Thin-Walled Structures 89 (2015) 212226

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Finite element formulation for lateral torsional buckling analysis

of shear deformable mono-symmetric thin-walled members
Arash Sahraei a, Liping Wu b, Magdi Mohareb a,n
a
b

Department of Civil Engineering, University of Ottawa, Ottawa, ON, Canada K1N 6N5
Candu Energy Inc., 2285 Speakman Drive, Mississauga, ON, Canada L5K 1B1

art ic l e i nf o

a b s t r a c t

Article history:
Received 26 April 2014
Received in revised form
5 October 2014
Accepted 25 November 2014
Available online 23 January 2015

A shear deformable theory and a computationally efcient nite element are developed to determine the
lateral torsional buckling capacity of beams with mono-symmetric I-sections under general loading.
A closed-form solution is also derived for the case of a mono-symmetric simply supported beam under
uniform bending moments. The nite element is then used to provide solutions for simply supported
beams, cantilevers, and developing moment gradient factors for the case of linear moments. The
formulation is shown to successfully capture interaction effects between axial loads and bending
moments as well as the load height position effect. The validity of the element is veried through
comparisons with other established numerical solutions.
& 2014 Elsevier Ltd. All rights reserved.

Keywords:
Thin-walled members
Finite element
Mono-symmetric sections
Shear deformable members
Lateral torsional buckling

1. Motivation
Wide ange mono-symmetric sections are commonly used as
girders in bridge structures. In buildings, they represent a viable
design alternative as exural members in cases such as roof
members where positive bending moments induced by gravity
load combinations involving gravity loads can be signicantly
larger than negative moments typically induced by wind uplift.
When such members are used in large span laterally unsupported
beams, their resistance is frequently governed by lateral torsional
buckling resistance. Relatively recently, design standards (e.g.,
CAN/CSA-S16-09 [1], ANSI/AISC 360-05 [2] and the subsequent
edition ANSI/AISC 360-10 [3]) have incorporated provisions for
quantifying the lateral torsional buckling resistance for simply
supported mono-symmetric members under general loading.
More complex cases involving continuous beams, cantilever suspended constructions, cantilevers, are beyond the scope of North
American design standards, although, as will be discussed in the
literature review, past research has tackled some of these issues.
The present study contributes to the existing body of knowledge
by developing a theory and nite element for the buckling analysis
of mono-symmetric sections. In a recent study, Wu and Mohareb
[4,5] developed a shear deformable theory and nite element
formulation for lateral torsional buckling of thin-walled members.

Corresponding author.
E-mail address: mmohareb@uottawa.ca (M. Mohareb).

http://dx.doi.org/10.1016/j.tws.2014.11.023
0263-8231/& 2014 Elsevier Ltd. All rights reserved.

The theory was limited to members with doubly symmetric

sections, and the resulting element exhibited slow convergence
characteristics, thus requiring a several hundreds of degrees of
freedom to model simple problems. Within this context, the
present study is intended to advance the work in [4,5] in two
respects; (a) it extends the developments to beams with monosymmetric sections, and (b) it devises an effective interpolation
scheme to accelerate the convergence characteristics of resulting
nite element.

2. Literature review
The present work is concerned with the lateral torsional
buckling of beams of mono-symmetric sections based a shear
deformable thin-walled theory. Thus, within the vast body of
research about lateral torsional buckling, the present review
focuses on the work related to beams of mono-symmetric crosssections (Section 2.1) and recent buckling solutions under shear
deformable theories (Section 2.2).
2.1. Lateral torsional buckling for members of mono-symmetric
cross-sections
Several studies have investigated the lateral torsional buckling
resistance of mono-symmetric I-beams. Using the nite integral
method, Anderson and Trahair [6] developed tables for the critical
loads of cantilevers and simply supported beams. Based on energy

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

List of symbols
ai to zi elements of stiffness matrices
^
a
of stiffness matrices
h^ i ito fi h elements
i
A ; B ; C matrices which are coefcients of quadratic eigenvalue problem
A
cross-sectional area
Ap
a pole
Ai i 1 to 8 integration constants
Bz
matrix relating displacement elds to integration
constants
C
section centroid
Dhh ; Dxh ; Dxx ; Dyk ;
Dyyk ; Dyh ; Dyh ;

properties of cross-section related to shear

Dxx0 y0 ; Dx0 y0

T
dz
E
Ez
G
H 
I 
I xx ; I yy
I px
I
J 
Kf
K G N
K G M
K G V
K G qy
K G qz
K G Q y

deformation
eld displacements
modulus of elasticity
diagonal matrix of exponential functions
shear modulus
matrix relating nodal displacements to integration
constants
identity matrix
moments of inertia of the cross-section about x-axis
and y-axis respectively
polar moment of inertia about x-axis
warping constant
St. Venant torsional constant
stiffness matrix due to exural stresses
geometric matrix due to normal forces
geometric matrix due to bending moments
geometric matrix due to shear forces
geometric matrix due to load position effect of the
distributed transverse load
geometric matrix due to load position effect of the
distributed axial load
geometric matrix due to load position effect of the
concentrated transverse load

solutions, Robert and Burt [7] developed a lateral torsional solution

for beams with boundary conditions similar to those reported in [6].
Both studies [6,7] focused on members under concentrated and
uniformly distributed loads. Using the Raleigh Ritz method, Wang
and Kitipornchai [8] developed buckling solutions for cantilevers and
simply supported beams [9] under concentrated and uniformly
distributed loads. Also, Kitipornchai et al. [10] investigated the effect
of moment gradient on the buckling resistance of simply supported
beams. Based on the stationarity condition of the total potential
energy, Kitipornchai and Wang [11] investigated the elastic lateral
torsional buckling resistance of the simply supported tee beams
under moment gradient. They showed that for inverted tee beams,
uniform bending moment is not the most severe loading case and the
cases involving high moment gradients ordinarily were more critical.
Using shell nite element analysis, Helwig et al. [12] modelled the
lateral buckling capacity of girders subject to transverse point and
uniformly distributed loads. Attard [13] developed solutions for
estimating elastic lateral torsional capacity of beams with monosymmetric and doubly symmetric sections and general boundary
conditions. Using Ritz and Galerkins methods, Mohri et al. [14]
developed an analytical model for estimating the lateral torsional
buckling resistance of simply supported beams under concentrated
and uniformly distributed loads. Andrade et al. [15] extended the
application of three-factor lateral torsional buckling formula in the

K s 
K sv 
l
L
Lz
mi
M1 ; M2
M xp z
N1 ; N2
Np z
O
qy ; qz
S0
SC
ub
huN iT
U
V
V 1; V 2
V yp z
x; y; z
yA

yb ; zb

zz
zs
s
b

213

stiffness matrix due to shear stresses

stiffness matrix due to Saint Venant shear stress
length of a nite element
span of the member
matrix of shape functions
roots of quadratic eigenvalue problem
internal bending moment at both ends of an element
resultant of the moments of the normal stresses
obtained from pre-buckling analysis
internal normal forces at both ends of an element
resultant of the normal stresses obtained from prebuckling analysis
origin of the Cartesian coordinates x, y and z
distributed load applied to a member acting along the
y- and z-direction, respectively
sectorial origin
shear center of the cross-section
lateral buckling displacement
vector of nodal displacements
internal strain energy
potential energy
internal shearing forces at both ends of an element
resultant of shear force component along y-direction
obtained from pre-buckling analysis
Cartesian coordinates
coordinate of the shear center along y-direction
constant
end moment ratio
load multiplier
total potential energy
variation of total potential energy
Buckling rotation angles about y, z axes, respectively
matrix of eigenvectors
normal stress along z direction
shear stress on the cross-section mid-surface
warping function or sectorial area of a cross-section
warping deformation (1/length)

Eurocode [16] to mono-symmetric cantilevers subject to uniformly distributed and concentrated transverse tip loads applied.
Their solution incorporated the effect of load height. Based on the
principle of stationarity of the second variation of the total
potential energy, Zhang and Tong [17] developed a new theory
for estimating the lateral torsional buckling capacity of cantilevers subject to concentrated and uniformly distributed loads and
uniform bending moments. Mohri et al. [18] developed linear and
nonlinear models to investigate into the lateral torsional buckling
capacity of simply beams under moment gradient. Using a
hyperelastic constitutive model, Attard and Kim [19] formulated
lateral torsional buckling solutions for shear deformable simply
supported beams subject to uniform bending moment. Using the
Generalized Beam Theory (GBT), Camotim et al. [20] modeled
beams with fork-type end supports under uniform moment, midspan point load, two-point loads, distributed load and linear
moments. They observed that among all loading conditions
including end moments and transverse loads applied at shear
center, the lowest critical buckling moments do not necessarily
correspond to uniform bending moment. Mohri et al. [21]
developed a non-linear model to investigate the effect of axial
forces on lateral torsional buckling resistance of simply supported I and H-sections. Their solutions involved concentrated
and uniformly distributed loads.

214

Table 1
Comparative studies on lateral torsional buckling of mono-symmetric I-beams.
Author(s)

Loading types

Effects captured

Simply
supported

Cantilever Concentrated
transverse load(s)

Uniformly
distributed load

Uniform
bending
moment

Axial
load

Distortional
effects

Shear
deformations

Pre-buckling
deformations

Closed- FEA Other numerical

form
methods

Stationarity of the total

potential energy
RayleighRitz

RayleighRitz

RayleighRitz

Mohri et al. [14]

Ritz and Galerkin

RayleighRitz

Helwig et al.
[12]

Zhang and Tong

[17]
Erkmen and

Mohareb [40]

Attard and Kim

[19]
Camotim et al.
[20]

Mohri et al. [21]

Erkmen [30]
Present study

Stationarity of the total

potential energy
Shell FEA

Andrade et al.
[15]

Mohri et al. [18]

Analysis type

Finite integral

Linear
moment

Solutions developed

Beam 3D FEA

Hyperelastic

Stationarity of the total

potential energy
Stationarity of the
complementary energy

Stationarity of the total

potential energy

LTBEAM,
Shell FEA,
and GBT

Galerkin

Beam 3D FEA

HellingerReissner
Stationarity of the total
potential energy

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

Anderson and
Trahair [6]
Bradford [29]
Roberts and
Burt [7]
Kitipornchai
et al. [10]
Wang and
Kitipornchai
[9]
Wang and
Kitipornchai
[8]
Attard [13]

Boundary condition
types

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

In addition to the above solutions, several nite element formulations have been developed for the lateral torsional buckling of monosymmetric sections. This includes the work of Krajcinovic [22] and
Barsoum and Gallagher [23] who developed a nite element for
buckling analysis based on the Vlasov thin-walled beam theory [24].
Based on the principle of stationarity of the second variation of the
total potential energy, Attard [25] developed two nite element
formulations for estimating lateral torsional buckling loads of beams.
Papangelis et al. [26] developed a computer program to predict
elastic lateral torsional buckling estimates of beams, beam-columns
and plane frames. Distortional effects in doubly and monosymmetric sections were also investigated in the work of Hancock
et al. [27], Bradford and Trahair [28] and Bradford [29]. Using the
HellingerReissner principle, Erkmen [30] developed a hybrid nite
element formulation for shear deformable elements. Lateral torsional
buckling solutions for web-tapered mono-symmetric beams were
investigated in [3136]. Also, solutions for laminated composite
include the work in [37,38]. Table 1 provides a comparative summary
of the most relevant studies. As shown in the table, the present study
aims at developing a general theory and nite element formulation
for the lateral torsional buckling analysis of mono-symmetric members. The solution captures warping and shear deformation effects
and excludes pre-buckling and distortional effects. It is applicable to
general boundary and loading conditions and incorporates the
destabilizing effects of axial loading, shearing force, and bending
moments.
2.2. Buckling solutions under shear deformable theories
Other shear deformable theories were also developed. This
includes the work of Erkmen and Mohareb [39] who developed a
complementary energy variational principle and formulated a nite
element (Erkmen and Mohareb [40]) for thin-walled members with
open cross-sections. In a subsequent study, focused on torsional
buckling of columns, Erkmen et al. [41] demonstrated that the elastic
torsional buckling of columns can be guaranteed to converge from
below.

215

4. Variational formulation
This section outlines the details of the variational formulation.
The treatment is similar to that presented in Wu and Mohareb [4]
for doubly symmetric sections. As such, only important milestones
are provided here and reader is referred to Wu and Mohareb [4]
for more a more thorough discussion of the methodology. A righthanded Cartesian coordinate system is adopted in which the Z-axis
is oriented along the axial direction of the member while X-axis
and Y-axis are parallel to major and minor principal axes of the
cross-section, respectively. The origin is taken to coincide with the
cross-section centroid Cxc 0; yc
 0 while pole Ap is taken to
coincide with the shear center SC xA 0; yA .
4.1. Problem description and notation
The member is assumed to be subjected to a uniformly
distributed transverse load qy applied at a distance yqy z from
the shear center and a uniformly distributed axial load qz acting at
distance yqz z from the origin. Under such external loads, the
member deforms from conguration 1 to 2 as shown in Fig. 1 and
undergoes displacements vp z, wp z and rotation xp z. As a
convention, subscript p represents pre-buckling displacement,
strain, and stress elds. The applied loads are assumed to increase
by a factor and attain the values qy and qz at the onset of
buckling (Conguration 3). Under the load increase, it is assumed
that pre-buckling deformations linearly increase to vp z, wp z
and p z. The section then undergoes lateral torsional buckling
(Conguration 4) manifested by lateral displacement ub , weak-axis
rotation yb , angle of twist zb and warping deformation b . Again,
as a matter of convention, subscript b denotes eld displacements,
strains, or stresses, occurring during the buckling stage (i.e., in
going from conguration 3 to 4) while superscripts n denotes the
total elds (i.e., in going from conguration 1 to 4).

3. Assumptions
The following assumptions have been adopted
1. The formulation is restricted to prismatic thin-walled members
with mono-symmetric sections consisting of segments parallel
to the principal axes,
2. Regarding shear/bending action, the cross-section remains
rigid in its own plane during deformation but does not remain
perpendicular to the neutral axis after deformation in line with
the Timoshenko theory. The hypothesis is further generalized
for torsion/warping action. Similar kinematics have been used
in buckling problems in Saade [42], Kollar [43] Back and Will
[44], Attard and Kim [19], Kim and Lee [37] and Lee [38], and
Wu and Mohareb [4,5].
3. The material is assumed to be linearly elastic and obeys the
Hookes law,
4. Strains are assumed small but rotations are assumed to be
moderate. Rotation effects are thus included in the formulation
by retaining the non-linear strain components,
5. The member buckles in an inextensional mode [45] which
means that throughout buckling, the centroidal longitudinal
strain and curvature in yz-plane remain zero. This signies that
the member is assumed to buckle under constant axial load and
bending moments, and
6. The solution neglects pre-buckling deformation effects.

Fig. 1. Different stages of deformation.

4.2. Kinematic relations

Under the kinematic assumptions postulated above, a point S
on the section mid-surface can be shown [4] to undergo total
displacements uns s; z; vns s; z; wns s; z given by


uns s; z ub z  ys yA zb z
vns s; z vp z xszb z
wns s; z wp z ysxp z xsyb z s b z
xs xp zzb z ysyb zzb z

1ac

Fig. 2 depicts the global coordinate system, displacements, and

sign conventions adopted in this study.

216

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

in which, U b is the internal strain energy due to normal stresses, U sv is

the internal strain energy due to SaintVenant shear stress, U s is the
internal strain energy due to shear stresses, V N is the destabilizing
term of the total potential energy due to normal forces, V M is the
destabilizing term of the total potential energy due to bending
moments, V V is the destabilizing term of the total potential energy
due to shear forces, V qy is the destabilizing term of the total potential
energy due to transverse load position effect, V qz is the destabilizing
term of the total potential energy due to longitudinal load position
effect. Under this condition, the second variations of the above energy
terms take the form

Z L

2

2
0
2 U b
EI yy yb EI 0b
dz
0

2 U sv

Fig. 2. Global coordinate system and displacement components.

Us G
2

4.3. Conditions of neutral stability

The condition of neutral stability is given by evoking the
variation of the second variation of the total potential energy ,
[4], i.e.,





1
1
2 2 U 2 V 0
2
2
2
in which, U is the internal strain energy and V is the load potential
energy gained by externally applied loads. The variation of their
second variation of U and V are given by
n R R nh 
i
h

i

12 2 U 12 0L A E zzb 2 zzp 2 zzb



Z
2
1 L  0 2
2
GJ zb dz:
G zsb zsp zsb dAdz
2 0





Z



L
2
1
1

2 V 2 U 2 V 
 qy yqy yA zb
2
2
2
0

i
3ab
qz yqz yb zb dzg
Eqs. (1ac) are differentiated with respect the appropriate
coordinates to yield the strain expressions. The rst variation
and second variations of strains are

zzb  x0yb 0b



dx 0
ub  yb h 0zb b
ds
h





 i
2
2 zzb u0b 2 2 y yA u0b 0zb x2 y  yA 2 0zb


0
0
2y yb zb yb zb





dy
dx

2 zsb 2
yb zb x yb 0yb  xh b 0yb
ds
ds
 




dx
dy
dk  0
yb 0b h b 0b 
ub zb

ds
ds
ds



dk  0
4ad
zb zb
k
ds


where, hs xs sin s  ys  yA cos s, ks xs cos s


ys  yA sins. From Eqs. (4ad) by substituting into the vibrah


2
2
2
tional expression 1=2 1=2 U V 0, and recal-

zsb

ling

the

pre-buckling

stress

expressions

zzp N p z=A Mxp zy=Ixx and zzp G zsp , one obtains

1 2

2

1 2
U b 2 U sv 2 U s 2 V N 2 V M 2 V V
2

2
2
V qy V qz 0

 0 2
GJ zb dz



0
Dxx u0b  Dxx yb Dxh zb Dxh b u0b



0
 Dxx u0b Dxx yb  Dxh zb Dxh b yb


0
0
Dxh u0b  Dxh yb Dhh zb Dhh b zb



0
Dxh u0b  Dxh yb Dhh zb Dhh b b dz

Z L


N p z 
0
 Au0b 2  2yA A u0b zb
2 V N 
A
0

 0 
 I xx I yy yA 2 A zb 2 dz

Z L
M xp z h
0
0
0
2 V M 
 2I xx yb zb yb zb 2I xx u0b zb
I xx
0

 0 2i
 I px  2yA I xx zb dz


2V yp z 
0
 Dyy yb zb Dyyk u0b zb  Dyk zb zb
Dyy
0

0
0
 Dxx0 y0 yb yb Dx0 y0 yb 0b Dyh b yb  Dyh b 0b dz

2 V V 

2 V qy 
2 V qz 

0
L
0


2 i
 qy yqy  yA zb dz

2qz yqz yb zb dz

6ah

in which, the following sectional properties have been dened.


R
R
R
R 
I yy A x2 dA I xy A xydA I A 2 dA I px A y x2 y2 dA
Z
Z
Z
Z  2
dx
2
Sx ydA Sx xdA Dhh h dA Dxx
dA
A
A
A
A ds
Z 
Z  2
Z
dx
dy
dy dk
dA Dyy
dA
dA Dyk k
Dxh h
ds
ds
A
A
A ds ds
Z  2
dy dk
dA
Dyyk
ds
A ds

Z
Z
Z
dy
dy
dx dy
Dyh xh
dA Dyh h dA Dxx0 y0 x
dA
ds
ds
A
A
A ds ds
Z
dx dy
Dx0 y0
dA
A ds ds
It is observed that for common cross-sections consisting
exclusively of segments parallel to x and y axes (such as a monosymmetric
all area integral terms containing the product

 I section),

dy=ds dx=ds vanish, i.e., Dxx0 y0 Dx0 y0 0.
As a verication of the validity of Eqs. (6ah), when the crosssection is doubly symmetric constants I xy ; I px ; Sx ; Sx ; Dxh ; Dyk ;
Dyh and Dyh vanish and the present variational statement revert to
that in Wu and Mohareb [4] when the coordinate system is taken to
be orthogonal.

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

217

4.4. Finite element formulation I

4.5. Finite element formulation II

The variational expressions in Eqs (6ah) consist of the buckling

displacement functions ub yb zb b and their rst derivatives
with respect to coordinate z. Thus, each of the assumed four
displacement functions needs to satisfy only C 0 continuity. By taking
two nodes per element, and adopting a linear interpolation
scheme

between the two nodal values, the displacement elds ub z yb z
zb z b ziT are related to the nodal displacements through
D
E
ub z yb z zb z b z
14
2( )
3
(
)
(
)
(
)
1
y1
u
z1
1


5
H b z 12 4 u
2
y2
z2
2

The element developed in Section 4.4 has a minimal number of

degrees of freedom (8 DOFs for the buckling solution) but will be
shown to exhibit slow convergence characteristics, thus needing
hundreds of elements to solve simple problems. Within this context,
the present section aims at developing an element which preserves
the low number degrees of freedom per element while accelerating its
convergence characteristics. This is achieved by adopting different
interpolation schemes for pre-buckling internal forces and buckling
displacement elds.

21

21

21

21

7ad
D
  E
1
z=L
z=L
is the vector of shape
in which, H b z 12
linear shape functions andu1 ; u2 ; y1 ; :::: 2 are the nodal displacement. In a similar manner, the pre-buckling stress resultants
N p z V yp z M xp z are linearly interpolated between the internal
forces N 1 ; N 2 ; V 1 ; V 2 ; M 1 ; M 2 at the nodes as obtained from the prebuckling analysis (Fig. 3b, d, f), i.e.,
D
E
Np z V yp z M xp z
2(
3
)
(
)
(
)
N1
V1
M1


5
H b z 12 4  N
V2
 M2
2


21

21

21

8ac
The resulting element is similar to that reported in [5] (and will
be subsequently referred to as the WM element), with two
differences. The present element is geared towards monosymmetric sections while the WM element is for doubly symmetric sections. Also, the present formulation is based on an
orthogonal coordinate system while the WM is based on general
non-orthogonal coordinates.

4.5.1. Approximation of pre-buckling internal forces

In general, the pre-buckling internal forces N p z, V yp z and
M xp z are non-constant functions. Under such conditions, the
closed-form solution of the governing neutral stability conditions
stemming from the above variational principle become unattainable. Thus, the non-constant internal forces obtained from prebuckling analysis (Fig. 3a, c, e) are approximated as piecewise
constant functions equal to the average values of the internal
forces. Thus, one can set
Np z  N p

N 1  N 2
V 1  V 2
; V yp z  V yp
; M xp z  M xp
2
2

M 1 M 2
2

9ac

in the variational statement (Eqs. (6ah)) within the subdomain Le

(Fig. 3b, d, f) of the element. When the number of elements is
sufciently large, the piecewise representation of pre-buckling internal
forces would approach that of actual internal force distributions and
the resulting approximate total potential energy expression of the
system will approach that based on Eqs. (6ah). This treatment will be
shown rather advantageous from a computational viewpoint and will
lead to desirable convergence characteristics.

Fig. 3. Internal forces for a beam-column: (a) normal forces within member, (b) idealized constant normal force within the element, (c) shearing forces within member,
(d) idealized constant shearing force within the element, (e) bending moments within member, and (f) idealized bending moment within the element.

218

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

4.5.2. Approximate equations of neutral stability

Under the approximations introduced in Section 4.5.1, Eqs. (6ah)
are integrated by parts and like terms are grouped together. Noting
that the functions ub , yb , zb and b are arbitrary, one recovers
the conditions of neutral stability

zb ; Le b iT Ai ; Bi ; C i ; Di T emi z . By substituting into Eq. (10), one

obtains the quadratic Eigen value problem

h i
  h i
m2i A mi B C

44


41

f0g41

11

10

in which, D d=dz d=Le d is the rst derivative of displacement

elds with respect to the dimensionless coordinate z=Le .
4.5.3. Formulating shape functions
It is proposed to generate shape functions which satisfy Eq. (10).
The presence of the unknowns N p ; V yp ; M xp makes such a solution
unattainable given that is unknown a-priori. As such, the terms
involving N p ; V yp ; M xp are assumed negligible. This assumption
turns out to be accurate for beams of practical dimensions.
Another issue arising when solving Eq. (10), is the need to
estimate. qy yqy  yA =GDxx When the load qy is a applied at the
shear center, i.e., yqy yA , it is clear that qy yqy  yA =GDxx would
vanish. Otherwise, the order of magnitude for maximum for
the distributed load qy can be estimated by equating the
lateral torsional buckling resistance of the beam C b 2 EI yy =2L2
r



x 2x 4 GJL2 = 2 EI yy I =Iyy , in which, C b 1:14, to the
external moments qy L2 =8 for a simply supported beam, and solving

for qy . It is clear that the proposed scheme will yield approximate

value for qy since it does not necessarily capture the end conditions of the element, yielding approximate shape functions. Nevertheless, the approximate functions thus obtained will be shown to have
superior convergence characteristics compared to that in formulation I.

h i  
h i
in which, matrices A , B and C are dened this time in
dimensional form as
2
3
0
GDxh
0
 GDxx
h i 6 0
0
0 7
EI yy
6
7
A 6
12
7
0
 GJ Dhh
0 5
4 GDxh
0
0
0
EI
2

6
  6 GDxx
B 6
6 0
4
 GDxh
2

0
h i 6
60
C 6
60
4
0

GDxx

GDxh

GDxh

GDhh

0
 GDxx
0

qy yqy  yA

GDxh

7
7
7
GDhh 7
5
0
0

3
0
GDxh 7
7
7
7
0
5
 GDhh

13

14

in which, i is the eigenvector corresponding the eigenvalue mi .

Eq. (11) can be expressed into the following equivalent linear
eigenvalue problem
02   h i
3
2h i
3 1(   )
 
mi i
0
B
C
A
0
@4
5 mi 4
5 A
 

0 81
i
 I  0
0 I 
81
88

4.5.4. Closed-form solution for the eld equations

The coupled system of equations (Eq. (10)) has constant
coef
cients. Its solution is assumed to take the form ub =Le ; yb ;

0
0

 GDxh

88

T

15

is the identity matrix. The

in which, I  Diag 1 1 1 1
above eigenvalue problem is observed to have four zero roots,

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

i.e., m1 m2 m3 m4 0. Thus, the closed-form solution takes

the form
ub A1 A2 z A3 z2 A4 z3 A5 em5 z A6 em6 z A7 em7 z A8 em8 z

yb B1 B2 z B3 z2 B4 z3 B5 em5 z B6 em6 z B7 em7 z B8 em8 z

zb C 1 C 2 z C 3 z2 C 4 z3 C 5 em5 z C 6 em6 z C 7 em7 z C 8 em8 z
b D1 D2 z D3 z2 D4 z3 D5 em5 z D6 em6 z D7 em7 z D8 em8 z
16ad
The remaining four roots mi (i 5 to 8) can either be obtained
by solving the right eigenvalue problem in Eq. (15) or from the
equivalent bi-quadratic characteristic equation
n


G  D2xh E2 GI I yy Dhh Dxx E2 GI I yy Dxx E2 GI I yy J m4
i
h
ED2xh G2 I yy J  EDhh Dxx G2 I yy J Dxx E2 I I yy qy yA m2

o
0
17
EGI yy qy yA D2xh Dhh Dxx
It is observed that for the special case of doubly symmetric
sections, one has Dxh yA 0 and the last coefcient of Eq. (17)
vanishes. In such a case, one obtains six repeated zero roots,
and Eq. (16ad) becomes an invalid solution. Thus, as stated in
Assumption 1, the present solution is restricted to monosymmetric sections. By substituting Eq. (16ad) and their derivatives into the eld equations (Eq. (10)), and performing algebraic
simplications, the 32 integration constants Ai  Di i 1:::8
can be reduced to eight
independent
constants Ai i 1:::8.
D
ET

T
The eld displacements dz ub z yb z zb z b z
T
are
thus
related
to
integration
constants
A


A1 A2 A3 A4 A5 A6 A7 A8 T through


dz Bz48 fAg81
18
where,

h
i
Bz48 F z44 44 Ez44 ; F z44
2
3
1 z z2
z3
6EI yy Dhh
6 0 1 2z 3z2 
7
6
GDxh 2  Dhh Dxx 7
6
7
6
7
0
60 0 0
7
4
5
 6EI yy Dxh
0 0 0
2
GDxh  Dhh Dxx
and 44

h 

 

  i

 
2

19ab





 T
 T
0 1 0 0 14 ,
p1 14 1 0 0 0 14


and,T
 , T p2 14
p3 14 0 0 1 0 14 , and p4 14 0 0 0 1 14 have
been dened. It is noted that when a section is doubly symmetric,
matrix H 88 becomes singular and the shape functions introduced in
Eq. (22) become unattainable. From Eqs. (23ad), by substituting into
Eqs. (6ah) and then Eq. (5), one obtains

 

un T K f K sv K s K G N K G M K G V K G qy


24
K G qz g un 0
in which, K f is the elastic stiffness matrix due to exural stresses,
K sv is the elastic stiffness matrix due to Saint Venant shear stresses,
K s is the elastic matrix due to the remaining shear stresses, K G N is
the geometric matrix due to normal forces, K G M is the geometric
matrix due to bending moments, K G V is the geometric matrix due to
shear forces, K G qy is the geometric matrix due to the distributed
transverse load and K G qz is the geometric matrix due to the
distributed axial load. These stiffness matrices are obtained from
 

K f ; K sv ; K s ; K G N ; K G M ; K G V ; K G qy ; K G qz
h
iT 
 1
M 1 ; M 2 ; M 3 ; M 4 ; M 5 ; M 6 ; M 7 ; M 8  H 88
H 1
88

25
in which, M 1  to M 8  are provided in Appendix A.
5. Examples
This section provides various buckling examples aimed at assessing the quality of the results, and illustrate its various features. All
examples assume steel material with E 200; 000 MPa and
G 77; 000 MPa and all the examples (excluding Example 7), are
related to the section illustrated in Fig. 4. Cross-sectional properties are I xx 5:6987  107 mm4 , I yy 1:42155  107 mm4 , A
8  103 mm2 , I 3:2080  1010 mm6 , I px 2:1056 109 mm5 ,
yA  58 mm, J 8:61867  105 mm4 , Dxx 5:600  103 mm2 ,
Dhh 5:71264  107 mm4 , Dyy 2:400  103 mm2 , Dxh
 2:52800 105 mm3 , and Dyk  72; 000 mm3 .
5.1. Example 1: Closed-form solution for a simply supported beam
under uniform bending moment

is the matrix of eigen-

vectors of the quadratic eigenvalue problem dened in Eq. (15) and

Ez44 Diag em5 z ; em6 z ; em7 z ; em8 z  is the diagonal matrix of exponential functions. The Integration constants Ai can be related to the

huN iT ub1 yb1 zb1 b1 u
nodal
displacements

A simply supported beam of length L with a mono-symmetric

cross-section is subject to uniform bending moment M x .All other
internal forces are assumed to vanish. By setting M xp z
M xp a 0; N z V y qy qz 0 in the governing differential equations (Eq. (10)) and solving the resulting coupled system of

b2yb2 zb2 b2 iT by setting z 0 and z L in Eq. (18) leading to

fuN g81 H 88 fAg81
in which,
"

F 044

H 88 
F l 44

44 E044


44 El 44

219

20
#
21

From Eq. (20) by solving for the integration constants and

substituting into Eq. (18), one obtains


22
dz Lz48 fuN g81
1
is a matrix of shape function, and
in which, Lz48 Bz48 H 88

T
 T
ub z L1 z 18 fuN g81 p1 14 Lz48 fuN g81


 
yb z L2 z T18 fuN g81 p2 T14 Lz48 fuN g81
 


zb z L3 z T18 fuN g81 p3 T14 Lz48 fuN g81
 T

T
b z L4 z 18 fuN g81 p4 14 Lz48 fuN g81
23ad

Fig. 4. Dimensions of the mono-symmetric cross-section.

220

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

5.2. Example 2: Mesh density analysis

differential equations, one obtains



dz Bz48 fAg81

26

The relevant boundary conditions are

0

ub 0 0;

EI yy yb 0 M xp zb 0 0;

ub L 0;

EI yy yb L M xp zb L 0;

zb 0 0;
zb L 0;

0b 0 0
0b L 0
27ah

From Eq. (26) by substituting into the displacement eld

equations in Eqs. (27ah), one obtains
p
7 d
M cr M xp1;2  b2a





2
a  E2 I yy I
 GE Dhh I yy Dxx I
G2 D2xh  Dhh Dxx
L
L


2


2
b EGI yy
EI 2Dxh x Dxx
 Gx D2xh  Dhh Dxx
L
L


2
c EG2 I yy
EI  D2xh Dhh Dxx Dxx J
 GJ D2xh  Dhh Dxx
L
L
2

28ae

d b  4ac

In Eq. (28ae), it can be veried that by setting Dxh x 0,

one recovers the critical moment expression M crd based the shear
deformable theory as provided in Wu and Mohareb [4], i.e.,
v
u
u
M crd 7 G t

Dhh Dxx E2 I I yy 2 Dxx E2 I I yy J 2 EDhh Dxx GI yy JL2

29

Dhh Dxx G2 L4 E2 I I yy 4 EDhh GI yy 2 L2 EDxx GI 2 L2

 2 
L EIyy
=2a
also, for a large spans L, one has  b
2
2x
(
)
(
)


p



2
2
2
2
2
d
p
=L EIyy
L EIyy
L EIyy
2a
 e -0,b e=2a
 e -0,
=
2
2
2
2x

(
2
2
L EIyy
2x

nh

2x

(
)
q  2 2
L EIyy
ed
 e -0,  2a=
 e -1,in
2

2x

which,

2 x

i o2 n
 

io
=L 2 EIyy =2
2x 4 GJL2 = 2 EI yy I =Iyy

Consider a 2 m span cantilever section subject to a vertical

concentrated load located at the tip and acting at the shear center.
The critical load as determined by the present formulation is
provided for various discretizations. Results based on the present
formulation (i.e., based on formulation II are compared to those
predicted by (a) the WM element, (b) the classical non-shear
deformable element by Barsoum and Gallagher [23] (referred to as
BG) and (c) the WM element (Table 2). Also, for comparison, a
solution was performed under the ABAQUS S4R shell element.
ABAQUS S4R solution yields the lowest buckling load estimate of
282.3 kN. This is due to the fact that the shell element captures
both the distortional and shear deformation effects and thus
provides the most exible representation of all solutions. The
present element and the WM element predict nearly equal
buckling loads of 287.2 kN and 288.0 kN, respectively. These values
are slightly higher than that based on the shell solution. While
shear deformation is captured in both formulations, they do not
account for distortional effect and thus they provide a slightly
stiffer representation for the member. The largest buckling load is
that based on the BG which predicts a buckling load of 329.5 kN.
This is expected since the BG element neglects shear distortional
and shear deformation effects and thus provides the stiffest
representation of the member among all solutions. Under the
present formulation, it is observed that no more than eight
elements are needed to attain convergence, in a manner similar
to the BG element, but in contrast with the signicantly larger
number of WM elements needed.
The present solution converges from below for the problem, i.e., a
coarser mesh tends to under-predict the buckling loads. This contrasts to the WM and BG elements which consistently converge from
above. However, convergence from below cannot be guaranteed. This
is illustrated by the considering a 5 m span simply supported beam
under end reverse moments (Table 3). For the case of a single
element, the approximation M xp z  M xp M 1  M 2 =2 introduced
in Eq. (9c) yields, M xp 0, thus vanishing the destabilizing term due
to bending moment, and the only destabilizing term remaining is the
due to shear (which is minor in the present 5 m span beam), yielding
a high buckling moment prediction of 286,800 kNm. A signicant
predictive improvement is obtained by taking eight elements.
Table 3
Convergence study for a simply supported
(span 5 m) under reverse end moments.

and

one can show that the critical moment expression in Eq. (28a)
approaches that of the classical solution [45], i.e.,
v
2
!3
u
2
u 2
2 EIyy 4
GJL
I

5
30ab
x 7 t x 4 2

M cl
EIyy Iyy
2L2

Number of elements

Buckling moment (kNm)

286:8  103
602.4
529.9
499.9
498.1

1
2
4
8
10

In Eqs. (28a) and (30), the positive sign is taken when the large
ange is in compression.

beam

Table 2
Mesh density study for cantilever under a concentrated load at the tip (span 2 m, ABAQUS critical load 282.3 kN).
Present study

WM element

BG element

Number of
elements

Buckling load
(kN)

Present study/
ABAQUS (%)

Number of
elements

Buckling load
(kN)

WM/ABAQUS
(%)

Number of
elements

Buckling load
(kN)

BG/ABAQUS
(%)

2
3
4
5
6
8
10

275.9
283.6
285.7
286.5
286.9
287.2
287.2

97.73
100.4
101.2
101.5
101.6
101.7
101.7

32
64
128
256

302.7
291.4
288.4
288.0

107.2
103.2
102.2
102.0

2
3
4
5
6
8

332.2
330.0
329.7
329.6
329.6
329.5

117.7
116.9
116.8
116.8
116.8
116.7

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

221

5.3. Example 3: Inuence of span on shear deformation effects

5.4. Example 4: Beam under linear bending moment

A cantilever is subject to a transverse concentrated load applied

at the shear center of the free end. Two spans are examined in this
example; 1000 and 4000 mm. The lateral torsional buckling load is
estimated based on four solutions: (1) The classical BG solution
which neglects shear deformations and distortional effects, (2) the
WM solution [5] which captures shear deformations but neglects
distortional effects, (3) the present formulation which also captures
shear deformations and neglects distortional effects and (4) the
ABAQUS shell analysis which considers both effects. As suggested in
Example 2, eight elements were used for the BG solution, 64
elements per meter were used for the WM element, and eight
elements were used for the present solution. In the ABAQUS model,
10 elements per top ange, four elements per bottom ange, 10
elements along the web height and 50 elements in the longitudinal
direction were taken to model the beam. The results are presented
in Table 4. As observed in Example 2, the ABAQUS shell element
solution provides a lowest buckling prediction. This is attributed to
the fact that the shell formulation is the only solution that captures
distortional effects, which tend to be more signicant in a short span
cantilever. This is illustrated in Fig. 5a where the web of the 1 m
span cantilever is observed to undergo minor distortion near the top
ange (relative to the shown straight reference line). In contrast, the
web for the 4 m span (Fig. 5b) cantilever is observed to essentially
undergo no distortion compared to the straight reference line. Since
both the WM element and the present element capture shear
deformation effects, their buckling load predictions are smaller than
those based on the BG element. As illustrated by results, the shear
deformation effect is more pronounced in the short span cantilever.
This is evident by the 24% difference observed between the buckling
load prediction based on the present element and the BG element.
The difference is only 6% for the longer span cantilever.

A simply supported mono-symmetric beam is subject to a

linear bending moment distribution as shown in Fig. 6a. A strong
axis moment M x is a applied at the left end and a moment vary
Mx at the other end where  1 r r 1. Two cases are considered: In Case 1, the larger ange is located in the top of the section
so that the larger ange is under compression (Fig. 6b) and in
Case 2, the smaller ange is in the top so that the smaller ange is
under compression (Fig. 6c). Spans were taken to vary from 1 m to
5 m. For Case (a) under uniform moments, i.e.,  1, the lateral
torsional buckling capacity is predicted based on three different
FEA solutions including present element (with eight elements
along the span), and the WM and BG elements, the closed-form as
provided in Eq. (28ae) as well as the classical solution as given
by Eq. (30ab) while retaining the positive sign. Table 5 shows
that the critical moment as predicted by the present FEA and
closed form solution agree well with the classical solution for
large spans (2 m and larger). For the 1 m span, where shear
deformation effects are signicant, the critical moment as predicted by the present shear deformable solution is less than that
predicted by the classical solution which neglects shear deformation effects. It is noted that the results based present element
nearly coincide the WM element with a signicantly lower
number of DOFs.
For non-uniform moments, the moment gradient factor is
dened as the ratio of the critical moment as predicted by the
present study to that of classical solution as given by Eq. (30a).
This ratio accounts for the end moment ratios and the span. The
results are depicted in Fig. 7 for Case 1 where moments induce
compression in the larger ange and Fig. 8 for Case 2 where the
smaller ange is under compression. For the longer spans, the
moment gradient factor is observed to be almost identical for 3 m,
4 m and 5 m spans (and larger spans not shown on the gure) in
the moment gradient range  1 r r 0 i.e., when the larger ange
is under entirely under compression. For shorter spans, smaller
moment gradient factors are obtained given that shear deformation effects gain signicance in such short spans. When moments
induce compression in the smaller ange, the moment gradient
factor monotonically increases with the end moment ration
(Fig. 8). In contrast, for the case where the larger ange is under
compression, the moment gradient factors peak around 0:5
for 3 m, 4 m, 5 m spans and close to 0:0 for the short
span beams.

Table 4
Buckling loads (kN) for a mono-symmetric cantilever beam under a tip vertical
concentrated load.
Span ABAQUS Present
(mm)
element

WM
BG
Present/
element element ABAQUS

WM/
BG/
ABAQUS ABAQUS

1000
4000

930.8
84.90

1.07
1.04

866.5
81.86

925.5
84.50

1223
89.70

1.07
1.03

1.41
1.10

Fig. 5. Distorted cross-section at free end: (a) L 1000 mm, (b) L 4000 mm.

222

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

Fig. 6. Simply supported beam under moment gradient (a) elevation, (b) cross-section for Case (1)moments M x induces compression in larger ange, and (c) cross-section
for Case (2)moments M x induces compression in smaller ange.

Table 5
Lateral torsional buckling loads (kN m) for a simply supported beam under uniform bending moment.
Span (m)

Present nite element (8 elements)

WM element (64 elements/m)
BG element (8 elements)
Closed-form solution present study(Eq. 28a)
Classical closed-form solution M cl (Eq. 30a)
Present/classical

4652
4658
5017
4647
5017
0.93

1441
1444
1466
1440
1466
0.98

767.8
768.9
773.0
764.5
772.9
0.99

508.4
509.0
510.0
508.2
510.0
1.00

376.4
376.6
376.9
376.0
376.9
1.00

Fig. 7. Moment gradient factor versus various end moment ratios and spans (m)for Case (1): larger ange under compression.

Fig. 8. Moment gradient factor versus various end moment ratios and spans (m) for Case (2): smaller ange under compression.

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

5.5. Example 5: Axial forcebending interaction

A simply supported member has a 4 m span and is subject to an
axial compressive force Q z and two equal end moments. In the
absence of uniform bending moment, the exuraltorsional buckling
load Q z0 obtained is 1580.6 kN while from Example 4, the buckling
moment M x0 in the absence of axial load is 508.4 kN m. In order to
develop the Q z  M x buckling interaction diagram, several load
combinations Q zi ; M xi i 1; :::; n are applied to the member and
the buckling eigenvalues are obtained
 for each case. This gives
i 1; :::; n critical load combinationsi Q zi =Q z0 ; i M xi =M x0 in which
each load combination has been normalized with respect to Q z0 and
M x0 . The resulting normalized interaction curve is depicted in Fig. 9.
As can be seen, unlike doubly symmetric sections, the diagram
is non-symmetric about the horizontal axis. This observation is in
line with what is observed in [21]. The higher critical moment
ratio M x =M x0 1 is obtained when the section is under pure
bending and when the top ange is under compression.

223

top ange loading, buckling loads are lower than that based on
shear center loading. Bottom ange loading is associated with a
stabilizing effect which increases the buckling load.
5.7. Example 7: Mono-symmetric I-girder
The present example illustrates the applicability of the formulation
for other types of mono-symmetric sections. A simply supported girder
(cross-section given in Fig. 10) is subject to a mid-span point load
applied at the shear center. Four spans are examined in this example;
2000 m, 4000 m, 6000 m and 8000 mm. The lateral torsional buckling
load estimated based on the present study are compared to those of
obtained from the classical BG element. Sectional properties are
I xx 1:2  109 mm4 , I yy 3:73  108 mm4 , A 3  104 mm2 ,
I 2:39  1013 mm6 , I px 1:70  1010 mm5 , yA  105 mm,
J 4:0  106 mm4 , Dxx 16; 000 mm2 , Dhh 1:44  109 mm4 ,
Dyy 14; 000 mm2 , Dxh  2:107  106 mm3 , and Dyk 4:27
105 mm3 .
As dicussed in previous examples, as the beam span increases,
shear deformation effects become less signicant. Consequently in
Table 7, the buckling load ratio varies from 0.82 at a span of 2 m to
0.97 at a span of 12 m.

Fig. 10. Dimensions of the I-girder cross-section.

Fig. 9. Normalized interaction diagram.

5.6. Example 6: Effect of load height position for a member under

concentrated transverse load
A cantilever spanning 5 m is subject to a concentrated transverse load applied at the tip. Three different load positions are
considered: (a) top ange, (b) shear center and (c) bottom ange.
The results are shown to agree well with those based on the shell
nite element analysis (Table 6). Due to the destabilizing effect of
Table 6
Load position effect on lateral torsional buckling estimates (kN) of a cantilever
beam under a tip vertical load.
Load position

(1) Present study

55.3
(2) ABAQUS
53.5
Percentage difference
3.3%
((1)  (2))/(2)

56.4
54.4
3.7%

68.0
63.2
7.6%

Table 7
Lateral torsional buckling loads (kN) for a simply supported beam under mid-span
point load.
Span (m)

Present solution

8:984  104 1:286  104 4154

1:098  105 1:432  104 4521

BG element
Present solution/BG
element

0.82

0.90

0.92

12

1921

705.1

2062

727.6

0.93

0.97

6. Summary and conclusions

1. A general shear deformable element was developed for buckling analysis of members with mono-symmetric sections.
2. Compared to the shear deformable WM element [5], the
number of degrees of freedom needed for convergence was
observed to reduce signicantly resulting.
3. A closed-form solution was derived for the buckling moments
of shear deformable mono-symmetric simply supported
beams under uniform bending moments.
4. Results obtained based on the present element and the WM
element, were found to be in close agreement.
5. For long spans, excellent agreement was obtained with ABAQUS
FEA shell results. For shorter spans, the present solution provides
higher buckling predictions compared to ABAQUS results, but

224

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

lower than those based on the classical BG element. This is a

natural outcome of the fact that ABAQUS shell model captures
shear deformation and distortional effects, thus providing the
most exible representation, while the present solution captures
shear deformation effects but not the distortional effect and the
classical solution captures neither effect.
6. Based on the present formulation, moment gradient factors
were developed for the mono-symmetric section investigated
in the study and were shown to depend upon the end moment
ratio as well as the span. Beyond certain span (3 m in the
present problem) when 1 r r0 and the lager ange is
under compression, the moment gradient factors were
observed to become independent of the span.
7. Interaction effects between moments and axial force as well as
the load height position effects were successfully captured
through the present element.

in which, function di;j are dened as




3i 3j mi mj eLmi Lmj  1
di;j
mi mj

A4

Elastic stiffness due to shear stresses

Matrix M 3  related to
2
0
0
0 0
6
0
0 0
6
6
0 0
6
6
6

6
M 3  6
6
6
6
6
6
6
Sym:
4

K s  is given by
0

e1

e2

e3

g1

f 1;2

f 1;3

g2

f 2;3
g3

0 7
7
7
0 7
7
e4 7
7
7
f 1;4 7
7
7
f 2;4 7
7
7
f 3;4 5
g4

A5

Appendix A. Matrices needed to determine stiffness matrices

in which, functions ei , f i;j , and g i are dened as

This appendix provides explicit expressions for matrices forming stiffness matrices.
In Eq. (25), the elastic stiffness matrix

components K f ; K sv ; K s  and geometric stiffness components
K G N ; K G M ; K G V ; K G qy ; K G qz were expressed as a function of
matricesM 1 ; M 2 ; :::; M 8 . In order to obtain elastic and geometric
stiffness matrices, rst, M 1 ; M 2 ; :::; M 8  matrices should be
calculated numerically as follows

ei

Elastic stiffness due to exural stresses

 
Matrix M 1  related to K f is given by
2
0 0
0
0
0
0
0
6
0
0
0
0
0
0
6
6
2
6
a
a
a
4L
6L
1
2
3
6
6
12L3 b1
b2
b3
6
M 1  EI yy 6
6
c1;1 c1;2 c1;3
6
6
c2;2 c2;3
6
6
6
Sym:
c3;3
4

0 7
7
7
a4 7
7
7
b4 7
7
c1;4 7
7
7
c2;4 7
7
c3;4 7
5
c4;4

in which, functions ai , bi and ci;j are dened as



ai 22i eLmi  1
!
1 eLmi Lmi 1
bi 62i mi

m2i
m2i



mi mj I yy 2i 2j I 4i 4j eLmi Lmj  1


ci;j
I yy mi mj

A1

d1;1

d1;2

d1;3

d2;2

d2;3
d3;3

and parameter is
 36Dhh E2 I yy 2 L

G Dxh 2  Dhh Dxx

A7

Geometric stiffness due to normal forces

A2ac

Elastic stiffness due to Saint Venant shear stress

to K sv  is given by





G eLmi mj  1

Dhh 4i 4j 3i 4j mi 3j 4i mj 3i 3j mi mj
mi mj


Dxx 2i 2j  1i 2j mi  1j 2i mj 1i 1j mi mj

Dxh  2i 4j  2j 4i 1i 4j mi  2j 3i mi 1j 4i mj

 2i 3j mj 1i 3j mi mj 1j 3i mi mj



G e2Lmi  1 2Dhh 3i 4i  2Dxx 1i 2i 2Dxh 1i 4i  2Dxh 2i 3i
gi
2

 2Lm

2
i
Gmi e
 1 Dxx 1i 2Dxh 1i 3i Dhh 3i 2

2



G e2Lmi  1 Dxx 2i 2  2Dxh 2i 4i Dhh 4i 2
A6ac

2mi
f i;j

 
and ij denotes the jth element of eigenvector i .

Matrix M 2  related
2
0 0
0
6
0
0
6
6
6
0
6
6
6
6
M 2  GJ 6
6
6
6
6
6
6
Sym:
4

6EI yy eLmi  12i  1i mi

mi

7
0 7
7
0 7
7
7
0 7
7
d1;4 7
7
7
d2;4 7
7
7
d3;4 7
5
d4;4

Matrix M 4  related to K G N is given by

2
0
0
0
0
0
0
0
2
3
6
L
L
L
h
h
h
1
2
3
6
6
4L3
3L4
6
k
k
k
1
2
3
3
2
6
6
9L5
6
n
n
n
1
2
3
5
M 4  N 6
6
p1 o1;2 o1;3
6
6
6
p2 o2;3
6
6
Sym:
p3
4

h4 7
7
7
k4 7
7
7
n4 7
7
o1;4 7
7
7
o2;4 7
7
o3;4 7
5
p4

A8

in which, functions hi , ki , ni , oi;j and pi are dened as




hi 1i 3i yA eLmi  1



Lmi
2 1i 3i yA Lmi e eLmi 1
ki
mi

A3
ni





3 1i 3i yA 2eLmi L2 mi 2 eLmi  2Lmi eLmi  2

mi 2



mi mj eLmi mj 1 A1i 1j Ay2A I xx I yy 3i 3j AyA 1i 3j 1j 3i


oi;j
A mi mj

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226






 
mi e2Lmi  1 21i 21i 3i yA 23i y2A
mi e2Lmi  1 I xx I yy 23i

2
2A

pi

A9ae

in which, functions xi;j and yi are dened as


L m mj
 3i 3j e i
 1 yA  yqy
xi;j
mi mj

Geometric stiffness due to bending moments

Matrix M 5  related
2
0
0
0
6
0
0
6
6
6
0
6
6
6
M 5  M 6
6
6
6
6
6
6
Sym:
4

yi

to K G M is given by
0

q1

q2

q3

r1
t1

r2
s1;2

r3
s1;3

t2

s2;3
t3

7
0 7
7
q4 7
7
7
r4 7
7
s1;4 7
7
7
s2;4 7
7
s3;4 7
5

A10

23i eLmi  1
mi




6EDhh Iyy 3i eLmi 1
1
eLmi Lmi  1

2
mi 2
mi 2
G Dxh  Dhh Dxx


1
2j 3i mi 2i 3j 2j 3i mj IIpxxx 3i 3j mi mj


B
2i 3j
C
Lmi mj

e
1 @
A
 1i 3j 1j 3i mi mj  23i 3j yA mi mj


si;j
mi mj
ti


 2Lm



e i  1 3i e2Lmi 1 21i mi  42i 23i mi yA

2
2Ixx

Ipx

2
3i mi

A11ad

Geometric stiffness due to shear forces

Matrix M 6  related
2
0
0
0
6
0
0
6
6
6
0
6
6
6
M 6  V 6
6
6
6
6
6
6
Sym:
4

vi;j

to K G V is given by
0

0
0

0
u1

0
u2

0
u3

w1

v1;2

v1;3

w2

v2;3
w3

7
0 7
7
0 7
7
7
u4 7
7
v1;4 7
7
7
v2;4 7
7
v3;4 7
5
w4

z1
a^ 1
b^

z2
a^ 2
b^

z3
a^ 3
b^

d^ 1

c^ 1;2
d^

c^ 1;3

0
0

c^ 2;3
d^
3

7
7
7
7
7
7
4 7
7
c^ 1;4 7
7
7
c^ 2;4 7
7
7
c^ 3;4 7
5
d^
z4
a^ 4
b^

zi

 3i yqz eLmi  1
mi

1
eLmi Lmi  1

2
2
mi
mi

1
0


2
Lmi
e
L mi 2  2Lmi 2
6EDhh I yy 3i yqz eLmi  1
^b 3 y @ 2 
A


i
3i qz
mi 3
mi 3
Gm D 2  D Dxx

a^ i  23i yqz

A13ac

Geometric stiffness due to distributed transverse load

Matrix M 7 related to K G qy is given by
2
0
0
0 0 0
0
0
6
0
0 0 0
0
0
6
6
6
0 0 0
0
0
6
6
0 0
0
0
6
M 7  qy 6
6
y1 x1;2 x1;3
6
6
y2 x2;3
6
6
6
Sym:
y3
4

0 7
7
7
0 7
7
7
0 7
7
x1;4 7
7
7
x2;4 7
7
x3;4 7
5
y4

hh


 yqz eLmi mj  1 2i 3j 2j 3i

A17ad

Load position matrix for concentrated transverse load

Dyy mi

xh

A12



D
eLmi mj 1 2i 3j 2j 3i  1i 3j mi  1j 3i mj Dyk
3i 3j mi mj
yy


mi mj

3i e2Lmi  1 Dyy 2i Dyy 1i mi Dyk 3i mi

A16

mi mj
 2Lm

 2i 3i yqz e i  1
d^ i
mi

Gmi Dxh 2  Dhh Dxx



to K G qz is given by

in which, functions zi , a^ i , b^ i , c^ i and d^ i are dened as

c^ i;j

6EDhh I yy 3i eLmi  1

wi

Matrix M 8  related
2
0
0
0
6
0
0
6
6
6
0
6
6
6
6
M 8  qz 6
6
6
6
6
6
6
Sym:
4

in which, functions ui , vi;j and wi are dened as

ui 

A15ab

2mi

Geometric stiffness due to distributed axial load

in which, functions qi , r i , si;j and t i are dened as

r i 63i



 3i 2 yA  yqy e2Lmi  1

t4

qi

225

When a member is subject to a concentrated transverse load Q y

applied at z zQ y and position yQ y relative to the shear center, the
load function

in Eq. (6g) can be demonstrated as qy z
Q y Dirac z  zQ y . Substituting this load function into Eq. (6g), one
can obtain a new geometric stiffness matrix K G Q y due to load
position effect relative to the shear center SC. Matrix M 9  related
to K G Q y is given by
2
3
0
0
0 0 0
0
0
0
6
0
0 0 0
0
0
0 7
6
7
6
7
6
0 0 0
0
0
0 7
6
7
6
0 0
0
0
0 7
7

6
6
7
M 9   Q y yQ y  yA 6
f^ 1 e^ 1;2 e^ 1;3 e^ 1;4 7
6
7
6
7
^
6
^ 2;3 e^ 2;4 7
f
e
2
6
7
6
7
6
Sym:
f^ 3 e^ 3;4 7
4
5
f^
4

A18
A14

in which, functions e^ i;j and f^ i are dened as

e^ i;j 3i 3j ezQ y mi mj
f^ i 3i 2 e2zQ y mi

A19ab

226

A. Sahraei et al. / Thin-Walled Structures 89 (2015) 212226

References
[1] Canadian Standards Association, (CSA), limit states design of steel structures,
CAN/CSA S16-09, Toronto, ON, Canada; 2009.
[2] American Institute of Steel Construction, (AISC). Specication for structural
steel buildings. (ANSI/AISC 360-05). Chicago, IL: AISC; 2005.
[3] American Institute of Steel Construction, (AISC). Specication for structural
steel buildings. (ANSI/AISC 360-10). Chicago, IL: AISC; 2010.
[4] Wu L, Mohareb M. Buckling of shear deformable thin-walled membersI.
Variational principle and analyticalsolutions. Thin Wall Struct 2011;49:
197207.
[5] Wu L, Mohareb M. Buckling formulation for shear deformable thin-walled
membersII. Finite element formulation. Thin Wall Struct. 2011;49:20822.
[6] Anderson JM, Trahair NS. Stability of monosymmetric beams and cantilevers.
J Struct Div, ASCE 1972;98(ST1):26986.
[7] Roberts TM, Burt CA. Instability of monosymmetric I-beams and cantilevers.
Int J Mech Sci 1985;27:31324.
[8] Wang CM, Kitipornchai S. On stability of monosymmetric cantilevers. Eng
Struct 1986;8:16980.
[9] Wang C, Kitipornchai S. Buckling capacities of monosymmetric Ibeams.
J Struct Eng 1986;112:237391.
[10] Kitipornchai S, Wang C, Trahair N. Buckling of monosymmetric Ibeams under
moment gradient. J Struct Eng 1986;112:78199.
[11] Kitipornchai S, Wang CM. Lateral buckling of tee beams under moment
gradient. Comput Struct 1986;23:6976.
[12] Helwig T, Frank K, Yura J. Lateraltorsional buckling of singly symmetric
I-beams. J Struct Eng 1997;123:11729.
[13] Attard MM. General non-dimensional equation for lateral buckling. Thin Wall
Struct 1990;9:41735.
[14] Mohri F, Brouki A, Roth JC. Theoretical and numerical stability analyses of
unrestrained, mono-symmetric thin-walled beams. J Construct Steel Res
2003;59:6390.
[15] Andrade A, Camotim D, Providncia e Costa P. On the evaluation of elastic
critical moments in doubly and singly symmetric I-section cantilevers.
J Construct Steel Res 2007;63:894908.
[16] Eurocode3 Design of Steel Structures, Part 1.1: General rules for buildings.
European Committee for Standartisation, Draft Document ENV 1993-1-1,
Brussels; 1992.
[17] Zhang L, Tong GS. Elastic exural-torsional buckling of thin-walled cantilevers.
Thin Wall Struct 2008;46:2737.
[18] Mohri F, Damil N, Potier-Ferry M. Linear and non-linear stability analyses of
thin-walled beams with monosymmetric I sections. Thin-Walled Struct
2010;48:299315.
[19] Attard MM, Kim M-Y. Lateral buckling of beams with shear deformationsa
hyperelastic formulation. Int J Solids Struct 2010;47:282540.
[20] Camotim D, Andrade A, Basaglia C. Some thoughts on a surprising result
concerning the lateraltorsional buckling of monosymmetric I-section beams.
Thin Wall Struct 2012;60:21621.
[21] Mohri F, Damil N, Potier-Ferry M. Buckling and lateral buckling interaction in
thin-walled beam-column elements with mono-symmetric cross sections.
Appl Math Modell 2013;37:352640.

[22] Krajcinovic D. A consistent discrete elements technique for thinwalled

assemblages. Int J Solids Struct 1969;5:63962.
[23] Barsoum RS, Gallagher RH. Finite element analysis of torsional and torsional
exural stability problems. Int J Numer Methods Eng 1970;2:33552.
[24] V.Z. Vlasov, Thin-walled elastic beams, 2nd ed., Program for scientic translation, Jerusalem, Israel; 1961.
[25] Attard MM. Lateral buckling analysis of beams by the fem. Comput Struct
1986;23:21731.
[26] Papangelis JP, Trahair NS, Hancock GJ. Elastic exuraltorsional buckling of
structures by computer. Comput Struct 1998;68:12537.
[27] Hancock GJ, Bradford MA, Trahair NS. Web distortion and exuraltorsional
buckling. J Struct Div ASCE 1980;106:155771.
[28] Bradford MA, Trahair NS. Distortional buckling of I-beams. J Struct Div ASCE
198135570.
[29] Bradford MA. Distortional buckling of monosymmetric I-beams. J Construct
Steel Res 1985;5:12336.
[30] Erkmen RE. Shear deformable hybrid nite-element formulation for buckling
analysis of thin-walled members. Finite Elements Anal Des 2014;82:3245.
[31] Andrade A, Camotim D, Dinis PB. Lateraltorsional buckling of singly symmetric web-tapered thin-walled I-beams: 1D model vs. shell FEA. Comput
Struct 2007;85:134359.
[32] Bradford MA. Inelastic buckling of tapered monosymmetric I-beams. Eng
Struct. 1989;11:11926.
[33] Bradford MA. Stability of tapered I-beams. J Construct Steel Res 1988;9:
195216.
[34] Trahair NS. Bending and buckling of tapered steel beam structures. Eng Struct
2014;59:22937.
[35] Yuan W-b, Kim B, Chen C-y. Lateraltorsional buckling of steel web tapered
tee-section cantilevers. J Construct Steel Res 2013;87:317.
[36] Gelera K, Park J. Elastic lateral torsional buckling strength of monosymmetric
stepped I-beams. KSCE J Civ Eng 2012;16:78593.
[37] Kim N-I, Lee J. Lateral buckling of shear deformable laminated composite
I-beams using the nite element method. Int J Mech Sci 2013;68:24657.
[38] Lee J. Lateral buckling analysis of thin-walled laminated composite beams
with monosymmetric sections. Eng Struct 2006;28:19972009.
[39] Erkmen RE, Mohareb M. Buckling analysis of thin-walled open membersa
complementary energy variational principle. Thin Wall Struct 2008;46:60217.
[40] Erkmen RE, Mohareb M. Buckling analysis of thin-walled open membersa
nite element formulation. Thin Wall Struct 2008;46:61836.
[41] Erkmen R, Mohareb M, Bradford M. Complementary energy based formulation
for torsional buckling of columns. J Eng Mech 2009;135:14206.
[42] Saad K, Espion B, Warze G. Non-uniform torsional behavior and stability of
thin-walled elastic beams with arbitrary cross sections. Thin Wall Struct
2004;42:85781.
[43] Kollr LP. Flexuraltorsional buckling of open section composite columns with
shear deformation. Int J Solids Struct 2001;38:752541.
[44] Back SY, Will KM. Shear-exible thin-walled element for composite I-beams.
Eng Struct 2008;30:144758.
[45] Trahair NS. Flexural-torsional buckling of structures. USA: CRC Press; 1993.