Thin-Walled Structures 119 (2017) 879–901

Contents lists available at ScienceDirect

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

Full length article

Local-distortional interaction in cold-formed steel beams: Behaviour, MARK

strength and DSM design
⁎
André Dias Martins, Dinar Camotim , Pedro Borges Dinis
CERIS, ICIST, DECivil, Instituto Superior Técnico, Universidade de Lisboa, Portugal

A R T I C L E I N F O A B S T R A C T

Keywords: This work presents and discusses numerical results concerning cold-formed steel simply supported beams sub-
Cold-formed steel beams jected to uniform bending and exhibiting (i) three cross-section shapes (lipped channels, zed-sections and hat-
Local-distortional interaction sections), and (ii) two end support conditions (differing in the warping and local displacement/rotation re-
Major/minor/skew bending straints, which are either free or fully prevented). A systematic numerical investigation is carried out, in order to
Shell finite element analysis
characterise the post-buckling behaviour and ultimate strength of beams experiencing more or less severe L-D
Elastic-plastic post-buckling behaviour
interaction effects. 43 geometries and 11 yield stresses are considered for each combination of cross-section
Direct Strength Method (DSM)
shape and support conditions, thus ensuring distinct (i) ratios between the local (McrL) and distortional (McrD)
critical buckling moments (0.50 ≤ McrD /McrL ≤ 2.00), and (ii) local or distortional slenderness values, ranging
from 0.50 to 3.50. The numerical results are obtained through ABAQUS shell finite element analyses and concern
the (i) post-buckling behaviour (elastic and elastic-plastic), ultimate strength and failure mechanisms of beams
previously selected to undergo considerable L-D interaction. Based on the acquired information, a first con-
tribution towards the Direct Strength Method (DSM) design of cold-formed steel beams undergoing different
“levels” of L-D interaction is presented and discussed.

1. Introduction in-depth knowledge on the structural response/performance of beams

Most cold-formed steel members display very slender thin-walled
open cross-sections, which makes them highly susceptible to buckling
phenomena involving cross-section deformations, namely local (L)
buckling (wall bending only – no fold-line motions) or distortional (D)
buckling (wall bending and cross-section distortion – fold line motions).
In fact, any of these instability phenomena may be critical and govern
the structural response of such members for either “short” (L instability)
or “intermediate” (D instability) lengths. Moreover, when the critical
buckling moments McrL and McrD are close (or, at least, not far apart),
the beams experience L-D interaction, generally occurring in beams
with short-to-intermediate lengths. Obviously, the behavioural features
associated with this coupling phenomenon must be taken into account
in the design of such members, since they may cause significant ulti-
mate strength erosion (particularly when McrL≈McrD) – otherwise, un-
safe designs might occur. Unfortunately, L-D interaction has been far
less studied than its local-global counterpart, particularly in beams (or
in beam-columns, for that matter). Indeed, there are only a few in-
vestigations dealing with L-D interaction in cold-formed steel beams.
However, it is consensual, amongst the technical and scientific com-
munities working with cold-formed steel structures, the need to acquire
affected by these coupling effects. This constitutes a necessary first step
in the path towards the development, calibration and validation of
design approaches able to handle such effects and to meet the appro-
priate requirements for the incorporation in cold-formed steel codes/
specifications.
Most of the available results concerning numerical/experimental
investigations or design proposals dealing with L-D interaction in cold-
formed steel members involve fixed-ended columns under uniform
compression – e.g., the works of Kwon and Hancock [1], Young et al.
[2] and Martins et al. [3–5]. Although research has already been con-
ducted on the behaviour of cold-formed steel beams, the authors are
only aware of very few studies addressing the influence of L-D inter-
action on that same behaviour. They consist of numerical and experi-
mental investigations − the former concern simply supported (i) lipped
channel beams exhibiting flange or web-triggered L-D interaction under
uniform major-axis bending and with 0.85 ≤ McrL /McrD ≤ 1.15 [6] or
McrL/McrD≈1.0 [7,8], and also (ii) zed-section beams with and without
intermediate stiffeners subjected to 4-point bending [9–11]. As for the
experimental investigations, it is worth noting (i) the tests conducted by
Bernard et al. [12,13] on simply supported thin-walled profiled steel
decks with and without intermediate (“v-shaped” [12] and “flat-hat”

⁎
Corresponding author.
E-mail addresses: andrerdmartins@ist.utl.pt (A.D. Martins), dcamotim@civil.ist.pt (D. Camotim), dinis@civil.ist.pt (P.B. Dinis).

http://dx.doi.org/10.1016/j.tws.2017.06.011
Received 6 May 2017; Accepted 13 June 2017
Available online 20 August 2017
0263-8231/ © 2017 Elsevier Ltd. All rights reserved.
A.D. Martins et al.
[13]) stiffeners under minor-axis bending (subsequently, the authors
compared their ultimate strength results with the predictions of several
design standards [14] and showed evidence of L-D interaction), (ii)
most of the tests reported by Wang and Zhang [15] on lipped channels
with several lip configurations (upright, inclined and return) involving
specimens under uniform or non-uniform bending, which showed clear
evidence of L-D interaction, and (iii) the studies carried out by Douty
[16] and Serrete and Peköz [17] on simply supported standing seam
roof panels with several configurations (these authors reported evi-
dence of L-D interactive failures in most specimens). However, a sys-
tematic investigation, aimed at identifying which combinations of the
ratios involving McrL or McrD and the yield moment lead to sizeable/
relevant L-D interaction effects, is still lacking – the purpose of this
work is to provide a first contribution towards filling this gap, based on
the knowledge acquired from the recently reported Generalised Beam
Theory (GBT) post-buckling results [18], which shed fresh light on the
mechanics of lipped channel beams undergoing L-D interaction.
Thus, the main objectives of this work are (i) to present and discuss
numerical results obtained for simply supported uniformly bent beams
affected by L-D interaction and (ii) to provide a first contribution to-
wards the development of an efficient Direct Strength Method (DSM)
design approach for such members. The beams analysed exhibit (i)
three cross-section shapes (lipped channels, hats and zeds) and (ii) two
end support conditions (addressed in Section 2). A systematic numerical
investigation is carried out, in order to characterise the post-buckling
behaviour and strength of beams experiencing more or less severe L-D
interaction effects. It involves beams (i) with 43 geometries, for each
combination of cross-section dimensions and end support conditions,
selected to ensure distinct ratios between McrD and McrL, namely 0.50 ≤
RDL ≤ 2.00 (RDL=McrD /McrL) and (ii) 11 yield stresses, such that a wide
slenderness range (0.50–3.50 interval) is covered. The numerical results
presented and discussed are obtained from ABAQUS [19] shell finite ele-
ment analyses and concern the beam (i) post-buckling behaviour, (ii)
ultimate strength and (iii) failure mode – special attention is paid to
assessing the ultimate strength erosion due to the coupling effects. Fi-
nally, the paper closes with some considerations about the impact of the
findings reported on the DSM-based design of cold-formed steel beams
experiencing different L-D interaction levels, as well as a few comments
about the work on this topic planned for the near future.
2. Buckling analysis – beam geometry selection
The identification/selection of uniformly bent beam geometries
prone to L-D interaction is quite straightforward, since short-to-inter-
mediate beams have similar local and distortional buckling moments
(or, at least, not far apart). As done in similar studies, such geometries
were selected by means of sequences of GBT buckling analysis using
code GBTUL (Bebiano et al. [20]). Seven types of beams (E = 210 GPa,
v = 0.30) are considered, combining (i) three cross-section shapes,
namely (i1) lipped channels (C) bent about the major-axis, (i2) hats bent
about the major-axis (HM) or minor-axis (Hm – lips under compression),
and (i3) zeds (Z) under skew bending causing uniform flange com-
pression (the worst case), (ii) two support conditions (termed here SCA
and SCB – the exception are the Hm-beams, for which only SCB are
considered): while the SCA beams are simply supported with respect to
major and minor-axis bending and have the end cross-section torsional
rotations prevented, the SCB beams differ in the fact that the end cross-
section warping and local displacements/rotations are also prevented –
physically, preventing these displacements/rotations corresponds to
rigidly attaching thick end plates to the beam end cross-sections.
In order to confirm/illustrate the assertion that short-to-inter-
mediate beams are prone to L-D interaction, Fig. 1(a1) shows, for the
C+SCA beam with bw = 100, bf = 65, bl = 12.5 and t = 1.0 mm, the
variation, with the length L (logarithmic scale), of the single half-wave
(Mb.1) and critical (Mcr) buckling moments. On the other hand,
Fig. 1(a2) provides GBT-based “approximate” buckling curves obtained
880
Thin-Walled Structures 119 (2017) 879–901
by considering three deformation mode sets, namely modes (i) 7–17
(local), (ii) 5+6 (distortional), and (iii) 3+4 (lateral-torsional: minor
axis-bending + torsion) – note that the cross-section discretisation in-
volves 9 intermediate nodes (3 in the web and flanges). Since the 7–17
modes yield practically the exact critical local buckling moment and the
5+6 modes lead to an approximate1 critical distortional buckling
moment, Fig. 1(a2) makes it possible to conclude that beams with
lengths 45 < L < 250 cm are highly prone to L-D interaction – this
fairly large length interval evidences the relevance of this coupling
phenomenon. Obviously, when global buckling is critical the solution
obtained with deformation modes 3+4 only is the exact lateral-tor-
sional buckling moment – note that there is a gap in the transition
between the “exact” and 3+4 curves, which corresponds to lengths of
beams prone to D-G or L-D-G interaction (coupling phenomena outside
the scope of this work – see, for instance, the works of the authors
[21,22] on lipped channel beams affected by D-G interaction). Finally,
Fig. 1(b) shows the local and distortional critical buckling modes of the
L = LDL = 50 cm beam, associated with practically coincident local and
distortional buckling moments – they exhibit 8 (local) and one (dis-
tortional) half-waves. Naturally, the post-buckling behaviour (either
elastic or elastic-plastic) of such beam is bound to be strongly affected
by L-D interaction.
The output of this selection are 43 beam geometries (labelled X1 to
X43, where “X” stands for either “C”, “HM”, “Hm” or “Z”) for each
combination of cross-section shape and end support conditions – they
can be found in Annex A (Tables A1–A7). All these beams (i) exhibit
RDL values in the range 0.50 ≤ RDL ≤ 2.00 and (ii) have global
buckling moments (McrG) much higher than (ii1) the local and dis-
tortional ones (McrG / Mcr. Max ≫ 1.0 – Mcr. Max=max{McrD; McrL}) and
(ii2) the yield moments (McrG / My. Max ≫ 1.0), thus ensuring that no
interaction with global (lateral-torsional) buckling occurs – the values
of the above two moment ratios are also given in Annex A. Local
buckling is almost always triggered by the compressed flange2 (most
common situation in practice) and, in order to study the effect of
strong L-D interaction, 26 beams were selected in the
0.85 < RDL < 1.15 range – the other 17 beams are obtained by
varying this ratio in 0.10/0.05 steps until 2.00 and 0.50, respectively,
making it possible to investigate “secondary (local or distortional)
bifurcation interactions” (Martins et al. [23]).
3. Post-buckling behaviour under local-distortional interaction
This section addresses the post-buckling behaviour of uniformly
bent beams affected by L-D interaction. Initially, brief descriptions of (i)
the shell finite element model adopted (Section 3.1) and (ii) the worst
initial imperfection shape considered, for each combination of cross-
section and support conditions (Section 3.2), are provided. Then, at-
tention is turned to the discussion of the post-buckling behaviour of
cold-formed beams under L-D interaction (Section 3.3). In particular,
three types of L-D interaction are addressed (for all the SCA and SCB C,
HM and Z-beams): (i) “true interaction” (Section 3.3.1), (ii) “secondary
local bifurcation interaction” (Section 3.3.2) and (iii) “secondary dis-
tortional bifurcation interaction” (Section 3.3.3). Finally, the Hm+SCB
beams are dealt with separately, since they have been much less studied
than the other ones in the past (Section 3.3.4).
3.1. Shell finite element analysis
The beam elastic and elastic-plastic post-buckling analysis were
determined by means of ABAQUS [19] shell finite element analysis
1
“Pure” distortional critical buckling modes often contain small (but not negligible)
contributions from local deformation modes.
2
The exceptions are some Hm beams, for which local buckling may be triggered by the
lips.
A.D. Martins et al.
Fig. 1. C+SCA beam (a1) Mcr vs. L curve (a2) “approximate” GBT solutions and (b) local/distortional critical buckling modes.
adopting models similar to those reported earlier [24]. Nevertheless,
the main characteristics of these models are briefly reviewed next:
(i) Discretisation. Beam mid-surface discretised into fine S4 element
(Abaqus nomenclature − 4-node isoparametric shell element with
the shear stiffness obtained by full integration)3 meshes.
(ii) Support Conditions. The SCA beams were modelled by imposing
null transverse displacements at the end cross-sections. In order to
avoid numerical difficulties related to the load application, both
end cross-sections are free to move axially (the axial rigid-body
translation is precluded by preventing the axial displacement of
the mid-span mid-web point). The SCB beams have their end cross-
sections attached to rigid plates,4 ensuring full warping and local
displacement/rotation restraints, whose external surfaces rest on
spherical hinges prevented from twisting [25]. Thus, the beam end
cross-sections are locally fixed and globally pinned (simply sup-
ported).
(iii) Loading. Equal major-axis (C and HM beams), minor-axis (Hm
beams) and skew (Z beams) bending moments applied at the end
sections, either (i) via nodal concentrated forces statically
equivalent to 1 kN cm (SCA beams) or (i2) directly on the rigid end
plate centroids (SCB beams).
(iv) Material Model. The carbon steel material behaviour, deemed iso-
tropic and homogeneous, is modelled as (i) linear elastic (buckling
analysis) or (ii) elastic-perfectly plastic (post-buckling analysis). In
the latter case, the well-known Prandtl-Reuss model is adopted – it
is based on J2-flow plasticity theory and combines von Mises's
yield criterion with its associated flow rule.
3.2. Initial geometrical imperfections
A crucial issue in investigations involving buckling mode interaction
is the identification of the most detrimental initial geometrical im-
perfection shape, i.e., that leading to the lowest beam strength. In this
regard, it is worth noting that a critical-mode imperfection study has
already been conducted, in [7], for a C+SCA beam with RDL = 1.00 −
it consisted of analysing the post-buckling behaviour of 24 beams
containing different linear combinations of adequately normalised
“pure” local and distortional buckling modes, and sharing the same
overall amplitude (0.1t). This study showed that, for beams with local
buckling triggered by the compressed flange, the most detrimental
3
Note that the mesh discretisation appearing in the ABAQUS beam deformed config-
urations shown in this document (Figs. 2–5) is a bit coarse – this was done to improve the
readability of these figures and also to reduce the document size (they are in eps format).
The actual meshes employed have elements with a length-to-width ratio roughly equal to
1.0.
4
In Figs. 2–5, showing ABAQUS deformed configurations of SCB beams, the attached
end plates are not displayed (although they were, obviously, considered in the simula-
tions). This was done to improve the “readability” of the figures along the beam length.
881
Thin-Walled Structures 119 (2017) 879–901
initial geometrical imperfection shape is the “pure” local one.5 This
same procedure was adopted for the remaining six beam types (C+SCB,
HM+SCA, HM+SCB, Z+SCA, Z+SCB, Hm + SCB) and the same output
was obtained: the “pure” local initial geometrical imperfections are the
most detrimental ones. Note, however, that this conclusion was drawn
from the analysis of beams with RDL = 1.00 and may not remain true
for beams with lower RDL values. Therefore, it was decided to analyse
also beams with RDL < 0.90 that contain “pure” distortional initial
geometrical imperfections involving (i) inward flange lip-motions (SCA
and SCB C, HM and Z-beams) or (ii) outward flange lip-motions (SCB
Hm-beams) (see [24]). For the remaining beams, “pure” local initial
geometrical imperfections were always considered. The amplitude of all
the initial geometrical imperfections considered in this work is equal to
0.1t.
3.3. Post-buckling behaviour of beams experiencing L-D interaction
3.3.1. True L-D interaction
Figs. 2(a1)-(a2) show the Mcr vs. (v+v0)/t post-buckling equilibrium
paths (v is the mid-span compressed/top flange-lip corner vertical dis-
placement and v0 the corresponding initial value) of RDL = 1.0 Z-beams
(Z24+SCA and Z33+SCB) with 4 critical slenderness [λcr=(My/
Mcr)0.5] values: 1.0, 2.5, 3.5 and ∞ – the last one stands for elastic
behaviour. On the other hand, Figs. 2(b1)-(b2) display beam deformed
configurations and plastic strain diagrams at the onset of collapse
concerning the λcr=1.0 and λcr=3.5 beams. The observation of these
results prompts the following remarks:
(i) As reported in [24], the SCA and SCB beam post-buckling beha-
viours are clearly distinct: while the former exhibits a fairly rapid
stiffness erosion/degradation until an elastic limit point is reached
(no elastic-plastic strength reserve), the latter involves a much
slower stiffness erosion/degradation and is associated with a re-
levant elastic-plastic strength reserve.
(ii) Despite the different behaviours associated with the two support
conditions, Fig. 2(b) shows that, regardless of the beam yield stress
(see Figs. 2(b1)-(b2) – Z-beams with λcr = 1.0; 3.5), both local and
distortional deformations are clearly visible in the compressed
flange at the onset of collapse, thus evidencing the occurrence of L-
D interaction. As in columns [3–5], the emergence of local and
distortional deformations at early loading stages characterises “true
L-D interaction”, causing relevant failure moment erosion (see
Section 4) and occurring for RDL≈1.0 – these assertions also apply
to SCA or SCB HM and C beams (the Hm beams are addressed se-
parately, in Section 3.3.4).
5
A similar study, for a RDL = 1.00 beam with local buckling triggered by the web (not
dealt with in this work) was also reported in [7].
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 2. RDL=1.0 Z-beam (a) elastic and elastic-plastic (λcr = 1.0, 2.5, 3.5) M/Mcr vs. (v+v0)/t equilibrium paths and (b) deformed configurations at the onset of collapse of λcr=1.0 and
λcr=3.5 beams for (1) SCA and (2) SCB.

Fig. 3. RDL=0.5HM beam (a) elastic and elastic-plastic (λD = 1.0, 2.5, 3.5) M/McrD vs. (v+v0)/t equilibrium paths and (b) deformed configurations at the onset of collapse of the λD = 2.5
and λD = 3.5 beams for (1) SCA and (2) SCB.

Fig. 4. RDL=2.0 C beam: (a) elastic and elastic-plastic (λL = 1.25, 2.5, 3.5) M/McrL vs. (v+v0)/t equilibrium paths and (b) deformed configurations at the onset of collapse of the λL =
1.25, λL = 2.5 beams, for (1) SCA and (2) SCB.

3.3.2. Secondary local bifurcation L-D interaction failing in distortional modes.

The post-buckling equilibrium paths McrD vs. (v+v0)/t displayed in
Figs. 3(a1)-(a2) concern RDL= 0.5HM beams (HM34+SCA and
HM23+SCB) with 4 distortional slenderness [λD = (My/McrD)0.5] va-
lues: again 1.0, 2.5, 3.5 and ∞. As for Figs. 3(b1)-(b2), they show the λD
= 2.5 and λD = 3.5 beam deformed configurations and plastic strain
diagrams at the onset of collapse. From the observation of these results
it is possible to draw the following conclusions:
(iii) However, the most slender (λD = 3.50) beam exhibits L-D inter-
(i) Generally speaking, the post-buckling results follow the trends
already reported in [21], in the context of uniformly bent beams
(ii) The beams with λD < 3.0 (see Fig. 3(b) for λD = 2.5) fail with no
trace of local deformations, i.e., no L-D interaction occurs before
failure. Indeed, their deformed configurations at the onset of col-
lapse exhibit virtually “perfect” distortional single half-waves (akin
to the critical distortional buckling mode) – indeed, the failure
modes is characterised by the presence of a “distortional plastic
hinge”, typically encountered in “pure” distortional collapses [21].
action, since local deformations develop mainly near the end cross-
sections, even if they are barely visible (see the deformed top

882
A.D. Martins et al.
Fig. 5. RDL=1.7 C+SCA-beams (a) elastic and elastic-plastic (λL = 1.5, 1.75, 3.0, 3.5) M/McrL vs. (v + v0)/t equilibrium paths and (b) deformed configurations of the λL = 1.75 and λL =
3.0 beams at stages I and II.
flanges in Figs. 3(b1)-(b2)). These deformations only emerge at
advanced loading stages, a feature that, like in columns [3–5],
characterises “secondary local bifurcation L-D interaction”. In this
type of L-D interaction the amplitude of the local deformations
remains quite small, even for high yield stresses. For this particular
RDL value, L-D interaction only develops if λD falls between 3.0 and
3.50 – once again, these conclusions apply also to the other beams
analysed (C and Z beams).
3.3.3. Secondary distortional bifurcation L-D interaction
Finally, Figs. 4(a1)-(a2) plot the McrL vs. (v+v0)/t post-buckling
equilibrium paths of RDL = 2.0 C beams (C43+SCA and C43+SCB)
with 4 local slenderness [λL=(My/McrL)0.5] values: 1.0, 2.5, 3.5 and ∞.
As before, the beam deformed configurations and plastic strain dia-
grams, at the onset of collapse, are shown in Figs. 4(b1)-(b2), for the
λL= 1.25 and λL= 2.5 beams. For these beams, it should be noted that:
(i) Once again, the SCA beam elastic equilibrium path exhibits a limit
point − in this particular case, this limit point is followed by a
minute “snap-back” phenomenon.
(ii) Surprisingly, the stockier beams (λL=1.25 − Fig. 4(b)) show dis-
tortional deformations at the onset of collapse. As discussed in
[18], this “unexpected” feature stems from the occurrence of
flange curling (e.g., [26–28]), which causes, since the early loading
stages, symmetric distortional deformations exhibiting a single
half-wave (like the major-axis bending deformations due to the
applied moment). Since the distortional critical buckling mode also
exhibits a single half-wave (nD=1) in these beams, it is virtually
impossible (with SFEA) to detect for which applied loading level
the distortional deformations due to L-D interaction come to into
play. Thus, in order to illustrate the development of “secondary
distortional bifurcation L-D interaction”, it is necessary to analyse
a beam with nD > 1. So, Fig. 5(a) shows the equilibrium path of a
RDL = 1.70+SCA beam with bw = 130, bf = 80, bl = 20, t =
0.963 and L = 1500 mm, which has nD = 2. As for Fig. 5(b), it
displays several beam deformed configurations. The observation of
these post-buckling results shows that:
(ii.1) The λL=1.75 beam exhibits several local half-waves to-
gether with a single distortional half-wave (see Fig. 5(b1)),
thus confirming that this beam is not affected by L-D inter-
action − since nD=2 and the half-wave single distortional
buckling moment (MbD.1) is higher than the critical one
(MbD.1/McrD.2=1.13). Moreover, the collapse mechanism
involves mostly yielding of the compressed/top flange-web
and flange-lip edges, typical of local failures – however, a
883
Thin-Walled Structures 119 (2017) 879–901
pronounced distortional half-wave is also visible.6 Therefore,
this beam does not exhibit a “pure” local collapse, due to the
heavy flange curling deformations, which also explains why
the currently codified DSM local strength curve is unable to
predict accurately the stocky beam failure moments (see
Section 4).
(ii.2) Conversely, the two post-buckling deformed configurations
displayed in Fig. 5(b2)-(b3), concerning the λL = 3.0 beam,
show that the distortional deformation pattern changes,
from one half-wave (due to flange curling) to two half-waves
(due to interaction with distortional buckling), along the
equilibrium path. This feature clearly indicates the occur-
rence of “secondary distortional bifurcation L-D interaction”.
In this case, the collapse mechanism involves yielding in the
compressed/top flange-lip edge, typical of distortional fail-
ures - recall (e.g. [18]) that distortional deformations prevail
in all the types of L-D interaction including the extreme case
of the "secondary distortional bifurcation L-D interaction".
(ii.3) Although slender beams with nD = 1 (e.g., the λL = 2.5
beam – deformed configurations displayed in Fig. 4(b)) are
also affected by “secondary distortional bifurcation L-D in-
teraction”, it is difficult (or even impossible) to distinguish
between the distortional deformations caused by flange
curling and L-D interaction by means of SFEA results.
(iii) “Secondary distortional bifurcation L-D interaction”, stemming
from the high local post-critical strength and large yield stresses,
occurs despite the substantial difference between McrL and McrD.
This implies the existence of a “border” separating L and L-D col-
lapses that depends on the values of RDL and My. Note that the
“amount” of distortional deformation appearing in this beam is
much larger than that of the local deformation occurring in the RDL
= 0.5HM beam (for the same λcr value). Obviously, this is due to
the different post-critical strengths associated with the critical bi-
furcation mode, which justifies the fact that higher ultimate
strength erosion takes place for beams undergoing “secondary
distortional bifurcation L-D interaction” (see Section 4).
3.3.4. Hat-section beams under minor-axis bending
This section deals with hat-section SCB beams acted by uniform
minor-axis bending (Hm+SCB beams) and covers all the topics ad-
dressed in Sections 3.3.1 to 3.3.3. Fig. 6(a) shows elastic and elastic-
6
At this stage, it is worth recalling that the flange curling and symmetric distortional
buckling deformation patterns are identical.
A.D. Martins et al.
Fig. 6. Hm+SCB beam (a) elastic and elastic-plastic (λcr = 1.0, 2.5, 3.5) M/Mcr vs. (v+v0)/t equilibrium paths and (b) λcr=1.0 and λcr=3.5 beam deformed configurations at collapse for
(1) RDL = 1.0, (2) RDL = 0.5, (3) RDL = 2.0.
plastic (λcr = 1.0, 2.5, 3.5) post-buckling equilibrium paths M/Mcr vs.
(v+v0)/t of three beams: (i) RDL= 1.0Hm33+SCB, (ii)
RDL=0.5Hm23+SCB and (iii) RDL=2.0Hm43+SCB beams. As for
Fig. 6(b), it shows deformed configurations of the λcr=1.0 and 3.5
elastic-plastic beams (amplified 5 and 2 times, respectively). The ob-
servation of all these results prompts the following comments:
(i) Except for the λcr = 1.0 beams, all the elastic-plastic equilibrium
paths depicted in Fig. 6(a) exhibit a failure limit point in the elastic
range, which occurs after a continuous and smooth stiffness ero-
sion/degradation that may start before or after the critical buckling
moment level has been reached – the onset of yielding occurs al-
ready in the descending branch. The strength erosion is due to the
development of the “secondary local or distortional bifurcation L-D
interaction” addressed in Sections 3.3.2 and 3.3.3 (see also items
(iii) and (iv) below). Moreover, the response of these beams is,
qualitatively, significantly different from those exhibited by the
SCB beams analysed earlier, which were either bent about the
major-axis or subjected to skew bending.
(ii) Like the beams analysed in Section 3.3.1, the RDL=1.0Hm33+SCB
beams are also affected by “true L-D interaction”, since clear local
and distortional deformations are visible in the flanges at the onset
884
Thin-Walled Structures 119 (2017) 879–901
of collapse (see Fig. 6(b1)). Although this coupling effect is
stronger in the λcr = 3.5 beam, both deformation types are already
perceptible in the stocky beam (λcr = 1.0).
(iii) On the other hand, the RDL=0.5Hm23+SCB beam exhibits a clear
“pure” distortional failure, i.e., no L-D interaction occurs. In fact, unlike
in the beams analysed in Section 3.3.2, no trace of local deformations
was found in the slender beams (e.g., the λcr = 3.5 beam in Fig. 6(b2)).
In this case, “secondary local bifurcation L-D interaction” does not
develop, because of the relevant strength degradation occurs well be-
fore the critical distortional buckling moment level is reached.
(iv) The behaviours of the RDL=2.0Hm43+SCB beams are qualitatively
similar to those presented in Section 3.3.3, i.e., (iv1) the stocky
beams (e.g., see the λcr = 1.0 beam in Fig. 6(b3)) fail in modes
exhibiting only local deformations, and (iv2) the slender beams
(e.g., see the λcr=3.5 beam in Fig. 6(b3)) show perceptible (but
small) distortional deformations at the onset of collapse, which
characterises “secondary distortional bifurcation L-D interaction”.
However, the distortional deformations are much smaller in the
Hm-beams than in the beams acted by major-axis/skew bending
analysed earlier – once again due to a premature strength loss
occurring in the elastic range (now after the critical buckling
moment level has been reached).
A.D. Martins et al.
4. Direct Strength Method (DSM) design
The Direct Strength Method (DSM), first developed by Schafer and
Peköz [29] based on an original idea from Hancock et al. [30], may be
viewed as an approach/procedure to establish methodologies for the
design of a wide variety of limit states – it should be noted that the
above pioneering work [29] included the application to cold-formed
steel beams. This versatility explains the worldwide popularity of the
method for the design of cold-formed steel members (but not limited to
them – see, for instance [31,32]). A major advantage of the DSM, when
compared to the traditional “Effective Width Method”, is the possibility
of taking explicitly into account buckling mode interaction. In this re-
gard, two strategies, based on an approach adopted for columns un-
dergoing L-D interaction (similar to an idea put forward by Schafer
[33]), can be followed: (i) the NLD approach (MNLD), already employed
in [6] and involving the replacement of My by MND,7 and (ii) the NDL
approach (MNDL), which consists of replacing My by MNL8 – one then
obtains, respectively,9
MND , λLD ≤ 0.776
MNLD = ⎧ −0.8 −0.8
⎨
⎩ (1 − 0.15λLD ) λLD MND , λLD > 0.776 (1)
MNL , λDL ≤ 0.673
MNDL = ⎧ −b −c
⎨ (1 − aλDL ) λDL MNL , λDL > 0.673 (2)
⎩
where λLD=(MND/McrL) is the local slenderness based on the distor-
0.5
tional strength and, similarly, λDL=(MNL/McrD)0.5 is the distortional
slenderness based on the local strength. As for the values of the para-
meters a, b and c, they can be found in [24]. The main goal of this
section is to provide a valid “path” to develop a rational DSM-based
design approach to handle L-D failures, thus extending the currently
available expressions (MN=min{MNLG; MND; MNG }) to MN=min{MNLG;
MND; MNG; MNLD}.
The results presented concern a representative sample of the beams
selected/identified in Section 2 − due to space limitations, it is not
feasible to present all the results obtained. However, this sample of
results, presented in Figs. 7–14, provides enough information to enable
assessing the quality of the ultimate strength estimates in beams ex-
hibiting the various RDL values considered. Specifically, it consists of
beams with RDL= 0.50–0.70–0.85–1.00–1.30–1.60–1.80–2.00: Figs. 7-
8, 9-10, 11-12 and 13-14 concern the C, HM, Z (both for SCA and SCB)
and Hm (only SCB) beams, respectively. They plot MU /My against the
critical slenderness (λL or λD) for the above 8 beam sets – obviously, the
two beams in each figure do not share the same geometry, i.e., they
have different web-to-flange and flange-to-lip width ratios (this issue is
discussed next). For each RDL value, the numerical MU /My values are
compared with their DSM predictions, namely (i) MNL or MND (design
curves currently codified and proposed in [24]) and (ii) MNDL, ac-
counting for L-D interaction – the MNLD curve was excluded from this
study, since it yields ultimate strength predictions very close to their
MNDL counterparts (and also to improve the readability of the figures10).
Nevertheless, the accuracy of the two approaches will be assessed and
7 local bifurcation L-D interaction” case (see item (iii) above), this
The currently codified DSM distortional curve is replaced by the curves proposed in
[24], since they were found to yield much more accurate bending strength predictions.
8
Note that the MNL curve prescribed by the current North-American specification [34]
is, in fact, the MNLG (local-global interaction). In this work, the MNL curve, concerning
“pure” local failures, is adopted. This curve is obtained by replacing Mne with My in the
MNLG expression appearing in [34] – since McrG is much higher than McrL (and McrD – see
Section 2 and Annex A), the differences between the MNL and MNLG curves are very small
for slender beams.
9
Both approaches include the additional plastic strength reserve (local and distor-
tional), based on the work of Shifferaw and Schafer [35], prescribed by the current North-
American specification [34] and dependent on the plastic moment. The determination of
the Z-section plastic moments (skew bending) was made following the procedure de-
scribed by Dwight [36] (the remaining cases are straightforward).
10
The exception are the Hm-beams, for which both the MNDL and MNLD curves are
depicted (only SCB beams were analysed).
885
Thin-Walled Structures 119 (2017) 879–901
compared later (in Fig. 16). The comparisons make it is possible to
extract several conclusions, which follow very similar trends for the
various cross-section types:
(i) First of all, as it would be logical to expect, all the numerical MU
/My values are well aligned along “Winter-type” curves.
(ii) Based on the results reported in [24], the currently codified DSM
distortional curve provides very substantial failure moment over-
estimations for both SCA and SCB beams with any type of cross-
section (see Figs. 7–14).
(iii) On the other hand, and generally speaking, the MND strength curve
proposed in [24] provides accurate ultimate strength estimates for
beams with RDL « 1.0 (e.g., RDL=0.50; 0.70 – see Figs. 7(a1)-(a2),
9(a1)-(a2), 11(a1)-(a2) and 13(a1)-(a2)). However, this accuracy is
heavily dependent on the flange-lip width ratio (bf /bl) and critical
distortional half-wave number,11 as unveiled in [24]. For instance,
the comparison between the MU /My values of the RDL = 1.0 beams
C1+SCB (bf /bl = 9.0) and C20+SCB (bf /bl = 6.7) shows that the
former (larger bf /bl ratio) are always higher. Indeed, the vast
majority of these beams fail in pure distortional modes, as only the
most slender ones exhibit minute local deformations, stemming
from “secondary local bifurcation L-D interaction” (see Fig. 3(b)) –
this means that there is virtually no ultimate strength erosion due
to this type of L-D interaction. As RDL increases up to 0.85
(Figs. 7(a3), 9(a3), 11(a3) and 13(a3)), the proposed MND strength
curve gradually ceases to provide safe ultimate strength estimates
for the slender beams, which exhibit visible local deformations
prior to collapse. This happens because such beam sequence cor-
respond to the transition between beams experiencing no L-D in-
teraction and beams experiencing first “secondary local bifurcation
interaction L-D interaction” (RDL=0.70) and then “true L-D inter-
action” (RDL=0.85). However, the failure moments of all the
beams affected by L-D interaction (either “secondary” or “true”)
are quite accurately predicted by the MNLD (and MNDL) values.
(iv) Naturally, neither the currently codified DSM NL nor the proposed
DSM ND strength curves are able to capture the ultimate strength
erosion exhibited by all the RDL=1.00 beams (see Figs. 7(a4)-(b4),
9(a4)-(b4), 11(a4)-(b4) and 13(a4)-(b4)) – this is because “true L-D
interaction” occurs along the whole slenderness range (see also
Fig. 2(b)). Once again, the NLD (and NDL) strength curve predicts
accurately the ultimate strength erosion inherent to this coupling
phenomenon.
(v) The currently codified MNL curve only provides accurate ultimate
strength estimates for stocky beams with RDL > 1.00 − the
number of accurate predictions grows (slowly) with RDL and λL,
i.e., as the beam collapse occurs progressively in modes exhibiting
more dominant local deformations (until there is virtually no trace
of L-D interaction) – see Figs. 7(a1)-(a4), 9(a1)-(a4), 11(a1)-(a4) and
13(a1)-(a4). On the other hand, the MNL (and MND – not shown in
those plots) design curve overestimates the failure moments of the
RDL=1.30–1.60–1.80–2.00 slender beams, thus providing clear
evidence of the occurrence of L-D interaction, in this case due to
“secondary distortional bifurcation”. Unlike in the “secondary
type of L-D interaction cannot be ignored in design. For such
beams the proposed DSM NLD (and NDL) approach also provides
high quality failure moment predictions.
(vi) There is a strong qualitative resemblance between the responses of
beams and columns affected by L-D interaction, which can be
summarised as follows. First, the amount of overestimation of the
11
Although this influence is negligible in the present work, since the critical distor-
tional half-wave number is equal to one in the overwhelming majority of the beams
analysed.
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 7. MU /My vs. (a) λL or (b) λD plots for C beams with (1)-(4) RDL=0.50–0.70–0.85–1.00 (SCA and SCB).

RDL < 1.00 beam failure moments, by the MND values, gradually switch from distortional failures to L-D interactive ones, first due to
grows as RDL and/or λD increase, because the L-D interaction ef- “secondary local bifurcation L-D interaction” and then to “true L-D
fects become progressively more significant. Indeed, there is a interaction”. Naturally, the number of accurate failure moment

886
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 8. MU /My vs. (a) λL or (b) λD plots for C-beams with (1)-(4) RDL=1.30–1.60–1.80–2.00 (SCA and SCB).

estimates, indicating distortional failures, grows as RDL increases. increases and λL decreases, because the L-D interaction effects
On the other hand, the amount of overestimation of the RDL > 1.00 become gradually less relevant − naturally, the number of accu-
beam failure moments, by the local design curve, diminishes as RDL rate estimates, indicating local failures, grows (slowly) with RDL.

887
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 9. MU /My vs. (a) λL or (b) λD plots for HM-beams with (1)-(4) RDL=0.50–0.70–0.85–1.00 (SCA and SCB).

888
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 10. MU /My vs. (a) λL or (b) λD plots for HM-beams with (1)-(4) RDL=1.30–1.60–1.80–2.00 (SCA and SCB).

889
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 11. MU /My vs. (a) λL or (b) λD plots for Z-beams with (1)-(4) RDL=0.50–0.70–0.85–1.00 (SCA and SCB).

890
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 12. MU /My vs. (a) λL or (b) λD plots for Z-beams with (1)-(4) RDL=1.30–1.60–1.80–2.00 (SCA and SCB).

891
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 13. MU /My vs. (a) λL or (b) λD plots for Hm-beams+SCB with (1)-(4) RDL=0.50–0.70–0.85–1.00.

892
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 14. MU /My vs. (a) λL or (b) λD plots for Hm-beams+SCB with (1)-(4) RDL=1.30–1.60–1.80–2.00.

893
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 15. Plots (a) MU /MNL vs. λL and (b) MU /MND vs. λD for 0.50≤ RDL≤ 2.00 (1) C, (2) HM, (3) Z, (4) Hm beams.

894
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Fig. 16. Plots of (a) MU /MNLD and (b) MU /MNDL vs. λD for 0.50≤ RDL≤ 2.00 (1) C, (2) HM, (3) Z, (4) Hm beams.

Indeed, there is a switch from “true L-D interaction” to “secondary
distortional bifurcation L-D interaction”, first, and to “pure” local
behaviour, later.
On the basis of the simply supported beam numerical failure
moments obtained in this work, the quality of their DSM-based pre-
dictions is assessed next, with the help of the plots (i) MU /MNL vs. λL
(Figs. 15(a1)-(a4)), (ii) MU /MND (curve from [24]) vs. λD (Figs. 15(b1)-
(b4)), (iii) MU /MNLD vs. λD (Figs. 16(a1)-(a4)) and (iv) MU /MNDL vs. λD
(Fig. 16(b1)-(b4)) – Figs. (1), (2), (3) and (4) correspond to the C, HM, Z

895
A.D. Martins et al.
and Hm beams, respectively. Note that (i) the C+SCA results reported
in [6]12 are also shown (the grey squares), and (ii) Fig. 15(b4), con-
cerning the Hm-beams, depicts also the currently codified MU /MND vs.
λD curve. The observation of these plots prompts the following com-
ments:
(i) The results reported in [6], concerning C+SCA beams in the small-
to-moderate slenderness range, follow the same trends as those
reported here – see Figs. 15(a1)-(b1) and 16(a1)-(b1).
(ii) The numerical-to-predicted failure moment ratio “clouds” (MU
/MNL, MU /MND, MU /MNLD and MU /MNDL) are very similar in the
vast majority of the eight pairs of plots displayed in Figs. 15 and
16. The exceptions are the MU /MND ratios of the SCA and SCB C
beams, which exhibit much better quality than those concerning
the other beam types considered – see the statistical indicators
presented in the above figures.
(iii) The pairs of plots in Figs. 15(a1)-(b4) clearly show that the cur-
rently codified DSM local and distortional strength curves are
unable to predict adequately the ultimate strength erosion due to
L-D interaction – the only fairly accurate MNL estimates concern
stocky beams with RDL > 1.00, failing in “pure” local modes.13,14
Indeed, both sets of predictions exhibit very poor indicators: mean
values well below 1.0 and very large standard deviations. Although
the indicators of the distortional design curve proposed in [24] are
visibly better, this curve still cannot cope with L-D interaction –
naturally, the most accurate estimates concern now stocky beams
with RDL < 1.00, failing in “pure” distortional modes. As was
clearly shown in [24], the currently codified DSM distortional
strength curve cannot predict adequately the failure moments of
the simply supported beams analysed in this work. For instance,
Fig. 15(b) clearly illustrates the poor performance of this design
curve, even for SCB beams.
(iv) The MNDL and MNLD approaches provide the best failure moment
estimations, an assertion that can be confirmed by merely looking
at Fig. 16 – both of them exhibit similarly good accuracy (see the
corresponding statistical indicators given in this figure). Never-
theless, better predictions are found for the SCA beams, due to the
less pronounced dependence on the flange-lip width ratio (bf /bl),
as reported in [24] – recall that none of the DSM approaches de-
pends on this ratio. It is still worth noting that all overestimations
concern beams with high flange-lip width ratios. Moreover, the
comparison between the ultimate strength predictions provided by
the DSM approaches developed in this work and that recently
proposed in [8], concerning exclusively RDL≈1.0C+SCA beams,
shows clearly that the former are much more accurate and much
less scattered. Moreover, it should be mentioned that the scatter of
the failure moments reported in [8] is probably due to the fact that
local buckling is web-triggered in part of the beams analysed (and
not always flange-triggered, as in this work) – it was clearly shown
in [7] that the post-buckling behaviour and strength are different,
both qualitatively and quantitatively, in the two cases.
(v) Based on the results obtained in this work, it appears that is not
necessary to identify a “border” beyond which L-D interaction
ceases to be relevant in RDL > 1.0 beams15 (i.e., the transition from
local failures to failures due to “secondary distortional bifurcation
12
These results concern simply supported lipped channel beams with RDL values
comprised between 0.85 and 1.20. However, 30 beams (out of 90) were excluded because
local bucking is triggered by the web (situation outside of the scope of this work).
13
This same conclusion was reached by the authors in the context of columns ex-
periencing L-D interaction (e.g.,[3–5]). Nevertheless, the quality of the indicators ob-
tained for the C beams is higher than that reported for the C columns − see item (ii)
above.
14
The occurrence of flange curling deformations precludes a better performance of the
MNL strength curve.
15
For columns, the authors recently showed that such identification is essential to
obtain accurate predictions of failure loads associated with L-D interactive modes [37].
896
Thin-Walled Structures 119 (2017) 879–901
L-D interaction”). Indeed, both the NLD and NDL approaches yield
good failure moment estimates (see also Figs. 8, 10, 12, and 14).
However, although both approaches provide similarly accurate
results for 0.50≤ RDL ≤2.0, the NDL approach should be adopted.
This is because the relevance of “secondary distortional bifurcation
L-D interaction” (affecting only the most slender beams) becomes
negligible as RDL increases well above 2.0, i.e., when the beam
behaviour tends to be “purely” local and, therefore, is perfectly
captured by the NDL approach – in such beams, λDL is always quite
small (below 0.673) and MNDL coincides with MNL (see Eq. (2)).
5. Conclusion
A numerical investigation on the influence of L-D interaction on the
post-buckling behaviour, ultimate strength and DSM design of cold-
formed steel simply supported uniformly bent beams was reported. The
beams analysed exhibited (i) three cross-section shapes (lipped chan-
nels bent about the major-axis, hats bent about the major or minor-axis
– lips compressed in the latter case – and zeds under skew bending
causing uniform flange compression) and (ii) two boundary conditions.
Initially, a beam geometry selection procedure was presented and
employed to identify simply supported beams exhibiting a wide variety
of ratios between (i) the distortional and local buckling moments, and (ii)
the yield moment and the non-critical (local or distortional) buckling
moment. Next, the elastic-plastic post-buckling behaviours of beams af-
fected by different types of L-D interaction were investigated, with the
purpose of shedding light on the underlying mechanics. Then, an ex-
tensive parametric study was carried out, by means of materially and
geometrically non-linear ABAQUS analyses, intended to gather failure
moment data concerning beams experiencing various types and “levels”
of L-D interaction. These data was subsequently used, in the last part of
the paper, to develop and assess the merits of DSM-based design ap-
proaches devised to handle failures involving this coupling phenomenon.
The comparison between the numerical failure moments and their DSM
estimates made it possible to draw some meaningful conclusions.
Amongst the various findings reported in this work, the following
deserve to be highlighted:
(i) There is a strong qualitative resemblance between the L-D coupling
effects in columns [3–5] and beams: two markedly different types
of L-D interaction were unveiled – “true L-D interaction” and
“secondary (local or distortional) bifurcation interaction”. The
former is associated with McrD /McrL ratios close to 1.0 and is
characterised by the emergence of both local and distortional de-
formations at early loading stages. The latter type is associated
with McrD /McrL values well above (“secondary distortional bi-
furcation”) or below (“secondary local bifurcation”) 1.0 – in such
beams, deformations patterns akin to the non-critical buckling
mode only emerge at relatively advanced loading stages, provided
that the yield stress is “high enough” to allow for it. Thus, the
above observations may be viewed as inherent to the coupling
phenomenon under consideration (L-D interaction) and remain
valid regardless of the particular combination of loading (columns,
beams or beam-columns) and support conditions – the latter “only”
influence the relevance of the L-D interaction effects.
(ii) The elastic and elastic-plastic post-buckling behaviours of SCA and
SCB beams are clearly distinct. Although is not possible to compare
directly these beams (or other members, for that matter) having
the same geometry and interaction level (i.e., McrD /McrL) and
different support conditions, some conclusions may still be drawn:
compared to the SCA beams, the SCB ones are (obviously) stiffer,
exhibit higher MU /Mcr and MU /My ratios (thus leading to much
more pronounced L-D interaction effects) and have significant
elastic-plastic strength reserve – the SCA beams often exhibit
elastic limit points or a minute elastic-plastic strength reverse.
(iii) As expected, the DSM NL and ND (currently codified or proposed in
A.D. Martins et al.
[24]) design curves cannot predict adequately the beam ultimate
strength erosion caused by L-D interaction. Nevertheless, the DSM
distortional strength curve proposed in [24] provides accurate es-
timates for beams experiencing L-D interaction due to a “secondary
local bifurcation” – since the local deformations are always minute,
no sizeable ultimate strength erosion occurs, which implies that this
type of L-D interaction can be ignored for design purposes.
(iv) The two DSM design approaches developed to account for L-D
interaction (NDL and NLD) were shown to handle adequately
simply supported beams with the three cross-section shapes and
two support conditions considered. The comparison between the
fairly extensive set of numerical failure moments gathered and
their estimates made it possible to conclude that both approaches
consistently provide good predictions for all the beams analysed,
which undergo different types and “levels” of L-D interaction –
these predictions are much more accurate and less scattered than
those recently reported in [8]. However, it should be noted that the
accuracy of the estimates is heavily dependent on the beam flange-
lip width ratio and critical distortional half-wave number, as
shown in [24] for beams failing in “pure” distortional modes (al-
though safe estimates are almost always obtained). This means that
the quality of such estimates could be even higher if these two
parameters were included in the DSM-based expressions – recall
that all the DSM expressions/approaches depend solely on the
ANNEX A. Data concerning the selected beams
See Annex Tables A1–A7 here.
Table A1
Selected lipped channel beams under major-axis bending with SCA: geometry, buckling moments and their relevant moment ratios (dimensions in mm and moments in kNcm).
897
Thin-Walled Structures 119 (2017) 879–901
slenderness (in this case, on both λD and λL), like in the traditional
“Effective Width Method”. Moreover, note also that the use of the
ND curves proposed in [24] is essential for the success of the
proposed DSM-based design approaches – basing them on the
currently codified ND curve leads to inaccurate predictions.
It should be underlined that all the beams analysed in this work
exhibit flange-triggered local buckling (the most common situation in
practice). As shown in [7], significantly different results would be ob-
tained for beams exhibiting web-triggered local buckling – this topic is
under investigation by the authors. Finally, one last word to mention
that the authors are also currently working on extending the results
reported in this work to beams subjected to non-uniform bending, a
loading condition much more common in practice. On the other hand,
an experimental program involving tests on lipped channel beams with
carefully selected geometries, to ensure very clear L-D interaction, is
also planned for the not too distant future.
Acknowledgements
The first author gratefully acknowledges the financial support of
FCT (Fundação para a Ciência e a Tecnologia – Portugal) through the
doctoral scholarship SFRH/BD/87746/2012.
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Table A2
Selected lipped channel beams under major-axis bending with SCB: geometry, buckling moments and their relevant moment ratios (dimensions in mm and moments in kNcm).

Table A3
Selected hat-section beams under major-axis bending with SCA: geometry, buckling moments and their relevant moment ratios (dimensions in mm and moments in kNcm).

898
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Table A4
Selected hat-section beams under major-axis bending with SCB: geometry, buckling moments and their relevant moment ratios (dimensions in mm and moments in kNcm).

Table A5
Selected zed-section beams under skew bending (uniformly compressed top flange) with SCA: geometry, buckling moments and their relevant moment ratios (dimensions
in mm and moments in kNcm).

899
A.D. Martins et al. Thin-Walled Structures 119 (2017) 879–901

Table A6
Selected zed-section beams under skew bending (uniformly compressed top flange) with SCB: geometry, buckling moments and their relevant moment ratios (dimensions in mm
and
moments in kNcm).

Table A7
Selected hat-section beams under minor-axis bending with SCB: geometry, buckling moments and their relevant moment ratios (dimensions in mm and moments in kNcm).

900
A.D. Martins et al.
References
[1] Y.B. Kwon, G.J. Hancock, Tests of cold-formed channels with local and distortional
buckling, J. Struct. Eng. (ASCE) 118 (7) (1992) 1786–1803.
[2] B. Young, N. Silvestre, D. Camotim, Cold-formed steel lipped channel columns in-
fluenced by local-distortional interaction: strength and DSM design, J. Struct. Eng.
(ASCE) 139 (6) (2013) 1059–1074.
[3] A.D. Martins, P.B. Dinis, D. Camotim, P. Providência, On the relevance of local-
distortional interaction effects in the behaviour and design of cold-formed steel
columns, Comput. Struct. 160 (November) (2015) 57–89.
[4] A.D. Martins, P.B. Dinis, D. Camotim, On the influence of local-distortional inter-
action in the behaviour and design of cold-formed steel web-stiffened lipped
channel columns, Thin-Walled Struct. 101 (April) (2016) 181–204.
[5] A.D. Martins, D. Camotim, P.B. Dinis, Behaviour and DSM design of stiffened lipped
channel columns undergoing local-distortional interaction, J. Constr. Steel Res. 128
(January) (2016) 99–118.
[6] N. Silvestre, P.B. Dinis, D. Camotim, Direct strength method for lipped channel
columns and beams affected by local-plate/distortional interaction, in: W.-.W. Yu,
R. Laboube (eds.), Proceedings of 18th International Specialty Conference on Cold-
Formed Steel Structures, Orlando, (26-27/10), pp. 17–37, 2006.
[7] P.B. Dinis, D. Camotim, Local/distortional mode interaction in cold-formed steel
lipped channel beams, Thin-Walled Struct. 48 (10–11) (2010) 771–785.
[8] M. Anbarasu, Local-distortional buckling interaction on cold-formed steel lipped
channel beams, Thin-Walled Struct. 98B (January) (2016) 351–359.
[9] M.R. Haidarali, D.A. Nethercot, Finite element modelling of cold-formed steel
beams under local buckling or combined local/distortional buckling, Thin-Walled
Struct. 49 (12) (2011) 1554–1562.
[10] M.R. Haidarali, D.A. Nethercot, Local and distortional buckling of cold-formed steel
beams with both edge and intermediate stiffeners in their compression flanges,
Thin-Walled Struct. 54 (May) (2012) 106–112.
[11] M.R. Haidarali, D.A. Nethercot, Local and distortional buckling of cold-formed steel
beams with edge-stiffened flanges, J. Constr. Steel Res. 73 (June) (2012) 31–42.
[12] E.S. Bernard, R.Q. Bridge, G.J. Hancock, Tests of profiled steel decks with v-stif-
feners, J. Struct. Eng. (ASCE) 119 (8) (1993) 2277–2293.
[13] E.S. Bernard, R.Q. Bridge, G.J. Hancock, Tests of profiled steel decks with flat-hat
stiffeners, J. Struct. Eng. (ASCE) 121 (8) (1995) 1175–1182.
[14] E.S. Bernard, R.Q. Bridge, G.J. Hancock, Design methods for profiled steel decks
with intermediate stiffeners, J. Constr. Steel Res. 38 (1) (1996) 61–88.
[15] H. Wang, Y. Zhang, Experimental and numerical investigation on cold-formed steel
C-section flexural members, J. Constr. Steel Res. 65 (May) (2009) 1225–1235.
[16] R.T. Douty, A Design Approach to the Strength of Laterally Unbraced Compression
Flanges, Bulletin 37, Engineering Experiment Station, Cornell University, Ithaca,
N.Y., 1962.
[17] R.L. Serrette, T. Peköz, Distortional buckling of thin-walled beams/panels. II: design
methods, J. Struct. Eng. (ASCE) 121 (4) (1995) 767–776.
[18] A.D. Martins, D. Camotim, R. Gonçalves, P.B. Dinis, On the mechanics of local-
distortional interaction in thin-walled lipped channel beams, Thin-Walled Struct.
(2017) accepted for publication.
[19] Simulia Inc., Abaqus Standard (version6.7-5), 2008.
[20] R. Bebiano, P. Pina, N. Silvestre, D. Camotim, GBTUL 1.0β – Buckling and Vibration
901
Thin-Walled Structures 119 (2017) 879–901
Analysis of Thin-Walled Members (DECivil/IST), Technical University of Lisbon,
2008.
[21] A.D. Martins, D. Camotim, R. Gonçalves, P.B. Dinis, Distortional-global interaction
in lipped channel beams − Part I: mechanics and elastic behaviour, in: Proceedings
of 8th European Conference on Steel and Composite Structures (Eurosteel 2017 –
Copenhagen, 13-15/9), 2017. (in press).
[22] A.D. Martins, D. Camotim, P.B. Dinis, Distortional-global interaction in lipped
channel beams − Part II: strength, relevance and DSM design, in: Proceedings of 8th
European Conference on Steel and Composite Structures (Eurosteel 2017 –
Copenhagen, 13-15/9), 2017. (in press).
[23] A.D. Martins, P.B. Dinis, D. Camotim, P. Providência, Local-distortional interaction
in pin-ended lipped channel beams: behaviour, strength and DSM design, CD-ROM
Proceedings of 8th International Conference on Advances in Steel Structures
(ICASS’2015 – Lisbon, 22-24/07), Paper 15, 2015.
[24] A.D. Martins, A. Landesmann, D. Camotim, P.B. Dinis, Distortional failure of cold-
formed steel beams under uniform bending: behaviour, strength and DSM design,
Thin-Walled Struct. 118 (September) (2017) 196–213.
[25] D. Camotim, P.B. Dinis, Distortional-global interaction in lipped channel columns,
Proc. Inst. Civil. Eng. (ICE)-Struct. Build. 166 (8) (2013) 381–391.
[26] E.S. Bernard, R.Q. Bridge, G.J. Hancock, Flange curling in profiled steel decks, Thin-
Walled Struct. 25 (1) (1996) 1–29.
[27] M. Lecce, K.J.R. Rasmussen, Nonlinear flange curling in wide flange sections, J.
Constr. Steel Res. 64 (7–8) (2008) 779–784.
[28] N. Silvestre, Nonlinear curling of wide single-flange steel panels, J. Constr. Steel
Res. 65 (3) (2009) 509–522.
[29] B.W. Schafer, T. Peköz, Direct strength predition of cold-formed steel members
using numerical elastic buckling solutions, in: W.W. Yu, R. Laboube (eds.).
Proceedings of the 14th International Specialty Conference on Cold-Formed Steel
Structures (St. Louis, 15-16/10), pp. 69–76, 1998.
[30] G.J. Hancock, Y.B. Kwon, E.S. Bernard, Strength design curves for thin-walled
sections undergoing distortional buckling, J. Constr. Steel Res. 31 (2–3) (1994)
169–186.
[31] D. Camotim, P.B. Dinis, A.D. Martins, Direct Strength Method (DSM) – a general
approach for the design of cold-formed steel structures, in: C. Yu (Ed.), Recent
Trends in Cold-Formed Steel Construction, Woodhead Publishing (Series in
Civil & Structural Engineering), Amsterdam, 2016, pp. 69–105.
[32] B.W. Schafer, Review: the direct strength method of cold-formed steel member
design, J. Constr. Steel Res. 64 (7–8) (2008) 766–778.
[33] B.W. Schafer, Local, distortional and Euler buckling in thin-walled columns, J.
Struct. Eng. (ASCE) 128 (3) (2002) 289–299.
[34] American Iron and Steel Institute (AISI), North American Specification (NAS) for
the Design of Cold-Formed Steel Structural Members (AISI-S100-16), Washington
DC, 2016.
[35] Y. Shifferaw, B.W. Schafer, Inelastic bending capacity of cold-formed steel mem-
bers, J. Struct. Eng. (ASCE) 138 (4) (2012) 468–480.
[36] J. Dwigth, Aluminium Design and Construction, Routledge, New York, 1999.
[37] A.D. Martins, D. Camotim, P.B. Dinis, On the direct strength design of cold-formed
steel columns failing in local-distortional interactive modes, (Submitted for pub-
lication), 2017.