Journal of Constructional Steel Research 59 (2003) 63–90

[Link]/locate/jcsr

Theoretical and numerical stability analyses of

unrestrained, mono-symmetric thin-walled
beams
F. Mohri a,b,∗, A. Brouki b, J.C. Roth b
a
IUT Nancy-Brabois, Département de Génie Civil, Université Nancy 1, 54601 Villers lès Nancy,
France
b
LPMM (UMR CNRS 7554) Université de Metz, Ile de Saulcy, 57045 Metz, France

Received 13 November 2001; received in revised form 18 January 2002; accepted 20 January 2002

Abstract

In this paper, a contribution to the overall stability of unrestrained thin-walled elements

with open sections is investigated. A model is developed for beam lateral buckling stability
analysis. According to the European steel code, Eurocode 3, an analytical solution is used for
checking the stability of laterally unrestrained beams. It depends on bending distribution, on
load height effect and on degree of the monosymmetry of the section called Wagner’s para-
meter. Coefficients C1, C2 and C3 are respectively affected to these parameters and are given
for some selected load cases. Based on the present model, these coefficients are recomputed
for simply supported beams. Ritz and Galerkin’s methods are applied. Following Galerkin’s
method, the coefficients C1 and C2 are the same as those adopted in Eurocode 3, but some of
Wagner’s proposed C3 coefficients are very different from those usually used. The different
solutions are first discussed and then compared to finite element results. ABAQUS software
is used in numerical simulations where three-dimensional beams with warping and shell
elements are chosen in modelling lateral buckling phenomena. The presented results are close
to shell element results but the regular solutions overestimate the lateral buckling resistance
of mono-symmetric I beams.
 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Beam element; Finite element; Eurocode3; Lateral buckling; Open section; Overall stability;
Shell element; Thin-wall; Warping

∗
Corresponding author. Tel.: (33) 3 83 91 22 20; fax: (33) 3 83 91 22 21..
E-mail address: [Link]@[Link] (F. Mohri).

0143-974X/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 0 2 ) 0 0 0 0 7 - X
64 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

Nomenclature
A Cross section area
a1 , a2, a3 Coefficients of a quadratic equation
Bw Bimoment
C1, C2 , C3 Coefficients of the analytical buckling moments
E Young’s modulus
ez Load height parameter
G Shear modulus
I0 Polar moment of inertia about shear centre
Iy , Iz Principal moment of inertia about y and z axes
Iw Warping constant
J St-Venant torsion constant
L Beam length
M0 Concentrated moment
Mb Buckling moment
Mb⫹ , M⫺ b Positive and negative buckling moments
My, Mz Bending moments about y and z axes
N Axial force
qz Distributed load in the z direction
Qz Concentrated load in the z direction
U Strain energy
Ul Linear part of the strain energy
Unl Second order part of the strain energy
W Work of the applied loads
u, v, w Shear displacement components of the shear centre C in the x, y
and z axes
uM, vM, wM Displacement components of point M in the x, y and z axes
v0 Displacement amplitude of v
x, y, z Principal co-ordinate of point M in the global Gxyz reference
yc, zc Shear co-ordinates in the Gyz reference
by, bz, bw Wagner’s coefficients
exx , exy , exz Axial and shear strain components
el Linear part of the axial strain
enl Second order part of the axial strain
y Coefficient of moment gradient
w Warping co-ordinate
⌸ Total potential
qx Torsion angle
q0 Torsion amplitude of qx
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 65

1. Introduction

Thin-walled elements with open sections are frequently used in steel structures
for their efficiency. Section shapes like I, T, L, C, and Z are commonly used in
engineering practices. With the improvement of the industrial processes in both hot-
rolled and cold-formed elements, these structures offer now, a competitive solution
to structural weight reduction and to high strength resistance. Most of thin-walled
beams have a good resistance in bending about the strong axis, but a low resistance
about the weak axis and also a low resistance in torsion. This is the reason why,
when an unrestrained element bends about its stiffer principal axis, it exhibits an
instability and may suddenly go in flexural–torsional behaviour (Fig. 1a).
The flexural–torsional behaviour of thin-walled beams is more complex since the
torsion is accompanied with warping (Fig. 1b) and also couplings between bending
and torsion. Vlasov’s model [1] developed for a small non-uniform torsion is com-
monly adopted in textbooks on thin-walled structures [2–4]. Always, the stability
analysis is reduced to the determination of buckling loads based on linear stability
and the recourse to design rules is always preferred in complex analyses, like those
adopted in the European steel code Eurocode 3 [5]. There is an extensive literature
devoted to thin-walled elements with an open section in both numerical and theoreti-
cal procedures. Finite element models for behaviour and stability analyses can be
found in [6–8]. Other works have been carried out on the stability of mono-symmetric
sections [9–11].
Following the European steel code Eurocode 3 [5], informative annex F, an ana-
lytical solution is formulated for the problem of lateral buckling of unrestrained
mono-symmetric I section beams. It is a function on boundary conditions, on bending
distribution, on the load height parameter and the degree of monosymmetry of the
section, currently related to Wagner’s coefficient. Some selected load cases are con-
sidered and corresponding coefficients C1, C2 and C3 are presented for each load
case. Most of these solutions have been inspired by the theoretical model developed
by Djalaly [12]. Many comparative studies between the regular and numerical sol-
utions have been investigated by Mohri et al. [13–15].
In the present work, a theoretical model is developed for the lateral buckling of

Fig. 1. (a) Flexural–torsional coupling in lateral buckling behaviour. (b) Section warping of I section
under torsion moment.
66 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

beams. The coefficients C1, C2 and C3 are recomputed for some load cases and new
coefficients are proposed for a simply supported beam with free warping. Compari-
son examples are considered using ABAQUS finite element software [17]. The limits
of the normative solutions used in Eurocode 3 are outlined for some load cases and
also for some section shapes.

2. A theoretical model for the stability analyses

2.1. Kinematics

A straight thin-walled element with an open section is pictured in Fig. 2 and a

direct rectangular co-ordinate system is chosen. Let us denote by x, the initial longi-
tudinal axis and by y and z the first and second principal bending axes . The origin
of these axes is located at the centroid G. The shear centre is denoted C, with its
co-ordinates (yc, zc) in Gyz. Consider M, a point on the section contour with its co-
ordinates (y, z, w) where w is the sectorial coordinate of the point introduced in
Vlasov’s model due to non-uniform torsion [1]. The fundamental assumptions of
this model are:

1. The contour of the cross section is rigid in its plane. This means that there is no
distortion deformation of the section.
2. There is no shear deformation in the mean surface of the section.

Based on the first two assumptions, the three displacement components of M can
be derived from those of the shear centre. The following linear relationships are
commonly used:
uM ⫽ u⫺yv’⫺zw’x⫺wq’x (1)

Fig. 2. An open section beam.

 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 67

vM ⫽ v⫺(z⫺zc)qx (2)
wM ⫽ w ⫹ (y⫺yc)qx (3)
where u, v and w are the displacement components of the shear centre point, respect-
ively in the x, y and z directions. qx is the twist angle. In these formulations,
(.)’denotes the derivative with respect to the x variable.
In the case of thin-walled beams, the components of the strain tensor which incor-
porate the large displacements, are reduced to the following ones and are denoted
by exx , exy and exz
1 1
exx ⫽ u’M ⫹ ((u’M)2 ⫹ (v’M)2 ⫹ (w’M)2)⬇u’M ⫹ ((v’M)2 ⫹ (w’M)2). (4)
2 2
In developing this relationship according to the displacement components (1–3), it
can be written as
exx ⫽ el ⫹ enl (5a)
in which
el ⫽ u’⫺yv⬙⫺zw⬙⫺wq⬘⬘x (5b)
1
enl ⫽ (v’2 ⫹ w’2 ⫹ R2qx’2) ⫹ (y⫺yC)w’qx’⫺(z⫺zC)v’q’x. (5c)
2
The other strain components are given by
1 ∂w
exy ⫽ ⫺ (z⫺zC ⫹ )q’x (6)
2 ∂y
1 ∂w
exz ⫽ (y⫺yC ⫹ )qx’. (7)
2 ∂z
in (5c), the expression of R2 is
R2 ⫽ (y⫺yC)2 ⫹ (z⫺zC)2. (8)
In the case of an elastic behaviour, the strain energy of the beam including the
torsion is

U⫽
1
2 冕冕
LA
Ee2xxdA dx ⫹
1
2冕
L
GJ(qx’)2dx. (9)

The last term in this relationship is the St-Venant torsion strain energy part. E and
G are respectively Young’s and shear moduli. J denotes the torsion constant. The
strain energy U is developed according to (5a) becoming

U⫽
1
2 冕冕
LA
E(e2l ⫹ 2elenl ⫹ e2nl)dA dx ⫹
1
2 冕
GJ(qx’)2dx.
L
(10)
68 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

In linear stability, the contribution of e2nl is omitted. In this case, the strain energy is

U⫽
1
2 冕冕
LA
Ee2l dAdx ⫹
1
2冕
L
GJ(qx’)2dx ⫹ 冕冕
LA
EelenldAdx ⫽ Ul ⫹ Unl (11a)

where Ul and Unl represent the linear and second order parts of the strain energy.
They are expressed by

Ul ⫽
1
2 冕冕
LA
Ee2l dA dx ⫹
1
2 冕
GJ(qx’)2dx
L
(11b)

Unl ⫽ 冕冕
LA
EelenldA dx. (11c)

Using relationship (5a) and the conditions of the principal axes in bending and
warping, the following expression is obtained for Ul

Ul ⫽
1
2 冕
(EA(u’)2 ⫹ EIz(v⬙)2 ⫹ EIy(w⬙)2 ⫹ EIw(q⬙x)2)dx
L
(12)

⫹
1
2 冕
GJ(q’x)2dx.
L

A denotes the cross section area, Iy and Iz are the principal inertial moments about
the principal bending axes. Iw is the warping constant. These geometric parameters
have the following relationships:

A ⫽ dA 冕
A
(13a)

冕
Iy ⫽ z2 dA
A
(13b)

冕
Iz ⫽ y2dA
A
(13c)

冕
Iw ⫽ w2dA.
A
(13d)

In the same manner, based on the relationships (5b, c) and the conditions of the
principal axes, the second order part of the strain energy Unl can be formulated as
a function of the initial section stress resultants and leads to
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 69

Unl ⫽ Unl(N) ⫹ Unl(My) ⫹ Unl(Mz) ⫹ Unl(Bw) (14a)

with
1 1
冕冉 I0
Unl(N) ⫽ N v’2 ⫹ w’2 ⫹ qx’2 ⫹ zcv’q’x⫺ycw’q’x dx
2 2
L
2 冊 (14b)

冕
Unl(My) ⫽ My(bzqx’2 ⫹ v⬙qx)dx
L
(14c)

冕
Unl(Mz) ⫽ ⫺ Mz(byqx’2⫺w⬙qx)dx
L
(14d)

Unl(Bw) ⫽ ⫺ bwqx’2dx冕 (14e)

in which, N denotes the axial force, My and Mz represent the bending moments. Bw
is the bimoment introduced in Vlasov’s model (Fig. 3). According to (5b) and the
geometric parameters (13a–13d), the section stress resultants are related to displace-
ment components by the following relationships

冕
N ⫽ Eel dA ⫽ EAu’;
A
(15a)

冕
My ⫽ Eelz dA ⫽ ⫺EIyw⬙;
A
(15b)

冕
Mz ⫽ ⫺ Eely dA ⫽ EIzv⬙;
A
(15c)

冕
Bw ⫽ ⫺ Eelw dA ⫽ EIwq⬙x.
A
(15d)

Fig. 3. Section stress resultants in a open section (Msv , St- Venant torsion moment).
70 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

In Eqs (14a–14d), by , bz , bw are Wagner’s coefficients. I0 is the polar moment of

inertia about the shear centre. They are defined as follows

by ⫽
1
2Iz 冕
y(y2 ⫹ z2)dS⫺yc;
S
(16a)

bz ⫽
1
2Iy 冕
z(y2 ⫹ z2)dS⫺zc;
S
(16b)

bw ⫽
1
2Iw 冕
w(y2 ⫹ z2)dS;
S
(16c)

Iy ⫹ Iz
I0 ⫽ ⫹ y2c ⫹ z2c . (16d)
A
The linear strain energy Ul combined with the second order strain energy Unl can
be used in the overall stability analyses of an open thin-walled element. In buckling
problems of bars, only axial loads are initially applied (N ⫽ 0 and My ⫽ Mz ⫽
Bw ⫽ 0). The expression of the strain energy is then reduced to
U ⫽ Ul ⫹ Unl(N). (17)
The buckling loads in pure bending, in pure torsion or in flexural–torsional can be
computed from (17). In the case of the lateral buckling of beams loaded initially
about the strong bending axis (y-axis here), the stress resultants are reduced to My
(My ⫽ 0 and N ⫽ Mz ⫽ Bw ⫽ 0). The strain energy is then
U ⫽ Ul ⫹ Unl(My). (18)
In the following sections, the lateral buckling is investigated. Relationship (18) is
then explored.

2.2. Lateral buckling of beams

Consider a straight beam under a uniformly distributed loads qz (Fig. 4). In order
to fulfil the conditions imposed in Eurocode 3 [5], a mono-symmetric I section is
chosen and the loads are applied along points P located at the height ez from the
shear centre. The external load work W is defined by the relationship

冕
W ⫽ qzwpdx
L
(19)

in which, wp is the vertical displacement of points P. Taking into account for second
order torsion terms, the expression of wp can be written as
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 71

Fig. 4. A monosymmetric I section beam and load height parameter (ez).

wp ⫽ w⫺ez(1⫺cosqx). (20)
When the circular function cos qx is approximated by cosqx ⫽ 1⫺q2x / 2, the final
expression of the external work is

冕
L
1
2
q2x
W ⫽ qzwdx⫺ qzez dx.
2 冕
L
(21)

According to relationships (18) and (21), the total potential of the beam in lateral
buckling behaviour is defined as

⌸ ⫽ U⫺W ⫽
1
2冕(EIyw⬙2 ⫹ EIzv⬙2 ⫹ GJq⬘2x ⫹ EIwq⬙x2)dx ⫹ My(qxv⬙
L
冕L
(22)

冕
⫹ bzq⬘2x )dx⫺ qzwdx ⫹
L
1
冕
L
q e q2 dx.
2 z z x

The total potential ⌸ can be written in two components which depend on displace-
ment couplings:

⌸(w) ⫽
1
2冕
L
冕
(EIyw⬙2)dx⫺ qzwdx
L
(23a)

⌸b ⫽
1
2 冕
(EIzv⬙2 ⫹ GJqx’2 ⫹ EIwq⬘⬘2
L
x )dx ⫹ 冕
My(qxn⬙ ⫹ bzqx’2)dx
L
(23b)

⫹
1
冕
q e q2dx.
2 z z x
L

The first term ⌸(w) is referred only to displacement w and corresponds to the
classical bending potential of the beam before buckling. It permits to lead the equilib-
72 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

rium equation in the vertical direction (fundamental state). In linear stability, this
term has no effect on the buckling loads of the beam. The second term ⌸b is the
additional potential in the buckled state. It allows us to give the equilibrium equations
in the lateral buckling state. The buckling loads are estimated from this expression.
In what follows, since we are concerned with lateral buckling, only the reduced form
of the potential (23b) is then explored. Equilibrium equations corresponding to the
buckled state are obtained from the stationary conditions of the potential (23b). In
the case of conservative loads, the following conditions yield

冕
d⌸b ⫽ (EIzv⬙dv⬙ ⫹ GJq⬘xdqx⬘ ⫹ EIwq⬙xdq⬙x)dx ⫹ My(v⬙dqx ⫹ qxdv⬙
L
冕
L
(24)

冕
⫹ 2bzq⬘xdqx⬘)dx ⫹ qzezqxdqxdx ⫽ 0
L

The variation of the total potential ⌸b is a function on the virtual displacements

and their derivatives. Integrating by part the relationship (24), one gets an expression
which depends only on the virtual displacements of the beam in lateral buckling
which are δv and δqx. After some calculations, the two lateral buckling equilibrium
equations are arranged to
EIzv(4) ⫹ (Myqx)⬙ ⫽ 0 (25a)
x ⫺(GJ ⫹ 2bzMy)q⬙x⫺2bzM⬘yq⬘x ⫹ qzezqx ⫹ Myv⬙ ⫽ 0
EIwq(4) (25b)
In these equations, expressions like v(4) denote the x-derivative of order 4 (e.g.
v(4) ⫽ d4v / dx4). The equilibrium equations (25a) and (25b) are coupled. However,
in the case of a simply supported beam, the equilibrium equation (25a) can be trans-
formed to
EIzv⬙ ⫹ Myqx ⫽ 0 (26a)
Using this relationship, the torsion equilibrium equation (25b) is then uncoupled to
v and yield
M2y
x ⫺(GJ ⫹ 2bzMy)q⬙x⫺2bzM⬘yq⬘x ⫹ qezqx⫺
EIwq(4) q ⫽0 (26b)
EIz x

2.3. Analytical expressions of buckling loads

Two principal methods are used in stability literature for computing analytical
solutions for the buckling loads. The recourse to Ritz and Galerkin’s methods is
usual. The Ritz method is based on the stationary conditions of the total potential.
Galerkin’s method is applied to the differential equilibrium equations. With the two
methods, the displacement modes are firstly approximated by some functions which
are compatible with the boundary conditions of the beam. In the present work, Ritz
method is applied to the beam potential given in (23b). Galerkin’s method is applied
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 73

to the torsion equilibrium equation (26b). Using the two methods and after integration
along the beam according to the adopted functions for the mode displacements, the
buckling loads or equivalent buckling moments are given by the solutions of a quad-
ratic equation written in the form

⫺a1M2b ⫹ Pz(a2ez ⫹ a3bz)Mb ⫹ P2z

Iw
Iz 冉 GJL2
1⫹ 2
p EIw 冊 (27)

where a1, a2 and a3 are coefficients. Mb is the unknown buckling moments of the
beam. Pz is Euler’s buckling load (Pz ⫽ p2EIz / L2). The solution of this equation for
the unknown Mb lets us now write the buckling moments in a compact form as fol-
lows:

Mb ⫽ C1
p2EIz
L2 冋
(C2ez ⫹ C3bz) ± 冪(C e ⫹ C b )
2 z 3 z
2
⫹ 冉
Iw
Iz
GJL2
1⫹ 2
p EIw
. 冊册 (28a)

The coefficients C1, C2 and C3 are given by

冪a
1
C1 ⫽ (28b)
1

a2
C2 ⫽ C1 (28c)
2
a3
C3 ⫽ C1 . (28d)
2
This relationship is similar to that adopted in Eurocode 3 (Annex F)[5]. The coef-
ficients C1, C2 and C3 depend on the load case applied to the beam. The lateral
buckling moment of the beam is then a function of bending distribution (C1), load
height effect ez (C2) and Wagner’s coefficient bz (C3), which define the degree of
monosymmetry of the section.
As an example, let us develop these two methods to the case of a simply supported
beam under a uniformly distributed load, pictured in Fig. 4. First, for such boundary
conditions, the displacement modes in bending and torsion are approximated by a
sinusoidal function
v qx
⫽ ⫽ sin p
v0 q0 L
x
冉冊 (29)

where v0 and q0 are the associated displacement amplitudes. According to the load
qz ,the expression for the bending moment My is
qzL x2
My ⫽ x⫺qz . (30)
2 2
Ritz’s method is applied to the potential (23b). According to relationships (29) and
(30) and after integration and some calculations, a quadratic expression is obtained
for the potential ⌸b . It can be written in matrix form as
74 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

冤 冥
k3

再冎
k1
2 v0
⌸b ⫽ {v0 q0}t (31)
k3 q0
k2
2

with
p4EIz
k1 ⫽ ;
4L3

p4
k2 ⫽ EIw 3 1 ⫹ 2
4L 冉
GJL2
p EIw
⫹
Mb p2⫺3
L 3 冊 冋冉 冊
bz ⫹ 2ez ; 册
k3 ⫽
L 冉
Mb p2 ⫹ 3
3
. 冊
For non-trivial solutions, the matrix determinant is reduced according to the compact
form (27) and leads to

⫺ 冉
2(p2 ⫹ 3) 2 2
3p2 冊 8
Mb ⫹ Pz 2ez ⫹
p 3p冋
4(p2⫺3)
2 冉 Iw
冊册
GJL2
bz Mb ⫹ P2z 1 ⫹ 2
Iz p EIw
⫽ 0, 冉 冊
where Mb is the buckling moment related to the maximal bending moment of the
beam, Mb ⫽ qzL2 / 8. By identification, according to the relationships (27) and(28)
the coefficients C1, C2 and C3 are then
3p2
C1 ⫽ ⫽ 1.15;
2(p2 ⫹ 3)

6
C2 ⫽ ⫽ 0.46;
p2 ⫹ 3

p2⫺3
C3 ⫽ ⫽ 0.53.
p2 ⫹ 3
In the same manner, Galerkin’s method is applied to the torsion equilibrium Eq.
(26b) and the relationships (29) and (30) are also used. After integration along the
beam, a quadratic equation is obtained and is reduced according to the relation-
ship (27):

⫺ 冉
8p4 ⫹ 360 2
15p 4 冊
8
Mb ⫹ Pz 2zq ⫹
p 3p2 冉
4(p2⫺3) Iw GJL2
bz Mb ⫹ P2z 1 ⫹ 2
Iz p EIw冊⫽0 冉 冊
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 75

By identification, the coefficients C1, C2 and C3 are then given by

冪
15p4
C1 ⫽ ⫽ 1.132;
8p ⫹ 360
4

8
C2 ⫽ C ⫽ 0.4589;
2p2 1

4(p2⫺3)
C3 ⫽ C1 ⫽ 0.525.
6p2
However, when a concentrated load Qz is applied to the beam at the position
(x ⫽ a), instead of a distributed load (Fig. 5), the load qz in the potential ⌸b (23b)
and in the torsion equilibrium equation (26b) is written as
qz ⫽ Qz⌬(x⫺a) (32)
where ⌬is Dirac’s function (⌬ ⫽ 1 if x ⫽ aand ⌬ ⫽ 0 elsewhere).
The same procedures can then be followed for the other load cases and the Ci
coefficients can be computed without any difficulty. They are not developed here,
but the coefficients corresponding to some principal load cases are summarised in
Table 1. The coefficients obtained with the two methods are presented. The equival-
ent coefficients adopted in Eurocode 3 [5] are noted for information and comparison
in the last columns.
One can observe, that in load case 1, the coefficients resulting from Galerkin’s
method are the same as those adopted in Eurocode 3 and that those obtained with
Ritz are good but C1 is approximate. In load cases 2 and 3, the coefficients C1 and
C2 obtained with Galerkin’s method are the same as Eurocode’s coefficients, but the
coefficients C3 , written in bold, are very different. Load case 4 is not treated in
Eurocode 3, but it is very important in engineering practices.
The improved coefficients corresponding to load case 2 (C1 ⫽ 1.36, C2 ⫽ 0.55
and C3 ⫽ 0.41) were presented for the first time without any developments by Mohri
and Brouki [13] and Balac and Kolekova [16]. Also, for this load case, Trahair [4]
used the Ritz method and obtained (C1 ⫽ 1.42, C2 ⫽ 0.58 and C3 ⫽ 0.42). One can
observe some differences between the improved coefficients and those of the Ritz

Fig. 5. (a) A beam under a concentrated load. (b) A beam under moment gradient.
76 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

Table 1
Ci coefficients corresponding to different load cases

Ritz method Galerkin Eurocode 3 [5]

No. Load case Mb C1 C2 C3 C1 C2 C3 C1 C2 C3

1 qzL2 /8 1.15 0.46 0.53 1.13 0.46 0.53 1.13 0.46 0.53

2 QzL / 4 1.42 0.58 0.42 1.36 0.55 0.41 1.36 0.55 1.73

3 QzL / 4 1.05 0.43 0.57 1.05 0.43 0.55 1.05 0.43 1.12

4 QzL / 3 1.11 0.51 0.49 1.095 0.499 0.481 – – –

method. With this in view, a comparative study was made by Mohri [13] and showed
that the improved coefficients (C1 ⫽ 1.36, C2 ⫽ 0.55 and C3 ⫽ 0.41) are closer to
Finite Element results than those resulting from the Ritz method (C1 ⫽ 1.42,
C2 ⫽ 0.58 and C3 ⫽ 0.42).
In the case of a beam under moment gradient (M0, yM0) (Fig. 5b), a similar
expression to (28a) is formulated in Eurocode 3 (annex F) and the coefficients C1
and C3 are proposed for some y values, varying from –1 to +1. The author’s interest
[15] focused on the validity of such analytical solutions when they are investigated
according to relation (29). First, one can check that for a beam under a moment
gradient, the coefficients C1 and C3 can be computed according to the displacement
modes (29) and the following expression for the bending variation My along the beam
x
冉 冊x
My(x) ⫽ M0 1⫺ ⫹ ⌿M0 .
L L
(33)

When Galerkin’s method is applied to the torsion equilibrium equation (26b), the
final expressions of the coefficients C1 and C3 as function of ⌿ are [15]:

冪a ;
1
C1 ⫽ (34a)
1

1⫹⌿
C3 ⫽ C1 ; (34b)
2
with
2p2⫺3 2 2p3 ⫹ 6
a1 ⫽ 2 (y ⫹ 1) ⫹ ⌿. (35c)
6p 6p2
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 77

According to these relationships, the numerical values of C1 and C3 for some values
of ⌿ are presented in Table 2. The equivalent coefficients adopted in Eurocode 3
[5] are also presented. For these load cases, it is then demonstrated that the coef-
ficients C1 and C3 adopted in Eurocode 3 are computed under the assumption of the
sinusoidal displacements mode shown in (29) which are symmetric.
Nevertheless, it is well known in stability literature, that mode displacements
depend strongly on load distribution and boundary conditions. Mode displacements
are symmetric when loads are symmetric and are not symmetric when the loads have
not symmetric distribution [9]. In this case, relationship (29) is acceptable for all
load cases pictured in Table 1. In the case of a beam loaded by the moment gradients
of Table 2, one can check that only the uniform bending corresponding to ⌿ ⫽ 1
is symmetric. All the other gradient moments (⌿ ⫽ 1) are not symmetric and the
displacement buckling modes are then not symmetric. In this case, the relationships
(29) and the corresponding buckling loads or moments are no longer valid. When
a beam is loaded under asymmetric loads, a realistic estimation of the buckling loads
can be obtained when the displacement modes are approximated according to Fourier
sine series with many terms [9,20], like

冘
n

v(x) ⫽ aifi(x) (35a)

i⫽1

冘
n

q(x) ⫽ bigi(x) (35b)

i⫽1

where ai and bi are the associated displacement amplitudes. fi and gi are sinusoidal
functions; ‘n’ is the number of terms needed. In these conditions, analytical solutions
are very difficult to obtain and only numerical solutions can be found. The use of the
relationship (29) in the case of a beam under gradient moments and the corresponding
analytical solutions has been extensively discussed in a previous paper [15]. It is

Table 2
C1 and C3 coefficients for a beam under moment gradient

Mb ⫽ M0 Galerkin Eurocode 3 [5]

⌿ C1 C3 C1 C3

1 1 1 1 1
0.75 1.141 0.999 1.141 0.998
0.5 1.324 0.993 1.323 0.992
0.25 1.564 0.977 1.563 0.977
0 1.881 0.94 1.88 0.94
–0.25 2.284 0.857 2.281 0.855
–0.5 2.711 0.678 2.704 0.676
–0.75 2.94 0.368 2.927 0.366
–1 2.766 0 2.75 0
78 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

demonstrated that the lateral buckling resistance of mono-symmetric I sections are

largely overestimated by Eurocode’s solutions. In what follows, the validity of the
coefficients C1, C2 and C3 will be investigated in the case of a beam under transverse
load cases shown in Table 1.

3. Numerical comparisons

In the preceding section, an analytical model was developed for lateral buckling
stability analysis of beams. The theoretical bases of the regular solutions have been
investigated and the coefficients C1, C2 and C3 are shown for many load cases. Ritz
and Galerkin’s methods have been used in the analysis. Following Galerkin’s
method, the coefficients C1 and C2 are the same as those adopted in Eurocode 3,
but some of Wagner’s proposed C3 coefficients are very different to the regular ones.
In what follows, in order to make easier the presentation of the numerical results,
the buckling moments computed from the analytical solutions (28) are first discussed
in the case of a bisymmetric and a mono-symmetric I section. After this, many
examples are treated and all the analytical solutions are compared to finite
element results.

3.1. Lateral buckling beam resistance

Following the theoretical relationship (28), there are two buckling moments Mb.
One is positive and the other is negative, denoted respectively Mb⫹ and M⫺
b . Their
expressions are

b ⫽ C1
M+b,M⫺
p2EIz
L2 冋
(C2ez (36)

⫹ C3bz) ± 冪(C e ⫹ C b )
2 z 3 z
2
⫹
Iw
Iz
GJL2
(1 ⫹ 2 ) .
p EIw 册
In bar buckling, the engineer is looking only for positive buckling loads which
are in compression. The buckling loads corresponding to traction are not retained.
However, in lateral buckling of beams, the instability can occur either under positive
or negative loads and in this case, the two corresponding buckling moments have a
physical interpretation (Fig. 6).
Let us compare the two buckling moments Mb⫹ and M⫺ b in the case of a bisym-
metric and mono-symmetric I section (Fig. 7a, b). For a bisymmetric I section, the
shear centre C is located at the centroid G and it is well known that Wagner’s
coefficient is zero for such section shape. The buckling moments depend only on
the two coefficients C1 and C2. When the loads are applied on the shear centre, the
load height parameter ez vanishes and one can verify that:

b ⫽ ± C1
M+b,M⫺
p2EI
L2 冋冪 冉Iw
Iz
GJL2
1⫹ 2
p EIw
. 冊册 (37a)
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 79

Fig. 6. Beam deformation under positive and negative lateral buckling loads.

Fig. 7. (a) Bisymmetric I section (zc ⫽ 0; bz ⫽ 0). (b) Mono-symmetric I sections (zc ⫽ 0; bz ⫽ 0).

The intensities of the buckling moments Mb⫹ and M⫺

b are then equal, so:

M+b ⫽ |M⫺
b |. (37b)
When the loads are not applied on the shear centre (ez ⫽ 0), the two buckling
moments are given by

b ⫽ C1
M+b,M⫺
p2EIz
L2 冋
(C2ez) ± 冪 (C2ez)2 ⫹
Iw
Iz 冉 GJL2
1⫹ 2
p EIw
. 冊册 (38a)

Then, the intensities of the buckling moments Mb⫹ and M⫺

b are not equal, so

M+b ⫽ |M⫺
b |. (38b)
For a mono-symmetric I section, Wagner’s coefficient is different from zero. The
buckling moments depend on all the coefficients C1,C2 and C3. Nevertheless, when
the loads are applied on the shear centre (ez ⫽ 0) , the two buckling moments are
given by

b ⫽ C1
M+b,M⫺
p2EIz
L2 冋
(C3bz) ± 冪 (C3bz)2 ⫹
Iw
Iz冉 GJL2
1⫹ 2
p EIw
. 冊册 (39)

One can check here that the intensities of the two buckling moments are not equal
and relationship (38b) is satisfied. In the same manner, when the loads are not applied
on the shear centre, all the coefficients intervene. It is evident that the intensities of
the two buckling moments are different and relationship (38b) also occurs.
80 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

3.2. Finite element validations

In numerical simulations, the buckling loads are computed using the

Abaqus/Standard finite element software [17]. The lateral buckling loads or equival-
ent buckling moments are computed from the solutions of the eigenvalue problem.
Three-dimensional beam elements including the warping deformations (B31OS) and
thin-walled shell elements (S8R5) are used. Two steel sections are chosen for the
analysis. The first section is bisymmetric, denoted section A and the second is mono-
symmetric, denoted section B. The dimensions and the geometric parameters used
in analytical calculations are summarised in Table 3. Geometrical characteristics,
such as Iy, Iz, Iw and the shear centre position are available in literature on thin-
walled structures for the current section shapes such as I , T and channel [1,4,18,19].
The other parameters that intervene in the stability and in non-linear analyses, like
Wagner’s coefficients, are rarely shown. In a previous paper, Mohri et al. [21]
presented an efficient numerical method for computing the geometric parameters for
any open section shape. The elastic modulus used in all the analytical and numerical
calculations are respectively, E ⫽ 210GPa and G ⫽ 80.77GPa.

3.2.1. Finite element beam results

First, let us comment on the numerical results coming from beam elements
(B31OS). The analytical and the numerical buckling loads corresponding to load
case 2 of Table 1 are presented in Table 4 for section A . Two load positions are
considered and the corresponding positive and negative buckling moments are
shown. One can observe that for this section, the regular results are concord with
the three-dimensional beam elements (B31OS). According to relationships (37) and
(38), the positive and negative buckling moments have the same intensity when the
load is applied on the shear centre and are different when the load is on the top flange.
The analytical and numerical buckling loads corresponding to the same load case
(load case 2) are presented in Table 5 for section B. The load is applied on the shear
centre. The first two columns are the regular solutions computed with Wagner’s
coefficient C3 ⫽ 1.73 adopted in Eurocode 3. The results computed with the pro-

Table 3
dimensions and geometric properties for the selected I sections

Section A: Section B: as section A, but the

h=300 mm; b=150 mm; bottom flange width is reduced to
tf=10.7; tw =7.1 mm. b⬘=75 mm.
A=51.88 cm2; A=43.86 cm2;
Iy=7998.98 cm4; Iy=6011.84 cm4;
Iz=602.71 cm4; Iz=339.39 cm4;
J=15.57 cm4; J=12.51 cm4;
Iw=125.93×103 cm6; Iw=27.985×103 cm6;
bz =0.0 bz =10.48 cm;
zc=8.6 cm
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 81

Table 4
Numerical and analytical buckling moments for section A (in kNm)

Load on shear centre Load on top flange

L(m) EC3 EC3 B31OS B31OS EC3 EC3 B31OS B31OS

(+) (⫺) (+) (⫺ ) (+) (⫺) (+) (⫺)

4 204.26 ⫺204.26 204.51 ⫺204.51 134.65 309.85 133.41 ⫺310.63

5 145.86 ⫺145.86 146.05 ⫺146.05 100.20 ⫺212.32 99.35 ⫺212.5
6 112.81 ⫺112.81 112.95 ⫺112.95 80.41 ⫺158.27 79.74 ⫺159.05
7 91.89 ⫺91.89 92.01 ⫺92.01 67.63 ⫺124.84 67.06 ⫺125.51
8 77.55 ⫺77.55 77.6 ⫺77.6 58.68 ⫺102.48 58.17 ⫺103.06

Table 5
Numerical and analytical buckling moments for section B (load on shear centre)

EC3; C3 ⫽ 1.73 Improved; C3 ⫽ 0.41 Abaqus (B31OS) Analytic; C3 ⫽ 0

L(m) Positive Negative Positive Negative Positive Negative Positive Negative

4 259.79 ⫺42.98 134.43 ⫺83.05 105.98 ⫺106.05 105.67 ⫺105.67

5 175.59 ⫺36.83 98.52 ⫺65.64 80.65 ⫺80.68 80.41 ⫺80.41
6 129.15 ⫺32.79 77.48 ⫺54.65 65.24 ⫺65.25 65.07 ⫺65.07
7 100.59 ⫺29.80 63.78 ⫺47.00 54.86 ⫺55.18 54.75 ⫺54.75
8 81.63 ⫺27.42 54.17 ⫺41.32 47.38 ⫺47.38 47.3135 ⫺47.3135

posed coefficient C3 (C3 ⫽ 0.41) are arranged in columns 3 and 4. Abaqus results
with B31OS beam elements are arranged in columns 5 and 6. One can observe that
according to relationship (39), the positive and negative analytical buckling moments
are different, but the numerical buckling moments resulting from the three-dimen-
sional beams (B31OS) have the same intensity. The Abaqus results are very surpris-
ing and it is easy to check that the corresponding numerical buckling moments can
be easily found using the analytical solution (39) with C3 ⫽ 0 and the geometric
parameters for section B, presented in Table 3 . The corresponding results are
reported in the two last columns. This simple example demonstrates that the use of
three-dimensional warping elements beam (B31OS) is not desirable for the stability
analysis of the mono-symmetric sections, since they do not take Wagner’s coefficient
into account. The recourse to other elements like shell elements is then highly rec-
ommended.

3.2.2. Finite element shell results

With shell elements, web and flange walls are modelled with S8R5. With such
elements many problems can affect the numerical results, like local buckling, section
distortion and the local effects of concentrated loads and boundary conditions. All
82 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

these problems are naturally ignored in beam theory. Nevertheless, most of these
phenomena can be reduced since we are concerned with slender beams.
In the present work, attention is on modes corresponding to overall buckling. The
first two overall buckling modes are pictured in Fig. 8a, b. As an example, the
analytical and numerical buckling moments for a beam with section A, under load
case 2 are shown in Table 6 for comparison. The load is applied on the top and
bottom flanges. The numerical buckling moments resulting from three-dimensional
beam elements (B31OS) and shell elements (S8R5) concord with the analytical sol-
utions. Shell elements are also tested in the case of mono-symmetric sections. With
this in view, the numerical and analytical buckling moments variation with length
L are shown in Table 7, for a beam with section B under load case 2. The load is
applied on the shear centre. One can observe here, that according to relationship
(38a), the positive and negative numerical buckling moments resulting from shell
elements are not equal and concord with the improved analytical solutions (C3 ⫽
0.41). For this load height, the difference between regular solutions (C3 ⫽ 1.73)
and numerical results is very impressive. In what follows, the stability analysis of

Fig. 8. (a) Lateral buckling of I beam: first overall buckling mode. (b) Lateral buckling of I beam:
second overall buckling mode.
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 83

Table 6
Load case 2: Beam and shell numerical results compared to the regular buckling moments of the beam
section A (in kNm)

Load on top flange Load on bottom flange

L(m) EC3 B31OS S8R5 EC3 B31OS S8R5

4 134.65 133.41 130.33 309.85 310.63 305.82

5 100.2 99.35 96.93 212.32 212.50 207.68
6 80.41 79.74 78.45 158.27 159.05 156.57
7 67.63 67.06 66.54 124.84 125.51 124.85
8 58.68 58.17 58.25 102.48 103.06 103.69

Table 7
Load case 2: numerical and analytical buckling moments for section B (in kNm; load on shear centre)

L(m) EC3; C3 ⫽ 1.73 Improved; C3 ⫽ 0.41 Abaqus (S8R5)

L(m) Positive Negative Positive Negative Positive Negative

4 259.79 ⫺42.98 134.43 ⫺83.05 130.89 ⫺78.34

5 175.59 ⫺36.83 98.52 ⫺65.64 97.36 ⫺63.45
6 129.15 ⫺32.79 77.48 ⫺54.65 77.41 ⫺53.99
7 100.59 ⫺29.80 63.78 ⫺47.00 64.40 ⫺47.40
8 81.63 ⫺27.42 54.17 ⫺41.32 55.32 ⫺42.52

mono-symmetric I sections will be discussed and only shell elements are then used
in finite elements simulations.

3.3. Stability of mono-symmetric I sections

The analytical solutions discussed in the preceding sections are compared only
with shell element results. The analytical solutions are computed according to
relationship (27) and corresponding coefficients C1, C2 and C3 presented in Table 1
for the different load cases. Also, the numerical results presented here concern only
the load cases for which the improved C3 coefficients are different from those
adopted in Eurocode 3 [5] (see load cases 2, 3 and 4). The mono-symmetric I section
B is considered in the analysis.
The numerical and analytical buckling moment variation with length L are plotted
in Figs. 9 and 10, for load case 2. Two load positions are considered. The regular
solutions are reported with C3 ⫽ 1.73. The proposed results are computed with
C3 ⫽ 0.41. The numerical and analytical buckling moment variation with length are
presented in Figs. 11 and Fig. 12, for load case 3. The load is applied on the top
and bottom flanges. The regular solutions are reported with C3 ⫽ 1.12 and the
84 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

Fig. 9. Numerical and analytical buckling moments variation versus length L, Section B, load on top
flange.

Fig. 10. Numerical and analytical buckling moments variation versus length l, section B, load on bottom
flange.

improved results are computed with C3 ⫽ 0.55. One can observe that for the two
load cases, the proposed solutions are in agreement with shell results and that the
buckling moments computed according to the Eurocode 3 solutions are very much
higher than the numerical solutions for all the considered lengths. The difference
between the regular and shell results is very remarkable and can reach 65% in load
case 1 and 33% in the load case 2.
Numerical and analytical buckling moment variation with length L are shown in
Figs. 13 and 14, for the beam loaded with 2 concentrated forces applied at x ⫽
L / 3 ( load case 4 of Table 1), for two load positions, respectively on the top and
bottom flanges. This case is not considered in Eurocode 3 [5]. The improved results
are computed with C1 ⫽ 1.095; C2 ⫽ 0.499 and C3 ⫽ 0.48. The analytical solutions
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 85

Fig. 11. Numerical and analytical buckling moments variation versus length L, section B, load on top
flange.

Fig. 12. Numerical and analytical buckling moments variation versus length l, section B, load on bottom
flange.

concord well with shell results. One can check that the buckling moments of this
load case are very much lower than the corresponding buckling moments of the load
case 3 in Table 1. In lateral buckling, this load case is then more important than
load case 3 for engineering practices.
86 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

Fig. 13. Numerical and analytical buckling moments variation versus length L, section B, load on top
flange.

Fig. 14. Numerical and analytical buckling moments variation versus length L, section B, load on bottom
flange.

3.4. Incidence on lateral buckling beam resistance

The results presented in the preceding section show that the buckling moments
corresponding to the regular solutions are very much higher than the improved ones
in the case of beams with mono-symmetric I sections. Let us consider the effect of
buckling moment on beam resistance. According to Eurocode 3 [5], the ultimate
beam resistance Mrd is a function on the lateral buckling moment Mb of the beam.
In the case of a cross section pertaining to Class 1, its expression is
Mrd ⫽ cLTMpl,y, (40)
where Mpl,y is the plastic section moment and cLT is the buckling factor. The buckling
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 87

Table 8
Bending moments resistance of a beam with section B, under load case 2.

Load on top flange Load on bottom flange

L(m) EC3 Improved EC3 Improved

(C3 ⫽ 1.73) (C3 ⫽ 0.41) (C3 ⫽ 1.73) (C3 ⫽ 0.41)

4 93.54 76.00 100.24 95.19

5 85.50 64.62 95.56 87.92
6 76.22 55.27 89.75 79.25
7 66.88 47.95 82.77 70.11
8 58.50 42.22 75.01 61.56

factor is computed from the slenderness ratio lLT ⫽ 冑Mpl,y / Mb and the imperfect
curve (a) [5].
As an example, the beam resistance under load case 2 is considered for two mono-
symmetric I sections. The first is section B considered previously. The other section
is a T section obtained from section B with no bottom flange. The plastic moments
of these sections under consideration, of a steel grade S235 (fy ⫽ 235MPa), are
Mpl,y ⫽ 108.8kNm for section B and Mpl,y ⫽ 65.62kNm for T section. The resistant
moments of the beam, corresponding to Eurocode 3 and the improved analytical
solutions, are presented in Table 8, for section B. One can observe that the bending
resistant moments are overestimated by the regular solutions. The difference can
reach 38% when the load is on the top flange and 20% when the load is on the
bottom flange. The bending resistant moments of the T section beam, corresponding
to Eurocode 3 and the improved analytical solutions, are exposed in Table 9. For
this section shape, the bending resistant moments are also overestimated by the reg-
ular solutions.
In previous work [15], the lateral buckling of beams under moment gradient was
investigated. Many numerical examples on mono-symmetric I sections were dis-

Table 9
Bending moments resistance of a beam with T section, under load case 2

Load on top flange Load on bottom flange

L(m) EC3 Improved EC3 Improved

(C3 ⫽ 1.73) (C3 ⫽ 1.73) (C3 ⫽ 1.73) (C3 ⫽ 0.41)

4 60.11 52.43 62.43 60.30

5 57.31 47.39 60.62 57.58
6 53.96 42.40 58.51 54.33
7 50.09 37.86 56.03 50.55
8 45.90 33.93 53.13 46.42
88 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

cussed. It was demonstrated that the regular solutions are correct only for the positive
moment gradients (0ⱕyⱕ1). But, when a beam is under a negative moment gradient
(⫺1ⱕyⱮ0), the regular solutions are unsafe and the bending moment resistance is
highly overestimated. The difference can reach 25% in the case of mono-symmetric
I section B and ⬎100% for the T section beam. This is contrary to the Eurocode
purposes which are to seek for economic and safe solutions.

4. Discussion and recommendations for the improved coefficients

The improved coefficients for checking the lateral buckling of unrestrained beams
with mono-symmetric I section are presented in Table 10. These coefficients permit
an analytical computation of the buckling moments according to the compact
relationship (27). An analytical relationship for C1 and C3 are formulated for the
case of a beam under moment gradient. They are given as a function of the moment
gradient coefficient y. These relationships are correct only for positive values of y
(0ⱕyⱕ1). When the beam is loaded under a negative moment gradient (⫺1ⱕ
yⱮ0), the relationship (27) is not correct and the recourse to the finite element
method or other numerical models are then recommended.
These coefficients, as those used in the informative annex F of the Eurocode 3
are computed from the linear stability model. They are formulated under the assump-
tion of small displacements and the effect of the pre-buckling deflections are neg-

Table 10
improved Ci coefficients corresponding to different load cases.

No. Load case Mb C1 C2 C3

1 qzL2 / 8 1.13 0.46 0.53

2 QzL /4 1.36 0.55 0.41

3 QzL /4 1.05 0.43 0.55

4 QzL /3 1.095 0.499 0.481

5 M0 C1 ⫽ 冑1 / a1a C3 ⫽ C1[(1 ⫹ ⌿) / 2]

2p2⫺3 2 2p3 ⫹ 6
a
a1 ⫽ 2
(y ⫹ 1) ⫹ ⌿ correct only for 0ⱕyⱕ1.
6p 6p2
 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90 89

lected. The load action passes through the shear centre with no initial torsion. They
are convenient for mono-symmetric I sections when the loads are applied along the
web line. They can also be used for checking the stability of laterally unrestrained
C section, when the loads act through the shear centre.

5. Conclusions

This paper has investigated the stability analysis of thin-walled elements with open
section. The developed model permits the study of bar buckling and lateral buckling
of unrestrained beams. The lateral buckling of mono-symmetric I beam is developed
and analytical solutions are formulated. The lateral buckling resistance of a beam is
a function of stress bending distribution, of load height parameter and of the degree
of monosymmetry of the section related to Wagner’s coefficient. Coefficients C1, C2
and C3 are computed for some selected load cases and compared to the usual coef-
ficients adopted in Eurocode 3. It is found that some coefficients are the same as
those adopted in Eurocode 3, but the C3 are very different for some load cases. In
numerical simulations, the lateral buckling loads or equivalent buckling moments are
computed from the solutions of the eigenvalue problem. Three-dimensional beams
including warping are used in the stability of bisymmetric I sections, but thin-walled
shell elements are preferred for the stability analysis of mono-symmetric I sections.
The comparison examples studied have shown that the improved solutions are in
agreement with shell element results and that most of the regular solutions overesti-
mate the real lateral buckling moments of mono-symmetric I sections and conse-
quently the resistance of the beams in lateral buckling behaviour.

References

[1] Vasov VZ. Thin walled elastic beams. Moscow; 1959. [French translation: Pièces longues en voiles
minces. Paris, Eyrolles; 1962].
[2] Bleick F. Buckling strength of metal structures. New York: McGraw-Hill, 1952.
[3] Timoshenko SP, Gere JM. Theory of elastic stability, 2nd ed. New York: McGraw-Hill, 1961.
[4] Trahair NS. Flexural-torsional buckling of structures. London: Chapman and Hall, 1993.
[5] Eurocode 3, Design of steel structures, Part 1.1: General rules for buildings. European Committee
for Standardisation, Draft Document ENV 1993-1-1, Brussels; 1992 [French translation: Calcul des
structures en acier et document d’application nationale. Partie 1.1 Règles générales et règles pour
le bâtiment. Paris: Eyrolles; 1997]
[6] Barsoum RS, Gallagher RH. Finite element analysis of torsional and torsional-flexural stability prob-
lems. International Journal for Numerical Methods in Engineering 1970;2:335–52.
[7] Bazant ZP, El Nimeiri M. Large-deflection spatial buckling of thin walled beams and frames. Journal
of Engineering Mechanics Division 1973;99(EM6):1259–81.
[8] Laudiero F, Zaccaria D. A consistent approach to linear stability of thin-walled beams of open
section. International Journal of Mechanical Sciences 1988;30(7):503–15.
[9] Kitipornchai S, Wang CM, Trahair NS. Buckling of monosymmetric I-beams under moment gradient.
Journal of Structural Engineering 1986;112(4):781–99.
[10] Conci A. Large displacement analysis of thin-walled beams with generic open section. International
Journal of Numerical methods in Engineering 1992;33:2109–27.
90 F. Mohri et al. / Journal of Constructional Steel Research 59 (2003) 63–90

[11] Wang CM, Kitipornchai S. On stability of monosymmetric cantilevers. Engineering Structures

1986;8:169–80.
[12] Djalaly H. Calcul de la résisitance ultime au déversement. Revue Construction Métallique, CTICM,
Paris 1974;1:58–77.
[13] Mohri F, Brouki A. Behaviour and instabilities of open thin-walled elements, Part 2: Instabilities. In:
Proceedings of 6th International Colloquium on Stability and Ductility of Steel Structures, Timisoara,
Romania, 9–11 September, 1999. p. 95–100.
[14] Mohri F, Brouki A, Roth JC. Lateral buckling of thin-walled beams under asymmetric loads. In:
Proceedings of the Conference Eurosteel’99, CVUT Praha, 26–29 May, 1999; p. 55–58.
[15] Mohri F, Brouki A, Roth JC. Déversement des poutres en I sous chargements asymétriques. Revue
Construction Métallique, CTICM, Paris 2000;2:41–52.
[16] Balac I, Kolekova Y. Stability of monosymmetric beams. In: Proceedings of 6th International Collo-
quium on Stability and Ductility of Steel Structures, Timisoara, Romania, 9–11 September 1999. p.
57–64
[17] Abaqus standard user’s manuel, v. 5.6. Pawtucket (RI): Hibbit Karlsson and Sorensen, Abaqus, 1999.
[18] Chages A. Principles of structural stability theory. New Jersey: Prentice-Hall, 1974.
[19] Murray NW. Introduction to the theory of thin-walled structures. Oxford: Clarendon Press, 1986.
[20] Brouki A. Déversement des poutres à parois minces et à sections ouvertes. Aspects numériques et
réglementaires. Doctorat dissertation, University of Metz; 1998
[21] Mohri F, Azrar L, Potier-Ferry M. Flexural-torsional post-buckling analysis of thin-walled elements
with open sections. Thin-Walled Structures 2001;39(11):907–38.