International Journal of Solids and Structures 47 (2010) 2825–2840

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

Lateral buckling of beams with shear deformations – A hyperelastic formulation

Mario M. Attard a,*, Moon-Young Kim b
a
School of Civil and Environmental Engineering, The University of New South Wales, Sydney 2052, Australia
b
Department of Civil and Environmental Engineering, Sungkyunkwan University, Suwon, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The equilibrium and buckling equations are derived for the lateral buckling of a prismatic straight beam.
Received 8 September 2009 A consistent finite strain constitutive law is used, which is based on a hyperelastic model for an isotropic
Received in revised form 17 May 2010 material. The kinematics of the cross-sectional deformations are based on a Timoshenko type beam dis-
Available online 19 June 2010
placement of the cross-sectional plane using Euler angles and two shear finite rotations coupled with
warping taken normal to the displaced plane. Also derived are the second order approximations to the
Keywords: displacements, curvatures, twist and internal actions. The constitutive relationships for the internal
Finite strain
actions reveal new coupling terms between the bending moments, torsion and bimoment, which are
Beams
Elastica
functions of the cross-sectional warping and shear deformations. New Wagner type nonlinear torsion
Lateral buckling terms are derived which are functions of the warping of the cross-sectional plane, and are coupled to
Hyperelastic the twisting and shear deformations of the cross-section. Solutions are determined for the lateral buck-
Shear deformations ling of a prismatic monosymmetric beam under pure bending and the flexural–torsional buckling under
axial compression. For the flexural–torsional buckling problem it is found that the Euler type column
buckling formula is consistent with Haringx’s column buckling formula while the torsional buckling for-
mula is different to conventional equations. The second variation of the total potential is also derived. The
effects of shear deformations are explored by examining the non-dimensional lateral buckling equation
for a simply supported beam.
Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved.

1. Introduction
Thin-walled structural components are widely used in many
industries including aerospace, building, aircraft and shipbuilding.
It important that the nonlinear and stability response of these
structural components, be accurately modelled and estimated by
Mohareb and Dabbas (2003) and later Emre Erkmen and Mohareb
(2008) showed that shear deformations can be very important
when estimating the lateral buckling of a long column supporting
a short wide flange cantilever. The inclusion of shear deformations
in the buckling analysis of axially loaded columns and sandwich
type columns with a soft shear core has received attention in the
literature. The major issue of contention involves the definition
of the constitutive relationship for the axial force and shear force
for one-dimensional beam theory. Two distinct methods for col-
umn buckling with shear have been presented, one essentially by
Engesser (1891) and the other by Haringx (1942). The controversy
has been discussed extensively by Zielger (1982), Reissner (1982),
Kardomateas and Dancila (1997), Bažant (2003), Bažant and
Beghini (2004, 2006), Bažant and Cedolin (1991), Beghini et al.
(2008), Aristizabal-Ochoa (2008), Blaauwendraad (2008). Attard
* Corresponding author. Tel.: +61 2 9385 5075; fax: +61 2 9385 6139.
E-mail address: m.attard@unsw.edu.au (M.M. Attard).
(2003a,b) and Attard and Hunt (2008a,b) applied a finite strain
constitutive model derived from a hyperelastic formulation to
the problem of column and sandwich column buckling and was
shown to be consistent with Haringx’s approach and Reissner’s
proposal for beam actions Reissner (1972).
Several authors have extended the theory of lateral buckling and
nonlinear analysis of thin-walled beams to include shear deforma-
tions, including Chang et al. (1996), Lee et al. (submitted for publi-
cation), Machado (2008), Machado and Cortínez (2005a,b), Reissner
(1989), Sapountzakis and Mokos (2008) and Simo et al. (1984). Lee
et al. (submitted for publication) provided spatial stability theory
for shear-flexible thin-walled beams using both the Engesser’s
and the Haringx’s buckling approach and compared the predictions.
Analytical solutions in dimensionless form were presented for the
spatial lateral buckling of simply supported thin-walled beams
having monosymmetric and non-symmetric cross-sections, and a
parametric study undertaken to compare the prediction using the
different approaches.
An outline of the contents of this paper is as follows. In Section
2, the kinematics for the lateral buckling problem are developed
using the Timoshenko beam analogy coupled with warping taken
to occur in a normal direction to the displaced cross-sectional
plane. The cross-sectional displacements without warping are first
defined using three finite Euler angles and two shear finite

0020-7683/$ - see front matter Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2010.06.012
2826 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

rotations. Many nonlinear and stability analyses for the lateral tor-
sional behaviour of beams are often based on assuming a Hookean
constitutive relationship between Green’s strain tensor and its con-
jugate stress, the second Piola–Kirchhoff stress tensor. Attard
(2003a) showed that such a finite strain constitutive relationship
is incorrect when including shear deformations for an isotropic
material and that the constitutive relationship should be derived
from a consistent finite strain hyperelastic formulation. In Section
3, the hyperelastic constitutive relations for stresses and internal
actions are developed using the strain energy density for a com-
pressible isotropic neo-Hookean material proposed in Attard and
Hunt (2004). The constitutive relationships are developed using
Lagrangian physical stresses defined with respect to the unwarped
displaced cross-sectional plane. Section 4 details the derivation of
the equations of equilibrium for elastic initially straight isotropic
prismatic beams under conservative loads, taking into account
the nonlinear effects of changes in geometry and bending shear.
The equilibrium equations derived are consistent with the equa-
tions presented by Reissner (1989) and the well-known vector rep-
resentation of the beam equilibrium equations. Terms normally
associated with the Wagner effect emerge from a consideration
of warping, and are coupled to the twisting and shear deformations
of the cross-section. The nonlinear theory presented can be used to
predict the static and stability response of thin-walled beams
when displacements are considered to be finite. Warping shear is
ignored here. A closed form solution for the problem of the lateral
buckling of a prismatic monosymmetric beam under pure bending
is presented in Section 5. Section 6 details the derivation for the
flexural–torsional buckling of a simply supported prismatic beam
under axial compression. Expressions for the second variation of
the total potential are developed in Section 7. Generalised non-
dimensional lateral buckling equations for a simply supported
beam are developed in Section 8. Shear deformations are shown
to be important for short spans and beams with soft flexible shear
rigidity. Finally, Section 9 provides a summary of the work in this
paper. Further detailed derivations needed to support the work in
this paper are presented in Appendices A,B,C,D,E,F.
2. Kinematics
3 axis (refer to Fig. 1). Unit vectors in the direction of the coordinate
axes x, y & z are denoted by i1, i2 & i3. Under bending, axial, shear
and torsion deformations excluding warping, it is assumed that
the cross-sectional shape remains unchanged (undergoes no strain
within the cross-sectional plane). The Timoshenko beam hypothe-
sis is adopted with two shear angles of rotation.
We will firstly consider the case of no warping. The directions of
the tangent base vectors in the deformed state can be described
using Euler angles defined as w, h & / as shown in Fig. 1. The Euler
angles are taken as functions of x only. The angle w is related to the
tortuosity of the deformed beam axis, see Love (1944). The angle h
is a bending rotation while / is a twisting angle taken here about
the shear centre with coordinates (ys, zs) and identified in Fig. 1.
The covariant tangent base vectors in the deformed state denoted
by g^i are defined by:
 
^i ¼ dj:i þ uj;i gj
g ð1Þ
Here, i and j are indices which take values of 1, 2 or 3 and are associ-
ated with the x-, y- and z-axes, respectively, dji is the Kronecker delta,
gj are the covariant initial tangent base vectors in the undeformed
state ðg1 ¼ i1 ; g2 ¼ i2 & g3 ¼ i3 Þ and uj;i represents the derivatives of
the uj displacement vector components with respect to the coordi-
nate corresponding to the index i. The convention is adopted where
a repeated index is used to imply summation. A bracketed index indi-
cates that the summation convention is suppressed. A scalar quantity
is denoted by a lowercase italic light symbol such as c or /. A bold
lower case symbol such as u is used to represent a vector while a bold
capital symbol such as R is used to distinguish a second order tensor.
The rotation tensor denoted by the symbol R, rotates a given
vector v about an axis parallel to a normalised axis vector u
(u  u = 1) through an angle 0 6 x 6 2p as depicted in Fig. 2. The
angle is defined as positive using the right hand screw rule where
the thumb of the right hand is extended in the direction of the axis
vector u and the closing fingers define a positive rotation. The rota-
tion vector R is given by:
R ¼ cos xI þ ½1  cos xu  u þ sin xu ð2Þ
To describe the displacement of the cross-section, first consider a
rotation of w about the unit vector i1 in the 1 or x-axis, such that

Consider a straight prismatic beam with the materials lines Rw ¼ cos wI þ ½1  cos wi1  i1 þ sin wi1  ð3Þ
aligned with a Cartesian rectangular coordinate system with origin Applying this rotation to the y- and z-axis gives:
at the centroid. Since the initial axis system is a Cartesian rectangu-
lar system there is no distinction between covariant and contravari- ^ ¼ Rw i3 ¼  sin wi2 þ cos wi3
b
ant tensor components. The longitudinal axis of centroids of the ð4Þ
^ ¼ Rw i2 ¼ cos wi2 þ sin wi3
h
undeformed beam is taken as the x or 1 axis. The cross-sectional
centroidal principal axes are taken as the y or 2 axis and the z or

y−ys
gˆ 1s y,2

t̂ n̂
φ
ĝ2
ω
θ Rv
x,1
v
ψ b̂
z − zs ĝ3
x,1
u
ψ about 1, θ about bˆ , φ about nˆ
bˆ ⋅ tˆ = 0 tˆ × bˆ = nˆ

z,3 O
Fig. 1. Tangent base vectors at the shear centre. Fig. 2. Rotation tensor R.
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2827

^ is orthonormal to b
Here, h ^ and both are in the original {i , i } plane. ^3  i1 ¼ u1;3 ¼ l3 ¼ sin h sin /
g
2 3
n o
The vectors h; ^ b
^ are the {i , i } vectors rotated about the i ^3  i2 ¼ u2;3 ¼ m3 ¼  sin w cos /  cos w sin / cos h
g ð15Þ
2 3 1

through an angle of w. Consider next a bending rotation h about ^3  i3 ¼ 1 þ u3;3 ¼ n3 ¼ cos w cos /  sin w sin / cos h
g
^ such that
the vector b
In the above, u1, u2 & u3 are displacement functions in the x-, y- and
^b
Rh ¼ cos hI þ ½1  cos hb ^ þ sin hb
^ ð5Þ z-directions, respectively. The comma notion symbolises differenti-
ation with respect to the proceeding variable. In Eqs. (13)–(15), the
^ and the unit vector i2 rotated to ^t,
The unit vector i1 is rotated to n displacement components are taken at the shear centre
such that (y = ys, z = zs). Eq. (13) leads to the following:

^ ¼ Rh i1 ¼ cos hi1 þ sin hh

n ^ ¼  sin hi1 þ cos hh
^ ^t ¼ Rh h ^ ð6Þ 1 þ us;x ¼ k1s ðcos h cos uo cos ao  sin h sin uo Þ
v s;x ¼ k1s ðsin h cos w cos uo cos ao
^ taken through the
Finally, a twist rotation / about the vector n þ cos w cos h sin uo  sin w cos uo sin ao Þ ð16Þ
shear centre is applied, defined by:
ws;x ¼ k1s ðsin h sin w cos uo cos ao
^ n
R/ ¼ cos /I þ ½1  cos /n ^ þ sin /n
^ ð7Þ þ sin w cos h sin uo þ cos w cos uo sin ao Þ
so that the tangent base vectors within the displaced cross-sec- where us(x) is the axial displacement of the shear centre and, vs(x)
^2 & g
tional plane ðg ^ 3 Þ are expressed by: and ws(x) are the displacements of the shear centre in the y- and z
-directions, respectively. The stretch of the shear centre axis k1s can
^ þ cos /^t g
^2 ¼ R/^t ¼ sin /b
g ^ ¼ cos /b
^ 3 ¼ R/ b ^  sin /^t ð8Þ be written as:
^ form a moving orthonormal triad frame. The qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The vectors n^ ; ^t & b
k1s ¼ ð1 þ us;x Þ2 þ v 2s;x þ w2s;x ¼ ðc ^1 Þ2 þ ðc ^ 2 Þ2 þ ðc ^ 3 Þ2 ð17Þ
beam shear deformations can be characterised by two rotations of
the normal vector n ^ that lie within the
^ about the vectors ^t & b Eq. (16) can be rearranged thus
plane of the cross-section. We select the first shear rotation as uo
about the vector b ^ and then a rotation a about the vector ^t. The k1s cos uo cos ao ¼ ð1 þ us;x Þ cos h þ ðv s;x cos w þ ws;x sin wÞ sin h
o
shear rotation tensors are then: k1s sin uo ¼ ð1 þ us;x Þ sin h þ ðv s;x cos w þ ws;x sin wÞ cos h
k1s cos uo sin ao ¼ v s;x sin w þ ws;x cos w
Ruo ¼ cos uo I þ ½1  cos uo b ^b ^ þ sin u b ^ ð9Þ
o ð18Þ
Rao ¼ cos ao I þ ½1  cos ao ^t  ^t  sin ao^t ð10Þ
with the shear angles defined by
^ 1s taken at
Applying the shear rotations, the tangent base vector g ð1 þ us;x Þ sin h þ ðv s;x cos w þ ws;x sin wÞ cos h
the shear centre (y = ys, z = zs) is: sin uo ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1 þ us;x Þ2 þ v 2s;x þ w2s;x
^1s ¼ k1s Rao Ruo n
^ ð19Þ
g
  v s;x sin w þ ws;x cos w
^ tan ao ¼
^ þ sin uo^t þ sin ao cos uo b
¼ k1s cos uo cos ao n ð1 þ us;x Þ cos h þ ðv s;x cos w þ ws;x sin wÞ sin h

¼c ^ þc
^1 n ^2 þ c
^2 g ^3
^3 g ð11Þ Expressions (18) and (19) give the normal and shear components of
the longitudinal stretch of the shear centre axis k1s, and the shear
Here k1s is the stretch of the shear centre axis and c
^1 ; c
^2 ; c
^3 are de- angles in terms of the displacement derivatives us,x, vs,x & ws,x and
fined by the Euler angles. The vector components in the x-, y- and z-direc-
tions, of the tangent base vector taken without shear deformations
c^1 ¼ k1s cos uo cos ao (taken normal to the plane of the cross-section, see Eq. (6)) can also
c^2 ¼ k1s sin uo cos / þ k1s sin ao cos uo sin / ð12Þ be used to characterise the shear deformations by noting that
c^3 ¼ k1s sin uo sin / þ k1s sin ao cos uo cos / 1þu  s;x ¼ 1 þ us;x  c1 ¼ k1s cos h

The term c ^1 represents the axial stretch component in the normal v s;x ¼ v s;x  c2 ¼ k1s sin h cos ww s;x ¼ ws;x  c3 ¼ k1s sin h sin w

direction n^ , while c
^2 & c ^3 are measures of the shear deformation ð20Þ
components taken in the tangent base vector directions in the de- where u s;x ; v
 s;x & w
 s;x are the deflection derivatives taken without
formed state, g ^2 & g^3 , respectively. At the shear centre axis, the
shear deformations, c1, c2 & c3 are shear deformations and  k1s is a
direction cosines (l1, l2, l3, m1, m2, m3, n1, n2 and n3) and the deriv- modified stretch defined by:
atives of the displacement functions, can be derived from the fol-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

lowing equations: k1s ¼ 1þu  s;x 2 þ v 2s;x þ w  2s;x ð21Þ
^1s  i1 ¼ 1 þ
g u1;1 ¼ k1s l1 ¼ k1s ðcos h cos uo cos ao  sin h sin uo Þ Eq. (20) leads to the following equations, which relate the bending
^1s  i2 ¼ u2;1 ¼ k1s m1 ¼ k1s ðsin h cos w cos uo cos ao
g angle and tortuosity angle at the shear centre:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ cos w cos h sin uo  sin w cos uo sin ao Þ w 2s;x þ v 2s;x
1þu s;x w s;x
^1s  i3 ¼ u3;1 ¼ k1s n1 ¼ k1s ðsin h sin w cos uo cos ao
g cos h ¼  sin h ¼ tan w ¼ ð22Þ
k1s k1s v s;x
þ sin w cos h sin uo þ cos w cos uo sin ao Þ
The geometric torsion or tortuosity of the shear centre axis is re-
ð13Þ lated to the derivative of the angle w with respect to x, and can
^2  i1 ¼ u1;2 ¼ l2 ¼  sin h cos /
g be derived from the above equations, hence:
^2  i2 ¼ 1 þ u2;2 ¼ m2 ¼  sin w sin / þ cos w cos / cos h
g ð14Þ  
dw d w  s;x v s;x w s;xx  w s;x v s;xx
¼ tan1 ¼ ð23Þ
^2  i3 ¼
g u3;2 ¼ n2 ¼ cos w sin / þ sin w cos / cos h dx dx v s;x w 2s;x þ v 2s;x
2828 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

Also from Eq. (22) we have for the curvature: The evolution of the Darboux type triad sets n ^ and
^ ; ^t & b
0qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 ^; g
n ^2 & g
^ 3 along the axis of the beam is governed by the equations:
dh d w 2s;x þ v 2s;x  
¼ tan1 @ A
~^t þ j
^ ;x ¼ j
n ^ ^t ¼ j
~ 0b ~n^ þ s
d/ ^
b
dx dx 1þu  s;x ;x
dx
   
d/ ^ ð36Þ
ðv s;x v s;xx þ w  s;xx Þð1 þ u
 s;x w s;x Þ  u  s;xx w 2s;x þ v 2s;x ^ ;x ¼ j
b ~ 0n
^  s t
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð24Þ dx
k2 w  2s;x þ v 2s;x ^ ;x ¼ jg^2  j g 0 ^3 g ^2;x ¼ jn^ þ sg
^3 g ^3;x ¼ j0 n
^  sg
^2
1s n
Integrating Eqs. (14) and (15) leads to expressions for the displace- These are associated with rotation vectors given by:
ment functions, that is:  
d/ ^
1 s n ~ 0^t þ j
^ j ~b sn^ þ j0 g^2 þ jg^3 ð37Þ
u ¼ us ðxÞ þ ðy  ys Þl2 þ ðz  zs Þl3 dx
u ¼ v s ðxÞ þ ðy  ys Þðm2  1Þ þ ðz  zs Þm3
2
ð25Þ
3
u ¼ ws ðxÞ þ ðy  ys Þn2 þ ðz  zs Þðn3  1Þ 2.2. Covariant tangent base vectors

The displacement function in vector form is: ^1 can

Using Eqs. (1) and (26), the covariant tangent base vector g
^2  i2 Þ þ ðz  zs Þðg
u ¼ us þ ðy  ys Þðg ^3  i3 Þ now be written in terms of the curvatures and torsion of the shear
^ ^ centre axis as:
¼ us þ ðy  y Þðsin /b þ cos /t  i2 Þ þ ðz  zs Þ
s
^ þy
^  sin /^t  i3 Þ ¼ us þ ~zb ~^t  ðy  ys Þði2 Þ  ðz  zs Þði3 Þ
^1 ¼ ½k1s cos uo cos ao  ðy  ys Þj þ ðz  zs Þj0 n
g ^
 ðcos /b
ð26Þ þ ðk1s sin uo  s~zÞ^t þ ðk1s sin ao cos uo þ sy ^
~Þb
in which, ^1  ðy  ys Þj þ ðz  zs Þj0 n
¼ ½c ^2  sðz  zs Þg
^ þ ½c ^2

~z ¼ ðy  ys Þ sin / þ ðz  zs Þ cos / ^3 þ sðy  ys Þg

þ ½c ^3 ð38Þ
~ ¼ ðy  ys Þ cos /  ðz  zs Þ sin /
y ð27Þ It is convenient to split the longitudinal stretch into its normal and
shear components. These components of longitudinal stretch can be
In the above, us is the displacement of the shear centre axis in space,
derived from Eq. (38) and are:
given by:
g ^ ¼ kn ¼ k1s cos uo cos ao  ðy  ys Þj þ ðz  zs Þj0
^1  n ð39Þ
us ¼ us i1 þ v s i2 þ ws i3 ð28Þ

^1  ^t ¼ kst ¼ k1s sin uo  s~z g

g ^¼k
^1  b sb
2.1. Curvatures and torsion
¼ k1s sin ao cos uo þ sy
~ ð40Þ
The curvatures will be calculated with respect to the moving where kn is the normal component of longitudinal stretch while kst
triad orthonormal vector system n ^ defined at the shear cen-
^ ; ^t & b & ksb are the tangential components of the longitudinal stretch, ta-
tre, as well as, the orthonormal frame n ^; g
^2 & g^3 . The moving triad ^ directions, respectively. The covariant tangent base
ken in ^t & b
orthonormal frame n ^ is analogous to a Darboux frame used
^ ; ^t & b vectors within the plane of the displaced cross-section were defined
in the differential geometry of surfaces. The deformation curva- in Eq. (8). Because we have assumed the cross-sectional shape does
tures with respect to the unit normal n ^ at the shear centre are not change during deformation, the following holds:
given by:
g ^2 ¼ 1 g
^2  g ^3  g
^3 ¼ 1 g
^2  g
^3 ¼ 0 ð41Þ
dh dw
j ¼ n^ ;x  g^2 ¼ cos / þ sin h sin / ð29Þ The first invariant Ik and the Jacobian J of the system are therefore:
dx dx
dh dw 2 2
j0 ^ ;x  g
¼ n ^3 ¼ sin /  sin h cos / ð30Þ Ik  2 ¼ ðkn Þ2 þ ðk1s sin uo  s~zÞ þ ðk1s sin ao cos uo þ sy
~Þ
dx dx
J ¼ kn ð42Þ
The deformation torsion about the unit normal to the cross-section
^ ; g^2 & g
with respect to the triad vectors n ^ 3 is defined by: The Jacobian J is related to the ratio of the volume change. Here the
d/ dw Jacobian is a function of the normal component of the longitudinal
s ¼ g^3;x  g^2 ¼ þ cos h ð31Þ stretch only.
dx dx
The curvatures can also be written with respect to the triad vectors
^ That is, the normal curvature is given by: 2.3. Warping
^ ; ^t & b.
n
dh Here we assume that warping of the cross-sectional plane is a
j~ ¼ n^ ;x  ^t ¼ j cos / þ j0 sin / ¼ ð32Þ
dx function of cross-sectional geometry, proportional to the deforma-
While the geodesic curvature is: tion torsion and occurs about the unit normal to the cross-sectional
dw plane. Thus, we augment Eq. (26) to give the displacement vector
j~ 0 ¼ n^ ;x  b^ ¼ j sin /  j0 cos / ¼ sin h ð33Þ as:
dx
^ the relative or geodesic
^ ; ^t & b,  
With respect to the triad vectors n ^  cos /^t
u ¼ us  ðy  ys Þ i2  sin /b
torsion is defined by  
^ þ sin /^t  sxðy; zÞn
 ðz  zs Þ i3  cos /b ^ ð43Þ
^ ;x  ^t ¼ s  d/
b ð34Þ
dx
where x(y, z) is a cross-sectional warping function. Note in the
Eqs. (31)–(33) produce for the rate of change of the Euler angles:
above, us is now the displacement of the shear centre axis only if
dh dw j~ 0 d/ j~ 0 the warping function is zero at the shear centre (the displacement
¼j
~ ¼ ¼s ð35Þ
dx dx sin h dx tan h of the shear centre is us  sxðys ; zs Þn
^ ). In Appendix A, expressions
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2829

for the displacements are truncated to second order so that compar- defined by the moving orthonormal triad frame n ^ The
^ ; ^t & b.
isons can be made to other lateral buckling formulations which are transformation between the second Piola–Kirchhoff stress tensor
based on second order approximations. The covariant tangent base components and the Lagrangian physical stresses can be estab-
vectors, which include warping, can now be derived from Eq. (43) lished using vector transformation. The normal stresses Snn are ta-
and are given by: ken normal to the cross-sectional plane (in the direction n ^ ) while
the tangential shear stresses Snb & Snt are taken within the cross-
^1x ¼ ðkn  xs;x Þn
g ^2  sðz  zs þ xjÞg
^ ½c ^2 ^ respectively. The transfor-
sectional plane in the directions ^t & b,
^3 þ sðy  ys þ xj0 Þg
þ ½c ^3 ð44Þ mation between the second Piola–Kirchhoff stress tensor compo-
nents and Lagrangian stresses are therefore:
g ^ þ cos /^t  sx;y n
^2x ¼ sin /b ^ 2  sx;y n
^¼g ^ ð45Þ
P11 g^1x þ P12 g^ 2x þ P13 g^3x ¼ Snn n ^
^ þ Snt^t þ Snb b ð57Þ
 
g ^  sin /^t  sx;z n
^3x ¼ cos /b ^3  sx;z n
^¼g ^ ^ g
ð46Þ P12 ¼ Snn n
^ þ Snt^t þ Snb b ^2x ð58Þ
The contravariant base vectors are:  
P13 ¼ Snn n ^ g
^ þ Snt^t þ Snb b ^3x ð59Þ
^1x ¼ n
Jg  2^t þ sx
^ þ sx ^¼n
 3b ^ þ sx;y g
^2 þ sx;z g
^3 ð47Þ
and hence:
^2x ¼ ðc
Jg ^2 þ sðz  zs ÞÞn^ þ ðkn  xs;x Þg ^2
  Snn ¼ P11 g ^ þ P12 g
^1x  n ^ þ P13 g
^ 2x  n ^
^ 3x  n
0 2
^3 s þ ðy  ys þ xj Þs g
þ c ^2 x;z
¼ P11 ðkn  xs;x Þ  P12 sx;y  P13 sx;z
 2

þ c ^2 s þ ðz  zs þ xjÞs g ^3 x;z ð48Þ Snt ¼ P11 g
^1x  ^t þ P12 g
^2x  ^t þ P13 g
^3x  ^t
  ð60Þ
¼ P k1s sin uo  sz þ P cos /  P13 sin /
11 12
^3x ¼ ðc
Jg ^3  sðy  ys ÞÞn ^ þ ðkn  xs;x Þg ^3
  Snb ¼ P11 g
^1x  b^ þ P12 g ^ þ P13 g
^ 2x  b ^
^ 3x  b
^2 s  ðz  zs þ xjÞs g2
þ c ^3 x;y  
  ¼ P11 k1s sin ao cos uo þ sy  þ P12 sin / þ P13 cos /
þ c ^3 s  ðy  ys þ xj Þs2 g
0 ^2 x;y ð49Þ
in which:
in which J is the volume invariant or Jacobian. The stretch of the
beam and the shear deformations are thus calculated as: ¼y
y ~0
~  xj z ¼ ~z þ xj
~ ð61Þ
2 2 2 Incorporating the constitutive relations given in Eq. (56), expanding
^1x ¼ ðk1 Þ ¼ ðkn  xs;x Þ þ ½c
^1x  g
g ^2  sðz  zs þ xjÞ
0 2 to second order in terms of the deformations, we can write the con-
^3 þ sðy  ys þ xj Þ
þ ½c ð50Þ stitutive relations as:
^2x ¼ 1 þ ðsx;y Þ2 g
^ 2x  g ^3x ¼ 1 þ ðsx;z Þ2 g
^3x  g ^3x ¼ s2 x;y x;z
^2x  g    
g Snn ¼ Eðkn  xs;x  1Þ þ ðG þ KÞ s c ^3 x;z þ s2 x
^2 x;y þ c 
ð51Þ  
Sn2 ¼ G c^2  sðz  zs þ x;y  xjÞ Sn3 ð62Þ
The cross-section no-longer remains planar because of warping. The ^3 þ sðy  ys  x;z þ xj0 ÞÞ
¼ Gðc
first invariant Ik and the volume invariant J are updated to including E 2Gg
where G ¼ 2ð1þ gÞ is the shear modulus, E ¼ 2G þ K, K ¼ ð12gÞ is the
warping and are related to the deformations by:
2
Lamé constant, E is the elastic modulus and g is the Poisson’s ratio.
Ik  2 ¼ ðkn  xs;x Þ2 þ ½c
^2  sðz  zs þ xjÞ The material parameter governing the normal stress is not the elas-
 
^3 þ sðy  ys þ xj0 Þ2 þ s2 x2;y þ x2;z
þ ½c ð52Þ tic modulus E as would be expected for a uniaxial stress state, as the
displacements assumed restrain the dilation of the cross-section
  shape (see Attard, 2003a; Attard and Hunt, 2008b). An approxima-
J ¼ kn  xs;x þ s c ^3 x;z þ s2 x
^2 x;y þ c  ð53Þ tion in beam theory is to replace E by E. The alternate shear stresses
in which Sn2 & Sn3 are taken in the g^2 & g
^ 3 directions, respectively, as indi-
cated in Fig. 4. The s2 x
 term in Eq. (62) gives rise to terms which
 ¼ x;z ðy  ys þ xj0 Þ  x;y ðz  zs þ xjÞ
x ð54Þ are analogous to the Wagner effect although unlike conventional
beam theory, the Wagner effect here is related to the warping of
the section (see Alwis and Wang, 1996; Attard, 1986).
3. Hyperelastic constitutive relations for stresses and internal
Integrating over the cross-sectional plane gives for the internal
actions
actions:
Z
The strain energy density function U for a compressible isotro- N¼ Snn dA ¼ EAðc
^1 þ ys j  zs j0  1Þ þ ðG þ KÞAsðzs c
~2  ys c
~3 Þ
pic neo-Hookean material (see Attard and Hunt, 2004) is given by: A

1 1 þ ðG þ KÞðIpo  J t Þs2 ð63Þ

U¼ GðtrðC  IÞ  2 ln JÞ þ Kðln JÞ2 ð55Þ
2 2 Z Z
Here, tr symbolizes the trace of a tensor and C is right Cauchy– Qt ¼ Snt dA ¼ GAk1s sin uo Qb ¼ Snb dA ¼ GAk1s sin ao cos uo
A A
Green deformation tensor. The constitutive relationship for a hyper-
Q 2 ¼ GAc
^2 Q 3 ¼ GAc
^3
elastic material can be established for the second Piola–Kirchhoff
stress tensor P by (see Attard and Hunt, 2004): Z ð64Þ
nn
oU M3 ¼ yS dA
P¼2 ¼ GI  ph C1 ph ¼ G  K ln J ð56Þ 
A
  
oC
¼ EIzz j  b1 s2  ðG þ KÞs C yy c
^2 þ C yz c
^3 ð65Þ
In the above, ph represents a hydrostatic stress. Z
   
A physical Lagrangian stress system is used here, defined with M2 ¼ zSnn dA ¼ EIyy j0  b2 s2 þ ðG þ KÞs C zy c^2 þ C zz c
^3 ð66Þ
respect to the cross-sectional plane before warping and directions A
2830 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

Z h i
The following resultants of the product of the shear stresses Sn2
Mt ¼ Sn3 ðy  ys  x;z þ xj0 Þ  Sn2 ðz  zs þ x;y  xjÞ dA
A with the coordinates y, z, x are then
R Z
¼ GJ t s þ 2GsðC xz j0 þ C xy jÞ þ GIx sððj0 Þ2 þ j2 Þ A
zSn2 dA
¼ zðz  zs þ x;y  xjÞ dA ¼ Iyy þ C zy ð77Þ
ð67Þ Gs A
Z R Z
    ySn2 dA
M- ¼ xSnn dA ¼ EIx s;x  U x s2  ðG þ KÞs C xy c^2 þ C xz c^3 A
¼ yðz  zs þ x;y  xjÞ dA ¼ C yy ð78Þ
A Gs A
ð68Þ R Z
A
xSn2 dA
¼ xðz  zs þ x;y  xjÞ dA ¼ C xy  Ix j ð79Þ
where Izz and Iyy are the second moment of areas about the y and z Gs A
axes, respectively, and Ipo is the polar second moment of area about We see that the warping related geometrical parameters are related
the shear centre, such that to the distribution of the shear stresses due to torsion. By examining
Z Z a state of loading which consists of only an axial force and bending
Izz ¼ y2 dA Iyy ¼ z2 dA Ipo moment internal actions (no twisting moment), the following
Z h A A expressions are derived in Appendix C:
i Z Z
¼ ðy  ys Þ2 þ ðz  zs Þ2 dA ð69Þ
A C yy ¼ yx;y dA ¼ 0 C yz ¼ yx;z dA ¼ Izz
ZA A
Z
In the above equations, M3 and M2 are the resultant bending C zy ¼ zx;y dA ¼ Iyy C zz ¼ zx;z dA ¼ 0 ð80Þ
moments about the deformed g ^3 & g
^ 2 axes passing through the A A
centroid, respectively; Mt is the Saint Venant twisting moment For a loading state that only involves torsion, the nonlinear consti-
component about the unit normal to the cross-section; Mx is the tutive equations, (63)–(68) and a consideration of equilibrium, indi-
bimoment;Jt is the Saint Venant torsion constant, defined by: cate that there would be nonlinear axial shortening and bending
Z curvature under torsion (see Attard, 1986). Appendix D contains
Jt ¼ ðx;y þ z  zs Þ2 þ ðx;z  y þ ys Þ2 dA ð70Þ the derivation of the nonlinear axial shortening and bending curva-
A
ture effects under a loading of only torsion.
N is the axial force acting at the centroid and normal to the de-
formed cross-sectional plane in the direction n ^ , while Qt and Qb 4. Virtual work
are the shear force resultants acting at the shear centre in the direc-
^ respectively. The shear forces Q & Q act in the direc-
tions ^t & b, The virtual work dW in terms of the second Piola–Kirchhoff
2 3
^
tions, g2 & g ^ 3 , respectively. Hence stress tensor is given by:
Z
8Q 2 ¼ Q t cos / þ Q b sin / Q 3 ¼ Q t sin / þ Q b cos / ð71Þ 1
dW ¼ trðPdCÞdV þ dW E ð81Þ
V 2
The warping related geometric parameters are defined by: with V being the volume in the undeformed state, kinematically
Z Z Z admissible variations denoted by the symbol d and the variation
C yy ¼ yx;y dA C yz ¼ yx;z dA C zy ¼ zx;y dA of the external work denoted by dWE. In component form, the trace
ZA ZA A
Z term in Eq. (81) is
C zz ¼ zx;z dA C xy ¼ xx;y dA C xz ¼ xx;z dA ð72Þ 1 1
trðPdCÞ ¼ Pij dg
^ij ¼ Pij g ^j
^ i  dg
A A A 2 2
 11 
¼ P g ^1x þ P12 g ^2x þ P13 g ^3x  dg ^1x
Using the de Saint-Venant (1855) theory of torsion as detailed in
 21 22 23

Love (1944), the following integral can be evaluated as (see also ^ ^
þ P g1x þ P g2x þ P g3x  dg ^ ^2x
Appendix B):  31 32 33

þ P g ^ 1x þ P g ^ 2x þ P g ^3x  dg^3x ð82Þ
Z Z
For beam theory, we ignore the terms associated with P , P23 & 22
x
 dA ¼ ½x;z ðy  ys Þ  x;y ðz  zs Þ dA ¼ Ipo  J t ð73Þ
A A P33, incorporating Eqs. (57)–(59) and taking account of the symme-
try of the stress tensor gives:
The other parameters are defined by: 1  
trðPdCÞ ¼ P11 g
^1x þ P12 g ^2x þ P13 g ^3x  dg ^1x
R R R 2
yx dA zx dA xx  dA E þ P12 g ^2x þ P13 g
^ 1x  d g ^ 3x
^ 1x  d g
b1 ¼ A
 b2 ¼ A
 U x ¼ A
 E ¼
E Izz E Iyy E Ix GþK  
¼ Snn n
^ þ Snt^t þ Snb b^  dg^ 1x
ð74Þ h  i
Z Z Z Z Z ^ g
þ Snn n ^ þ Snt^t þ Snb b ^ 2x g ^ 2x
^1x  dg
x;y dA ¼ zs A x;z dA ¼ ys A x dA ¼ 0 yx dA ¼ 0 zx dA ¼ 0 h  i
A A A A A
þ Snn n ^ g
^ þ Snt^t þ Snb b ^ 3x g ^ 3x
^1x  dg ð83Þ
ð75Þ
Substituting Eqs. (44), (45), (46), (47), (48), (49) and (53), taking the
Appendix B examines the expressions for the parameters in ðk xs Þ
variations, assuming that n J ;x  1, integrating over the area,
Eq. (74) for thin-walled open beams. The geometric parameters
incorporating Eqs. (63)–(68) and using the condition that
in Eq. (72) also appear in Chapter 10 of Vlasov (1959). We can give
physical meaning to these warping related geometric parameters Snt k1s sin ao cos uo ¼ Snb k1s sin uo ð84Þ
by considering the expressions for the shear stresses in Eq. (62) reduces the virtual work equation to:
and restricting our attention to torsion only, hence we have for Z L

^1 þ ys j  zs j0 Þ þ Q 2 dc
Ndðc ^2 þ Q 3 dc
^3
Sn2 & Sn3: dx þ dW E
0 þM 2 dj þ M 3 dj þ ðM t þ M wagner Þds þ M x ds;x
0

n2 n3 0
S ¼ Gsðz  zs þ x;y  xjÞS ¼ Gsðy  ys  x;z þ xj Þ ð76Þ ð85Þ
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2831

with p;x þ dp ¼ 0
Z ð99Þ
 nn   ^1x ðys ; zs Þ  p þ dm ¼ 0
m;x þ g
M wagner ¼ S c ^3 x;z þ sx
^2 x;y þ c  dA ð86Þ
A
The equilibrium equations developed in Eqs. (88)–(93) are consis-
where Mwagner denotes what has been traditionally been associated tent with Eq. (99) and the equations presented by Reissner
with the Wagner effect although here there are added shear terms ^1x ðys ; zs Þ is taken at the shear centre and the
(1989). In Eq. (99), g
and the physical derivations have depended on the presence of warping of the shear centre has been assumed to be zero. The equi-
cross-section warping. Eq. (85) results in the following equilibrium librium equations (87)–(93) can also be recast, in terms of the inter-
equations: nal force and bending moment components taken about the
^2 & g
g ^ 3 , hence:
Q t k1s sin ao cos uo ¼ Q b k1s sin uo ð87Þ
Q 2c
^3 ¼ Q 3 c
^2 ð100Þ
dN
 Q tj ~ 0 þ pn ¼ 0
~  Q bj ð88Þ
dx dN
 Q 2 j þ Q 3 j0 þ pn ¼ 0 ð101Þ
  dx
dQ t d/
þ Nj~  Qb s  þ pt ¼ 0 ð89Þ
dx dx dQ 2
þ Nj  Q 3 s þ p2 ¼ 0 ð102Þ
  dx
dQ b d/
þ Nj~0 þ Qt s  þ pb ¼ 0 ð90Þ dQ 3
dx dx  Nj0 þ Q 2 s þ p3 ¼ 0 ð103Þ
dx
dMe1
e 2j
M ~M e 3j
~ 0 þ mn ¼ 0 ð91Þ e1
dx dM
 ðM 2  Nzs Þj þ ðM 3 þ Nys tÞj0 þ mn ¼ 0 ð104Þ
  dx
dMe3
e 1j
þM e 2 s  d/ þ mb
~0 þ M dðM 3 þ Nys Þ e 0
dx dx  M 1 j þ ðM 2  Nzs Þs þ m2 ¼ Nc
^2  Q 2 c
^1 ð105Þ
dx
¼ Nk1s sin uo  Q t k1s cos ao cos uo
¼ ð1 þ us;x ÞðP y cos w þ Pz sin wÞ þ Px ðv s;x cos w þ ws;x sin wÞ dðM 2  Nzs Þ e
þ M 1 j  ðM 3 þ Nys Þs þ m3 ¼ Nc
^3 þ Q 3 c
^1 ð106Þ
ð92Þ dx

e2  
dM e 1j e3 d/ 5. Lateral buckling of a simply supported prismatic
þM ~M s þ mt
dx dx monosymmetric beam under pure bending
¼ Nk1s sin ao cos uo þ Q b k1s cos ao cos uo
 Consider a simply supported statically determinate beam of
¼ ð1 þ us;x ÞðP y sin w þ P z cos wÞ þ P x ðv s;x sin w þ ws;x cos wÞ sec h
length L under pure bending denoted by M o3 taken about the z-axis.
ð93Þ The initial conditions are that N, Qt, Qb = 0, the bending curvature is
Mo M o3 y
in which constant given by jo ¼ EIzz3 , the initial stress state is Snn
o ¼ Izz and
the initial stretch of the shear centre axis derived from Eq. (63) is
e 1 ¼ M t þ M wagner  dM x
M ð94Þ ko1s ¼ 1  ys jo . The buckling load can be established by looking at
dx the equilibrium equations (88)–(93) under small perturbations or
The terms Px, Py & Pz are internal force results taken in the x, y & z, variations about the initial loaded state. The variation symbol d
respectively. Eq. (87) is always zero if the constitutive relation (64) will be used to indicate small perturbations. Firstly, since there
is used. There are, therefore six equations of equilibrium. Eqs. (88)– are no resultant internal forces at any cross-section under pure
(90) represent the equilibrium of the rate change of the forces along bending if the problem is statically determinate, then under per-
the member while Eqs. (91)–(93) represent the moment increment turbations we have
equilibrium equations. ) dN ¼ 0 dQ t ¼ 0 ) duo ¼ 0 dQ b ¼ 0 ) dao ¼ 0 ð107Þ
The internal action force vector p and resultant moment vector
m are defined with components given in Eq. (63)–(68) and are An important conclusion here is that there are no shear deforma-
tions and no shear forces, involved with the pure bending buckling
^
^ þ Q t^t þ Q b b
p ¼ Nn e 1n
m¼M ^þM e 3b
e 2^t þ M ^ problem. Using the constitutive relations in Eqs. (66) and (67), Eqs.
ð95Þ (91) and (93) become:
^ þ Q 2g
¼ Nn ^2 þ Q 3 g
^3 e 1n
¼M ^ þ M2 g
^2 þ M3 g^3
dd Me1  
¼ EIyy jo  M o3 dj0 ð108Þ
in which dx
e 2 ¼ M2 cos /  M 3 sin /  Nzs ddj0 e 1 jo  M o ds ¼ 0
M EIyy þ dM 3 ð109Þ
ð96Þ dx
e 3 ¼ M2 sin / þ M3 cos / þ Ny
M s
where
Z o

where e 1 ¼ GJ ds  EIx ds;xx  M3

dM t
 Þ dAds
ðy x
zs ¼ ys sin / þ zs cos / y
s ¼ ys cos /  zs sin / ð97Þ A Izz
 o  

¼ GJt  M3 E b1 ds  EIx ds;xx ð110Þ
The distributed force and moment intensity vectors are depicted by
dp and dm, respectively, with components: Differentiating Eq. (108) with respect to x and substituting into Eq.
(109) gives:
dp ¼ pn n ^
^ þ pt^t þ pb b ^
^ þ mt^t þ mb b
dm ¼ mn n
ð98Þ EIyy
2 e1
d dM Mo3 e
^ þ p2 g
¼ pn n ^2 þ p3 g
^3 ^ þ m2 g
¼ mn n ^2 þ m3 g^3   2
þ d M 1  Mo3 ds
EIyy
 1 Mo3 dx EIzz
EIzz
The well-known vector representation of the beam equilibrium
equations are: ¼ 0 Provided EIyy – EIzz ð111Þ
2832 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
Using Eq. (110) we have the following uncoupled fourth order dif- np  EIEIyyyy  EIx nLp 2 þ GJ t
ferential equation: L 1 EI EIyy < EIzz &
M cr ¼ zz
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   2 
2 3  
EIx np 2 GJ t þ EIx nLp < EIzz
 o    1  EIGJzzt  EI L
EIyy EIzz M 3 E b1 GJ t EIx 5 zz

ds;xx 4    M o3 ð121Þ
EIyy
1  EIzz M3 o EI zz EI zz EI zz

  For the case EIyy = EIzz, the relevant equations are again Eqs. (109)
EIyy EIx GJ M o E b1 and (108) which reduce to:
þ ds;xxxx    dsM o3 1  t þ 3 ¼0 ð112Þ
EI
1  yy M o EIzz EIzz o
EIzz 3 ddj0 e 1 M3  Mo ds ¼ 0
EIyy þ dM 3 ð122Þ
dx EIzz
The associated equation for the lateral curvature is: e
dd M 1  
  ¼ GJt  M o3 E b1 ds;x  EIx ds;xxx ¼ 0 ð123Þ
M o3 E b1  GJt ds;x þ EIx ds;xxx dx
0
dj ¼   ð113Þ Applying the boundary conditions and looking at the determinate of
EI
M o3 1  EIyyzz the system of equations, non-trivial buckling moment solutions ex-
ist only for monosymmetric sections and are:
The solution to the standard differential equation Eq. (112), is then: ( np2 )
GJt GJ t  EIzz GJ t þ EIx L
Mcr ¼ ; ; EIyy ¼ EIzz ð124Þ
ds ¼ C 1 sinðk1 xÞ þ C 2 cosðk1 xÞ þ C 3 sinhðk2 xÞ þ C 4 coshðk2 xÞ E b1 E b1 E b1
ð114Þ The first two solutions in the above equation are independent of the
beam length and hence would indicate localised buckling at the sec-
2
dj0 ¼ ½C 1 cosðk1 xÞ  C 2 sinðk1 xÞk1 ðfk1 þ eÞ þ ½C 3 coshðk2 xÞ tion where the moment reaches these critical values. Further dis-
2 cussion about localised torsional buckling can be found in Attard
þ C 4 sinhðk2 xÞk2 ðfk2 þ eÞ ð115Þ
and Lawther (1989). In the lateral buckling analysis of beams, sec-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ondary warping or thickness warping is usually ignored. For angle
2 2
2 bþb  4ac 2 b þ b  4ac and tee beams where the contour warping is zero, the secondary
ðk1 Þ ¼ ðk2 Þ ¼ ð116Þ
2c 2c warping is, however, small but finite. Secondary warping was
  shown in Attard and Lawther (1989), to have a significant effect
 2 GJ t E b1 on the lateral buckling load of simply supported beams under mo-
a ¼ M o3  M o3 1
EIzz EIzz ment gradient.
 
EIyy EIzz E b1 GJ t  2 EIx
b¼  Mo3   M o3
EIyy
1  EIzz EI zz EI zz EIzz ð117Þ 6. Flexural–torsional buckling of a simply supported prismatic
beam under axial compression
EIyy EIx M o3 E b1  GJt EI
c¼  e¼   f ¼  x 
EI
1  EIyyzz
EI
Mo3 1  EIyyzz
EI
Mo3 1  EIyyzz Consider a simply supported beam of length L, whose ends are
prevented from twisting but are free to warp and bend. The beam
is loaded with an axial force No applied at each end of the beam at
The boundary conditions for the simply supported beam are that d
the position of the centroid causing only an axial initial stress
j0 = 0 & ds,x = 0 at x = 0 & L. Using these boundary conditions, the Snn No
o ¼ A . There is therefore an initial axial deformation given by:
determinant of the system of resulting equations is then:
 o  No
h i2 No ¼ EA c ^o1 ¼ 1 þ
^1  1 ) c ð125Þ
EA
sinhðk2 LÞ sinðk1 LÞf 2 ðk1 Þ2 ðk2 Þ2 ðk1 Þ2 þ ðk2 tÞ2 ¼0 ð118Þ
The boundary conditions are:
The solutions to the above equation are: djð0Þ ¼ djðLÞ ¼ 0 dj0 ð0Þ ¼ dj0 ðLÞ ¼ 0 ds;x ð0Þ ¼ ds;x ðLÞ ¼ 0
ð126Þ
np inp 2
k1 ¼ k2 ¼ b  4ac ¼ 0 k1 ¼ 0 k2 ¼ 0 ð119Þ
L L Applying variations to the force equilibrium equations (101)–(103)
and incorporating the constitutive relationships (63) and (64) we
where n is an integer. The appropriate equation for the buckling
conclude the following:
moment ðEIyy – EIzz Þ is therefore:  
ddN
¼0 ^1;x þ ys dj;x  zs dj0;x ¼ 0
dc ð127Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dx
b  4ac np2 np4 np2
2
2 bþ ddQ 2 No
ðk1 Þ ¼ ¼ ) c b þa¼0 þ No dj ¼ 0 ) dc ^2;x ¼ dj ð128Þ
2c L L L dx GA
o
) ddQ 3 N
   No dj0 ¼ 0 ) dc ^3;x ¼ dj0 ð129Þ
EI EI np4  o 2 GJ t E b1 dx GA
 yy x þ M3  M o3 1 Now turning to the variations of the equilibrium equations (104)–
EI
1  EIyyzz L EIzz EIzz
(106), substituting the constitutive relations (65)–(68), incorporat-
8 9 ing Eqs. (86) and (94) and the above results, Eqs. (127)–(129), we
< EI EI    
 o 2 EIx =np2
yy zz o E b1 GJ t derive:
   M3   M3 ¼0
: 1  EIyy EIzz EIzz EIzz ; L
EIzz e1
dd M
þ No zs dj þ No ys dj0 ¼ 0
ð120Þ dx
 
ðIpo  J t Þ No
Eq. (120) is a cubic equation for the buckling moment and is inde- ) GJt ds;x  EIx ds;xxx þ No ds;x þ No zs 1  dj ð130Þ
A GA
pendent of any shear rigidity. Standard solutions are available for  o
the cubic problem. For the case of a cross-section with b1 ¼ 0, Eq. N
þ No ys 1  dj0 ¼ 0
(120) reduces to a quadratic equation with the classical solution. GA
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2833

dðdM 3 þ dNys Þ 7. Second variation of total potential/work

 No zs ds ¼ No dc ^o1
^2  dQ 2 c
dx     ð131Þ
2 E Closed form solutions for beams with moment gradient can only
) EIzz dj;xx þ ðNo Þ  1  No dj ¼ No zs ds;x
G be derived using approximate methods. Approximate solutions to
the buckling problem can be obtained by exploiting numerical
dðdM 2  dNzs Þ methods such as the Rayleigh–Ritz method or the finite element
 No ys ds ¼ No dc ^o1
^3 þ dQ 3 c method. These methods can be used to determine the buckling load
dx     ð132Þ by either manipulating the variation of the equilibrium differential
E
) EIyy dj0;xx þ ðNo Þ2  1  No dj0 ¼ No ys ds;x equations or by using the second variation of the total potential.
G
Appendix F contains the linearized lateral buckling equations for
Eqs. (130)–(132) are differential equations in the three variables dj, a prismatic monosymmetric beam under transverse loading. A sta-
dj0 & ds. These equations are uncoupled if the shear centre coin- tic equilibrium configuration is considered stable if the second var-
cides with the centroid. Adopting here a similar technique used iation of the total potential with respect to all kinematically
by Vlasov (1959), we first select the following eigenfunctions: admissible variations is positive definite. An analogous approach
based on the second variation of work was proposed by Bažant
npx npx npx
dj ¼ A sin dj0 ¼ B sin ds;x ¼ C sin ð133Þ and Cedolin (1991). The second variation of the total potential/work
L L L is analogous to taking a variation of the equilibrium equations and
then applying the principle of virtual displacements transforming
which satisfy the boundary conditions (126) and where A, B and C
the equilibrium equations into variational form, see Vlasov (1959).
are scalar coefficients, and n is an arbitrary positive integer. Assum-
The virtual work expression derived in Eq. (85) is rewritten
ing axial compression with No = P, substituting Eq. (133) into the
here:
three Eqs. (130)–(132), the determinate of the resulting system of
equations is Z L" ^1 þ ys j  zs j0 Þ þ Q 2 dc
#
Ndðc ^2 þ Q 3 dc
^3
     dW ¼ dx þ dW E
 2
  0 þM 2 dj0 þ M3 dj þ ðMt þ M wagner Þds þ Mx ds;x
 EIzz þ P GE  1 þ P 0 Pzs 
     ð138Þ
  
 0 2 E
EIyy þ P G  1 þ P Pys ¼0
 
 
      2 ð I J Þ  Consider a prismatic beam with conservative loads, which in the
 P
Pzs 1 þ GA P
Pys 1 þ GA GJ t þ EIx nLp  P A 
po t

unbuckled state has initial internal actions N o ; Q o2 ; Q o3 ; Mo2 & Mo3

ð134Þ o M o3 y M o2 z
and an initial longitudinal stress Snn N
o ¼ A  Izz
þ Iyy
prior to instabil-
This leads to a fifth order polynomial in P, which can be solved for ity. The loading causes no torsion prior to buckling. The initial axial
the critical buckling load. When the shear centre coincides with the deformation, bending curvature and shear deformations prior to
centroid, the buckling load is uncoupled and consists of either the instability are not taken into account in this paper but are left for
modified Haringx column buckling load PHar about either of the future examination. Taking a variation of Eq. (138) and incorporat-
principal axis y or z (see Attard and Hunt, 2008a), or the torsional ing the initial conditions gives for the second variation of work. That
buckling load Px defined below: is:

2 3
1
Z 2
dNdð^1 þ ys  zs 0 Þ þ 12 dQ 2 d^2 þ 12 dQ 3 d^3 þ 12 dM 2 d 0 þ 12 dM 3 d
c j j c c j j Z Z
L 6 7
d2 W ¼ 6
4 þ 12 dM t d þ 12 dMx d ;x þ No d2 ð^1 þ ys  zs 0 Þ þ Q o2 d2 ð^2 Þ þ Mo3 d2
s s c j j c j7
5dx  q  d2 u dS ð139Þ
0 R   S
þQ o3 d2 ð^3 Þ þ Mo2 d2 0 þ 12 A Snn
c j ^ ^
o d d 2 ;y þ 3 ;z þ So
nn
s cx  ðd Þ2 dA
cx x s

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi!
PHar 1 4n2 Peuler E In the above, q is the vector of conservative surface tractions acting
¼   1 1 1 ð135Þ on the boundary surface defined by the domain S. Incorporating the
EA 2 1  GE EA G
constitutive law from Eqs. (63)–(68), and using Eqs. 73, 74, 75 and
n2 p2 EIzz n2 p2 EIyy (80), we have:
Peuler ¼ 2
or ð136Þ
L L2
n2 p2 EI
GJ t þ L2
x
ðIpo  J t Þ
Px ¼ r 2po ¼ ð137Þ
r2po A

21 2 2 2 3
2
^1 þ ys j  zs j0 ÞÞ þ 12 GAðdc
EAðdðc ^3 Þ þ 12 EIyy ðdj0 Þ2
^2 Þ þ 12 GAðdc
Z L6
6 þ 12 EIzz ðdjÞ2 þ 12 GJt ðdsÞ2 þ 12 EIx ðds;x Þ2 þ No d2 ðc ^1 Þ
7
7 Z Z
2 6 7
d W¼ 6  o o
 2   o o
 2 0 1 7 dx  q  d2 u dS ð140Þ
0 6 þ M 3 þ N ys d j  2 dðc 1 ^3 Þds þ M 2  N zs d j  2 dðc ^2 Þds 7 S
4   5
þQ o2 d2 ðc ^3 Þ þ 12 No IpoAJt  Mo3 b1 E  M o2 b2 E ðdsÞ2
^2 Þ þ Q o3 d2 ðc

The expression for the torsional buckling load Px is different to con- Using Eqs. (160), (161) and (158), we can write for the first variation
ventional equations with the polar second moment of area replaced of the deformation terms to first order:
by Ipo  Jt. As shown in Attard (2003a) torsional buckling is impos-
dj ¼ dx3;x dj0 ¼ dx2;x ds ¼ dx1;x ds;x ¼ dx1;xx
sible for a cross-section which does not warp such as circular sec- ð141Þ
tions with Ipo = Jt. dc
^1 ¼ dus;x dc
^2 ¼ dc2 dc
^3 ¼ dc3
2834 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

The second variation deformation terms are required to second or- 8. Generalised non-dimensional lateral buckling equation
der for consistency, that is:
We now derive a generalised non-dimensional equation for the
d2 j ¼ dx1 dx2;x  dx3 dus;xx  dx3;x dus;x
lateral buckling of a prismatic simply supported monosymmetric
 
d2 j0 ¼ dx1 dx3;x  dx2 dus;xx  dx2;x dus;x beam b2 ; zs ¼ 0 under either uniform bending or transverse load-
d2 c
^2 ¼ dx1 dc3 d2 c
^3 ¼ dx1 dc2 ð142Þ ing. The method adopted here follows that in Attard (1990). The
second variation of total potential for this problem is derived from
1 1
^1 ¼ ðdx2 Þ2 þ dx2 dc3 þ dx3 dc2 þ ðdx3 Þ2
d2 c Eq. (146) and is:
2 2 2
2 2 2 2
3
Z L GAðdc3 Þ þ EIyy ðdx2;x Þ þ GJ t ðdx1;x Þ þ EIx ðdx1;xx Þ
1 6 7
d2 u1 ffi ðy  ys Þdx1 dx2 þ ðz  zs Þdx1 dx3 d2 W ¼ 6 þ2Q o2 ðdc3  dx2 Þdx1 þ M o3 ð2dx1;x dx2  dc3 dx1;x Þ 7dx
2 0 4 2
5
1 M o3 b1 E ðdx1;x Þ þ qoy ey dx21
 xðdx2;x dx3  dx2 dx3;x Þ
2 1
1 1   þ P oy ay dx
 21 ð147Þ
2 2
d u ffi ðz  zs Þdx2 dx3  ðy  ys Þ dx21 þ dx23 2
2 2 ð143Þ o
with Py being a transverse force applied in the y-direction at mid-
þ xdx1;x dx2 span and x  1 being the twist at the mid-span. The following substi-
1 1   tutions are firstly made:
d2 u3 ffi ðy  ys Þdx2 dx3  ðz  zs Þ dx21 þ dx22
2 2 x Mcr m;X
 xdx1;x dx3 X¼ M o3 ¼ M cr mðXÞ Q o2 ¼ 
L L ð148Þ
dc3 ¼ c
3 hðXÞ dx1 ¼ x  1 gðXÞ dx2 ¼ x  2 f ðXÞ
The load potential expression is then:

2 3
Z Z Z Z q1 ðy  ys Þdx1 dx2 þ q1 ðz  zs Þdx1 dx3  12 q1 xðdx2;x dx3  dx2 dx3;x Þ
6   7
q  d2 u dS ¼ 4 þ 12 ½q2 ðz  zs Þ þ q3 ðy  ys Þdx2 dx3  12 q2 ðy  ys Þ dx21 þ dx23 5 dS ð144Þ
S S  
 12 q3 ðz  zs Þ dx21 þ dx22 þ ðq2 xdx2  q3 xdx3 Þdx1;x

Here, we consider conservative tractions, which produce only bend-
ing and shear (no bimoments or torsion prior to buckling), and are
applied either at the member ends or are uniformly distributed
along the length of the beam. Incorporating Eqs. (141) and (142)
into Eq. (140), and evaluating Eq. (144) at the boundaries gives:
where m(X) is a function describing the bending moment distribu-
tion; x 2 is the variation of the lateral bending rotation of a charac-
teristic point along the beam; c 3 is the variation of the shear
rotation evaluated at a characteristic point; f(X), g(X) & h(X) are
functions describing the distribution of the deformations which

2 3
EAðdus;x þ ys dx3;x  zs dx2;x Þ2 þ GAðdc2 Þ2 þ GAðdc3 Þ2 þ EIyy ðdx2;x Þ2
6 7
6
6 þEIzz ðdx3;x Þ2 þ GJ t ðdx1;x Þ2 þ EIx ðdx1;xx Þ2 þ 2Q o2 dc3 dx1  2Q o3 dc2 dx1 7
7
6 n o 7
Z L6 þNo ðdx2 Þ2 þ 2dc3 dx2 þ 2dc2 dx3 þ ðdx3 Þ2 7  o   
L
1 6    7 M 3 þ No ys dx1 dx2  M o2  No zs dx1 dx3
6 7
d2 W ¼ 6  Mo3 þ No ys 2dx1 dx2;x þ 2dx3 dus;xx þ 2dx3;x dus;x þ dc3 dx1;x 7dx þ    
2 0 6    7 þ 12 Q o2 ay dx21 þ dx23 þ 12 Q o3 az dx21 þ dx22 0
6 þ Mo2  No zs 2dx1 dx3;x  2dx2 dus;xx  2dx2;x dus;x  dc2 dx1;x 7
6   7
6 I J 7
6
4 þ No poA t  Mo3 b1 E  M o2 b2 E ðdx1;x Þ2 7
5
o
 2 2
 o
 2 2

þqy ey dx1 þ dx3 þ qz ez dx1 þ dx2
ð145Þ

where qoy & qoz are transverse uniformly distributed loads acting satisfy the kinematic boundary conditions. The term Mcr refers to
through the shear centre in the y- and z -directions, respectively, the critical buckling moment. Substituting Eq. (148) into (147),
and act at a distance, ey & ez, respectively, from the shear centre. we have for the second variation of the total potential
The terms ay & az refer to the distance in the y- or z-direction, respec- 2 EIyy 32 3
tively, of the applied boundary forces whose line of action passes L
C5 0 Mcr C 6 x
2
1
through the shear centre. We can reduce Eq. (145) further by ½x
2  1 6
c3 x 4 0 GALC 7
76  7
M cr C 4 54 c3 5 ð149Þ
2
restricting ourselves to the case of no initial axial force and ignoring Mcr C 6 M cr C 4 C x
1
the axial displacement terms. Hence, Eq. (145) simplifies to:

2 3
GAðdc2 Þ2 þ GAðdc3 Þ2 þ EIyy ðdx2;x Þ2 þ EIzz ðdx3;x Þ2 þ GJt ðdx1;x Þ2
Z L6 7

1 6 þEIx ðdx1;xx Þ2 þ 2Q o2 ðdc3  dx2 Þdx1 þ Mo3 ð2dx1;x dx2  dc3 dx1;x Þ 7     L
d2 W ¼ 6 7dx þ 1 Q o ay dx2 þ dx2 þ 1 Q o az dx2 þ dx2 ð146Þ
2 6 2Q o ðdc þ dx Þdx  Mo ð2dx dx þ dc dx Þ 7 2 2 1 3
2 3 1 2
0 4 3 2 3 1 2 1 3;x 2 1;x 5 0
 o   o  
 2 o
 2 2
 o
 2 2

 M 3 b1 E þ M2 b2 E ðdx1;x Þ þ qy ey dx1 þ dx3 þ qz ez dx1 þ dx2
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2835

in which To examine the importance of shear deformations during lateral

buckling, consider Eq. (151) but with the parameter W set to zero,
GJ EIx M cr  
C¼ C2 þ 3 C1  b E C 3 þ P oy aoy þ qoy Leoy C 7 hence
L L L 1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z 1 Z 1
M cr L ð1 þ K 2 A4 Þ
C1 ¼ ðg ;XX Þ2 dXC 2 ¼ ðg ;X Þ2 dXC 3 pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ A1 p ð154Þ
0 0 EIyy GJ ð1 þ SA6 Þ
Z 1 Z 1 Z 1
1
¼ mðg ;X Þ2 dXC 4 ¼ m;X gh dX þ mhg ;X dX ð150Þ We see that the shear parameter decreases the buckling capacity.
0 0 0 2
Z 1 Z 1
For a stiff shear rigidity GA or a long span the S parameter would
C5 ¼ ðf;X Þ2 dXC 6 ¼ ðmgÞ;X f dXC 7 be much less than unity and shear deformations would have little
0 0 impact on the capacity. Only for very short spans with flexible shear
Z 1 Z 1
2 rigidity would shear deformations have an impact on the buckling
¼ g 2 dXC 8 ¼ h dX
0 0
capacity.

Setting the determinant of (149) to zero gives for the critical load:
9. Summary
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A1 p W W2 þ 4ð1 þ K 2 A4 Þð1 þ SA6 Þ A consistent finite strain hyperelastic constitutive model for a
Mcr L
pffiffiffiffiffiffiffiffiffiffiffiffi
ffi¼ ð151Þ linearly elastic isotropic material was used to develop the equilib-
EIyy GJ 2ð1 þ SA6 Þ
rium and buckling equations for the lateral buckling of a prismatic
straight beam which included shear deformations. Several assump-
with the following defined parameters: tions were made and are summarised here.
p2 EIx p2 EIyy
K2 ¼ 2
S¼  The cross-sectional shape was assumed to be maintained during
GJL GAL2 deformation.
sffiffiffiffiffiffiffiffiffiffiffiffiffi
  p2 EIyy  The stresses associated with maintaining the cross-sectional
W ¼ b1 E A3  ay A7  ey A8 ð152Þ shape P22, P23 & P33 were ignored.
GJL2
 The assumed cross-sectional displacements were based on
pffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffi three finite Euler angles and two shear finite rotations.
C2 C5 C3 C5 C1  Warping was taken normal to the displaced cross-section plane.
A1 ¼ A3 ¼ A4 ¼ 2
pC 6 pC 6 C2 p C2  The constitutive relations were derived assuming the volume
sffiffiffiffiffiffi sffiffiffiffiffiffi invariant J was close to unity.
C 24 C 5 4 C5 8C 7 C 5  The constitutive relations were expanded to second order in
A6 ¼ A7 ¼ A ¼ ð153Þ
p2 C 26 C 8 pC 6 C 2 8 pC 6 C 2 terms of the curvatures, torsion, shear rotations and stretch.
p2 EIyy
 The material parameter2G + K was replaced by the elastic mod-
The non-dimensional shear parameter S ¼ GAL2
was introduced by ulus E.
Lee et al. (submitted for publication) to study parametrically the  The deformation measure ðkn Jxs;x Þ was taken as close to unity.
influence of the shear stiffness on the lateral buckling load. For most
beam sections made of an isotropic homogenous material, the shear Second order approximations to the displacements, curvatures
parameter will be very small, varying between about 0.0001–0.1. and twist were also derived. Lagrangian physical stresses were de-
For example, the S parameter for a rectangular section with a Pois- fined with respect to the unwarped displaced cross-sectional
son’s ratio of 0.3 would be 2.14(b/L)2 where ‘‘b” is the width of the plane. Because warping was taken normal to the displaced cross-
section. The buckling coefficients A1, . . ., A8 can be approximated by sectional plane, the constitutive relationships for the bending mo-
solving the buckling problem numerically and integrating using the ments, torsion and bimoment revealed new coupling terms which
derived eigenmode shape. Table 1 lists the buckling coefficients are functions of the cross-sectional warping and shear deforma-
estimated using trigonometric functions, which satisfy the bound- tions. Terms normally associated with the Wagner effect were
ary conditions and the Rayleigh–Ritz approach for three loading derived from a consideration of warping and were coupled to the
conditions. twisting and shear deformations of the cross-section. A closed form

Table 1
Buckling coefficients.

Mcr A1 A3 A4 A6 A7 A8 Loading
Mcr 1 1 1 – – – Mcr Mcr

qoy L2 1.13 0.53 1 0.66 – 0.92 q oy

8

ey

Poy L 1.36 0.65 1 1.06 1.11 – P yo

4

ay
2836 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

solution for the lateral buckling load for a prismatic monosymmet-

ric beam under pure bending was presented. An important conclu-
sion was that the lateral buckling load under pure bending for
simply supported boundary conditions, involves no shear. The flex-
ural–torsional buckling problem under axial compression was also
solved. The Euler type column buckling formulas were consistent
with Haringx’s column buckling formula which included shear.
The torsional buckling formula was different to conventional equa-
tions with the polar second moment of area replaced by Ipo  Jt. The
torsional buckling formula agrees with Attard (2003a) where it
was shown that torsional buckling is impossible for a cross-section
which does not warp such as for a circular section. The second var-
iation of the total potential was also derived. The effects of shear
deformations were explored by examining the non-dimensional
lateral buckling equation for a simply supported beam. The lateral
buckling capacity of most isotropic prismatic straight beams are
unaffected by shear deformations. Shear deformations do become
important for short spans and soft flexible shear rigidity. The effect
of pre-buckling deformations, warping shear and buckling under
torsional loading have not been examined in this paper but are Fig. 4. Direction of physical Lagrangian stresses.

subjects for future work.

nents to second order in the 1, 2 & 3 directions are then obtained

Appendix A. Second order approximation for displacement
from Eq. (43), that is:
components and internal actions
u1 ffi us  ðy  ys Þðx3 ð1  us;x Þ  x1 x2 Þ
Here the expressions for the displacements and internal actions
þ ðz  zs Þðx2 ð1  us;x Þ þ x1 x3 Þ
will be truncated to second order so that comparisons can be made

to other lateral buckling formulations which are based on second 1

 xðy; zÞ x1;x þ ðx2;x x3  x2 x3;x Þ
order approximations. Truncated expressions for the displace- 2
 
ments are also useful for finite element buckling formulations. 1
u ffi v s  ðz  zs Þ x1  x2 x3
2
Consider Eqs. (20)–(22). The angles h & w can be written as: 2 ð158Þ
0qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 1  2 
w 2s;x þ v 2s;x    ðy  ys Þ x1 þ x23 þ xðy; zÞx1;x x2

h ¼ tan 1 @ A w ¼ tan1 ws;x ð155Þ 2  
1þu  s;x v s;x 1
u3 ffi ws þ ðy  ys Þ x1 þ x2 x3
2
For small rotations ðv s;x ; w
 s;x Þ we also assert that the twist about the 1  2 2

1 axis is denoted by x1 such that approximately:  ðz  zs Þ x1 þ x2  xðy; zÞx1;x x3
2
x1 ¼ / þ w ð156Þ The shear angles uo & ao and shear deformations c
^2 & c
^3 can also be
approximated using Eqs. (18) and (19), by:
Also, we set:
c2 x3  c3 x2 c2 x2 þ c3 x3
 s;x x3 ¼ v s;x
us;x ¼ u  s;x ) c2 ¼ v s;x  x3
x2 ¼ w uo ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ao ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð159Þ
ðx2 Þ2 þ ðx3 Þ2 ðx2 Þ2 þ ðx3 Þ2
c3 ¼ ws;x þ x2 ð157Þ

Which for small rotations ðv s;x ; w

 s;x Þ represent rotations about the 1 1
c^1 ¼ k1s cos uo cos ao ffi 1 þ us;x þ ðx2 Þ2 þ x2 c3 þ x3 c2 þ ðx3 Þ2
z- and y-axes, respectively (see Fig. 3). The displacement compo- 2 2
c^2 ¼ k1s sin uo cos / þ k1s sin ao cos uo sin / ffi c2 þ c3 x1
c^3 ¼ k1s sin uo sin / þ k1s sin ao cos uo cos / ffi c3  c2 x1
ð160Þ
Likewise, the curvature and torsion defined above in Eqs. (29)–(31),
can also be approximated.

j ffi x3;x ð1  us;x Þ  x1 x2;x  x3 us;xx

j0 ffi x2;x ð1  us;x Þ þ x3;x x1  x2 us;xx ð161Þ
1
s ffi x1;x þ ðx3 x2;x  x2 x3;x Þ
2
The constitutive relationships (63)–(68) can be simplified to second
order terms as:
0 1
ð1  ys x3;x þ zs x2;x Þus;x  x2;x ðys x1 þ zs Þ
B   C
N ffi EA@ x3;x ðzs x1  ys Þ þ 12 x22 þ x23 A
ð162Þ
þx2 ðc3 þ zs us;xx Þ þ x3 ðc2  ys us;xx Þ

Fig. 3. Rotations at the shear centre.

þðG þ KÞAðzs c2  ys c3 Þx1;x þ ðG þ KÞðIpo  Jt Þx21;x
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2837

z-axis is denoted by a. The magnitude of the perpendicular distance

from the tangent at the contour, to the shear centre (see Fig. 5) is
represented by:
hðn; pÞ ¼ n þ hð0; pÞ ¼ ðy  ys Þ cos a  ðz  zs Þ sin a ð168Þ
The magnitude of the perpendicular distance from the normal at the
contour to shear centre is given by:
rðpÞ ¼ ðy  ys Þ sin a þ ðz  zs Þ cos a ð169Þ
From Eqs. (168) and (169) the rectangular coordinates can be ex-
pressed in terms of the curvilinear coordinates as:
ðy  ys Þ ¼ hðn; pÞ cos a þ rðpÞ sin a
ð170Þ
ðz  zs Þ ¼ hðn; pÞ sin a þ rðpÞ cos a
The radial distance R from the shear centre to the any point within
the cross-section can be written in either the curvilinear coordinate
system or the rectangular system, that is:

R2 ¼ rðpÞ2 þ hðn; pÞ2 ¼ ðy  ys Þ2 ðz  zs Þ2 ð171Þ

The warping function for a thin-walled open member (Attard, 1984)
Fig. 5. Cross-section of a thin-walled beam showing curvilinear profile coordinate is given by
system. Z p
xðn; pÞ ¼ hð0; pÞ dp  nrðpÞ ð172Þ
D
Q 2 ffi GAðc2 þ c3 x1 Þ Q 3 ffi GAðc3  c2 x1 Þ ð163Þ Rp
Consisting of a contour warping term D hð0; pÞ dp and a thickness
  warping term nr(p). The function x
 defined in Eq. (54) can be sim-
M3 ffi EIzz ð1  us;x Þx3;x  x2;x x1  x3 us;xx  b1 x21;x plified using Eqs. (170) and (171), and assuming that the thin-
walled cross-section is composed of straight elements so that
 ðG þ KÞx1;x ðC yy c2 þ C yz c3 Þ da
  ð164Þ dp
¼ 0, we can write:
M2 ffi EIyy ð1  us;x Þx2;x þ x3;x x1  x2 us;xx  b2 x21;x
x
 ¼ x;z ðy  ys Þ  x;y ðz  zs Þ
þ ðG þ KÞx1;x ðC zy c2 þ C zz c3 Þ
¼ hðn; pÞðx;z cos a þ x;y sin aÞ þ rðpÞðx;z sin a  x;y cos aÞ
 
1 ¼ hðn; pÞx;p  rðpÞx;n
M t ffi GJ t x1;x þ ðx3 x2;x  x2 x3;x Þ þ 2GðC xz x1;x x2;x    
2 da 2
þ C xy x1;x x3;x Þ ð165Þ ¼ hðn; pÞ hð0; pÞ 1  n  n þ rðpÞ
dp
  ¼ hðn; pÞ2 þ rðpÞ2  2nhðn; pÞ
1
M - ffi EIx x1;xx þ ðx3 x2;xx  x2 x3;xx Þ  U x x21;x  ðG
2 ¼ R2  2nhðn; pÞ ð173Þ
þ KÞx1;x ðC xy c2 þ C xz c3 Þ ð166Þ
Integrating over the area gives:
The expression for the Wagner component of the twisting moment Z Z
Mwagner, Eq. (86), can be reduced to second order using Eqs. (160), x
 dA ¼ ½R2  2nhðn; pÞ dA ¼ Ipo  J t ð174Þ
(161), (72), (75) and (80), as well as the expression for the normal A A
stress in Eq. (188). Hence:
  Note for a thin-walled open section the torsion constant is defined
 R
Ipo  J t by Jt ¼ A 2nh dA.
Mwagner ffi E M3 b1  M 2 b2 þ Mx U x þ N
A The parameters defined in Eq. (74) can be determined using Eq.
 
1  (173), therefore:
 x1;x þ x3 x2;x  x2 x3;x Z Z
2
  E Izz b1 ¼ yx  dA ¼ yðR2  2nhÞ dA
M x C xy
 M 2  Nzs þ ðc2 þ x1 c3 Þ AZ A

Ix
  ¼ y3 þ yz2 dA  2Izz ys
M x C xz
 M 3 þ Nys þ ðc3  x1 c2 Þ ð167Þ ZA
Ix
 2nhðh cos a þ r sin a þ ys Þ dA
Z A 
Appendix B. Monosymmetric parameters for a thin-walled open ¼ y3 þ yz2 dA  2Izz ys
A
beam Z
 ys J t  2nhðh cos a þ r sin aÞ dA ð175Þ
Consider a thin-walled open profile as shown in Fig. 5. The mid- A

dle surface of the cross-section denotes the contour of the cross- The term in brackets relates to the conventional definition for the
section. A curvilinear coordinate system p, n is shown, where p is monosymmetric parameter b1.
the profile coordinate measured from a pole D on the middle sur- Z
1
face and n is the normal thickness coordinate. The wall thickness is b1 ¼ y3 þ yz2 dA  ys ð176Þ
t(p). The angle between the tangent at the contour and the positive 2Izz A
2838 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

The other parameters in Eq. (74) are: Using the above and Eqs. (32) and (33) we can relate the torsion to
 Z  the deflections in the axial, as well as transverse y- and z-directions,
E Iyy b2 ¼  z3 þ y2 z dA þ 2Iyy zs þ zs J t hence:
ZA dw
þ 2nhðh sin a þ r cos aÞ dA ð177Þ j sin /  j0 cos / ¼ sin h
A
dx
Z Z   v s;x wq
s;xx  w  s;x v s;xx
E Ix U x ¼ xR2 dA þ 2nhx dA ) b1 sin /  b2 cos / s2 ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð185Þ
A A k1s w  s;x þ v 2s;x
2

dh
Appendix C. Wagner twisting moment j cos / þ j0 sin / ¼
dx
Eq. (86) for the twisting moment produced by the normal stres-  
) b1 cos / þ b2 sin /
ses is rewritten here:
Z
 
ðv s;x v s;xx þ w  s;xx Þð1 þ u
 s;x w  s;x Þ  u  s;xx w 2s;x þ v 2s;x
s 2
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k21s w  2s;x þ v 2s;x
 nn   ð186Þ
M wagner ¼ S c ^3 x;z þ sx
^2 x;y þ c  dA ð178Þ
A

We can derive certain equations to be satisfied by the warping func- Appendix E. Expression for normal stress
tion by examining this expression. Consider the case of normal
stresses only involving an axial force N at the centroid and variable Using Eqs. (62)–(68), we can write the expression for the nor-
bending about the z-axis at the shear centre, M3 which only causes a mal stress as
transverse and axial displacement of the shear centre. There will
N M3 y M2 z Mx x  
also be a shear Q2 associated with the variable bending moment Snn ¼  þ  þ E yb1 þ zb2  xU x s2
M3 and thus, ðc ^2 – 0c
^3 ¼ 0Þ. Hence, there must be bending about A Izz Iyy Ix

the centroidal y axis of M2 = Nzs. The normal stress is then given by I  Jt 2

þ ðG þ KÞsðzs c
^ 2  ys c^3 Þ  ðG þ KÞ po s
A
N M 3 y Nzs z
Snn ¼  þ ð179Þ ðG þ KÞys   ðG þ KÞzs  
A Izz Iyy  C yy c
^2 þ C yz c
^3  C zy c
^2 þ C zz c
^3
Izz Iyy
Substituting this into the second order twisting moment defined in ðG þ KÞxs  
 C xy c^2 þ C xz c
^3
Eq. (178) and assuming that this should be zero, we have: Ix
 
  Z  Z þ ðG þ KÞ sc^2 x;y þ sc ^3 x;z þ s2 x
 ð187Þ
Nzs M
M wagner ¼ c
^2 Iyy þ zx;y dA ^2 3
c yx;y dA ¼ 0
Iyy A Izz A Using Eqs. (160), (161), (75), (72) and (80), the expression for the
ð180Þ normal stress can be reduced to second order as:
N M3 y M2 z Mx x  
Using Eq. (75) we can conclude that Snn ffi  þ  þ E yb1 þ zb2  xU x ðx1;x Þ2
A Izz Iyy Ix
Z Z
h
yx;y dA ¼ 0 zx;y dA ¼ Iyy Ipo  Jt
ð181Þ  ðG þ KÞ ðx1;x Þ2 þ ðG þ KÞ x ðx1;x Þ2
A A A
   

xC xy xC xz
Similarly, by considering bending about the y-axis we can conclude: þ x1;x c2 zs þ z þ x;y  þ x1;x c3 ys  y þ x;z 
Ix Ix
Z Z ð188Þ
zx;z dA ¼ 0 yx;z dA ¼ Izz ð182Þ
A A

Appendix F. Linearized lateral buckling equations for a

Appendix D. Nonlinear axial shortening and bending under prismatic monosymmetric beam under transverse loading
torsion
The buckling analysis of statically determinate prismatic
Consider the case of a beam under pure torque, where there is monosymmetric beams subjected to transverse loads such as con-
no resultant axial force, bending moments or shear forces, but only centrated forces, or uniformly distributed load, is analysed here by
a twisting moment. From Eqs. (63)–(68), we have: developing the linearized governing equations. Firstly, the rela-
tionship between the axial force and shears with the internal force
N ¼ EAðk1s þ ys j  zs j0  1Þ þ ðG þ KÞðIpo  J t Þs2 ¼ 0 resultants in the x-, y- and z-direction is determined and linearized.
  Thus we have:
M 3 ¼ EIzz j  b1 s2 ¼ 0
  N ¼ Px cos h þ Py sin h cos w þ Pz sin h sin w
M 2 ¼ EIyy j0  b2 s2 ¼ 0 ð183Þ
  ffi Px þ Py x3  Pz x2 ð189Þ
M - ¼ EIx s;x  U x s2
Q 2 ¼ Px sin h cos / þ Py ðcos h cos w cos /  sin w sin /Þ
M t ¼ GJt s Q 2 ¼ 0 Q 3 ¼ 0
þ Pz ðcos h sin w cos / þ cos w sin /Þ
Therefore, there must be nonlinear second order axial shortening ffi Px x3 þ Py  Pz x1 ð190Þ
and curvatures induced in the beam given by:
Q 3 ¼ Px sin h sin / þ Py ð cos h cos w sin /  sin w cos /Þ
j ¼ b1 s2 j0 ¼ b2 s2

þ Pz ð cos h sin w sin / þ cos w cos /Þ
ðIpo  J t Þ
k1s  1 ¼   ys b1 þ zs b2 s2 ð184Þ
E A ffi Px x2  Py x1 þ Pz ð191Þ
 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840 2839

In some instances, it is more convenient for problems which are this additional term would vanish in agreement with our earlier
statically determinate to deal with the transverse force resultants derivation. Numerical solutions can be derived to Eq. (203).
Px, Py, Pz rather than the internal actions. The equilibrium equations
(101)–(106) are now linearized to first order in displacements, References
hence:
Alwis, W.A.M., Wang, C.M., 1996. Wagner term in flexural–torsional buckling of
dN thin-walled open-profile columns. Engineering Structures 18 (2), 125–132.
 Q 2 x3;x þ Q 3 x2;x þ pn ¼ 0 ð192Þ Aristizabal-Ochoa, J., 2008. Slope-deflection equations for stability and second-
dx
order analysis of Timoshenko beam–column structures with semi-rigid
connections. Engineering Structures 30, 3394–3395.
dQ 2 Attard, M.M., 1984. The Elastic Flexural–Torsional Response of Thin-Walled Open
þ Nx3;x  Q 3 x1;x þ p2 ¼ 0 ð193Þ
dx Beams. University of New South Wales, Australia.
Attard, M.M., 1986. Nonlinear shortening and bending effect under pure torque of
dQ 3 thin-walled open beams. Thin-Walled Structures 4, 165–177.
 Nx2;x þ Q 2 x1;x þ p3 ¼ 0 ð194Þ Attard, M.M., 1990. General non-dimensional equation for lateral buckling. Thin-
dx Walled Structures 9, 417–435.
Attard, M.M., 2003a. Finite strain – beam theory. International Journal Of Solids and
dMe1 Structures 40, 4563–4584.
 ðM 2  Nzs Þx3;x þ ðM 3 þ Nys Þx2;x þ mn ¼ 0 ð195Þ Attard, M.M., 2003b. Finite strain – isotropic hyperelasticity. International Journal
dx Of Solids and Structures 40, 4353–4378.
Attard, M.M., Hunt, G.W., 2004. Hyperelastic constitutive modeling under finite
dðM3 þ Nys Þ e strain. International Journal of Solids and Structures 41, 5327–5350.
 M 1 x2;x þ ðM 2  Nzs Þx1;x þ m3 Attard, M.M., Hunt, G.W., 2008a. Column buckling with shear deformations – a
dx hyperelastic formulation. International Journal of Solids and Structures 45,
¼ Nc2  ð1 þ us;x ÞQ 2 ¼ Px ðc2 þ x3 Þ  Py ð1 þ us;x Þ  Pz x1 ð196Þ 4322–4339.
Attard, M.M., Hunt, G.W., 2008b. Sandwich column buckling – a hyperelastic
formulation. International Journal of Solids and Structures 45, 5540–
dðM2  Nzs Þ e 5555.
þ M 1 x3;x  ðM 3 þ Nys Þx1;x þ m2
dx Attard, M.M., Lawther, R., 1989. Effect of secondary warping on lateral buckling.
Engineering Structures 11, 112–118.
¼ Nc3 þ ð1 þ us;x ÞQ 3 Bažant, Z.P., 2003. Shear buckling of sandwich, fiber-composite and lattice columns,
¼ Px ðc3 þ x2 Þ þ Pz ð1 þ us;x Þ  Py x1 ð197Þ bearings and helical springs: paradox resolved. ASME Journal of Applied
Mechanics 70, 75–83.
Bažant, Z.P., Beghini, A., 2004. Sandwich buckling formulas and applicability of
Let us now consider a monosymmtric beam with transverse loading
standard computational algorithm for finite strain. Composites: Part B 35, 573–
Poy such that Px, Pz = 0 with bending about the z-axis producing M o3 . 581.
Eqs. (189)–(191) reduce to: Bažant, Z.P., Beghini, A., 2006. Stability and finite strain of homogenized structures
soft in shear: sandwich or fiber composites, and layered bodies. International
N ffi P y x3 Q 2 ffi Py Q 3 ffi Py x1 ð198Þ Journal of Solids and Structures 43, 1571–1593.
Bažant, Z.P., Cedolin, L., 1991. Stability of Structures. Oxford University Press, New
Perturbations of the linearized equilibrium equations are consid- York.
ered. Eq. (196) becomes Beghini, A., Bažant, Z.P., Waas, A.M., Basu, S., 2008. Initial postcritical behavior of
sandwich columns with low shear and transverse stiffness. Composites: Part B
o 39, 159–164.
dM 3
¼ Poy ð199Þ Blaauwendraad, J., 2008. Timoshenko beam–column buckling. Does Dario stand the
dx test? Engineering Structures 30, 3389–3393.
Chang, S.P., Kim, S.B., Kim, M.Y., 1996. Stability of shear deformable thin-walled
Substituting Eq. (198) and the constitutive relationship (163) into space frames and circular arches. Journal of Engineering Mechanics 122, 844–
the shear equilibrium equation in the lateral direction, Eq. (194), 854.
becomes de Saint-Venant, B., 1855. Memoire Sur La Torsion Des Prisms. Mémoire Academic
Science of Savants Étrangers, Paris 14, 233–560.
Q 3 ¼ Py x1 ) GAdc3 ¼ Poy dx1 ð200Þ Emre Erkmen, R., Mohareb, M., 2008. Buckling analysis of thin-walled open
members – a complementary energy variational principle. Thin-Walled
Structures 46, 602–617.
Incorporating the constitutive relationships (165) and (166) into the Engesser, F., 1891. Die Knickfestigkeit Gerader St Be. Zeutralblatt des
twisting moment rate equilibrium equation (195) we have: Bauverwaltung 11, 483.
Haringx, J.A., 1942. On the buckling and lateral rigidity of helical springs.
GJt dx1;xx  EIx dx1;xxxx þ M o3 dx2;x þ mn Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen
45, 533.
d    o  Kardomateas, G.A., Dancila, D.S., 1997. Buckling of moderately thick orthotropic
¼ E b1 M3 dx1;x þ Mo3 dc3 ð201Þ
dx columns: comparsion of an elasticity solution with the Euler and Engesser/
Haringx/Timoshenko formulae. International Journal of Solids and Structures
Substituting the constitutive law, Eq. (164) and making use of Eq. 34, 341–357.
(199), the lateral bending rate equilibrium equation (197) is then: Lee, J.S., Lee, K.C., Kim, M.Y., Attard, M.M., submitted for publication. Spatial stability
of shear-flexible thin-walled beams based on Engesser’s and Haringx’s theories.
  Mo dx1 Thin-Walled Structures.
EIyy dx2;xx  M o3 dx1 ;x ¼ 0 ) dx2;x ¼ 3 þ C1 ð202Þ Love, A.E.H., 1944. A Treatise on the Mathematical Theory of Elasticity. Dover, New
EIyy York.
Machado, S., 2008. Non-linear buckling and postbuckling behavior of thin-walled
where C1 is a constant of integration. Eq. (201) is now transformed beams considering shear deformation. International Journal of Non-Linear
using the above equation and (200), resulting in an differential Mechanics 43, 345–365.
equation in one displacement unknown: Machado, S., Cortínez, V., 2005a. Lateral buckling of thin-walled composite
bisymmetric beams with prebuckling and shear deformation. Engineering
 o  Structures 27, 1185–1196.
M3 dx1
GJt dx1;xx  EIx dx1;xxxx þ M o3 þ C 1 þ mn Machado, S., Cortínez, V., 2005b. Non-linear model for stability of thin-walled
EIyy composite beams with shear deformation. Thin-Walled Structures 43, 1615–
" o
# 1645.
d Py dx1 Mohareb, M., Dabbas, A., 2003. Lateral stability of partially restrained cantilever
¼ E b1 M o3 dx1;x  M o3 ð203Þ supports, 2–5.
dx GA
Reissner, E., 1972. On one-dimensional finite-strain beam theory: the plane
problem. Journal of Applied Mathematics and Physics 23, 795–804.
This is similar to the classical lateral buckling equation with the Reissner, E., 1982. Some remarks on the problem of column buckling. Ingenieur-
P oy dx1 Archiv 52, 115–119.
added term Mo3 GA
due to a consideration of shear. We see that
Reissner, E., 1989. Lateral buckling of beams. Computers and Structures 33, 1289–
if there is no transverse load P oy ¼ 0 as in the case of pure bending, 1306.
2840 M.M. Attard, M.-Y. Kim / International Journal of Solids and Structures 47 (2010) 2825–2840

Sapountzakis, E., Mokos, V., 2008. Shear deformation effect in nonlinear analysis of
spatial beams. Engineering Structures 30, 653–663.
Simo, J.C., Hjelmstad, K.D., Taylor, R.L., 1984. Numerical formulations of elasto-
viscoplastic response of beams accounting for the effect of shear. Computer
Methods in Applied Mechanics and Engineering 42, 301–330.
Vlasov, V.Z., 1959. Thin-Walled Elastic Beams. Translated from second
Russian ed., Israel Program for Scientific Translation, Jerusalem, Israel,
1961.
Zielger, H., 1982. Arguments for and against Engesser’s buckling formulas.
Ingenieur-Archiv 52, 105–113.