Journal of Constructional Steel Research 67 (2011) 151–163

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

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

Elastic lateral stability of I-shaped cellular steel beams

Amr M.I. Sweedan ∗
Department of Civil and Environmental Engineering, U.A.E University, P.O. Box 17555, Al-Ain, United Arab Emirates

article info abstract

Article history: In this paper, the lateral stability of cellular steel beams is numerically investigated. The study is carried
Received 8 August 2010 out using three-dimensional finite element modeling of simply supported I-shaped cellular steel beams
Accepted 25 August 2010 with a broad spectrum of cross-sectional dimensions, span lengths and web perforation configurations.
Stability analyses are carried out for beams subjected to equal end moments, mid-span concentrated loads
Keywords: and uniformly distributed loads. Finite element results reveal that, unlike the case of conventional beams
Cellular
with solid webs, the moment-gradient coefficient Cb is significantly influenced by the beam geometry and
Perforated
Steel beams
slenderness. In addition, the Cb coefficient of cellular beams depends on the web perforation configuration.
Lateral torsional buckling Moment-gradient coefficient values that fluctuate closely to those values recommended by design codes
Finite element are associated with pure elastic lateral torsional buckling (LTB) deformations. As the beam slenderness
decreases, the web distortion increases, leading to the lateral distortional buckling (LDB) mode, which is
associated with lower Cb values than code-recommended ones. Severe reduction in the Cb coefficient to
values less than 1.1 is noticed for shorter-span beams where the response is dominated by non-lateral
local buckling modes.
A simplified approach is developed to enable accurate prediction of a moment modification factor
κLB for cellular beams. The proposed κLB factor is provided by an empirical formula that is derived based
on the best fit of the finite element results related to lateral buckling (LTB and LDB) modes only. The
proposed approach allows for accurate and conservative evaluation of the critical moment associated
with the lateral torsional/distortional buckling of cellular beams. Several numerical examples are worked
out to illustrate the application of the proposed procedure.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Under general loading conditions, the bending moment distri-
bution may vary along the beam span. As a result, the compres-
sive stresses developed in the compression flange of the beam are
expected to have variable intensity along the span. Current de-
sign codes account for the influence of the moment gradient on
the beam moment-carrying capacity through a modification coeffi-
cient that is refereed to as the moment-gradient coefficient, Cb . This
coefficient relates the nominal moment Mn for a beam subjected
to a specific transverse load to the corresponding critical buckling
moment Mo-cr of the same beam under a uniform moment:
Mn = Cb Mo-cr .
Table 1 provides a comparison between various values of moment-
gradient coefficients adopted by three major international design
codes and standards for beams subjected to mid-span concen-
trated loads and uniformly distributed loads. Such codes are the
American steel construction manual AISC 2005 [1], the new Eu-
ropean standards EC3 [2] and the Australian design code SA [3].
∗ Tel.: +971 50 2338970; fax: +971 3 7623154.
E-mail addresses: amr.sweedan@uaeu.ac.ae, asweedan@hotmail.com.
0143-974X/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2010.08.009
As indicated by Eq. (1), all codes assign a Cb value of unity for
beams subjected to equal end moments (i.e., uniform moment dis-
tribution along the span of the beam). From a practical design
standpoint, the tabulated values are very close. It is, however,
interesting to note that the European code [2] provides slightly
higher estimates of the critical moment for beams subjected to
mid-span concentrated loads. Meanwhile, the American code [1]
recommends a slightly higher Cb value for beams subjected to uni-
formly distributed loads.
2. Literature review
(1) I-shaped steel sections are extensively used as main structural
elements (e.g., beams and columns) in various building structures.
It is advantageous to have web perforations, especially in beam el-
ements, to allow for the passage and installation of piping, duct-
works and electrical conduits without increasing the floor-to-floor
height (Fig. 1). Besides, web perforations result in a significant re-
duction in the amount of structural steel used. Two configurations
of web perforations are commonly used in engineering practice:
hexagonal perforations in castellated beams and circular perfora-
tions in cellular beams. Although the former configuration is more
common, the latter has recently become widely used in building
152 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
Fig. 1. Practicality of cellular beams in constructing various elements of main
structural systems.
construction. Precise manufacturing of cellular sections in an eco-
nomical way is rather a cumbersome task that involves cutting
the web of a solid rolled section and reassembling it by shifting
and welding the section components, as depicted by Fig. 2(a) and
(b), respectively. However, the current advancement in computer-
controlled cutting and welding technology allows for highly
precise and cost-effective production of cellular sections. A ma-
jor architectural advantage of cellular beams is that their appeal-
ing aesthetical appearance makes them essential elements in the
construction of exposed structures. Despite the fact that the aug-
mentation of the perforated section height enhances the in-plane
structural characteristics, the discontinuity in the beam cross-
section due to the presence of web openings may have a severe
penalty on the load-carrying capacity of cellular beams in the case
of failure by lateral–torsional buckling prior to the attainment of
their full capacity. Such impact depends mainly on the geometri-
cal configurations of web openings [4–7]. The widespread use of
castellated beams as structural elements in multistory buildings,
commercial and industrial buildings, stadiums and parking garage
structures has prompted several investigations into their structural
behavior. However, investigations related to the behavior of cellu-
lar beams are quite rare. For both perforation configurations, the
non-uniformity in cross-section properties due to the existence of
web openings increases the level of complexity of the behavior and
(a) Computerized cut of the original solid
web.
Fig. 2. Manufacturing of cellular beams (http://www.asdwestok.co.uk).
Table 1
Comparison between moment-gradient coefficient values recommended by
international design codes.
Case of loading American European Australian code
code (AISC standards (AS4100 [3])
2005 [1]) (EC3 [2])
Mid-span concentrated 1.35 1.365 1.32
load
Uniformly distributed 1.14 1.132 1.13
load
the associated potential failure modes. Failure of this special type
of beam is controlled by several modes, among which the stability-
related ones can be classified as follows [8,9].
1. Lateral–torsional buckling failure.
2. Web post-buckling due to excessive shear stresses.
3. Vertical buckling of web posts.
Experimental studies have been conducted to investigate the
behavior and stability of castellated beams by Kerdal and Nether-
cot [8] and Nethercot and Kerdal [10]. Further numerical investi-
gations have been carried out by Mohebkhah [11] on the inelastic
lateral torsional buckling of castellated beams. Such studies
provided quantitative data that revealed that the change in the
slenderness of castellated beams due to the castellation has a sig-
nificant influence on the moment-gradient coefficient used in the
design of beams subjected to flexure, which in turn affects the
strength and stability of such beams. Redwood and Demirdjian [12]
and Zaarour and Redwood [13] studied the web buckling of cas-
tellated beams both experimentally and theoretically. Recently,
experimental investigations of the distortional buckling of castel-
lated beams have been conducted by Zirakian and Showkati [14]
and Zirakian [15]. The experimental measurements were then in-
corporated into several extrapolation techniques to provide more
accurate predictions of the critical buckling loads. The previous
review reveals that a considerable volume of research has been di-
rected towards investigating the response and behavior of castel-
lated beams. In contrast, very limited studies have been directed to
investigating the behavior and response of cellular beams. Chung
et al. [9] conducted analytical and numerical analyses to assess the
load-carrying capacity of cellular beams related to the Vierendeel
failure mechanism. The coupled influence resulting from the com-
bined effect of the moment and shear on steel beams with web
openings of various shapes has been also investigated numerically
(b) Re-assembly by welding to form a cellular member.
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
by Chung et al. [16]. The study focused on flexural failure associated
with the Vierendeel mechanism along with shear failure. Results of
this finite element investigation have then been used to develop a
generalized moment–shear interaction curve for determining the
load-carrying capacities of steel beams with web openings of var-
ious shapes and sizes [17]. The influence of circular web openings
on the overall stability of axially loaded cellular sections has been
studied numerically by Sweedan et al. [18]. Their investigation has
led to the development of a simplified design approach and corre-
sponding design charts to allow for easy prediction of the buckling
capacity of cellular columns.
The current study is motivated by the lack of information per-
taining to elastic lateral torsional/distortional buckling of non-
composite cellular steel beams. The elastic stability of such beams
is typically a concern during the construction stage when lateral
bracing elements are not yet installed. As such, the unbraced length
of the cellular beam is long enough for elastic instability to occur.
Furthermore, the increase in the flexure capacity of cellular beams
makes it attractive for designers to use them with longer spans.
With such relative increase in the span length, the issue of elastic
stability becomes more pronounced. The study investigates the in-
fluence of the discontinuities in the cellular beam cross-section on
the moment-gradient coefficient and consequently on the bending
capacity of such beams. The investigation is conducted numerically
by developing a detailed three-dimensional finite element model
to simulate the nonlinear flexural–torsional behavior of cellular
beams. An extensive parametric study is carried out to assess the
impact of variation of cross-section dimensions, beam slenderness,
and web opening sizes and spacings on the lateral stability and the
corresponding load-carrying capacity of cellular steel beams.
3. Assumptions and finite element modeling
A three-dimensional (3D) finite element model is developed to
simulate the behavior of cellular steel beams having an I-shaped
cross-section. Modeling is conducted using the general-purpose
finite element software package ANSYS [19]. Geometrical details of
analyzed beams are simulated using the four-node shell element
(SHELL181). This element has six degrees of freedom at each
node, three translations and three rotations, which enable explicit
simulation of various buckling deformations. Besides, SHELL181 is
suitable for modeling thin to moderately thick shell structures.
3.1. Material model and meshing considerations
Numerical computations are conducted for cellular steel beams
that are assumed to be constructed of linear elastic material with
Young’s modulus E = 2.1 × 105 MPa and Poisson’s ratio ν =
0.3. Each I-shaped beam is characterized by its span L, flange
width bf , flange thickness tf , web height hw and web thickness tw .
Web perforations are circular in shape, with diameter dh , and are
uniformly spaced at distance s along the span of the beam. Flanges
are modeled with 12 elements across the width bf . Meanwhile,
the element size is reduced by 20% along the web height hw to
provide better simulation of the deformations associated with
lateral, distortional or local web buckling modes. The size of the
elements along the span of the beam is restricted not to exceed
twice the size of the element across the flange. Several mesh
configurations are attempted until the above-provided limitations
are set after providing convergence of the predicted buckling load
within reasonable execution time. A typical mesh configuration is
shown in Fig. 3 for a cellular beam having a span L = 8820 mm
with bf = 170 mm, tf = 17 mm, hw = 300 mm, tw = 10 mm,
dh = 180 mm and s = 315 mm. It should be noted that, although
the thickness of the elements does not appear in the figure, it is
used by ANSYS to generate the stiffness matrix of the modeled
beam.
153
Fig. 3. Typical finite element mesh for a cellular beam.
3.2. Boundary conditions and load application
The general geometry and coordinate system of a typical cellu-
lar beam is presented in Fig. 4(a). The beam geometry is defined
using a global Cartesian coordinate system with its origin located
at the mid-height of the cross-section at the right end of the beam.
As indicated in Fig. 4(a), the depth of the beam is directed along
the Y -axis while its longitudinal axis coincides with the Z -axis. The
current study considers beams with simply supported end condi-
tions imposed at both ends of analyzed beams. Vertical displace-
ments (Uy ) and out-of-plane displacements (Ux ) are restrained at
the central node of the cross-section at both ends, while one end
only is restrained against longitudinal displacements (Uz ). Twist-
ing rotations at beam ends are prevented by restraining the flanges’
tip points against out-of-plane displacements (Ux ), as shown in
Fig. 4(a). Meanwhile, all nodes, including those at the beam ends,
are left unrestrained against major-axis rotation (Rx ), minor-axis
rotation (Ry ) and warping displacements.
Beams with various geometrical parameters are analyzed un-
der three different cases of loading to allow for evaluating vari-
ous moment-gradient coefficients. These cases of loading include
pure end moments, mid-span concentrated loads and uniformly
distributed loads. Developed finite element models include bear-
ing stiffener plates in the vicinity of concentrated loading and at
support reactions to avoid web crippling resulting, associated with
localized web yielding. In order to avoid any undesirable localized
web deformations and stress concentration, the end moment load-
ing is simulated with couple forces in the form of uniform loading
applied at the top and bottom flanges of the beam end sections.
In this loading scheme, the top flange is subjected to compressive
longitudinal forces, while the bottom flange is subjected to tensile
forces opposing the compressive ones, as presented in Fig. 4(b). An
idealized application of transverse loading requires application of
such loads at the shear center of the section to avoid destabilizing
or stabilizing effects of loading applied above or below the shear
center, respectively. However, the presence of web perforations
does not allow for direct application of transverse loads at the shear
center. A typical solution to overcome such a problem is adopted by
applying half the load at the top flange and the other half at the bot-
tom flange, with both portions acting along the same direction, as
shown in Fig. 4(c) and (d) for uniformly distributed and mid-span
concentrated loads, respectively. As a result, the torques produced
by both load portions, when acting on the deformed geometry of
the beam, will be counteracting each other. This eliminates any sta-
bilizing/destabilizing effects on the beam and consequently leads
to an overall response that simulates that of loading applied at the
shear center of the beam.
154 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163

(a) Boundary conditions. (b) End moment (M). (c) Uniform load (W ). (d) Mid-span
load (P ).

Fig. 4. Boundary conditions and different applied loads on the beam.

4. Verification of the finite element model

Table 2
Comparison between finite element and theoretical buckling moments for plain-
To verify the performance of the 3D finite element model de- webbed beams.
scribed in the previous section, the model is employed to conduct (1) (2)
bf tf hw tw L Mcr-FE Mo-cr |1%|a
an eigenvalue analysis of various beams to identify their critical (mm) (mm) (mm) (mm) (m) (kN m) (kN m) (%)
moment values that correspond to the lateral torsional buckling
200 10 450 10 13.5 65.29 65.05 0.37
mode of failure. The verification phase starts with evaluating the 200 10 450 10 9.0 112.70 112.18 0.47
critical buckling moment of plain-webbed (i.e., without holes) sim- 160 10 450 10 13.5 41.94 41.82 0.27
ply supported beams having different geometrical dimensions and 160 10 450 10 9.0 69.10 68.83 0.40
subjected to uniform end moments. Selected beam sections are as- 120 10 450 10 13.5 24.72 24.68 0.17
120 10 450 10 9.0 39.01 38.90 0.29
sumed to have an average web height to thickness ratio (hw /tw ) of
80 10 450 10 13.5 12.48 12.46 0.11
45, while the flange width to thickness ratio (bf /tf ) varies between 80 10 450 10 9.0 19.06 19.03 0.18
extreme values of 8 to 20, as shown in Table 2. Each beam cross-  
a
|∆|% =  (2)−( 1) 
∗ 100%.
section is analyzed twice, considering a span-to-depth ratio (L/hw )

(1) 

of 20 and 30. The finite element predictions are then compared to

the analytical solution of the critical buckling moment Mo-cr given
by [20]

π C2 π 2
 
Mo-cr = EIy C1 1+ , (2)
L C 1 L2

where L is the unbraced span of the beam and Iy is the minor-axis

moment of inertia of the cross-section. In Eq. (2), C1 and C2 are the
torsional and warping rigidities, respectively, which are obtained
as follows:
C1 = GJ ,
C2 = ECW ,
where G is the shear modulus of elasticity, J is the torsional con-
stant of the cross-section, E is Young’s modulus of elasticity and
Cw is the warping torsional constant of the cross-section. The re-
sults of the quantitative comparison between analytical solutions
and finite element predictions are summarized in Table 2, indicat-
ing an excellent agreement, with an absolute maximum relative
error that does not exceed 0.5%.
As is alluded to in Section 2, no previous experimental or
numerical results related to lateral torsional buckling of cellular
beams are available in the literature. Mohebkhah [11] reported
numerical predictions of critical buckling moments of castellated
beams subjected to uniform moments, mid-span concentrated
loads and uniformly distributed loads. In order to provide further
validation of the performance of the finite element model devel-
oped in the current study for cellular beams, a modified version
of the model is produced for a CPE140 castellated beam section
Fig. 5. Castellation profile of the CPE140 beam for verification of the finite element
(3) model.
(4)
with span lengths and web castellation that match those reported
by Mohebkhah [11], as shown in Fig. 5. Particular cross-section di-
mensions of the castellated section are bf = 73 mm, tf = 6.9 mm,
hw = 196.2 mm and tw = 4.7 mm. Such a beam is analyzed us-
ing the developed model for three different span lengths, 2.94, 3.36
and 4.20 m, to ensure elastic buckling failure. Geometrical detail-
ing of the castellation configuration is shown in Fig. 5. A summary
of results obtained under various cases of loading, along with a
comparison to values reported by Mohebkhah [11], is presented in
Table 3. The tabulated values reveal a very close agreement be-
tween both sets of results, with an absolute maximum relative
error of 2.32%. Such agreement confirms the adequacy of the devel-
oped finite element model in predicting the buckling moments of
perforated web beams under various loading conditions. It is worth
mentioning that the mesh used in the current study is relatively
finer than the one used by Mohebkhah [11].
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163 155

Table 3 to allow for investigating the influence of change in hole spacing

Validation of the finite element model for beams with castellated webs. on the buckling behavior, excluding any impact from variation in
(1) (2)
Case of loading L (m) Mcr-FE (kN m) Mcr-FE (kN m) |1%|a (%) the number of web holes along the beam span.
(current study) (Mohebkhah [11]) The general-purpose finite element program ANSYS [19] is
2.94 18.68 18.32 1.93 utilized to solve the eigenvalue problem describing the buckling of
End moments 3.36 15.50 15.40 0.65 the modeled beams. The nonlinear behavior is taken into account
4.20 11.52 11.60 0.69
in the buckling problem formulation by considering the first-order
Mid-span 2.94 24.23 24.11 0.50 and second-order terms of the deformations encountered. This is
concentrated 3.36 20.87 20.66 1.01 reflected in the two components constituting the overall stiffness
load 4.20 15.51 15.62 0.71
matrix of the beam, as indicated below:
2.94 20.88 20.53 1.68
Uniform load 3.36 17.37 17.50 0.75 ([K ] + ωi [S ]) {φ}i = {0}, (5)
4.20 12.93 13.23 2.32
where [K ] is the conventional small-deformation stiffness matrix,
a
|∆|% = | (2)−(
(1)
1)
| ∗ 100%. [S] is the stress stiffness matrix, ωi is the ith eigenvalue and φi is the
ith eigenvector. The lowest eigenvalue obtained corresponds to the
Table 4 critical elastic load (or moment) while the associated eigenvector
Variation of span-to-web-height ratio for various hole spacings. represents the fundamental mode shape at buckling.
s/hw Number of web
perforations
6. Results and discussion
1.05 1.575 2.10

10.50 15.75 21.00 10 In this section, buckling moment results obtained from the
12.60 18.90 25.20 12 comprehensive parametric analysis are discussed. For cases of
14.70 22.05 29.40 14
both uniform and concentrated transverse loading, the buckling
16.80 25.20 33.60 16
L/hw -value 21.00 31.50 42.00 20
moment value is normalized with respect to the corresponding
25.20 37.80 50.40 24 critical moment that results from a uniform end moment case
29.40 44.10 58.80 28 of loading to obtain the Cb coefficient, as indicated by Eq. (1).
33.60 50.40 67.20 32 Variation of the Cb coefficient obtained is investigated by its
37.80 56.70 75.60 36
being plotted against a non-dimensional factor αLB that relates
the warping rigidity of the beam C2 to its torsional rigidity C1 in
5. Finite element analysis of cellular beams accordance with the following relation:

The developed 3D finite element model is utilized to investigate π C2
the influence of the beam geometry and perforation configuration αLB = , (6)
L C1
on its lateral stability. The study covers three main cases of loading:
pure end moments, mid-span concentrated load and uniformly where all parameters involved are as previously defined in Eqs.
distributed load. The cross-section dimensions of the analyzed (2)–(4). In view of Eq. (6), determination of the non-dimensional
beams are selected to cover, and extend beyond, the practical range factor αLB requires evaluating the torsional parameters J and Cw of
of geometric parameters of such beams. For possible application the I-shaped beam section. The values of the torsional constant J
of the results obtained to various rolled and built-up sections, all and the warping torsional constant Cw for plain-webbed I shaped
geometrical parameters are presented in non-dimensional form. sections (i.e., without web holes) can be approximated by [20,21]
The study considers sections with web dimensions defined by − bt 3
hw /tw values that vary between 30 and 80. The flange dimensions J = (7)
are assigned a bf /tf ratio that ranges from 10 to 20. Meanwhile, 3
the web perforations are characterized by hole diameter to web tf b3f h2o
height (dh /hw ) values that vary between 0.5 and 0.8. The spacing Cw = , (8)
between consecutive web holes as a ratio of the web height s/hw 24
varies between 1.05 and 2.1. These limits correspond to a wide where b represents the width of each element of the I-section,
range of hole spacing-to-diameter ratios s/dh that extends from 1.3 t is the thickness of the same element, bf is the width of the
to 4.2. A major factor that is considered in the current study is the flange, tf is the thickness of the flange and ho is the height between
influence of the span on the buckling capacity and the associated centers of both flange plates (i.e., ho = hw + tf ). For the special
mode shape. While the behavior of long-span beams is expected geometry of cellular beams considered in the current study, net
to be governed by the lateral torsional buckling mode, the buckling dimensions are arbitrarily adopted in the calculation of the cross-
capacity of beams with relatively shorter span may be reduced due section properties. These dimensions correspond to a net section
to the simultaneous occurrence of global and local buckling modes. located at the center of the web hole. As such, the Jnet parameter is
This coupled effect may be more pronounced in cellular beams due obtained by applying Eq. (7) to a net section that excludes a portion
to the existence of web perforations, as will be explained in more of the web having height equal to the diameter of the opening:
detail in the next section. In order to explore the influence of the
2 1
beam span on its buckling characteristics, each set of cross-section Jnet = (hw tw3 − dh tw3 ).
bf tf3 + (9)
dimensions with a specific perforation configuration is analyzed 3 3
nine times for various span-to-web-height ratios L/hw , as shown The same approach is followed when the cross-sectional area or
in Table 4. As presented in Table 4, for the basic case of s/hw inertia needs to be considered. Meanwhile, the warping torsional
equal to 1.05 the L/hw ratio varies between 10.50 and 37.8. This constant Cw is obtained following Eq. (8), which is almost
L/hw range corresponds to the practical range of span-to-depth unaffected by the presence of the web opening [22].
ratio of steel beams having medium to long spans. For higher The main advantage of presenting the variation of the Cb coeffi-
s/hw values, the L/hw ratio is increased accordingly to maintain cient with respect to the non-dimensional factor αLB is that results
the same number of web perforations for a specific beam cross- of the finite element analysis indicate that cellular beams with the
section with different s/hw values. This condition is implemented same αLB and having the same perforation configuration ((dh /hw )
156 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
Table 5
Variation of the moment-gradient coefficient Cb with non-dimensional factor αLB .
Beam propertiesa
bf (mm) tf (mm) hw (mm) tw (mm) bf /tf
140 14 350 7 10
170 17 425 8.5 10
200 20 500 10 10
150 15 250 5 10
120 8 250 5 15
220 10 450 5 22
a
dh /hw = 0.7 and s/hw = 1.05 are kept constant for all beams.
and (s/hw )) possess the same Cb coefficient under the same case
of loading. In other words, for two beams having different cross-
section dimensions, the Cb coefficient will be equal as long as both
beams have the same αLB value irrespective of the specific dimen-
sions of the web (hw , tw ) and the flange (bf , tf ). Table 5 sum-
marizes sample results of several analyses that are conducted to
confirm the variation of the moment-gradient coefficient Cb with
the suggested non-dimensional factor αLB . It should be noted that,
for all the cases reported in Table 5, the perforation configuration
is kept constant with dh /hw = 0.7 and s/hw = 1.05. The first three
rows of the table imply that beams with various cross-sections
have the same αLB factor if all of these sections have equal bf /tf ,
hw /tw , hw /bf , dh /hw and s/hw ratios. The tabulated data in these
three rows confirm that beams with different cross-section dimen-
sions can still have equal Cb coefficient as long as all beams have
the same αLB value. Furthermore, a comparison between the re-
sults provided in the first three rows and those presented in the last
three rows provides additional confirmation of the correlation be-
tween the Cb coefficient and the non-dimensional parameter αLB .
6.1. Variation of the moment-gradient coefficient Cb with cellular
beam dimensions
Figs. 6 and 7 present sample results to demonstrate the vari-
ation of the Cb coefficient against the non-dimensional factor αLB
for simply supported beams subjected to mid-span concentrated
loads and uniformly distributed loads, respectively. The results
presented pertain to the particular case of cellular beams with
bf /tf = 10 and s/hw = 1.05. Meanwhile, various hw /tw values
that vary between 30 and 80 are considered along with several
web perforation sizes defined by a dh /hw ratio that ranges from 0.5
to 0.8. In the figures provided, each curve includes several points
that correspond to a wide span range that extends from a relatively
short-span case defined by L/hw value of 10.50 (highest αLB value)
to a long-span case with L/hw equal to 37.8 (lowest αLB value). A
comparison between the results provided in Figs. 6 and 7 reveals
that, similar to the regular case of beams with solid webs, beams
subjected to mid-span concentrated loads can sustain higher mo-
ments than those supporting uniformly distributed loads. This is
attributed to the fact that in the former case of loading the high-
moment region is localized around the mid-span, whereas in the
latter case relatively high-moment values are affecting longer re-
gions of the beam’s span.
A general trend of variation in the Cb coefficient with respect
to αLB can be noticed under both cases of loading shown in Figs. 6
and 7. This trend shows that the Cb coefficient has its maximum
value for slender beams with longer spans that are characterized
by lower αLB values. A reduction in the Cb value is demonstrated for
beams having shorter spans (i.e., with higher αLB values). This trend
can be explained in view of the various buckling modes associated
with various Cb values along each curve. Careful investigation of
the buckling modes reveals that Cb values that fluctuate closely
to those values recommended by design codes are solely related
to pure elastic lateral torsional buckling (LTB) deformations, as
αLB Cb (Mid-span load) Cb (Uniform load)
hw /tw hw /bf L (m)
50 2.5 7.35 0.613 1.340 1.114
50 2.5 8.93 0.613 1.341 1.115
50 2.5 10.50 0.613 1.341 1.115
50 1.667 10.50 0.318 1.359 1.127
50 2.083 5.25 0.900 1.332 1.108
90 2.045 9.45 1.332 1.308 1.085
shown in Fig. 8(a). As the beam slenderness decreases (i.e., the αLB
value increases), the level of web distortion during lateral buckling
increases, leading to formation of the lateral distortional buckling
(LDB) mode, where local web distortion and lateral buckling
occur simultaneously, as shown in Fig. 8(b). Such an LDB mode
corresponds to a reduction in the moment-carrying capacity of
the beam, as indicated by the associated lower Cb values. As the
αLB value increases, a progressive reduction in the Cb coefficient
is noticed, where Cb values less than unity are obtained. Finite
element results indicate that this significant reduction in the Cb
values is associated with non-lateral buckling modes such as the
local web buckling mode depicted by Fig. 8(d). This response occurs
in short-span beams where shear deformations become dominant
and lateral deformations are difficult to be triggered. The plots
presented in Fig. 6 show that, for beams subjected to mid-span
concentrated loads, Cb values that are equal to or less than 1.1 are
associated with non-lateral buckling modes. Meanwhile, for beams
subjected to uniformly distributed loads, the formation of local
non-lateral buckling modes is observed at Cb values less than 1.0,
as presented in Fig. 7. The transition stage between lateral buckling
modes (Fig. 8(a) and (b)) and non-lateral local buckling modes
(Fig. 8(d)) is controlled by the interaction between these two types
of modes, as shown in Fig. 8(c). Along this transition stage, lateral
deformations gradually decrease with progressive introduction of
localized deformations until the overall buckling mode is governed
by local non-lateral deformations.
6.2. Influence of beam geometry on the moment-gradient coefficient
Cb
Figs. 6 and 7 demonstrate that, for a particular hole size, beams
with slender web plates (i.e., with higher hw /tw values) are more
prone to shear deformations and, therefore, are controlled by non-
lateral local buckling modes at lower αLB values than beams with
stockier web plates. It can be also observed from Figs. 6 and 7
that web plates with bigger hole size (higher dh /hw ) experience
higher shear deformations. Such behavior is mostly dominated by
local web buckling over a wider span range. Lateral buckling modes
affect only slender beams with significantly low αLB values.
Fig. 9(a) and (b) and Fig. 9(c) and (d) correspond to cellular
beams with bf /tf = 10 and dh /hw = 0.8 subjected to mid-span
concentrated loads and uniformly distributed loads, respectively.
These figures show that beams with widely spaced web holes
(i.e., with higher s/hw values) possess higher shear stiffness. As
such, smaller shear deformations are introduced in the web, and
the buckling response is controlled by lateral buckling (LTB or
LDB) modes. For mid-span concentrated load application, Fig. 9(a)
shows that with the increase in s/hw to a value of 1.575, almost all
the beams experience a lateral torsional buckling (LTB) mode with
the exception of one case with high αLB values and hw /tw = 80
that experiences web local buckling failure. With further increase
in s/hw to a value of 2.1, the response becomes fully controlled
by lateral torsional buckling (LTB) modes, as depicted by Fig. 9(b).
Beams that are subjected to uniformly distributed loads with
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
(a) dh /hw = 0.5.
(c) dh /hw = 0.7.
Fig. 6. Variation of the Cb coefficient with the αLB parameter for I-beams with bf /tf = 10 and s/hw = 1.05 subjected to mid-span concentrated load.
s/hw = 1.575 and hw /tw = 80 are also shown to undergo a
web local buckling mode. In addition, a few other beams with
various hw /tw values and high αLB are shown to undergo lateral
distortional buckling (LDB) modes, as presented in Fig. 9(c). Once
the holes become widely spaced (s/hw = 2.1), all beams fail by
lateral torsional buckling (LTB) modes as shown in Fig. 9(d).
The influence of the flange aspect ratio on the moment-gradient
coefficient Cb of cellular beams is shown in Fig. 10(a)–(c) for a
representative case of cellular beams having dh /hw = 0.6 with
various perforations spacing of s/hw = 1.05, 1.575 and 2.10,
respectively. In these figures, the variation of the moment-gradient
coefficient Cb is presented with respect to the non-dimensional
factor αLB for the uniform and concentrated loading cases. Fig. 10(a)
indicates a clear reduction in the Cb coefficient for cellular beams
with closely spaced web perforations (s/hw = 1.05) especially
with higher αLB values (shorter spans). This reduction results from
web distortions that become more pronounced in beams with
shorter spans. Fig. 10(a) reveals also that beams with flanges of
smaller aspect ratio (bf /tf = 10) experience more reduction
in the Cb coefficient. Meanwhile, for beams with flange plates of
higher aspect ratios (bf /tf = 15 and 20) the reduction in the
Cb coefficient takes place at relatively higher αLB values (i.e., for
beams with shorter spans). This observation may be attributed to
the fact that bigger flange plates provide more effective torsional
157
(b) dh /hw = 0.6.
(d) dh /hw = 0.8.
restraints for the perforated webs, which consequently minimizes
the encountered reduction in the Cb coefficient.
A comparison between the plots provided in Fig. 10(a) and (b)
indicates that as s/hw increases from 1.05 (Fig. 10(a)) to 1.575
(Fig. 10(b)), less reduction in the Cb coefficient takes place for
cellular beams with similar αLB values. The same observation can
be reached by comparing the results reported in Fig. 10(b) (s/hw =
1.575) and those presented in Fig. 10(c) (s/hw = 2.10). The slight
reduction in the Cb coefficient observed in cellular beams with
widely spaced web perforations is attributed to the fact that such
beams experience smaller web distortions and therefore provide
higher torsional resistance.
7. Prediction of the critical moment for lateral buckling of
cellular beams
The previous discussion indicates that the Cb coefficient of cel-
lular beams may have a lower value than those recommended
by current design codes for lateral torsional buckling (LTB) of
solid beams. In particular, this reduction in Cb value occurs in
cases when the stability is controlled by lateral distortional buck-
ling (LDB). In such cases, using conventional Cb values may re-
sult in an unconservative design of cellular beams. Furthermore,
158 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163

(a) dh /hw = 0.5. (b) dh /hw = 0.6.

(c) dh /hw = 0.7. (d) dh /hw = 0.8.

Fig. 7. Variation of the Cb coefficient with the αLB parameter for I-beams with bf /tf = 10 and s/hw = 1.05 subjected to uniformly distributed load.

the Cb values that are currently available in design codes de- the considered hole size range (dh /hw = 0.5–0.8) and is, therefore,
pend mainly on the boundary conditions and applied load con- not considered as a variable in the suggested formula. Meanwhile,
figuration only. Finite element results reveal that the critical the κLB value depends on the loading configuration. For the range
buckling moment of cellular beams is also influenced by the beam of geometrical parameters and types of loading considered in the
geometry, slenderness and web perforation configuration. This current study, the following formula is suggested for evaluating the
section reports on a simplified procedure that allows for more ac- moment modification factor κLB :
curate prediction of the critical buckling moment of cellular beams
−0.03
  
under various cases of loading. It should be noted that the pro- hw
κLB = + [β − 0.002(λ − 10)] ψ,

(10)
posed procedure applies only to lateral buckling (LTB and LDB) λψ tw
modes of cellular steel beams. Thus, failures resulting from lo- where
cal web buckling are not considered in the development of this
procedure. The suggested procedure depends on introducing a bf
λ= (11)
moment modification factor κLB that allows for evaluating the tf
critical lateral buckling moment of cellular beams subjected to
1.10 for end moments

uniform moments, mid-span concentrated loads and uniformly
distributed loads. The suggested moment modification factor is β = 1.45 for mid-span concentrated loads (12)
provided by an empirical formula that is determined based on best 1.21 for uniform loads
fit of the finite element results related to lateral buckling (LTB and and
LDB) modes only. Careful examination of the correlation between 2
the suggested κLB factor and various parameters considered in the
  
s s
ψ = 0.89 + 0.15 − 0.028 . (13)
current study indicates that it varies with the two main parame- hw hw
ters defining the beam cross-section geometry (hw /tw and bf /tf ).
Once the moment modification factor κLB is identified, the critical
In addition, a trend of variation is shown between the moment
buckling moment of cellular beams can be evaluated as
modification factor κLB and the web hole spacing (s/hw ). The pro-
posed κLB factor is also shown to have insignificant dependence on Mcr = κLB Mo−cr , (14)
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163 159

(a) Pure LTB mode. (b) LDB mode.

(c) Interaction of LDB and web buckling modes. (d) Local web buckling mode.

Fig. 8. Various lateral and non-lateral buckling modes of cellular beams: (1) indicates a section at mid-span and (2) indicates a section at support.

in which Mo−cr (net ) is obtained in accordance with Eq. (2) for a net Table 6
section located at the center of the web hole as previously ex- Geometrical properties of cellular beams of illustrative examples.

plained in Section 6. It is worth mentioning that for the particular Beam bf tf hw tw L/hw dh /hw s/hw
case of uniform moments, and unlike the conventional Cb coeffi- designation (mm) (mm) (mm) (mm)
cient for solid beams, the value of the moment modification fac- CB1 170 17 300 10 12.6 0.7 1.05
tor κLB is not equal to 1. This is due, in part, to the fact that the CB2 450 22.5 640 16 29.4 0.6 2.10
value of Mo−cr (net ) used is calculated for a net section which does
not provide the actual critical moment value of the cellular beam.
of the critical moment of simply supported cellular beams sub-
In addition, the κLB factor reflects the influence of the reduced web
jected to end moments, mid-span concentrated loads or uniformly
stiffness on the distortional buckling resistance of cellular beams
through its dependence on the web slenderness hw /tw , flange plate distributed loads. The cellular steel beams are assumed to have
dimensions bf /tf , spacing between web holes s/hw and web hole Young’s modulus E = 2.1 × 105 MPa and Poisson’s ratio ν = 0.3.
size, which is reflected in the Iy-net , Jnet and Mo-cr values. Two different beams are considered, namely CB1 and CB2, to rep-
Fig. 11(a) and (b) show the variation of the moment modifica- resent various cases of cross-section dimensions, span-to-depth
tion factor κLB with hw /tw for the two extreme cases of cellular ratio and web perforation size and spacing. The cross-section di-
beams having bf /tf = 10 and s/hw = 1.05 and 2.10, respectively. mensions and perforation configurations of the beams considered
The graphs confirm the validity of the suggested formula as the are summarized in Table 6. A detailed description of the procedures
maximum relative percentage error does not exceed 2%. The mo- involved in evaluating the critical moment of cellular beams CB1
ment obtained based on Eq. (14) represents the nominal value of and CB2 is presented in the following subsections.
the elastic critical buckling moment related only to lateral buck-
ling (LTB and LDB) modes of cellular steel beams. The correspond- 8.1. Beam CB1 subjected to a uniform bending moment distribution
ing elastic design moment resistance can be readily obtained by
multiplying the outcome of Eq. (14) by the appropriate resistance bf
1. Based on Eq. (11): λ = = 170
= 10.
reduction factor specified by design codes according to the load tf 17
combination under consideration by the designer engineer. 2. Substituting for s/hw = 1.05 in Eq. (13): ψ = 0.89 +
0.15(1.05) − 0.028(1.05)2 = 1.017.
8. Illustrative numerical examples 3. A value of β = 1.10 is extracted from Eq. (12) for a cellular
beam subjected to a uniform bending moment distribution.
Numerical examples are presented herein to explain the appli- 4. Given that hw = 300 and tw = 10, the κLB factor is evaluated
cation of the procedure proposed in Section 7 for the prediction using Eq. (10) as
160 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163

(a) s/hw = 1.575 mid-span concentrated load. (b) s/hw = 2.10 mid-span concentrated load.

(c) s/hw = 1.575 uniformly distributed load. (d) s/hw = 2.10 uniformly distributed load.

Fig. 9. Variation of the Cb coefficient with αLB parameter for I-beams with bf /tf = 10 and dh /hw = 0.8 under various loading conditions.

−0.03
  
300 6. Then, the torsional and warping rigidities (C1 and C2 ) are calcu-
κLB =
10 ∗ 1.017 10 lated using Eqs. (3) and (4), respectively, as
√
+ [1.10 − 0.002(10 − 10)] 1.017 = 1.021. E 2.1 ∗ 105
C1 = GJnet = Jnet = ∗ 586,806.67
2(1 + ν) 2(1 + 0.3)
5. The net section properties are evaluated in accordance with
= 47,395,923,076.92 N mm2
Section 6 as follows:
C2 = ECW = 2.1 ∗ 105 ∗ 349,705,907,042
1
Iy(net ) = [(hw − dh )tw + 3
2tf b3f ] = 73,438,240,478,750,000 N mm4 .
12
1 7. The critical moment of the net section Mo-cr (net ) is readily eval-
= [(300 − 210)103 + 2 ∗ 17 ∗ 1703 ] uated using Eq. (2) as
12
= 13,927,666.67 mm4 .

π C2 π 2
 
Mo-cr = EIy(net ) C1 1+ = 445.3 kN m.
Using Eq. (9), L C 1 L2
2 1 2 8. Finally, Eq. (14) is employed to estimate the critical moment of
Jnet = bf tf3 + (hw tw3 − dh tw3 ) = (170 ∗ 173 ) the cellular section as
3 3 3
1 Mcr = κLB Mo−cr (net ) = 1.021 × 445.3 = 454.46 kN m.
+ (300 − 210)10 = 586,806.67 mm4 .
3
3
Using Eq. (8), 8.2. Beam CB1 subjected to a mid-span concentrated load
tf b3f h2o 1
Cw = = (17 ∗ 1703 ) ∗ (300 + 17)2 The previous steps are repeated after updating the β factor in
24 24 step (3) to be = 1.45, which is used in step (4) to produce a κLB
= 349,705,907,042 mm6 . value of 1.373.
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163 161

(a) s/hw = 1.05. (b) s/hw = 1.575.

(c) s/hw = 2.10.

Fig. 10. Variation of the Cb coefficient with αLB parameter for I-beams with different flange width-to-thickness ratio bf /tf and dh /hw = 0.6 under various loading conditions.

(a) s/hw = 1.05. (b) s/hw = 2.10.

Fig. 11. Numerical and predicted κLB factor for I-beams with bf /tf = 10 under various loading conditions.

Then, step (8) is applied to obtain the critical moment of the Then, step (8) is applied to obtain the critical moment of the
cellular section under a mid-span concentrated load as Mcr = κLB cellular section under a uniformly distributed load as Mcr = κLB
Mo−cr (net ) = 1.373 × 445.3 = 611.60 kN m. Mo−cr (net ) = 1.131 × 445.3 = 503.84 kN m.

8.3. Beam CB1 subjected to a uniformly distributed load 8.4. Beam CB2 subjected to various cases of loading

The procedures followed in Section 8.1 are followed after up- The general procedure outlined in Section 8.1 is followed for
dating the β factor in step (3) to be = 1.21, which is used in step cellular beam CB2 for which λ = 20 and s/hw = 2.1 and for
(4) to produce a κLB value of 1.131. ψ = 1.082. For these beam cross-section dimensions, the net
162 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
Table 7
Comparison between the proposed model predictions and finite element results.
(1) (2)
Beam Case of Mcr (kN m) Mcr
designation loading (proposed (kN m)
method) (FEM)
End moments 454.46 457.01
CB1 Concentrated 611.60 612.07
load
Uniform load 503.84 504.22
End moments 1093.21 1086.44
CB2 Concentrated 1465.90 1476.21
load
Uniform load 1210.34 1226.76
a
|∆|% = | (2)−( 1)
| ∗ 100%.
(2)
section properties are evaluated as Iy(net ) = 341,806,131.33 mm4 ,
J(net ) = 3766,712.83 mm4 and CW = 37,495,623,779,297 mm6 .
Table 7 summarizes the critical moment values assessed using
the proposed procedure for the considered cellular beams CB1 and
CB2. Also presented in the table are the critical moment values
evaluated using finite element modeling of both cellular beams
under the three different cases of loading. The tabulated values
confirm the adequacy and accuracy of the proposed procedure;
the maximum absolute relative error is within the 2% range, as
previously indicated in Section 7.
9. Summary and conclusions
The current study investigates the elastic lateral buckling
of doubly symmetric I-shaped simply supported cellular beams
under various loading conditions. The investigation is conducted
numerically using three-dimensional finite element modeling to
identify the critical buckling moment and the corresponding
buckling mode for a huge number of cellular beams that covers the
potential range of practical geometric dimensions and perforation
configurations of I-shaped cellular beams. The model developed
consists of the type and number of elements, and it allows for
capturing different possible deformation patterns including in-
plane flexural, out-of-plane flexural, web distortional and localized
shear deformations. A comprehensive finite element analysis is
carried out to identify the critical buckling moment of this special
type of beam along with the corresponding fundamental buckling
mode. The results of such analyses are utilized to assess the
variation of the moment-gradient coefficient Cb with the cellular
beam geometry and perforation configurations under various cases
of loading. Variation of Cb is investigated relative to a non-
dimensional factor αLB that relates the warping rigidity to the
torsional rigidity of cellular beams. The main conclusions that may
be extracted from the study are as follows.
• Cellular beams subjected to mid-span concentrated loads pro-
vide higher moment-carrying capacity than those supporting
uniformly distributed loads. This behavior is similar to that of
plain-webbed conventional beams.
• In general, the reduction in moment-gradient coefficient (Cb )
is more pronounced in shorter-span beams where excessive
shear stresses are developed leading to local web distortions
that occur simultaneously with lateral buckling.
• Quantitative assessment of the moment-gradient coefficient
(Cb ) shows that Cb values that are slightly higher or lower
than recommended code values for plain-webbed beams are
associated with the pure elastic lateral torsional buckling
(LTB) mode of long-span beams. Buckling of intermediate-
span beams is controlled by the lateral distortional buckling
(LDB) mode, where web distortion is coupled with lateral
deformations. Meanwhile, buckling of short-span beams is
dominated by the high level of web distortion associated with
the high shear stresses induced in the web plates. In this case
no lateral buckling occurs and significant reduction in the Cb
|1%|a (%) values takes place. This behavior is characterized by Cb values
less than 1.1 for beams subjected to mid-span concentrated
loads. Meanwhile, for beams subjected to uniformly distributed
0.56 loads, this behavior occurs at Cb values less than 1.0.
0.08 • Cellular beams with slender web plates (i.e., having higher
0.07
hw /tw values) and cellular beams with big hole size (dh /hw ) are
more prone to shear deformations and, therefore, are mostly
0.62
governed by local web buckling. Such beams are less likely to
0.70
experience lateral buckling (LTB or LDB) modes unless they
1.34 have significantly long spans.
• Widely spaced web hole (high s/hw values) configurations lead
to higher shear stiffness and consequently no or small web
distortions. In this case, the beam response is controlled by
lateral buckling (LTB or LDB) modes.
• Cellular beams with relatively small flange plates experience
more reduction in the moment-gradient coefficient (Cb ) than
those with bigger flange plates. Bigger flange plates provide
more effective torsional restraints of the perforated webs
which, consequently, minimizes the encountered reduction in
the Cb coefficient.
The study is extended to provide a moment modification factor κLB
that allows for easy evaluation of the critical lateral buckling mo-
ment of cellular beams subjected to uniform moments, mid-span
concentrated loads and uniformly distributed loads. An empirical
formula is provided for the evaluation of the modification factor
κLB for various geometrical and loading conditions. The suggested
formula is determined based on best fit of the finite element results
related to lateral buckling (LTB and LDB) modes only. The formula
is proven to provide a powerful tool for obtaining accurate predic-
tions of the κLB factor with a maximum relative percentage error
in the order of 2%. Finally, a detailed application of the proposed
approach has been described in detail through a set of illustrative
numerical examples.
Acknowledgement
The work presented in this paper was partially supported by
the Research Affairs at the United Arab Emirates University under
contract no. 02-01-7-11/09.
References
[1] American Institute of Steel Construction (AISC). ANSI/AISC 360-05 specifica-
tion for structural buildings. 13th ed. Chicago (USA); 2005.
[2] European Committee for Standardization. EN 1993-1-1 Eurocode 3 (EC3).
Design of steel structures — part 1-1: general rules and rules for buildings;
2005.
[3] Standards Australia (SA). AS 4100 Steel Structures. Sydney (Australia); 1998.
[4] Lawson R.M.. Design for openings in the webs of composite beams. CIRIA/Steel
Construction Institute, CIRIA Special Publication and SCI Publication 068;
1987.
[5] Darwin D. Steel and composite beams with web openings. In: Steel design
guide series no. 2. Chicago (IL, USA): American Institute of Steel Construction
AISC; 1990.
[6] Redwood RG, Cho SH. Design of steel and composite beams with web openings.
Journal of Constructional Steel Research 1993;25:23–41.
[7] Ochlers DJ, Bradford MA. Composite steel and concrete structural members:
fundamental behaviour. Pergamon; 1995.
[8] Kerdal D, Nethercot DA. Failure modes for castellated beams. Journal of
Constructional Steel Research 1984;4(4):295–315.
[9] Chung KF, Liu TCH, Ko ACH. Investigation on Vierendeel mechanism in steel
beams with circular web openings. Journal of Constructional Steel Research
2000;57:467–90.
[10] Nethercot DA, Kerdal D. Lateral–torsional buckling of castellated beams.
Structural Engineering Journal, London 1982;60B(3):53–61.
[11] Mohebkhah A. The moment-gradient factor in lateral–torsional buckling on
inelastic castellated beams. Journal of Constructional Steel Research 2004;60:
1481–94.
[12] Redwood R, Demirdjian S. Castellated beam web buckling in shear. Journal of
Structural Engineering, ASCE 1998;124(10):1202–7.
 A.M.I. Sweedan / Journal of Constructional Steel Research 67 (2011) 151–163
[13] Zaarour W, Redwood R. Web buckling in thin webbed castellated beams.
Journal of Structural Engineering, ASCE 1996;122(8):860–6.
[14] Zirakian T, Showkati H. Distortional buckling of castellated beams. Journal of
Constructional Steel Research 2006;62:863–71.
[15] Zirakian T. Lateral–distortional buckling of I-beams and the extrapolation
techniques. Journal of Constructional Steel Research 2008;64:1–11.
[16] Liu TCH, Chung KF. Steel beams with large web openings of various shapes and
sizes: finite element investigation. Journal of Constructional Steel Research
2003;59:1159–76.
[17] Chung KF, Liu TCH, Ko ACH. Steel beams with large web openings of
various shapes and sizes: an empirical design method using a generalised
moment–shear interaction curve. Journal of Constructional Steel Research
2003;59:1177–200.
163
[18] Sweedan AMI, El-Sawy KM, Martini MI. Identification of the buckling capacity
of axially loaded cellular columns. Journal of Thin-Walled Structures 2009;
47(4):442–54.
[19] ANSYS User’s Manual Release 12 0 1. Canonsburg (Pennsylvania); 2009.
[20] Timoshenko SP, Gere JM. Theory of elastic stability. 2nd ed. New York (NY):
McGraw-Hill Book Company; 1961.
[21] Salmon CG, Johnson JE, Malhas FA. Steel structures: design and behavior,
emphasizing load and resistance factor design. 5th ed. Upper Saddle River
(New Jersey): Pearson Prentice Hall Publishers; 2009.
[22] Kohnehpooshi O, Showkati H. Numerical modeling and structural behavior of
elastic castellated section. European Journal of Scientific Research 2009;31(2):
306–18.