen at the lower left corner. The corresponding distribution
of the static moment Sz (s) = 0 z(s)t(s)ds has been shown in Fig. B.
t(s)
B=S
z
a
x
Qy
2 2
3
a 3
a
y
Figure A Geometry of the symmetric rectangular thin-walled single-cell section.
3 2
2
a t
B=S
z
1 2 x
8
a t 1 2
a t
8
s y
3 2
2
a t
Figure B Distribution of the static moment Sz (s) in the symmetric rectangular thin-walled single-cell section.
Next, the following line integrals are calculated:

I
ds 1 8a
= (a + 3a + a + 3a) = , (b)
2 t t t
I
Sz (s)
ds
1 t

1 2 1 2 1 3 3 2 1 1 3
= a t a 3a a2 t a2 t a a2 t a 3a a2 t = 6a3 . (c)
t 3 8 2 2 2 3 8 2 2
(continued)

From Eq. (2101) then follows that

8a Qy 9 Qy
H0 5 3 (6a3 ) = 0 H0 = . (d)
t 3
a t 20 at
18 18
B=S
z
x
3 3
Qy
18 18
y
Figure C Shear stresses in the symmetric rectangular thin-walled single-cell section with = Qy /(40at).
The distribution of shear stresses xs (x, s) follows from Eq. (298a) and has been indicated in Fig. C.
The arrows indicate the positive direction of the shear stresses. These will be referred to as the shear flow
in what follows, due to the analogous behaviour of water flowing down through a system of pipes.
As shown in Example 2.11, a symmetric thin-walled single-celled profile exposed to a shear

force acting along the line of symmetry will have a symmetric distribution of shear stresses.
Especially, the shear stress at the line of symmetry vanishes. Hence, if the origin of arc-length
coordinate is placed at the line of symmetry we must have H0 (x) = 0, entailing that a single-
celled symmetric profile can be analysed from the static equations alone.
For a thin-walled profile with N cells, arc-length parameters sj are introduced for all cells,
orientated clock-wise as shown in Fig. 215. The origins O1 , O2 , . . . , ON for the arc-length
coordinates are chosen arbitrarily. The shear forces per unit length H0j (x) at the origins are
unknown and must be determined from geometrical conditions in addition to the static shear
flow equations. Similarly to Eq. (299) these are determined by the conditions that the warping
must be continuous along all the peripheries j of the cells.
At first, the static moments Sy and Sz for the open profile in Fig. 215 are determined.
Especially, the variation of Sy (x, sj ) and Sz (x, sj ) along j as a function of sj are registered.
Next, the calculation of the shear force per unit length Hj (x, sj ) within each cell can be arranged
as illustrated in Fig. 216. Firstly, the shear force per unit length Hsj (sj ) of the open, branched
profile is calculated, orientated in the direction of the arc-length coordinate sj . This is given by
Grashofs formula (274), i.e.
Qz Qy
Hsj (sj ) = Sy (sj ) Sz (sj ). (2103)
Iy Iz

xs H0N
0 S Qz
ON
Qy
H01
N
j
O1 sN x sj
B j
z
s1 1 Oj
sk k
H0j
Ok
H0k
Figure 215 Closed multi-cell section exposed to bending.
N H
N
j
H1
1 Hk Hj
N H
sN
j
Hs1
1 Hsk Hsj
N H
0N
j
H01
1 H0k H0j
Figure 216 Arrangement of calculations of shear forces per unit length within cells.

In each cell, the initial conditions induce a constant shear force per unit length H0j . Due to
the chosen sign convention, the net shear force on the periphery segment jk between cell j and
cell k becomes Hsj H0k . Hence, on this segment the total shear force per unit length becomes
Hj (sj ) = Hsj (sj ) + H0j H0k . (2104)
Continuity requires that Eq. (2101) is fulfilled for all N cells, i.e.
Hj (sj )
I
dsj = 0, j = 1, 2, . . . , N. (2105)
j t(sj )
Insertion of Eq. (2104) leads to the following N linear equations for the determination of
H01 , H02 , . . . , H0N :
N
dsj Hsj (sj )
X Z I
(H0j H0k ) = dsj , j = 1, 2, . . . , N. (2106)
k=1 jk t(sj ) j t(sj )
The coefficient matrix in Eq. (2106) is identical to this for the determination of Prandtls stress
function Sj in the cells in the corresponding St. Venant torsion problem. With H0j determined
from Eq. (2106), Hj (sj ) can be calculated from Eq. (2104). Finally, the shear stresses xs (sj )
follow from Eq. (298a) with positive sign when acting in the direction of sj .
Example 2.12 Shear stresses in a single-symmetric double-cell section

The profile shown in Fig. A is identical to that considered in Example 2.11. However, a partition with the
wall thickness t has been included as shown in the figure. The bending centre B is placed as indicated.
Because the section is symmetric around the z-axis the indicated coordinate system with origin in B is
a principal axis coordinate system. The section is still exposed to a shear force Qy acting through the
shear centre S. Due to the symmetry the shear centre is placed on the z-axis (yS = 0). The zs will be
determined as a part of the solution. The cross-sectional area and inertial around the z-axis become
7 3
A = 9at, Iz = a t. (a)
4
13
9
a
a
t(s) t(s) t(s) 2
S B
z
a
x
Qy
a
2
y
a 2a
Figure A Geometry of the rectangular thin-walled double-cell section.

(continued)

O1 O2
1 H s1 s2 2 H
1 2
1 2
Figure B Definition of the origins of the arc-length coordinates in the rectangular thin-walled double-cell section.
1 2
2
a t a2 t
Sz (s1 ) Sz (s2 )
1 2 1 2 1 2
8
a t 8
a t 8
a t 1 2
a t
8
1 2
2
a t a2 t
Figure C Distribution of the static moments Sz (s1) and Sz (s2) in the rectangular thin-walled double-cell section.
The origin of the arc-length coordinates are chosen as shown in Fig. B. The static moments Sz (s1 )
and Sz (s2 ) are next calculated with the sign and magnitude as indicated in Fig. C. Especially, it is seen
that Sz (s1 ) = Sz (s2 ). The shear force per unit length in the open profile is given by Eq. (2103):
Qy
Hsj (sj ) = Sz (sj ), j = 1, 2. (b)
Iz
Next, Eq. (2106) provides
I
a a a a Hsj (s1 )
(H10 H20 ) + (H10 0) + (H10 0) + (H10 0) = ds1 , (c)
t t t t 1 t(s1 )
I
2a a 2a a Hsj (s2 )
(H20 0) + (H20 0) + (H20 0) + (H20 H10 ) = ds2 . (d)
t t t t 2 t(s2 )
The right-hand sides are calculated by insertion of Eq. (b) with the distribution of Sz (sj ) shown in
Fig. C. The results become
a2 t a a2 t a2 t 2 a2 t s a2 t
I
Hsj (S1 ) 4 Qy 2 4 Qy
ds1 = a a a = , (e)
1 t(s1 ) 7 a3 t2 3 8 2 2 2 3 8 2 2 7 t
2 a2 t
I
Hsj (s2 ) 4 Qy 2a 2 2 2a 2 12 Qy
ds2 = 3 2 a t + a a2 t + a a2 t + a t a = . (f)
2 t(s2 ) 7a t 2 3 2 3 8 7 t
(continued)

Then, the following solutions are obtained for the initial values of the shear forces per unit length:
4 Qy 12 Qy
4H 10 H20 = , H10 + 6H20 = , (g)
7 a 7 a
which implies that
24 Qy 88 Qy
H10 = , H20 = . (h)
322 a 322 a
Next, Hj (sj ) is calculated from Eqs. (2103) and (2104) along both peripheries 1 and 2 . Finally, the
shear stresses xs (sj ) are computed from Eq. (298a). Due to the constant wall thickness this reduces to
1
xs (sj ) = Hj (sj ), j = 1, 2. (i)
t
The flow of the shear stresses xs has been shown in Fig. E.
68 24 96
88 88
88
23 88 23
23
68 24 96
Figure D Distribution of the shear stresses in the rectangular thin-walled double-cell section with = Qy /(322at).
13
9
Q4 Q1
a
2
S B
z
a Q5 Q3 Q2
x
Qy
y a
2
Q4 Q1
80
1449
Figure E Shear forces in the wall segments of the rectangular thin-walled double-cell section. (continued)

Next, the shear forces Q1 , Q2 , Q3 , Q4 and Q5 in the wall segments shown in Fig. E are calculated.
These become:
1 1 Qy 12
Q1 = (96 88) 2a t = Qy , (j)
2 322 at 483
2 1 Qy 167
Q2 = (96 + 23) a t = Qy , (k)
3 322 at 483
2 1 Qy 191
Q3 = (112 + 23) a t = Qy , (l)
3 322 at 483
1 1 Qy 33
Q4 = (68 24) a t = Qy , (m)
2 322 at 483
2 1 Qy 125
Q5 = (68 + 23) a t = Qy . (n)
3 1288 at 483
It is seen that Q2 + Q3 + Q5 = Qy . The static equivalence (283) of the shear stresses is then fulfilled.
The position zs of the shear centre follows from Eq. (295):

1 14 4 13 80
zs = (Q4 Q1 )a Q2 a + Q3 a + Q5 a zs = a. (o)
Qy 9 9 9 1449
The position of the shear centre has been illustrated in Fig. E.
Example 2.13 Distribution of shear stresses in open and closed sections due to bending and
St. Venant torsion
Q
Q Q
Q Q
Mx Mx
Figure A Shear flow in open and closed sections due to transverse shear forces and torsional moments.
Fig. A shows the shear flow in open and closed thin-walled sections exposed to a transverse shear force Q
or a torsional moment M . For closed sections the shear stresses xs (x, s) are uniformly distributed over
the wall thickness for both loadings. For open section this is only the case in the case for the shear force
loading, whereas the shear stresses for St. Venant torsion is linearly varying in the thickness direction
around the mid-line of the wall.

2.4 Summary 91
2.4 Summary
The computation of shear stresses in beams has been the focus of this chapter with detailed
explanation of the theory for thin-walled sections subjected to torsion and/or bending. A brief
summary of the main findings is given in the following.
Uncoupling of bending and torsion requires that the shear force acts through the so-called shear
centre. The shear centre always lies on a line of symmetry within a symmetric cross-section.
Hence, the position of the shear centre coincides with that of the bending centre for double-
symmetric sections.
St. Venant torsion is characterised by the fact that the beam is allows to warp freely, leading
to a homogeneous twist of the cross-section along the beam axis. The homogenous torsion
problem can be defined in terms of the warping function or, alternatively, in terms of Prandtls
stress function.
Shear stresses due to homogeneous torsion vary linearly over the thickness of the wall in open
thin-walled profiles. However, in a closed thin-walled section with one or more cells, the
shear stresses due homogeneous torsion is homogeneous over the thickness.
Homogeneous torsion induces warping in a beam. The warping increases linearly with the tor-
sional moment. However, in a circular profile (a cylinder as well as a tube) there is no
warping.
The torsional stiffnesses of open and closed cross-sections are different . Thus, a closed cross-
section has a much higher torsional stiffness than an open cross-section with a similar cross-
sectional area. Likewise, the ultimate strength of a closed section is higher than that of a
similar open section.
Shear stresses from bending can generally be analysed by means of the so-called stress func-
tion. However, this requires the use of a numerical scheme, e.g. the finite-element method.
Grashofs formula defines the shear stresses from bending in open thin-walled sections. For
closed thin-walled sections, Grashofs formula must be adjusted be an additional term that
Warping takes place due to bending since the shear stresses are accompanied by shear strains.
For a rectangular cross-section, the beam warps into an S-shape.
Shear flow is a graphical interpretation of the direction in which the shear stresses are acting on
the cross-section of a beam. In the case of bending, this forms an analogy to water streaming
down a system of pipes.
Internal walls in a section will not increase the torsional stiffness and strength significantly.
However, the inclusion of an internal wall oriented in the direction of the shear force will
provide an increase of the shear strength and stiffness.
The theory presented in this chapter can be used for the determination of the shear stresses
and warping deformations in beams subjected to any combination of bending and homogeneous
torsion. However, if torsion is prevented, e.g. at one end of the beam, another theory must be
applied as described in the next chapter.


References
Krenk, S. (2001). Mechanics and analysis of beams, columns and cables (2nd ed.). Springer Verlag.
SaintVenant, B. (1855). Memoire sur la torsion des prismes. Mem. Acad. Sci. Savants Etrangers 14, 233
560.
Timoshenko, SP (1921). On the correction for shear of the differential equation for transverse vibrations of
static bars. Philosophical Magazine, Series 6 41, 744746.
Vlasov, V.Z. (1961). Thin-walled elastic beams. Israel Program for Scientific Translations.
93
94 References

ISSN 1901-7278

Beams 3 D

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Beams 3 D

Caricato da

Copyright:

Formati disponibili

Elastic Beams in Three Dimensions

Lars Andersen and Sren R.K. Nielsen

DCE Lecture Notes No. 23

Elastic Beams in Three Dimensions

Lars Andersen and Sren R.K. Nielsen

Printed in Denmark at Aalborg University

ISSN 1901-7286 DCE Lecture Notes No. 23

Aalborg, August 2008 Lars Andersen and Sren R.K. Nielsen

DCE Lecture Notes No. 23

1 Beams in three dimensions 1

2 Shear stresses in beams due to torsion and bending 45

DCE Lecture Notes No. 23

1.2 Equations of equilibrium for spatial beams

Figure 11 Beam in referential state.

DCE Lecture Notes No. 23

equilibrium of the differential beam element:

1.2.1 Section forces and stresses in a beam

Elastic Beams in Three Dimensions

Figure 13 Stresses and stress resultant on a cross-section of the beam.

Figure 14 Components of the stress tensor.

DCE Lecture Notes No. 23

in the (x, y, z)-coordinate system given as

Moment equilibrium of the cube shown in Fig. 14 requires that

Hence, is a symmetric tensor.

1.2.2 Kinematics and deformations of a beam

sin tan , (110)

ux (x, y, z) = wx (x) + zy (x) yz (x), (111a)

uy (x, y, z) = wy (x) zx (x), (111b)

Elastic Beams in Three Dimensions

Figure 15 Deformation components in beam theory.

DCE Lecture Notes No. 23

Deformed bottom flange

Elastic Beams in Three Dimensions

Figure 18 Kinematics of Bernoulli-Euler beam theory.

DCE Lecture Notes No. 23

Elastic Beams in Three Dimensions

Figure 19 Decomposition of cross-section deformation into bending and shear components.

Figure 110 Definition of curvature.

1.2.3 Constitutive relations for an elastic beam

DCE Lecture Notes No. 23

Elastic Beams in Three Dimensions

1.3 Differential equations of equilibrium for beams

DCE Lecture Notes No. 23

1.3.1 Governing equations for a Timoshenko beam

whereas mechanical boundary conditions imply the prescription of N , Qy , Qz , My and Mz ,

Elastic Beams in Three Dimensions

1.3.2 Governing equations for a Bernoulli-Euler beam

d2 My /dx2 dQz /dx + dmy /dx = 0

dwz (x0 ) dwz (x0 )

dMz (x0 ) dMy (x0 )

DCE Lecture Notes No. 23

1.4 Uncoupling of axial and bending deformations

1.4.1 Determination of the bending centre

Elastic Beams in Three Dimensions

Figure 111 Translation of coordinate system.

DCE Lecture Notes No. 23

Then the relation between {N, My , Mz } and {N , My , Mz } and {, y , z } and { , y , z }

where C denotes the constitutive matrix,

Likewise, the constitutive relation in the (x , y , z )-coordinate system reads

where the constitutive matrix has the form

Elastic Beams in Three Dimensions

Iyz = Iyz yB (AzB ) zB (AyB ) + yB zB A = Iyz yB zB A. (152c)

Figure 112 Single-symmetric cross-section.