You are on page 1of 10

ARTICLE IN PRESS

Thin-Walled Structures 46 (2008) 466–475


www.elsevier.com/locate/tws

Elastic distortional buckling of doubly symmetric I-shaped


flexural members with slender webs
Tadeh Zirakian
Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697-2175, USA
Received 14 September 2007; received in revised form 30 October 2007; accepted 6 November 2007
Available online 20 February 2008

Abstract

Intermediate length I-section members, particularly those with slender webs and stocky flanges, have been shown to buckle in a
distortional mode, in which the web distorts and the flanges displace sideways with less twist than for the lateral–torsional buckling
mode. This, in turn, reduces the torsional rigidity of the member, and hence results in a lower buckling strength. The web distortional
flexibility is not addressed explicitly in standards for flexural design of steel I–section members. However, in the 2005 AISC specification
it has been tried to account for the influence of web distortional flexibility on the lateral–torsional buckling resistance for slender-web
I-section members through implicit use of J ¼ 0 in the lateral–torsional buckling equations. This paper evaluates the amount of
effectiveness of this consideration within the elastic limit. By comparing the AISC code predictions with the accurate finite strip analysis
(FSA) distortional buckling solutions as well as the theoretical predictions of two elastic distortional buckling formulae developed by
other researchers, it is found that the assumption of J ¼ 0 for slender-web members results in overconservative estimate of the buckling
strength in some cases which may cause economic losses.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Stability; Buckling; I-beams; Distortion; Elasticity; Structural design

1. Introduction Finite strip analysis (FSA) of an I-section beam in


bending (Fig. 1) reveals the three aforementioned funda-
Steel I-sections may be subject to one of the three generic mental buckling modes. As it is seen, the wavelength of DB
types of buckling; namely local, distortional, or lateral– is intermediate between that of LB and long half-
torsional. Local buckling (LB) is characterized by localized wavelength LTB which places it firmly in the practical
distortions of the cross-section over a short wavelength in range of member lengths.
the absence of lateral translation. In lateral–torsional Although many studies have been performed separately
buckling (LTB), on the other hand, the cross-sections of on local and LTB, and codes of practice for the design
the member translate and twist as rigid bodies. Distortional of structural steelwork contain the relevant clauses, DB is
buckling (DB) is a buckling mode in which lateral less well known and has long been ignored in design
deflection and twists are combined with changes in the specifications. Work in the last 25 years or so has added
cross-sectional shape. This arises generally from web much to our understanding of this type of buckling
distortion because in most I-sections, flanges are compara- and after many years of relative neglect, DB has recently
tively stocky and any flange distortions are small. Web been the subject of a good deal of research and this work
distortion allows the flanges to deflect laterally with has now reached a point where design procedures suitable
different angles of twist, reduces the effective torsional for inclusion in codes of practice are beginning to emerge.
resistance of the member and consequently reduces the A conspicuous example of this is the recent develop-
resistance to buckling [1]. ments such as those of the AISI Standard for cold-formed
steel design [2]. The 2005 American Institute of Steel
Tel.: +1 818 522 5997; fax: +1 818 500 8042. Construction’s Specification for Structural Steel Buildings
E-mail address: zirakian@uci.edu [3] which provides an integrated treatment of allowable

0263-8231/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2007.11.001
ARTICLE IN PRESS
T. Zirakian / Thin-Walled Structures 46 (2008) 466–475 467

Nomenclature Specification

aw the ratio of two times the web area in M nBradford nominal elastic distortional buckling strength
compression due to application of major axis predicted by Bradford’s equation
bending moment alone to the area of the M nFSA nominal elastic distortional buckling strength
compression flange components predicted by the finite strip method (CUFSM)
A cross-sectional area M nLTB nominal elastic lateral–torsional buckling
Aw web area strength
bf flange width M nWangetal nominal elastic distortional buckling strength
bf t ; bf b widths of top and bottom flanges predicted by Wang et al.’s equation
Cb moment-gradient factor for lateral–torsional My yield moment about the axis of bending
buckling rt radius of gyration of the flange components in
d full nominal depth of the section flexural compression plus one-third of the web
E modulus of elasticity of steel, 200,000 MPa area in compression due to application of major
f factor given by Eq. (14) axis bending moment alone
Fcr elastic critical stress Rpg bending strength reduction factor
Fcrd elastic distortional buckling stress Sx elastic section modulus about the axis of
F crðJ¼0Þ elastic lateral–torsional buckling stress deter- bending
mined by Eq. (5), which is the same as Eq. (3) tf flange thickness
with J taken equal to zero tf t ; tf b thicknesses of top and bottom flanges
Fy specified minimum yield stress of the type of tw web thickness
steel being used a web to cross-sectional area ratio
G shear modulus of elasticity of steel g non-dimensional buckling moment with web
h clear distance between flanges when welds are distortion
used ḡo non-dimensional buckling moment given by
hc twice the distance from the centroid to the Eq. (15)
inside face of the compression flange lpcompact limiting slenderness parameter for compact
ho distance between flange centroids flange
Iy moment of inertia of the cross-section about the lrnoncompact limiting slenderness parameter for noncom-
plane of the web pact web
J St. Venant torsional constant G flange thickness parameter
K̄ beam parameter given by Eq. (16) n Poisson’s ratio
Lb laterally unbraced length x beam slenderness parameter
Lr limiting laterally unbraced length for the limit F flange widths to thicknesses parameter
state of inelastic lateral–torsional buckling C web slenderness parameter
Mcr elastic critical buckling moment O flange width parameter
M nAISC nominal elastic distortional buckling strength
predicted by the AISC

stress design (ASD) and load and resistance factor design where
(LRFD) methods, also contains specific rules for control- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ling the errors associated with neglecting web distortional  2ffi
C b p2 E J Lb
flexibility in their I-section member flexural resistance F cr ¼ 1 þ 0:078 . (3)
ðLb =rt Þ2 S x ho r t
equations.
Based on the AISC specification [3] provisions of
Chapter F, for doubly symmetric I-section members when However, in Section F5 of the specification which applies
the web satisfies the noncompact limit: for doubly symmetric slender-web I-shaped members when
Lb4Lr, the nominal elastic LTB moment is calculated by
sffiffiffiffiffiffi
E M nAISC ¼ Rpg F crðJ¼0Þ Sx , (4)
lrnoncompact ¼ h=tw p5:70 , (1)
Fy
where
the nominal elastic LTB moment is calculated by
C b p2 E
F crðJ¼0Þ ¼ . (5)
M nLTB ¼ F cr S x , (2) ðLb =rt Þ2
ARTICLE IN PRESS
468 T. Zirakian / Thin-Walled Structures 46 (2008) 466–475

buckling curve Rpg in the 2005 AISC specification. Conversely, if the web
14 is sufficiently thick, or if longitudinal stiffeners are pro-
12 vided, web bend buckling will not occur before yielding [5].
10 This factor is based upon fundamental studies by Basler [6].
Anyway, in calculating the elastic LTB moment of slender-
Mcr /My

8
web I-section members (Eq. (4)), the bending strength
6
reduction factor is also applied which is directly tied to the
4 local bend buckling of the web. Nonetheless, Rpg may
2 implicitly account for web distortion effects in cross-
0 sections with large web slenderness, i.e. hc/tw. This issue is
10 100 1000 10000 also investigated in this paper in order to find the
Buckling length (mm) effectiveness of this factor in comparison with that of
consideration of J ¼ 0.
buckled shapes for increasing half-wavelength
White and Jung [7] reported a study on the effect of web
distortion on the buckling strength of steel I-section
members. In their paper, they evaluated the implications
of the use of Eq. (3), or the implied conservative use of
Eq. (5), relative to rigorous elastic DB solutions for doubly
and singly symmetric I-section members with compact
or noncompact webs, subjected to major-axis bending
150 mm (LB) 1750 mm (DB) 8000 mm (LTB) moment. They also evaluated the effectiveness of the
AASHTO [8] and AISC [3] rules for controlling the errors
Fig. 1. Finite strip analysis of an I-beam in bending (h0 ¼ 300 mm,
bf ¼ 90 mm, tw ¼ 2.5 mm, tf ¼ 10 mm). associated with neglecting web distortional flexibility in
their I-section member flexural resistance equations.
Finally, they concluded that for many compact- or non-
Rpg is the bending strength reduction factor: compact-web I-sections, the use of J ¼ 0 in the LTB equa-
sffiffiffiffiffiffi! tions results in excessive conservatism, and for slender-web
aw hc E I-section members, the use of F crðJ¼0Þ is prudent to avoid
Rpg ¼ 1   5:7 p1:0, (6)
1200 þ 300aw tw Fy the need for explicit consideration of web distortion effects.
This paper focuses on DB of doubly symmetric I-shaped
aw and rt are defined as members with slender webs in the elastic range of structural
response, and further evaluates the amount of effectiveness
hc tw of the 2005 AISC [3] rules by comparing the code
aw ¼ p10, (7) predictions with FSA DB solutions as well as the
bf t f
theoretical predictions of two elastic DB formulae.

bf 2. Finite strip analysis


rt ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , (8)
12ððho =dÞ þ ð1=6Þaw ðh2 =ho dÞÞ
In this study, finite strip DB solutions are developed in
order to evaluate the accuracy of the code predictions.
and These solutions are conducted using the FSA software,
sffiffiffiffiffiffiffiffiffiffiffiffi CUFSM [9]. CUFSM is an open source stability analysis
E program which was originally written by Schafer [10] to
Lr ¼ prt . (9)
0:7F y support research on the behavior and design of cold-
formed steel members. The software employs the semi-
Those sections that violate Eq. (1) are referred to as analytical finite strip method to provide solutions for the
slender-web I-sections and for these types of sections, the cross-sectional stability of thin-walled members. Moreover,
AISC specification [3] bases the LTB resistance on Eq. (3), CUFSM allows all elastic buckling modes of a structure to
but the St. Venant torsional constant J is taken equal to be quantified and examined that this, in turn, results in
zero (Eq. (5)). In fact, the implicit use of J ¼ 0 in Section better understanding of elastic buckling behavior of the
F5 is intended to account for the influence of web thin-walled members. In particular, it should be noted that
distortional flexibility on the LTB resistance for slender- the finite strip method is generally a useful tool for
web I-section members [4]. performing parametric studies due to its simple require-
In addition, if the web is slender, it may undergo bend ments on input and the speed of the solution.
buckling, leading to a decrease of web contribution to the The applicability of CUFSM for studying the DB of
moment resistance. This loss of effective strength is I-section beams was studied and reported by the author in
introduced through the bending strength reduction factor another paper [11], in which the convergence and accuracy
ARTICLE IN PRESS
T. Zirakian / Thin-Walled Structures 46 (2008) 466–475 469

of the FSM solutions were verified by calibrating against bf b =ho ¼ 0:5 and the following range of proportions:
the results of other published studies in this respect and
also a discussion was made on the FSA of I-section beams. tf bf b bf b
1p p4; 1p p4; and 10p p20. (11)
Finally, it was demonstrated that the software can be tw bf t tf
successfully used for steel I-section members.
Based on the analyses of the previous aforementioned Using Bradford’s equation, the ratio of Fcrd/Fcr obtained
study, 40 elements were considered in this study for the for an I-beam in uniform bending can be used as an
analysis of I-beams so that high accuracy would be insured estimate for Fcrd/Fcr of the same beam subjected to a
in the analysis. The material properties adopted for each moment gradient. Fcr can be calculated based on estab-
beam are E ¼ 200,000 MPa, v ¼ 0.3, G ¼ 0.385E, and lished elastic LTB equations (Eq. (3) herein), including the
Fy ¼ 345 MPa that varies in one case. All of the solutions moment gradient parameter Cb, and then to apply an
are of simply-supported beams subjected to uniform adjustment such as Eq. (10) to determine the corresponding
bending moment and no transverse web stiffeners are DB strength.
included in the finite strip models; that is, the web is The second buckling formula which is used in this
modeled as unstiffened. Furthermore, all of the members research, was proposed by Wang et al. [13] for predicting
considered in this study have compact flanges and slender the moment distortion factor for monosymmetric beam-
webs, and the DB strengths are all smaller than Fy. Lastly, columns having equal flange thicknesses and under
it must be added that the web, flange, or coupled LB did combined uniform bending and axial force. In fact, the
not occur in any case. application of regression analysis to parametric studies
Ultimately, using FSA the elastic buckling response of an yielded a simple empirical formula for the distortion factor,
I-section beam in bending is shown in Fig. 1(a), where the g=ḡo as
buckling moment Mcr is plotted in a non-dimensionalized f C½Fa þ 10ð1  aÞ
form, being divided by the yield moment My. The three g=ḡo ¼ 1  p1:0, (12)
axF2
fundamental buckling modes are also shown in Fig. 1(b).
The FSM analysis demonstrates that the considered where in the case of doubly symmetric beams-columns with
I-section beam has no distinct distortional minimum while the axial force equal to zero, we have
LB and long half-wavelength buckling are clear. However,
M nWangetal Lb
the buckling mode for 1750 mm long I-beam, shown in g ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi , (13)
Fig. 1(b), is classified as either ‘local-lateral coupling’ or DB EI y GJ
mode since both translation and rotation of the fold line of
the member are accompanied by simultaneous distortion of f ¼ 0:180  0:759O þ 1:516O2  1:453O3 þ 0:516O4 , (14)
the cross-section. It is important to note that the distor-
tional mode shown involves distortion of web of the cross- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
section and rigid response of the flanges [11]. 2
ḡo ¼ p ð1 þ K̄ Þ, (15)

3. Elastic distortional buckling formulae and


" #1=2
In order to evaluate the 2005 AISC code predictions of p2 EI y h2o
K̄ ¼ , (16)
the elastic DB resistances for the targeted unstiffened-web 4GJL2b
I-section member geometries in a more comprehensive
way, in addition to the application of the FSA models, the also, in accordance with Wang et al.’s [13] definitions in
theoretical predictions of two elastic DB formulae devel- their paper, the independent geometric parameters may be
oped by Bradford [12] and Wang et al. [13] have also been non-dimensionalized as follows:
considered in this study.
bf t tf t bf þ bf b
Using the results of a finite element DB analysis, O¼ ; G¼ ; F¼ t ,
Bradford [12] developed a design equation (Eq. (10)) for bf t þ bf b tf t þ tf b tf t þ tf b
the reduced buckling stress Fcrd of monosymmetric I-beams hc Aw Lb
C¼ ; a¼ ; x¼ . ð17Þ
with equal-thickness flanges accounting for distortion of tw A bf t þ bf b
the cross-section.
The parameters O, G, and F describe the geometry and
F crd 490ðtf =bf b Þðtf =tw Þð1  0:560ðbf t =bf b ÞÞ the degree of asymmetry of the flanges. C and a describe
¼1 . (10)
F cr E=F cr the web slenderness, and x indicates the beam slenderness,
and ultimately the practical values for these parameters lie
The nominal elastic DB moment using Bradford’s in the ranges [13]:
equation M nBradford is obtained by the product of Fcrd and
Sx Eq. (10) is based on regression analysis from finite 0pOp1; 0pGp1; 5pFp15,
element parametric studies of I-section members with 10pCp150; 0:1pap0:4; 15pxp80. ð18Þ
ARTICLE IN PRESS
470 T. Zirakian / Thin-Walled Structures 46 (2008) 466–475

4. Assessment of the AISC code predictions unbraced lengths (Lb) of the beams are all larger than Lr,
so that the elastic behavior is provided. Moreover, the
4.1. General ratios of nominal LTB moments to yield moments are
given in the first column of Table 3, and the occurrence of
As it was previously mentioned, the main objective of elastic buckling is quite obvious in all cases.
this paper is to evaluate the effectiveness of the 2005 AISC Table 3 summarizes the elastic DB results as well. The
design rules in case of elastic DB of doubly symmetric ratios of the AISC code predictions, FSA results, and
I-section members with slender webs, since F crðJ¼0Þ is theoretical predictions of Eqs. (10) and (12) to LTB
generally believed to be a conservative estimate of the DB solutions are shown in the second, third, fourth, and
capacity. fifth columns of the table, respectively. In general, the
To clarify this, it has been tried to present a quantitative reductions in elastic LTB resistance due to the web
measure of accuracy of the 2005 AISC code predictions distortion increase as the beam length decreases since
through comparison with the FSA solutions as well as distortion has larger effect on shorter beams. In other
theoretical predictions of two previously developed for- words, the disparity between the elastic distortional
mulae for elastic DB. Accordingly, four case studies are and LTB solutions decreases as the length increases. This
performed in such a fashion that in each case one beam fact is clearly demonstrated by the FSA and theoretical
dimension or property is varied while the others are kept predictions of Bradford’s (Eq. (10)) and Wang et al.’s
constant, and the accuracy of the code predictions is (Eq. (12)) formulae in comparison with the elastic LTB
assessed in that specific case. solutions (Table 3). As the beam lengthens, the finite strip
The cross-sectional dimensions, length, and yield method as well as Eqs. (10) and (12) indicate reductions
strength considered in each case are summarized in ranging from 6% to 4%, 10% to 2%, and 10% to 4%,
Table 1, and the varying parameter is clearly demonstrated respectively. In contrast, the AISC code predictions
in each case. Moreover, the cross-section dimensional indicate a disparate trend, i.e. by increasing of the beam
ratios viz. bf/tf, h/bf, h/tw, tf/tw, Lb/ho as well as Lr and Lb of length, the reductions increase from 12% to 40% and the
the considered beams are given in each individual case. DB moments are reduced significantly below the LTB
Since all of the beams in this study have slender webs and moments. As a result, it is found that the AISC code
compact flanges in accordance with the noncompact-web equations tend to give significantly conservative results
(Eq. (1)) and compact-flange (Eq. (19)) limits specified in relative to the FSA as well as theoretical DB solutions as
the AISC specification [3], the beam length increases, and unfortunately this is a
sffiffiffiffiffiffi conspicuous weakness of Eq. (5) since it yields irrational
bf E results in such a well-established case, which may bring
lpcompact ¼ p0:38 , (19)
2tf Fy about economic losses as well.
therefore, the h/tw ratios of the beams are larger than the
code limiting slenderness values for noncompact webs
Table 2
(Eq. (1)), and also the bf/2tf ratios of the beams are Dimensional ratios and lengths of I-beams (Case 1)
smaller than the code limiting values for compact flanges
(Eq. (19)). Cross-section bf =tf h=bf h=tw tf =tw Lb =ho Lr Lb
number (mm) (mm)

4.2. Case 1: variation of length B1-1 12.0 3.1 150.0 4.0 7.7 5989.5 6000
B1-2 12.0 3.1 150.0 4.0 8.3 5989.5 6500
B1-3 12.0 3.1 150.0 4.0 9.0 5989.5 7000
In this case, the accuracy of the AISC code predictions B1-4 12.0 3.1 150.0 4.0 9.6 5989.5 7500
for elastic DB is assessed as a result of variation of length, B1-5 12.0 3.1 150.0 4.0 10.9 5989.5 8500
while all the cross-sectional dimensions of the beams are B1-6 12.0 3.1 150.0 4.0 12.8 5989.5 10,000
kept constant (Table 1). The cross-section dimensional B1-7 12.0 3.1 150.0 4.0 15.4 5989.5 12,000
ratios as well as Lr and Lb of the studied members are listed B1-8 12.0 3.1 150.0 4.0 17.9 5989.5 14,000
B1-9 12.0 3.1 150.0 4.0 20.0 5989.5 15,600
in Table 2. As indicated in Table 2, the considered

Table 1
Summary of beam dimensions and yield strengths considered for the four case studies

Case number h (mm) hc (mm) ho (mm) d (mm) tw (mm) bf (mm) tf (mm) Lb (mm) Fy (MPa)

1 750 760 780 800 5 240 20 Varies 345


2 990 1000 1021 1042 Varies 250 21 6250 345
3 800 810 835 860 5 Varies 25 10,280 345
4 850 860 880 900 5 240 20 Varies Varies
ARTICLE IN PRESS
T. Zirakian / Thin-Walled Structures 46 (2008) 466–475 471

Table 3
Comparison of the AISC code predictions with the FSA and theoretical predictions (Case 1)

Cross-section number M nLTB =M y M nAISC =M nLTB M nFSA =M nLTB M nBradford =M nLTB M nWangetal =M nLTB

B1-1 0.78 0.88 0.94 0.90 0.90


B1-2 0.68 0.87 0.94 0.92 0.91
B1-3 0.60 0.85 0.94 0.93 0.92
B1-4 0.53 0.84 0.94 0.93 0.92
B1-5 0.43 0.80 0.94 0.95 0.93
B1-6 0.33 0.75 0.95 0.96 0.94
B1-7 0.25 0.69 0.95 0.97 0.95
B1-8 0.20 0.64 0.96 0.98 0.96
B1-9 0.17 0.60 0.96 0.98 0.96

Table 4
Dimensional ratios and lengths of I-beams (Case 2)

Cross-section number bf =tf h=bf h=tw tf =tw Lb =ho Lr (mm) Lb (mm)

B2-1 11.9 4.0 137.5 2.9 6.1 5980.5 6250


B2-2 11.9 4.0 145.6 3.1 6.1 6007.6 6250
B2-3 11.9 4.0 154.7 3.3 6.1 6039.2 6250
B2-4 11.9 4.0 165.0 3.5 6.1 6071.8 6250
B2-5 11.9 4.0 176.8 3.8 6.1 6099.8 6250
B2-6 11.9 4.0 190.4 4.0 6.1 6133.3 6250
B2-7 11.9 4.0 206.3 4.4 6.1 6166.7 6250
B2-8 11.9 4.0 225.0 4.8 6.1 6196.6 6250
B2-9 11.9 4.0 247.5 5.3 6.1 6230.9 6250

Table 5
Comparison of the AISC code predictions with the FSA and theoretical predictions (Case 2)

Cross-section number M nLTB =M y M nAISC =M nLTB M nFSA =M nLTB M nBradford =M nLTB M nWangetal =M nLTB

B2-1 0.70 0.92 0.96 0.94 0.93


B2-2 0.70 0.91 0.96 0.93 0.93
B2-3 0.71 0.91 0.95 0.93 0.92
B2-4 0.72 0.90 0.95 0.92 0.91
B2-5 0.72 0.90 0.95 0.92 0.90
B2-6 0.73 0.89 0.94 0.91 0.89
B2-7 0.74 0.88 0.93 0.90 0.88
B2-8 0.74 0.88 0.93 0.89 0.86
B2-9 0.75 0.87 0.92 0.88 0.84

4.3. Case 2: variation of web thickness As it is seen in Table 5, by increasing of the web
slenderness, the reductions increase in all cases, which vary
DB resistance of an I-beam is specifically related to its from 8% to 13% for the code predictions, 4–8% for the
web height to thickness ratio (h/tw), which is also known FSA results, 6–12% for Bradford’s equation, and ulti-
as web slenderness. It has been demonstrated that the mately 7–16% for Wang et al.’s equation. In general, the
effects of cross-sectional distortion become more pro- elastic DB predictions of the AISC specification are
nounced as the web slenderness increases. In this case, conservative relative to both the FSA and Bradford’s
the AISC code predictions are assessed as a result of equation predictions. However, the amount of the con-
variation of web slenderness. To achieve this, web servatism is not that remarkable (albeit a slightly more
thickness tw is varied from 7.2 to 4.0 mm while the other conservative results relative to the FSA solutions). In spite
cross-sectional dimensions are kept constant (Tables 1 of minor conservatism of the code predictions relative to
and 4). Furthermore, one constant length (6250 mm) is those of Wang et al.’s equation at lower web slendernesses,
considered for all beams, and as it is shown in Table 4, by increasing of the web slenderness the Wang et al.’s
this length is larger than the calculated parameter Lr of equation has conversely presented slightly conservative
all beams. Finally, the results of this case study are predictions relative to the code predictions. All in all, it
summarized in Table 5. seems that the AISC code predictions are in good
ARTICLE IN PRESS
472 T. Zirakian / Thin-Walled Structures 46 (2008) 466–475

9
8
7

Reduction (%)
6
5
4
3
2
1 Rpg
J=0
0
130 140 150 160 170 180 190 200 210 220 230 240 250 260
hc /tw

Fig. 2. Reductions for Rpg and J ¼ 0 due to variation of web slenderness.

Table 6
Dimensional ratios and lengths of I-beams (Case 3)

Cross-section number bf =tf h=bf h=tw tf =tw Lb =ho Lr (mm) Lb (mm)

B3-1 16.0 2.0 160.0 5.0 12.3 10,276.7 10,280


B3-2 13.3 2.4 160.0 5.0 12.3 8512.8 10,280
B3-3 10.7 3.0 160.0 5.0 12.3 6750.8 10,280
B3-4 8.0 4.0 160.0 5.0 12.3 4990.5 10,280
B3-5 5.3 6.0 160.0 5.0 12.3 3233.9 10,280

Table 7
Comparison of the AISC code predictions with the FSA and theoretical predictions (Case 3)

Cross-section number M nLTB =M y M nAISC =M nLTB M nFSA =M nLTB M nBradford =M nLTB M nWangetal =M nLTB

B3-1 0.82 0.85 0.92 0.90 0.90


B3-2 0.59 0.80 0.92 0.92 0.90
B3-3 0.40 0.74 0.91 0.93 0.90
B3-4 0.25 0.64 0.92 0.94 0.90
B3-5 0.14 0.49 0.94 0.95 0.90

agreement with the FSA as well as the theoretical results in equations, which is generally satisfactory in spite of some
this case. minor conservatism in comparison with the FSA and the
As it was discussed before, in addition to the use of J ¼ 0 other theoretical results.
in the theoretical LTB equation of slender-web I-section
members specified in the 2005 AISC specification, web 4.4. Case 3: variation of flange width
distortion effects in cross-sections with large web slender-
ness may be somewhat controlled as a result of application In this case, the flange width is varied from 400 to
of the bending strength reduction factor Rpg. Since the 133 mm while the other dimensions are kept constant
bending strength reduction factor directly accounts for the (Tables 1 and 6), and the code predictions are evaluated as
variations of the web slenderness, so its effectiveness in a result of this variation. In general, the buckling moment
comparison with that of consideration of J ¼ 0 in case of starts at a very low level for the lowest flange width and
DB of slender-web I-section members is investigated in this then sharply increases with a rising gradient as the flange
case study. Fig. 2 is included to demonstrate the nature of widens. This is because of the increase in the minor axis
the reductions induced separately by the two aforemen- and torsional rigidity of the section.
tioned factors, i.e. Rpg and J ¼ 0. This figure illustrates that Table 7 shows the ratios of the LTB moments to yield
by increasing of the web slenderness an approximately moments as well as ratios of the various DB predictions to
constant reduction (7.8–7.4%) is obtained by taking the LTB moments. Based on the results of the FSA and DB
torsional constant J equal to zero, while the reductions formulae, it is found that by decreasing of the flange
induced by Rpg vary from 0.1% to 6.0%. In fact, a width, the influence of distortion of the cross-section is
combination of the two reductions is induced by the code slightly reduced, and consequently the distortional and
ARTICLE IN PRESS
T. Zirakian / Thin-Walled Structures 46 (2008) 466–475 473

Table 8
Yield strengths, flange and web slendernesses, dimensional ratios, and lengths of I-beams (Case 4)

Cross-section number Fy (MPa) Flange lpcompact Web lrnoncompact bf =tf h=bf h=tw tf =tw Lb =ho Lb ¼ Lr (mm)

B4-1 250 10.7 161.2 12.0 3.5 170.0 4.0 7.9 6969.2
B4-2 290 10.0 149.7 12.0 3.5 170.0 4.0 7.4 6470.7
B4-3 345 9.1 137.2 12.0 3.5 170.0 4.0 6.7 5932.6
B4-4 415 8.3 125.1 12.0 3.5 170.0 4.0 6.1 5409.1
B4-5 485 7.7 115.7 12.0 3.5 170.0 4.0 5.7 5003.6
B4-6 550 7.2 108.7 12.0 3.5 170.0 4.0 5.3 4698.6
B4-7 620 6.8 102.4 12.0 3.5 170.0 4.0 5.0 4425.4
B4-8 690 6.5 97.0 12.0 3.5 170.0 4.0 4.8 4195.0

Table 9
Comparison of the AISC code predictions with the FSA and theoretical predictions (Case 4)

Cross-section number M nLTB =M y M nAISC =M nLTB M nFSA =M nLTB M nBradford =M nLTB M nWangetal =M nLTB

B4-1 0.79 0.88 0.95 0.93 0.91


B4-2 0.78 0.89 0.95 0.92 0.91
B4-3 0.76 0.90 0.94 0.91 0.90
B4-4 0.75 0.90 0.94 0.89 0.89
B4-5 0.75 0.91 0.94 0.87 0.88
B4-6 0.74 0.91 0.94 0.85 0.87
B4-7 0.74 0.91 0.94 0.84 0.86
B4-8 0.73 0.91 0.94 0.82 0.86

LTB moments become closer, i.e. the reductions decrease In the United States, the structural steels used for
from 8% to 6% for the FSA results and 10–5% for the construction are designated by the American Society of
theoretical predictions of Bradford’s equation, while Wang Testing and Materials (ASTM). In general, steel used in
et al.’s equation provides no significant variation in the ordinary constructions normally has values of Fy that
reduction. As it is seen in Table 7, a totally different trend range from 248 to 345 MPa, although higher strength steels
is indicated by predictions of the AISC code equations, i.e. are becoming more common. In this study, a wide range of
the reductions in the elastic buckling resistance increase yield strengths, i.e. from 250 MPa (used in riveted, bolted,
significantly from 15% to 51% as the flange width and welded buildings and bridges) to 690 MPa (used in
decreases. In this case also, the conservatism associated bridges), are considered, which are tabulated in Table 8.
with the use of the AISC code equations relative to the Included in the table are also the calculated limiting
application of finite strip method as well as the theoretical slenderness values for compact flanges and noncompact
DB formulae is found to be remarkably large. webs as well as the aforementioned beam dimensional
ratios and lengths. The grey-shaded slenderness values
4.5. Case 4: variation of yield strength correspond to the considered flange and web slendernesses
for all members, so that the constant cross-sectional
All of the three previously discussed cases target dimensions for all beams are determined such that the
members with Fy ¼ 345 MPa. Although some design beams to have compact flanges and slender webs. More-
specifications give specific limits on yield strengths aimed over, as it is shown in the table, the members correspond-
at controlling the unconservative errors due to the neglect ing to this case study have Lb ¼ Lr. Furthermore, it should
of web distortion effects, in order to gain a more complete be noted that by increasing of the yield strength, the values
understanding of the implications of web distortion as well of Lr decrease.
as evaluation of the AISC code predictions as a result of Looking at the DB results for this case study in Table 9,
variation of the yield strength, this particular case study it is found that based on the FSA results as well as the
was performed. White and Jung [7], also, reported a limited theoretical predictions of Eqs. (10) and (12), the strength
number of FEA studies conducted on beams with identical reductions due to the web distortional flexibility for beams
cross-sections and Fy ¼ 345 and 485 MPa to gauge the with identical cross-sections tend to increase as a result of
effect of the yield strength on the various predictions. Their increasing of the yield strength. However, the reductions
studies showed very similar reductions due to web are very similar for all yield strengths based on the FSA
distortional flexibility for both Fy ¼ 345 versus 485 MPa. results which vary from 5% to 6%, while the amount of
Consequently, they concluded that the errors associated this variation for Bradford’s and Wang et al.’s equations is
with the web distortional flexibility appear to be very from 7% to 18% and 9% to 14%, respectively. On the
similar for either Fy ¼ 345 or 485 MPa. contrary, the reductions as a result of application of the
ARTICLE IN PRESS
474 T. Zirakian / Thin-Walled Structures 46 (2008) 466–475

12
Rpg
10 J=0

Reduction (%)
8

0
200 250 300 350 400 450 500 550 600 650 700
Fy (MPa)

Fig. 3. Reductions for Rpg and J ¼ 0 due to variation of yield strength.

AISC code equations become smaller as the yield strength Table 10


increases, i.e. these reductions vary from 12% to 9%. All in Evaluation of the AISC code predictions
all, the code predictions are found to be slightly con- Case number The average conservatism of the AISC code
servative relative to the FSA solutions over the full range predictions relative to
of the yield strengths. However, these results indicate that
the reduction in the slender-web I-section member elastic FSA (%) Bradford’s Eq. Wang et al.’s
(10) (%) Eq. (12) (%)
LTB resistance due to web distortion is by and large very
similar for members of the same cross-section and different 1 18.6 18.3 17.2
yield strengths. 2 5.0 1.7 0.0
In the final analysis, the nature of the reductions induced 3 23.6 24.0 21.9
4 4.4 2.9 1.8
separately by the two parameters Rpg and J=0 is compared
in Fig. 3, since the amounts of both reductions are influenced
by the variation of the yield strength. As it is seen in the
figure, the reductions induced by Rpg show an increasing servative results relative to the FSA and other theoretical
trend from 0.7% to 4.6% as the yield strength increases, DB strengths in the first and third cases, while in the second
while in contrast the reductions induced by the use of J ¼ 0 and fourth cases the average amount of conservatism is not
decrease from 11.1% to 4.5% by increasing of the yield that remarkable. On the average, some minor unconserva-
strength. It is found that the reductions induced by Rpg are tism is even found relative to the predictions of Bradford’s
similar to the reductions obtained from the FSA and other and Wang et al.’s equations in the fourth case.
theoretical results in their increasing trend, while the total In addition, it should be emphasized that both the DB
reduction induced by the code equations (sum of the two formulae, i.e. Eqs. (10) and (12), were developed based on
reductions induced by Rpg and J ¼ 0) reveals a decreasing regression analysis of results of parametric studies for
trend in contrast to that of the FSA and theoretical results. members spanning a narrower range of design parameters
than those considered here. Therefore, only some I-beams
5. Discussion considered in this research satisfy the Eqs. (10) and (12)
limits (Eqs. (11) and (18)) which were used in their
In Section 4 the AISC code predictions in case of elastic development. However, these formulae are utilized in this
DB of slender-web I-section members relative to those of research together with the accurate finite strip analyses
the FSA and the theoretical formulae were investigated since it is believed that they provide significant insight
through four case studies. It was demonstrated that the into the behavior. Moreover, the predictions of the
code predictions unfortunately exhibit a totally different two formulae are quite close in most cases, and also
trend in some cases in comparison with that of the FSA the agreement between the predictions of the FSA and
and the other theoretical results. Moreover, despite some Eqs. (10) and (12) is by and large satisfactory. Lastly, from
scatter due to the application of various methods and the table provided in Section 5, it is evident that the three
formulae, the AISC code predictions are conservative average values provided for each case are generally in good
compared to the FSA and other theoretical results. The agreement with each other.
average amounts of conservatism of the AISC code As a whole, although the use of J ¼ 0 with the aim of
predictions relative to the FSA results as well as the controlling the unconservative errors due to the neglect of
predictions of Bradford’s and Wang et al.’s formulae in web distortion effects is generally an acceptable approach,
each case study are presented in Table 10. As it is seen in but still further theoretical and experimental work remains
the table, the AISC code equations give significantly con- for developing more efficient and optimum equations for
ARTICLE IN PRESS
T. Zirakian / Thin-Walled Structures 46 (2008) 466–475 475

use as a specification check to ensure against unconserva- efficient design rules which either prevent the occurrence or
tive and excessive conservative errors due to the effect of take into account the influence of distortional buckling,
the web distortional flexibility. Particularly, experimental and ultimately some economic benefits can also be gained
work is of great importance in this regard. Unfortunately, by developing further optimum design rules.
despite extensive theoretical studies on DB of I-beams,
little experimental work is found in the technical literature, Acknowledgments
while the distortional behavior of a structural member can
be investigated more efficiently and accurately through The author is grateful to Professor D.W. White from
experimental research. As an example, in accordance with Georgia Institute of Technology and Professor C.M. Wang
the findings in a recently reported experimental research by from National University of Singapore for their precious
Zirakian and Showkati [14,15] on DB of steel I-section help in providing some research reports and articles.
members, web distortion may also be influenced by some
other factors, e.g. initial geometric imperfections, etc., and
consequently the structural member may have different References
distortional behaviors. In addition, it is quite obvious that
[1] Bradford MA, Ronagh HR. Generalized elastic buckling of
experimental work is necessary for further validation of the restrained I-beams by FEM. J Struct Eng ASCE 1997;123(12):
theoretical results. Thus, further research is needed to 1631–7.
advance the ideas in this paper further and improve the [2] American Iron and Steel Institute. North American specification for
accuracy and effectiveness of the code equations. the design of cold-formed steel structural members. Washington, DC:
AISI; 2007.
[3] American Institute of Steel Construction. Specification for structural
6. Conclusions steel buildings. Chicago, IL: AISC; 2005.
[4] American Institute of Steel Construction. Commentary on the
This paper evaluates the accuracy of the 2005 AISC code specification for structural steel buildings. Chicago, IL: AISC; 2005.
predictions in case of elastic distortional buckling of [5] Baber TT, Simons DC. The impact of the AASHTO LRFD design
doubly symmetric I-shaped members with slender webs. code on bridge stiffness and strength: part I: methods and design
comparisons. Research report no. FHWA/VTRC 07-CR2. Depart-
The code predictions given by the AISC specification were ment of Civil and Environmental Engineering, University of Virginia,
compared with the accurate finite strip analysis as well as 2007.
the theoretical solutions of two previously developed [6] Basler K. Strength of plate girders in bending. J Struct Div ASCE
formulae for elastic distortional buckling of I-section 1961;87(ST4):153–81.
members. [7] White DW, Jung SK. Effect of web distortion on the buckling
strength of noncomposite discretely-braced steel I-section members.
Four case studies were considered and the comparison Eng Struct 2007;29:1872–88.
showed varying degrees of accuracy for the code predic- [8] American Association of State and Highway Transportation
tions. However, the 2005 AISC code equations were Officials. AASHTO LRFD bridge design specifciations. 3rd ed.
generally shown to provide conservative strength estimates Washington, DC: AASHTO; 2004.
[9] Schafer BW. CUFSM 3.12. Elastic buckling analysis of thin-walled
for slender-web I-section members relative to the FSA and
members by finite strip analysis; 2006. /www.ce.jhu.edu/bschafer/
the other theoretical solutions, and even the amount of this cufsmS.
conservatism was found to be rather dramatic in some [10] Schafer BW. CUFSM 2.5. Users manual and tutorials; 2001.
cases. /www.ce.jhu.edu/bschafer/cufsmS.
For slender-web members, the current AISC resistance [11] Zirakian T. Elastic distortional buckling analysis of steel I-beams
equations are based on the assumption of J ¼ 0 and the using CUFSM. In: Proceedings of the 2007 annual conference of the
Canadian Society of Civil Engineering, CSCE, 2007.
influence of the web distortional flexibility is not addressed [12] Bradford MA. Distortional buckling of monosymmetric I-beams.
explicitly in the standard. As it was shown in this paper, the J Construct Steel Res 1985;5:123–36.
use of J ¼ 0 in the theoretical lateral–torsional buckling [13] Wang CM, Chin CK, Kitipornchai S. Parametric study on
equations gives a conservative estimate of the buckling distortional buckling of monosymmetric beam-columns. J Construct
strength in case of slender-web doubly symmetric I-section Steel Res 1991;18:89–110.
[14] Zirakian T, Showkati H. Distortional buckling of castellated beams.
members. J Construct Steel Res 2006;62(9):863–71.
Further studies, both experimental and theoretical, are [15] Zirakian T, Showkati H. Experiments on distortional buckling of
required in order to enable the establishment of more I-beams. J Struct Eng ASCE 2007;133(7):1009–17.

You might also like