Computers and Structures 89 (2011) 422–434

Contents lists available at ScienceDirect

Computers and Structures

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

Post-buckling behaviour and strength of cold-formed steel lipped channel

columns experiencing distortional/global interaction
Pedro B. Dinis, Dinar Camotim ⇑
Department of Civil Engineering and Architecture, ICIST/IST, Technical University of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: This paper reports the results of a numerical investigation concerning the elastic and elastic–plastic post-
Received 18 November 2009 buckling behaviour of cold-formed steel lipped channel columns affected by distortional/global (flexural–
Accepted 23 November 2010 torsional) buckling mode interaction. The results presented and discussed were obtained by means of
Available online 23 December 2010
analyses performed using the finite element code ABAQUS and adopting column discretisations into fine
4-node isoparametric shell element meshes. The columns analysed (i) are simply supported (locally/glob-
Keywords: ally pinned end sections that may warp freely), (ii) have cross-section dimensions and lengths that ensure
Cold-formed steel lipped channel columns
equal distortional and global (flexural–torsional) critical buckling loads, thus maximising the distor-
Shell finite element analysis
Distortional and global buckling
tional/global mode interaction effects, and (iii) contain critical-mode initial geometrical imperfections
Distortional/global mode interaction exhibiting different configurations, all corresponding to linear combinations of the two ‘‘competing’’ crit-
Elastic and elastic-plastic post-buckling ical buckling modes. After briefly addressing the lipped channel column ‘‘pure’’ distortional and global
Column strength post-buckling behaviours, one presents and discusses in great detail a fair number of numerical results
concerning the post-buckling behaviour and strength of similar columns experiencing strong distor-
tional/global mode interaction effects. These results consist of (i) elastic (mostly) and elastic–plastic
non-linear equilibrium paths, (ii) curves or figures providing the evolution of the deformed configura-
tions of several columns (expressed as linear combination of their distortional and global components)
and, for the elastic–plastic columns, (iii) figures enabling a clear visualisation of (iii1) the location and
growth of the plastic strains and (iii2) the characteristics of the failure mechanisms more often detected
in the course of this research work.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Most cold-formed steel members display very slender thin-
walled open cross-sections, a feature making them highly suscep-
tible to several instability phenomena, namely local, distortional
and global (flexural or flexural–torsional) buckling – see
Fig. 1(a)–(d). Moreover, depending on the member length and
cross-section shape/dimensions, any of these buckling modes can
be critical. However, since several commonly used cold-formed
steel member geometries may lead to rather similar distortional
and global buckling stresses, the corresponding post-buckling
behaviour (elastic or elastic–plastic), ultimate strength and failure
mechanism are likely to be strongly affected by the interaction be-
tween these two buckling modes.
It has been well known for quite a long time that cold-formed
steel members exhibit stable local and global elastic post-buckling
behaviours with clearly different post-critical strength reserves:
rather high in the first case and quite low in the second. On the
⇑ Corresponding author. Tel.: +351 21 8418403; fax: +351 21 8497650.
E-mail address: dcamotim@civil.ist.utl.pt (D. Camotim).
other hand, fairly recent studies have shown that (i) the distortional
post-buckling behaviour fits somewhere in the middle of the two
previous ones (in kinematic and strength terms) and (ii) exhibits
a non-negligible asymmetry with respect to the sense of the
flange-stiffener motion (outward or inward) – e.g., see the works
of Kwon and Hancock [1], Prola and Camotim [2], Camotim and Sil-
vestre [3] or Silvestre and Camotim [4].
Concerning the mode interaction phenomena that may affect
the column post-buckling behaviour and strength, those stemming
from the nearly simultaneous occurrence of local and global buck-
ling are, by far, the better understood – this is attested by the fact
that their effects are already taken into account by virtually all cur-
rent hot-rolled and cold-formed steel design codes, either through
the well-known ‘‘plate effective width’’ concept (e.g., [5]) or by
means of the much more recent (but increasingly popular) ‘‘Direct
Strength Method’’ (e.g., [6,7]). On the other hand, the influence of
local/distortional mode interaction effects on the post-buckling
behaviour and strength of lipped channel columns has attracted
the attention of several researchers in the recent past (e.g., [8–
13]) – it is worth noting that some of the investigations carried
out have already led to the development and calibration of novel

0045-7949/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2010.11.015
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
Fig. 1. Lipped-channel column (a) local, (b) distortional, (c) flexural–torsional and (d) flexural buckling mode (cross-section) shapes.
t t
E = 210 GPa
bw
ν = 0.3
bf bs
bw bf bs t bw.eff
(mm) (mm) (mm) (mm)
(a) 150 110 17.5 2.4 (b)
Fig. 2. Lipped channel column (a) cross-section dimensions and elastic constants, and (b) effective cross-section geometry and dimensions (according to EC3 and for
fy = 355 MPa).
applications (design curves) of the Direct Strength Method (e.g.,
[9,12–14]).
However, there are very few studies addressing the influence of
the distortional/global buckling mode interaction on the post-
buckling behaviour and ultimate strength of cold-formed steel col-
umns (or any other members, for that matter). Indeed, besides a
preliminary version of the current paper [15], the authors are only
aware of a very recent publication that reports an experimental
study on fixed-ended cold-formed stainless steel lipped channel
columns [16]. Nevertheless, one should mention that (i) Eurocode
3 (Part 1–3) includes a procedure to jointly account for global, local
(through an ‘‘effective width’’ approach) and distortional (through
a ‘‘reduced thickness’’ approach) buckling, mainly based on work
done by Thomasson [17] and Höglund [18], and that (ii) the
North-American and Australian/New Zealander Specifications
adopt an ‘‘effective width’’ approach to handle distortional/global
interactive buckling, based on experimental work carried out by
Desmond et al. [19]. However, none of the above studies addresses
explicitly the mechanics of the mode interaction phenomenon
dealt with in this paper – this is probably the reason why those de-
sign approaches are extremely conservative.
Therefore, the aim of this work is to present and discuss a set of
numerical results concerning the (i) post-buckling behaviour (elas-
tic and elastic–plastic), (ii) ultimate strength and (iii) failure mode
nature of cold-formed steel lipped channel simply supported col-
umns affected by distortional/global (flexural–torsional) mode
interaction. In order to enable a thorough assessment of all possi-
ble mode interaction effects, one analyses columns with (i) the
cross-section dimensions and material properties given in
Fig. 2(a), leading to a distortional buckling load meaningfully
(about 20%) lower than its local counterpart, thus ensuring that lo-
cal/distortional interaction effects are not relevant, and (ii) a length
selected to guarantee the coincidence between the distortional (D
– multiple half-waves) and global (G – single half-wave) buckling
loads, thus maximising the distortional/global interaction effects.1
In order to provide an idea of how the Eurocode 3 design approach
handles a column with this cross-section, Fig. 2(b) shows the associ-
1
Moreover, the selection of the column cross-section dimensions, obtained
through ‘‘trial-and-error’’ buckling analyses, also satisfies two additional conditions:
(i) competing (‘‘pure’’) buckling modes with odd half-wave numbers, so that the
maximum deformation occurs at mid-span (although this feature is by no means
essential, it renders the presentation of the results much easier) and (ii) no higher-
order distortional buckling mode ‘‘close’’ to the distortional and global modes under
consideration – this last condition was not easy to enforce (for the cross-section
dimensions chosen, there is a 12% buckling load gap).
423
tred
bw.eff fy = 355 MPa
bf.eff ceff
bf.eff ceff t tred Aeff
(mm) (mm) (mm) (mm) (mm) (mm2 )
92.6 104.1 16.3 2.4 1.29 648.4
ated ‘‘effective area’’, which has been calculated for a yield stress
fy = 355 MPa and must be considered in the safety checking against
global (flexural–torsional) buckling – it corresponds to a 33%
cross-section area reduction (Aeff/A = 0.67).
A fairly large number of columns are analysed and they only dif-
fer in the initial geometrical imperfection configuration. The vari-
ous configurations consist of linear combinations of the
competing distortional and global buckling mode shapes with
amplitudes (mid-span flange-lip corner vertical displacements) of
(i) 10% of the wall thickness t (distortional mode) and (ii) L/1000
(global mode), values that are often adopted for cold-formed steel
members and fall below the typical allowable geometrical toler-
ances, namely (i) b/500, prescribed by ECCS [20] for plates, and
(ii) L/750, global tolerance recently stipulated in Europe. Although
investigating the influence of the initial imperfection amplitudes
on the column post-buckling behaviour and strength is outside of
the aim and scope of this work, a very limited imperfection-sensi-
tivity study is included in the paper – for further insight on this to-
pic, the interested reader is referred to investigations carried out (i)
by Schafer and Peköz [21] and Dubina and Ungureanu [22], in the
context of cold-formed steel members, and (ii) by Maiorana et al.
[23], concerning steel girder webs. All numerical results presented
were yielded by finite element analyses carried out in the code ABA-
QUS [24] that (i) adopt member discretisations into fine 4-node iso-
parametric shell element meshes (preliminary convergence/
accuracy studies showed that it suffices to discretise the cross-sec-
tion mid-line into 24 finite elements – 10 in the web, 6 in each
flange and 1 in each stiffener – Fig. 4(b) illustrates the meshes
adopted, which correspond to an element width approximately
equal to 15 mm and a length/width ratio roughly equal to 1), (ii)
model the simply supported conditions by imposing null trans-
verse displacements at all end section nodes, as illustrated in
Fig. 3(a) (note also that, to preclude a spurious longitudinal rigid-
body motion, the axial displacement was also prevented at the
mid-span mid-web node) and (iii) simulate the axial compression
loading through compressive forces p, uniformly distributed along
both column end-section mid-lines, as depicted in Fig. 3(b) (the
load parameter is P = pA/t, where A and t are the column cross-sec-
tion area and wall thickness). Detailed accounts of all the relevant
modelling issues can be found in previous works by Dinis et al. [11]
and Dinis and Camotim [25].
Column buckling analyses are performed at the outset, in order (i)
to select the column length that maximised the D/G interaction, and
also (ii) to obtain the associated buckling mode shapes, required to
define the initial geometrical imperfections. Next, one addresses
424 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
G
x3 x2
x1
(a)
u1=u2=0
Fig. 3. Lipped channel column (a) end-section support conditions and (b) loading procedure adopted in the shell finite element simulations.
Pb (kN)
400
300
200
100
LD=76cm
LD/G=222cm
0
6 10 100
(a)
Fig. 4. (a) Column buckling curves and (b) distortional and global (flexural–torsional) buckling mode shapes of the column with LD/G = 222 cm.
the pure distortional and global elastic post-buckling behaviours,
which are deemed not affected by D/G interaction. Finally, the
numerical results concerning the column post-buckling behaviour
and ultimate strength under D/G interaction are presented and dis-
cussed. They comprise (i) several elastic (mostly) and elastic–plastic
non-linear (post-buckling) equilibrium paths, (ii) curves and figures
providing the evolution, along the elastic paths, of the deformed con-
figurations of several columns (expressed as combinations of their
distortional and global components) and, for the elastic–plastic col-
umns, (iii) figures enabling a clear visualisation of (iii1) the location
and growth of the plastic strains and (iii2) the characteristics of the
failure mechanisms detected in the course of this research work –
by analysing members with different yield stresses, it is also possible
to assess how the D/G mode interaction effects vary with the yield-
to-critical stress ratio.
2. Buckling behaviour – length selection
The curves depicted in Fig. 4(a) provide the variation, with the
column length L (logarithmic scale), of (i) the ABAQUS critical load
Pcr and (ii) the single-wave buckling load Pb.1, provided by CUFSM
[26] finite strip analyses. As for Fig. 4(b), it shows the ABAQUS distor-
tional and flexural–torsional buckling mode shapes of the column
with L = 222 cm. These buckling results prompt the following
remarks:
(i) The ABAQUS buckling curve exhibits three distinct zones, cor-
responding to (i1) 1–4 half-wave local buckling, (i2) 1–3 half-
wave distortional buckling and (i3) single half-wave global
buckling.
(ii) The black and white dots identify the practically coincident
minimum single half-wave distortional critical loads yielded
by the CUFSM and ABAQUS analyses (Pcr.D = 203.7 kN), for
LD = 76 cm.
p=σt
u1=u2=0
p=σt
(b)
ABAQUS
CUFSM
DM
L(cm)
GM
1000
(b)
(iii) As clearly shown in Fig. 4(a), the LD/G = 222 cm has virtually
identical distortional and global critical loads
(Pcr.D = 203.6 kN and Pcr.G = 203.9 kN), associated with three
and single half-wave buckling modes, respectively (see
Fig. 4(b)). Obviously, the post-buckling behaviour and ulti-
mate strength of such column will be highly affected by dis-
tortional/global mode interaction.
3. Distortional and global post-buckling behaviours
Numerical results concerning the pure distortional and global
column post-buckling behaviours (i.e., without D/G mode interac-
tion) are first presented. Fourteen columns with seven lengths
are analysed: (i) LD = 76 cm (Pcr = 203.6 kN), (ii) LG.1 = 300 cm
(Pcr = 119.9 kN), (iii) LG.2 = 400 cm (Pcr = 72.0 kN), (iv) LG.3 = 450 cm
(Pcr = 58.8 kN), (v) LG.4 = 500 cm (Pcr = 49.3 kN), (vi) LG.5 = 600 cm
(Pcr = 26.6 kN), and (vii) LG.6 = 900 cm (Pcr = 20.9 kN) – note that
columns LG.1 and LG.6 have critical loads equal to, respectively,
59% and 10% of that associated with D/G interaction: LD/G = 222 cm
(Pcr.D = 203.6 kN). For each length, the analyses involve columns
containing either positive or negative critical-mode geometrical
imperfections, i.e., (i) LD columns having distortional imperfections
of magnitude ±10% of the wall thickness t (positive and negative
mean mid-span inward and outward flange-lip motions, respec-
tively), and (ii) LG columns with global imperfections having mid-
span web chord rotations equal to ±0.016 rad2 (positive and negative
mean clockwise and counter-clockwise rotations, respectively).
Fig. 5(a) and Fig. 5(b) show the upper portions (P/Pcr > 0.6) of
the post-buckling equilibrium paths (i) P/Pcr vs. v/t (v is the mid-
span top flange-lip corner vertical displacement), for the two LD
2
Note that, in the LG.1 columns, this magnitude corresponds to a global initial
imperfection associated with mid-span flange-lip corner vertical displacements equal
to L/1000 (i.e., 3 mm).
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434 425

LD Columns P / Pcr v0=±0.1t β0=±0.016rad P / Pcr LG Columns β=+0.03rad

1.2 1.2
v=−4.0 t LG.1
1 LG.2 1 LG.2

LG.1 LG.1
v=+3.8 t
0.8 0.8
LG.2
v<0 v>0 β<0 β >0
0.6 0.6
⏐v⏐/ t ⏐β⏐(rad)
20 10 0 10 20 0.6 0.3 0 0.3 0.6
(a) (b)
Fig. 5. Curves providing the column (a) distortional (P/Pcr vs. v/t) and (b) global (P/Pcr vs. b) elastic post-buckling behaviours.

Pb (kN )
300
ABAQUS L(cm) 222 300 900
200 GBTUL − 6 modes
(nw =1)
GBT
Buckling
100 +
Mode (nw=1)
LD/G= 222cm
0 L (cm)
200 1000
(a) GBT
pi 2 4 6
1.0 6 Deformation
4
Mode
0.5 2
0
L (cm)
200 1000 (c)
(b)
Fig. 6. GBT-based column buckling: (a) Pb vs. L curves (L > 100 cm), (b) modal participation diagrams (nw = 1) and (c) in-plane shapes of 3 buckling (L = 222, 300, 900 cm) and
deformation modes.

columns, and (ii) P/Pcr vs. b (b is the mid-span web chord rotation),
for the twelve LG columns – also shown are a few deformed config-
urations of column mid-span cross-sections at advanced post-
buckling stages. These post-buckling results lead to the following
comments:
(i) The LD and LG.4–LG.6 columns exhibit the expected stable dis-
tortional and global post-buckling behaviours (e.g., [27]):
while the LD columns exhibit a fair amount of post-critical
strength reserve and a clearly visible asymmetry (the inward
column is a bit stiffer), the LG.4–LG.6 ones display little post-
critical strength (the LG.6 columns are a little stiffer) and per-
fectly symmetric equilibrium paths.
(ii) However, the LG.1–LG.3 columns exhibit an unexpected unsta-
ble symmetric post-buckling behaviour – the equilibrium
path limit points occur for 0.87 6 P/Pcr 6 0.94 and the peak
load takes place for progressively higher b values as the col-
umn length increases (LG.1 ? LG.3).
(iii) In order to try to understand why the LG columns exhibit
two different post-buckling behaviours, it was decided to
perform buckling analyses of those columns using General-
ised Beam Theory (GBT – e.g., [28,29]), hoping that its modal
features would shed new light on this issue – indeed, this
was the case. The curves shown in Fig. 6(a) provide the var-
iation of the buckling load Pb with the column length L
(L P 222 cm): (iii1) one yielded by ABAQUS analyses and
already shown in Fig. 4(a), and (iii2) the other obtained from
GBT analyses carried out in GBTUL [30,31] and including 6
single half-wave deformation modes (4 global and 2 distor-
tional) – dashed curve. As for Fig. 6(b), it displays the GBT-
based column modal participation diagram (for single half-
wave buckling), providing the contributions of each GBT
deformation mode to the column buckling modes.3 Finally,
Fig. 6(c) shows the buckling mode shapes yielded by the
GBT analyses for columns with lengths L = 222, 300, 900 cm,
as well as the in-plane shapes of the 3 deformation modes
that participate in them. The following conclusions can be
drawn from these GBT buckling results:
(iii.1) The GBT curve descending branch involves, in fact,
two distinct buckling modes: (iii1) distortional–flex-
ural–torsional (2 + 4 + 6), for 222 < L < 700 cm, and
(iii2) flexural–torsional (2 + 4), for 700 6 L < 2000cm
– note that the participation of the anti-symmetric
distortional mode 6 progressively fades as the col-
umn length grows, until it vanishes for L  700 cm.
(iii.2) Since the six LG column lengths are located inside the
two intervals identified in the previous item, their
buckling modes have different natures: while (iii1)
the LG.6 column buckling mode is flexural–torsional
(participation of 70% and 30% from modes 2 and 4,
respectively), (iii2) the LG.1–LG.5 columns buckle in
distortional–flexural–torsional modes with varying
participations from modes 2, 4 and 6, namely 44%,
31% and 25% (LG.1), 55%, 36% and 9% (LG.2), 58%, 36%
and 6% (LG.3), 60%, 37% and 3% (LG.4), and 63%, 36%
and 1% (LG.5).
3
The participation of a given deformation mode in a column buckling mode is
obtained on the basis of the corresponding mid-span cross-section deformed
configurations (e.g., [28]).
426 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
(iii.3) The level of participation of mode 6 in the column
critical buckling mode provides the explanation for
the different post-buckling behaviours exhibited by
the LG.1–LG.6 columns. Indeed, its presence (iii1)
reduces the column post-critical strength (see the
column sequence LG.6–LG.1) and (iii2) is responsible
for the surprising unstable post-buckling behaviour
of the LG.1–LG.3 columns (note that the participation
of mode 6 is higher than 5% for these columns) – col-
umn LG.4 may be viewed as the ‘‘transition’’ between
the stable and unstable post-buckling behaviours
(see Fig. 5).
(iii.4) At this stage, it worth noting that the contribution of
mode 6 to the column critical buckling mode would
remain virtually undetected in the absence of the
GBT analysis – this statement can be clearly attested
by looking at the member and cross-section
deformed configurations shown in Figs. 4(b) and
6(c), obtained by means of ABAQUS shell finite element
and CUFSM finite strip analyses.4
(iii.5) Since the content of item (iii.1) may have considerable
implications in the design of cold-formed steel col-
umns affected by D/G coupling, it was decided to
investigate this matter further, by analysing lipped
channel columns with other cross-section geometries
and also susceptible to this interaction phenomenon.
The results obtained confirm both (iii1) the participa-
tion of mode 6 in the so-called ‘‘global’’ column buck-
ling modes and (iii2) the unstable post-buckling
behaviour exhibited by the columns whose critical
buckling mode has a visible participation of this defor-
mation mode. This statement is illustrated by the
buckling analysis of columns having bw = 110 mm,
bf = 90 mm, bs = 10 mm and t = 1 mm (C1 column),
and bw = 100 mm, bf = 70 mm, bs = 10 mm and
t = 1.5 mm (C2 column). These columns (iii1) buckle
in distortional–flexural–torsional modes for
272 < L < 700 cm (C1 column) and 154 < L < 400 cm
(C2 column), (iii2) have a mode 6 contribution to the
LD/G column critical buckling mode of 29% (C1 column)
and 38% (C2 column), and (iii3) exhibit an unstable
post-buckling behaviour for 272 < L < 500 cm (C1 col-
umn) and 154 < L < 300 cm (C2 column), i.e., roughly
for columns with LD/G < L < 2LD/G – note that, for this
length range, the contribution of mode 6 to the critical
buckling mode is always higher then 5% (as is the case
for the columns with the cross-section geometry given
in Fig. 2(a)).
(iii.6) In the GBT nomenclature, the LG.1–LG.5 column critical
buckling mode is designated as ‘‘mixed’’, since it
combines (‘‘mixes’’) deformation modes of different
natures (2 + 4 and 6 – global and distortional). Note
that the existence of a ‘‘mixed’’ buckling mode does
not imply the occurrence of a buckling mode interaction
phenomenon – the latter corresponds to the simulta-
neous (or nearly so) occurrence of two or more criti-
cal buckling modes (regardless of whether they are
‘‘mixed’’ or not).
(iii.7) In the L = 222 cm column, two critical buckling
modes occur simultaneously: (iii1) a three half-wave
pure symmetric distortional mode and (iii2) a single
half-wave ‘‘mixed’’ distortional–flexural–torsional
4 5
The sole ‘‘sign’’ of the presence of the anti-symmetric distortional mode 6 is a
practically imperceptible web double-curvature bending (see Fig. 6(c)) that can only
be detected if one knows about it beforehand.
mode, combining modes 2 (6%), 4 (19%) and 6 (55%)
(i.e., mode 6 is predominant) – the latter was termed
‘‘global’’ earlier, a designation that will be retained
for simplicity. Therefore, the D/G mode interaction
phenomenon investigated in this work really
involves the two competing critical buckling modes
just described.
4. Post-buckling behaviour under D/G mode interaction
One now investigates the elastic and elastic–plastic post-buck-
ling behaviours of simply supported columns with L = 222 cm
(Pcr = 203.6 kN), which are strongly affected by the interaction be-
tween nearly coincident distortional (three half-waves) and ‘‘glo-
bal’’ (single half-wave) buckling modes.5
4.1. Initial geometrical imperfections
A very important issue in mode interaction investigations is to
assess how the initial geometrical imperfection shape influences
the post-buckling behaviour and strength of the structural system
under scrutiny. Indeed, the commonly used approach of including
critical-mode imperfections ceases to be well defined, due to the
presence of two competing buckling modes that may be combined
arbitrarily. Thus, in order to obtain column equilibrium paths that
(i) cover the whole D/G critical-mode imperfection shape range
and (ii) can be meaningfully compared, one adopts the following
approach, which accounts for the fact that the two competing
(‘‘pure’’) buckling modes exhibit odd half-wave numbers:
(i) To determine the ‘‘pure’’ critical buckling mode shapes, nor-
malised to exhibit unit mid-span flange-lip corner vertical
displacements: (i1) a distortional mode with vD = 1 mm,
associated with a mid-web flexural displacement (measured
with respect to the web chord) of wD = 0.265 mm, and (i2) a
‘‘global’’ one also with vG = 1 mm, implying a mid-span web
chord rotation equal to bG = 0.005 rad.
(ii) To scale down the pure modes, thus leading to the following
magnitudes for the distortional and ‘‘global’’ imperfections:
vD.0 = 0.1t and vG.0 = L/1000 (in this case, 0.1t = 0.24 mm, L/
1000 = 2.22 mm). At this point, it is worth mentioning that,
in order to assess the influence of the initial imperfection
amplitudes on the column elastic post-buckling behaviour
under D/G interaction, a few results were obtained with
other imperfection magnitudes.
(iii) A given initial geometrical imperfection shape is obtained as
a linear combination of these scaled buckling modes shapes,
with coefficients CD.0 and CG.0 satisfying the condition
(CD.0)2 + (CG.0)2 = 1. A better visualisation and ‘‘feel’’ of the
initial imperfection shape is obtained by considering the
unit radius circle drawn in the CD.0–CG.0 plane that is shown
in Fig. 7(a): each ‘‘acceptable’’ imperfection shape lies on this
circle and corresponds to an angle h, measured counter-
clockwise from the horizontal (CD.0) axis and defining a
CG.0/CD.0 ratio (CD.0 = cos h and DG.0 = sin h). Fig. 7(b) provides
the pure D and G initial imperfection shapes (h = 0°, 90°, 180°
and 270°) – note that (iii1) h = 0° and h = 180° correspond to
inward and outward flange-lip motions, and that (iii2) h = 90°
and h = 270° are associated with clockwise and counter-
clockwise cross-section rotations. In this work, one consid-
ers (mostly) initial imperfections corresponding to 15°
intervals.
As just seen, this interaction really involves a three half-wave symmetric
distortional buckling mode and a ‘‘mixed’’ buckling mode combining anti-symmetric
distortion, major axis flexure and torsion (distortional–flexural–torsional mode).
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434 427

vG.0=L/1000 vD.0 =0.1t

0°
β G.0 90° CG.0 wD.0
90°
1
r =1 θ = 0º θ = 180º
180° 0 θ 0°
−1 1
CD.0
vD.0 =− 0.1t vG.0=− L/1000
−1
270°
− wD.0 180° −β G.0 270°
θ = 90º θ = 270º

(a) (b)
Fig. 7. (a) Initial imperfection representation in the CD.0–CG.0 plane and (b) initial imperfection shapes associated with h = 0°, 90°, 180° and 270°.

θ=180 θ=0 θ=0;180

θ
0 26
θ=179 15
θ=179
1
θ=165
θ θ
1 27
θ= 27 ; 45 15 45
165 ; 90 26 90
0≤θ≤180 P / Pcr P / Pcr 0≤θ≤180
θ=0 θ=27−179
1 1 θ=0;180
θ=180 θ
0.8 θ 0.8 1
1 …
θ … 26 (c1)
27 26 θ
… 0.6 0.6 27 θ=1−26
179 …
179 β (rad)
w<0 w>0 ⏐w⏐/ t β >0
(a) 0.4 (b) 0.4
1 2 0 0.4
(c2)
2 1 0 0.2

Fig. 8. (a) P/Pcr vs. w/t and (b) P/Pcr vs. b equilibrium paths (0° 6 h 6 180° columns), and (c) limit deformed shapes of the (c1) h = 27–179° and (c2) h = 1–26° columns (curve
descending branches).

4.2. Elastic mode interaction (i) All column equilibrium paths (P/Pcr vs. w/t and P/Pcr vs. b)
exhibit well-defined limit points, which occur almost always
Initially, one presents post-buckling results concerning columns for quite small w/t and b values. The exceptions are the
with 13 initial imperfection shapes corresponding to 0° 6 h 6 180° h = 0°, 180° column equilibrium paths (pure distortional ini-
and separated by 15° intervals6 – moreover, in order to clarify some tial imperfections), which have limit points associated with
behavioural aspects, the h = 1°, 26°, 27°, 179°, 181°, 359° columns (i1) slightly larger w/t values and (i2) null b values – they cor-
were also analysed. respond to a singular post-buckling behaviour that will be
Fig. 8(a) and Fig. 8(b) shows the upper parts of the most repre- addressed further ahead.
sentative column post-buckling equilibrium paths (i) P/Pcr vs. w/t, (ii) The comparison between the equilibrium paths P/Pcr vs. b of
where w is the mid-web flexural displacement at mid-span (mea- the (ii1) h = 90° column (pure ‘‘global’’ initial imperfections
sured with respect to the web chord), and (ii) P/Pcr vs. b, where b is with b0 = 0.005 rad), shown in Fig. 8(a), and (ii2) LG.1 column
the mid-span web chord rigid-body rotation. As for Fig. 5(c), it (b0 = 0.016 rad), depicted in Fig. 5(b), shows clearly the
shows the limit deformed configurations of the h = 27–179° and adverse effect (strength erosion) due to the D/G mode inter-
h = 1–26° columns, which correspond to the curve descending action. Indeed, in spite of the considerably smaller initial
branches (advanced post-buckling stages). In order to shed some imperfection magnitude (about one third), the limit point
light on issues raised by the close scrutiny of the curves shown of the h = 90° column equilibrium path corresponds to a
in Fig. 8(a) and Fig. 8(b), additional post-buckling results are pre- lower applied load level (P/Pcr = 0.81 vs. P/Pcr = 0.87).
sented in Figs. 9 to 14 – they consist of (i) equilibrium paths previ- (iii) The equilibrium paths displayed in Fig. 8(a) can be grouped
ously displayed in Fig. 8(a), complemented by illuminating column in three categories, each one corresponding to a different
mid-span cross-section deformed configurations, (ii) ultimate load post-buckling behaviour – they are identified next and
values and (iii) figures providing the mode coupling evolution addressed separately in the sequel:
along those equilibrium paths. The close observation of the post- (iii.1) h = 0°, 180° equilibrium paths, corresponding to pure
buckling results presented in Figs. 8(a)–(c) leads to the following distortional imperfections. As mentioned before,
remarks (and also further results): these two columns exhibit a singular post-buckling
behaviour, characterised by the fact that their
cross-sections exhibit no rigid-body rotations (i.e.,
b = 0).
6
Except for the 181° and 349° columns, addressed next, no post-buckling results (iii.2) 1° 6 h 6 26° equilibrium paths, associated with pre-
concerning columns with initial imperfection shapes defined by 180° < h < 360° are dominantly distortional imperfections with outward
presented - since the column post-buckling behaviour is symmetric with respect to
the deformed configuration global component sign, regardless of whether there is
outer half-waves.
mode interaction or not (see Fig. 5(b)), the 0° 6 h 6 180° column results provide all (iii.3) 27° 6 h 6 179° equilibrium paths, corresponding to all
the necessary information. the remaining imperfection shapes.
428 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434

27 ≤θ ≤ 179 CG
40
θ =60 θ =45
0.45 θ =90
20 θ =30
1 θ =135
θ =27
CD θ =165
D
θ =179
-80 -60 -40 -20 0 20

Fig. 9. Mode coupling ratio CG/CD evolution along the equilibrium paths (27° 6 h 6 179° columns).

90 120
Pu /Pcr θ (º) Pu / Pcr θ (º) Pu / Pcr
1
27 0.928 105 0.803
30 0.909 120 0.807
0.8 45 0.864 135 0.819
60 0.836 150 0.839
27 θ (º) 75 0.817 165 0.870
0.6 75 105 135
0 30 60 90 120 150 180 90 0.807 179 0.920

Fig. 10. Variation of Pu/Pcr with the initial imperfection shape (27° 6 h 6 179° columns).

θ=179,181 P / Pcr θ=1,359

1
II
III 2 θ =1 1 2 3 4
I 3
1
0.8
IV
4 1 2 3 4
θ =359
(a) 0.6 ⏐w⏐/ t

2 1 0 1 2

P / Pcr
θ =179 I II III IV
1

0.8
θ =359
θ=1
0.6 θ =181 I II III IV
θ=179 β (rad)
θ =181
(b) 0.4
0.5 0.25 0 0.25 0.5 (c)
Fig. 11. (a) P/Pcr vs. w/t and (b) P/Pcr vs. b equilibrium paths and (c) mid-span cross-section deformed configuration evolution (h = 1°, 359° and h = 179°, 181° columns).

CG 1 ≤θ ≤ 26
span web bending – since the 27°6h6179° category
40 includes the vast majority of equilibrium paths, thus
θ (º) Pu / Pcr providing a more meaningful characterisation of the
20 1 0.944 D/G interaction effects, it will be addressed first.
θ =26 15 0.937 Then, one tackles the 1°6h626° category, which con-
θ =15 26 0.928
CD stitutes another particular case.
θ =1 (a) (b)
0 20 40 60 4.2.1. Columns with 27°6h6179° initial geometrical imperfections
These equilibrium paths correspond to initial geometrical
Fig. 12. (a) Evolution of the mode coupling ratio CG/CD and (b) variation of Pu/Pcr
with the initial imperfection shape (1° 6 h 6 26° columns).
imperfections whose global components involve always clockwise
cross-section rigid-body rotations (in view of the global post-buck-
ling symmetry, it is not necessary to consider counter-clockwise
At this stage, it is worth noting that the first two
rotations)7 – such component is either (i) the only one (h = 90°) or
equilibrium paths concern a rather unexpected D/G
(ii) combined with a distortional one exhibiting inward
interactive behaviour, which will be the last to be
dealt with. Moreover, the equilibrium paths belong-
7
ing to the last two categories merge into common Recall that the GBT buckling analysis showed that these ‘‘global’’ components
combine (i) a clockwise torsion rotation, (ii) a downward major axis bending
post-buckling curves associated with (iii1) clockwise
displacement and (iii) downward flange-lip motions associated with anti-symmetric
mid-span web chord rotations and (iii2) either in- distortion. Since they are all ‘‘linked’’ through the buckling mode shape, it suffices to
ward (27°6h6179°) or outward (1°6h626°) mid- mention only the torsion rotation b (the most ‘‘visible’’ one).
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
Fig. 13. (a) P/Pcr vs. w/t equilibrium paths and (b) mid-span cross-section deformed configuration evolution (h = 26° and h = 27° columns).
(27°6h<90°) or outward (90°<h6179°) mid-span flange-lip motions.8
The joint observation of all these equilibrium paths leads to the fol-
lowing remarks:
(i) They all merge into a common curve, which corresponds to a
deformed configuration that may be deemed associated with
a ‘‘coupled buckling mode’’ shape. In order to provide a bet-
ter visualisation of this concept and, at the same time, quan-
tify the amount of ‘‘coupling’’, Fig. 9 shows the evolution,
along those equilibrium paths, of a ‘‘mode coupling ratio’’
defined as CG/CD and relating the amplitudes of the column
deformed configuration global and distortional components.
These amplitudes are obtained from the mid-span web (i1)
chord rotation b and (i2) mid-point flexural displacement
w, adopting the following simplifying assumptions:
(i.1) The column deformed configuration can be com-
pletely expressed as a linear combination of the glo-
bal and distortional buckling mode shapes,
normalised with respect to vG.0 = 2.22 mm (L/1000)
and vD.0 = 0.24 mm (0.1t), respectively – values corre-
sponding to bG.0 = 0.0111 rad and wD.0 = 0.0564 mm.
(i.2) The b and w values associated with a given deformed
configuration stem exclusively from its global and
distortional components.
(i.3) In view of the above assumptions, one readily deter-
mines the global and distortional participation fac-
tors, by means of the expressions CG = b/bG.0 and
CD = w/wD.0. Then, all CG/CD curves depart from a
unit-radius circle on the CG–CD plane (see detail),
with initial slopes that depend on the imperfection
shape.
The CG/CD curves shown in Fig. 9 provide valuable
information about the column D/G mode interaction
behaviour (for 27°6h6179°). Indeed, as post-buck-
ling progresses, all of them tend to a practically
straight line with slope defined by DCG  0.45DCD
(negative CD values stand for distortional deformed
configuration components with outward mid-span
flange-lip motions – see Figs. 7(a) and (b)).9 Then, it
seems reasonable to view this straight line as the rep-
resentation of the column ‘‘coupled buckling mode’’,
8
Although one always quotes the nature of the mid-span flange-lip motions (for
simplicity reasons), it should be noted that the key feature of the initial geometrical
imperfections is the nature of the flange-lip motions of the two outer half-waves - it
provides the explanation for several qualitative aspects of the column D/G interactive
behaviour.
9
When interpreting this coupled buckling mode shape, the reader should be
always aware of the global and distortional buckling mode normalisations - they
correspond to mid-span flange-lip corner vertical displacements equal to 2.22 mm
(global) and 0.24 mm (distortional), i.e., the former is more than nine times the later.
429
which combines participations of (i1) a three half-
wave distortional component with outward mid-span
flange-lip motions and (i2) a single half-wave global
component that are roughly equal to one and two
thirds of the total deformed configuration (31% and
69%, to be precise) – this coupled buckling mode shape
can be visualised in Fig. 8(c1).10
(ii) All equilibrium paths exhibit a limit point prior to merging
into the ‘‘common curve’’ and one observes that, generally
speaking, the column limit load decreases as the initial
imperfection global component becomes more dominant –
this can be confirmed by looking at the table included in
Fig. 10, which provides the variation of the column ultimate
load ratio Pu/Pcr with its initial imperfection shape (i.e., with
h). However, one observes that the minimum ultimate load
(Pu/Pcr = 0.803) occurs for the h = 105° column, i.e., the one
combining 96.5% and 25.9% of the pure global and distor-
tional imperfections11 – nevertheless, one must mention that
the Pu/Pcr value remains practically constant for
90° 6 h 6 120° (0.5% between the minimum and maximum
values – see curve detail in Fig. 10), i.e., analysing a column
with a pure global imperfection (h = 90°) will certainly provide
a rather accurate estimate of the minimum Pu/Pcr value – in
other words, the pure global imperfection may be viewed,
for practical purposes, as the most detrimental one, in the
sense that it maximises the strength erosion due to the D/G
interaction.12
(iii) Although it is evident that the initial imperfection global
component plays a crucial role in the column post-buckling
behaviour (note that all equilibrium paths exhibit limit
points), it is also obvious that there must exist a plausible
explanation for the qualitative and quantitative differences
exhibited by the 90° < h 6 179° and 1° 6 h < 90° column
equilibrium paths, clearly visible in Fig. 8(a) and Fig. 8(b)
– (iii1) while the former tend to the common curve in a
‘‘regular’’ fashion (the amplitudes of both the global and
distortional components grow monotonically), (iii2) the
latter either tend to that same common curve ‘‘irregularly’’
(27° 6 h < 90° – occurrence of distortional component
10
The mid-span cross-section deformed configuration associated with this coupled
buckling mode shape can also be viewed in Figs. 11(c) (h = 179° – IV) and 13(b)
(h = 27° – IV).
11
It is interesting to notice that 105° is very close to the average between 27° and
179°.
12
Even if this assessment concerns only the 27° 6 h 6 179° interval, it is valid for
any h – as shown in Fig. 8(a) and Fig. 8(b) and discussed ahead, the h = 0°, 180° and
1° 6 h 6 26° column ultimate loads are higher.
430 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
reversals) or do not tend to it at all (1° 6 h < 27° – they
tend to another common curve, which will be addressed
later).
(iv) All the distinct post-buckling behaviours described in the
previous item stem from the influence exerted by the outer
half-wave flange-lip motions (distortional feature) on the
major axis flexure (global feature). Indeed, (iv1) while
inward flange-lip motions reduce the cross-section major
moment of inertia, thus facilitating the occurrence of the
corresponding flexure, (iv2) the outward flange-lip motions
have precisely the opposite effect. This assessment justifies
(or is confirmed by) the following facts:
(iv.1) The ‘‘regularity’’ of the 90° < h6179° column equilib-
rium paths, due to the converging effects of the initial
imperfection global and distortional components.
(iv.2) The distortional component amplitude reversals
occurring in the 27° < h 6 90° column equilibrium
paths, due to the opposing effects of the initial imper-
fection global and distortional components, with the
former prevailing over the latter.
(iv.3) The ‘‘peculiarity’’ of the 1° < h 6 26° column equilib-
rium paths, due to the opposing effects of the initial
imperfection global and distortional components,
with the latter prevailing over the former.
(iv.4) The ‘‘equality/symmetry’’ between the equilibrium
paths and mid-span cross-section deformed configu-
rations concerning the column pairs h = 179°, 181°
and h = 1°, 359°, clearly shown in Figs. 11(a)–(c) –
for clarity, the first three cross-section deformed con-
figurations of each column are amplified 20, 5 and 3
times, respectively. It is worth noting that the initial
imperfections in each pair have the same distortional
component and opposite-sign global components.
4.2.2. Columns with 1° 6 h 6 26° initial geometrical imperfections
These equilibrium paths correspond to initial imperfections
whose global components involve always clockwise cross-section
rigid-body rotations, combined with a distortional one exhibiting
inward (1° 6 h 6 26°) mid-span flange-lip motions. The joint
observation of all these equilibrium paths leads to the following
remarks:
(i) Like those dealt with previously (27°6h6179°), they all
merge into a common curve – this merging occurs only in
their descending branches (after the limit points) and the
common curve corresponds to the deformed configuration
shown in Fig. 8(c2).13 However, the characteristics of this
deformed configuration change continuously as post-buck-
ling progresses, as attested by the equilibrium paths shown
in Fig. 8(a) and Fig. 8(b) (particularly the latter) and also
by the evolution of the mode coupling ratio CG/CD displayed
in Fig. 12(a) – indeed, one notices that the amplitude of the
deformed configuration distortional component starts
decreasing at a growing rate along the P/Pcr vs. w/t equilib-
rium path descending branches (see Fig. 8(a)). This also
implies a global component growth rate increase, clearly vis-
ible in the corresponding P/Pcr vs. b equilibrium paths (their
common curve becomes almost horizontal – see Fig. 8(b)).14
13
The mid-span cross-section deformed configuration associated with this coupled
buckling mode shape can also be viewed in Figs. 11(c) (h = 1° – 4) and 13(b) (h = 26° –
4).
14
Since the outer half-wave outward flange-lip motions decrease, the occurrence of
major axis flexure becomes ‘‘easier’’, which explains the global component (i.e., b)
growth rate increase.
Thus, no coupled buckling mode shape can be inferred from
(or linked to) the common curve concerning these equilib-
rium paths.
(ii) Whenever the initial outer half-wave outward flange-lip
motions are large enough (i.e., for h < 27°), their post-buck-
ling growth (amplification) ‘‘retards’’ the dominant appear-
ance of the (destabilising) deformed configuration global
component. This leads to slightly higher limit loads (see
table in Fig. 12(b) and compare its values with those appear-
ing in Fig. 10), that tend to occur for visibly larger w values
(e.g., compare the limit point locations of the h = 1° and
h = 179° column P/Pcr vs. w/t equilibrium paths).
(iii) The amplitude reversals of the column deformed configura-
tion distortional components cease abruptly, as shown in
Fig. 8(a). Indeed, no further equilibrium states could be
detected, most likely because the column outer half-wave
flanges-lips were about to ‘‘snap’’ from an outward position
to an inward one – obviously, such dynamic behaviour could
not be captured by the ABAQUS (static) geometrically non-lin-
ear analyses carried out.
(iv) To enable a better grasp of the qualitative and quantitative
differences between the two sets of column equilibrium
paths dealt with (27° 6 h 6 179° and 1° 6 h 6 26°), Fig. 13
shows the h = 26°, 27° column (iv1) P/Pcr vs. w/t equilibrium
paths and (iv2) mid-span cross-section deformed configura-
tion evolution. Despite the tiny initial imperfection differ-
ence, the two column post-buckling behaviours are quite
distinct – e.g., (iv1) the equilibrium path shapes and (iv2)
the distortional components of the mid-span cross-section
deformed configurations along the descending branches
(the h = 26° and h = 27° column flange-lip assemblies move
inward and outward, respectively).
4.2.3. Columns with h = 0°, 180° initial geometrical imperfections
Finally, one addresses the equilibrium paths of the columns
containing pure distortional initial imperfections, as they exhibit
rather peculiar post-buckling behaviours, both characterised by
(i) the highest limit loads (Pu/Pcr = 0.945, 0.961, for h = 0°, 180°,
respectively), (ii) much less pronounced limit points (see
Fig. 7(a)) and (iii) the total absence of torsional rotations (see
Fig. 8(b)). This is rather surprising, since (i) there are no D/G mode
interaction effects of the type found for all other initial imperfec-
tion shapes (no torsional rotations) but, at the same time, (ii) the
two column equilibrium paths are not similar to the purely distor-
tional ones shown in Fig. 5(a) (occurrence of limit points).
Then, in order to understand the nature of the singular column
post-buckling behaviour described above, it is convenient to look
at Fig. 14(a) and Fig. 14(b), showing (i) the two column equilibrium
paths, (ii) mid-span cross-section deformed configurations (ampli-
fied twice) at three equilibrium states and also (iii) the column
overall deformed configurations at the descending branch equilib-
rium states. Indeed, it is possible to observe that:
(i) The h = 0° and h = 180° column P/Pcr vs. w/t equilibrium
paths are slightly different – in particular, the former col-
umn exhibits a lower limit load.
(ii) While the first mid-span cross-section deformed configura-
tions (1–I) are ‘‘purely distortional’’, the second and third
ones (2–3 and II–III) indicate the presence of minor axis flex-
ure (web in tension) – this is clearly confirmed by the col-
umn overall deformed configurations shown in Fig. 14(b),
which correspond to the equilibrium states 3–III.
(iii) Therefore, the h = 0° and h = 180° columns are affected by a
different type of distortional/global mode interaction phe-
nomenon, which (iii1) involves only minor axis flexure and
(iii2) is not caused by the closeness of two buckling loads.
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434 431

θ=180 P / Pcr θ=0

1
III II 2 3 θ =0 1 2 3
I 1
0.8

θ =180 I II III

0.6 ⏐w⏐/ t θ =0 θ =180

1 0 1 (a) (b)
Fig. 14. (a) P/Pcr vs. w/t paths and mid-span cross-section deformed configuration evolution, and (b) post-peak deformed configurations (h = 0°, 180° columns).

IMP(D0–D5;G1) P /Pcr Pu /Pcr

IMP(D0–D5;G2) 1 G0
0.8 G3 G2 G1 (θ=0)
θ=105
0.8 L/750 L/1000 L/1250 0
IMP(D1–D5;G0)
D5 0.2t 0.768 0.797 0.817 0.916
D4 0.15t 0.770 0.799 0.820 0.929
0.6 D3 0.125t 0.771 0.801 0.821 0.936
IMP(D0–D5;G3)
D2 0.1t 0.773 0.803 0.823 0.944
w<0 w>0 ⏐w⏐/ t D1 0.075t 0.774 0.804 0.825 0.953
0.4
3 2 1 0 1 2 3 D0 0 0.775 0.806 0.827
(θ=90)

Fig. 15. Variation of (i) the P/Pcr vs. w/t paths and (ii) Pu/Pcr with the initial imperfection amplitudes, for h = 105°, 90° and 0° columns.

Indeed, this mode interaction stems from the occurrence of
horizontal shifts (towards the web) of the column cross-sec-
tion effective centroids, responsible for the development of
minor axis flexure (the axial compression becomes gradually
more eccentric) – these effective centroid shifts stem from
the progressive ‘‘weakening’’ (axial stiffness drop) of the
flange-lip assemblies, due to the presence of rather high lon-
gitudinal compressive normal stresses (e.g., [32]).15
(iv) The lower limit load exhibited by the h = 0° column is due to
the fact that it is associated with two half-waves involving
outward flange-lip motions, which develop higher compres-
sive stresses than their inward counterparts (e.g., [2,4]).16
(v) This distortional/global mode interaction did not occur in
the column single half-wave distortional post-buckling pre-
sented in Fig. 5(a) (the equilibrium paths exhibit no limit
points) – this is due to the much shorter column length
(about one third of that considered now), which corresponds
to a much higher global (minor axis flexure) buckling load
and, therefore, precludes the occurrence of interaction (the
cross-section ‘‘weakening’’ is insufficient to trigger it).
4.2.4. Influence of the initial imperfection amplitudes
In order to make it possible to assess the influence of the initial
imperfection amplitudes on the column elastic post-buckling
behaviour addressed above, Fig. 15 shows (i) the P/Pcr vs. w/t equi-
librium paths and (ii) the variation of Pu/Pcr for columns with
imperfection shapes corresponding to h = 105° (the most detrimen-
tal ones) and 15 different amplitudes. The initial imperfection
magnitudes stem from the combination of (i) five distortional
imperfection amplitudes (vD.0 = 0.075t, 0.1t, 0.125t, 0.15t, 0.2t –
amplitudes D1 to D5) and (ii) three ‘‘global’’ ones (vG.0 = L/1250,
15
This type of distortional/global mode interaction is qualitatively similar to the
well-know local-plate/global one, which is currently incorporated in virtually all steel
design codes through the ‘‘plate effective width’’ concept.
16
This limit load difference only exists because the number of distortional half-
waves is odd, which brings about the different behaviours exhibited by the outward
and inward ones. Moreover, this difference will obviously decrease as the (odd) half-
wave number increases.
L/1000, L/750 – amplitudes G1 to G3). The equilibrium paths asso-
ciated with each imperfection combination are identified by desig-
nations of the type IMP(Di; Gj) – note that the h = 105° column
previously analysed corresponds to IMP(D2; G2). In order to provide
a better feel of the column imperfection-sensitivity, Fig. 15 also in-
cludes equilibrium paths associated with pure distortional (h = 0°)
and pure global (h = 90°) initial imperfections – they are identified
by the designations G0 and D0, respectively. The observation of all
these equilibrium paths and ultimate strength values leads to the
following remarks:
(i) Varying the initial imperfection distortional and global
amplitudes does not change the column post-buckling behav-
iour qualitatively. Indeed, all equilibrium paths still (i1) exhi-
bit elastic limit points, (i2) merge into one of two common
descending branches (depending on whether the initial
imperfection is pure distortional or has a global component).
Moreover, the lower ultimate loads continue to belong to the
h = 105° columns, even if the differences with respect to their
h = 90° column counterparts remain minute (<0.3%).
(ii) Quantitatively speaking, the comparison between the results
concerning the h = 90° (pure global imperfections) and h = 0°
(pure distortional imperfections) columns shows that the
global imperfection amplitude has a larger impact on the
(elastic) ultimate strength. Indeed, the ultimate load reduc-
tions associated with the imperfect amplitude increases
from G1 to G3 and from D1 to D5 are equal to 7% and 4%,
respectively.
(iii) Concerning the h = 105° columns, whose initial imperfec-
tions are predominantly global (distortional and global com-
ponents with 25.9% and 96.5% of their maximum
amplitudes), their behaviours are naturally quite close to
those exhibited by their h = 90° counterparts. This is con-
firmed by the fact that increasing the initial imperfection
global amplitude from G1 to G3, while keeping vD.0 – 0 fixed,
also causes an ultimate load drop of about 7%. On the other
hand, varying the initial imperfection distortional amplitude
from D1 to D5, for a fixed vG.0 – 0, has almost no effect on
the ultimate load (it always drops by less than 1%). The
432 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434

Fig. 16. (a) P/Pcr vs. w/t elastic–plastic equilibrium paths of the h = 90°, 26° and 0° columns, and (b) plastic strain and deformed configuration evolution for fy/rcr  1.1.

previous assertions are illustrated in the zoomed portion of the h = 26° and (i3) the pure distortional (h = 0°) ones, and (ii) exhib-
Fig. 15, showing three equilibrium path sets, each of them iting different yield-to-critical stress ratios. As for Fig. 16(b1)–(b3),
corresponding to a different vG.0 value and comprising 5 they concern the columns with fy/rcr  1.1 and display nine plastic
paths (one per vD.0 value) – the differences are much smaller strain diagrams, corresponding to equilibrium states located along
within each set than between separate sets. their post-buckling paths (as indicated in Fig. 16(a1)–(a3)) and
including the column collapse mechanisms. Finally, Table 1 provides
the column ultimate load ratios (Pu/Pcr) associated with the h and fy
4.3. Elastic–plastic mode interaction
combinations considered in this study. The observation of these re-
sults prompts the following remarks:
In this section, a few results dealing with the elastic–plastic
post-buckling behaviour of simply supported lipped channel col-
(i) Out of the twelve column set analysed (i1) only those with fy/
umns experiencing D/G mode interaction are presented. These re-
rcr  1.1 exhibit (a minute) elastic–plastic strength reserve
sults concern columns (i) containing mostly the 13 initial
and (i2) those with fy/rcr  2.5 remain elastic up until the
imperfection shapes dealt with before (0° 6 h 6 180 – 15° intervals
ultimate (limit) load is reached – moreover, in the columns
and critical mode amplitudes of vD.0 = 0.1t and vG.0 = L/1000) and
with fy/rcr  1.7, the onset of yielding triggers the column
(ii) exhibiting three yield-to-critical stress ratios, namely fy/
failure.
rcr  1.1, 1.7, 2.5, which correspond to yield stresses equal to
(ii) In all h = 27–179° columns, yielding starts at the bottom lip
fy = 235, 355, 520 MPa, respectively – recall that rcr = 209.5 MPa.
mid-span zone, as illustrated in Fig. 16(b1) (diagram I –
For comparative purposes, some elastic results obtained earlier
h = 90° column). Collapse occurs almost immediately after
are presented again – they may be viewed as corresponding to
and is caused by the full yielding of the mid-span bottom
an infinite yield stress (i.e., fy = fy/rcr = 1).
web-flange corner, leading to the formation of a ‘‘distortional
Fig. 16(a1)–(a3) show the upper portions (P/Pcr > 0.6) of twelve
plastic hinge’’ that precipitates the collapse of the mid-span
equilibrium paths P/Pcr vs. w/t, describe the post-buckling behav-
bottom flange-lip assembly (see diagram II in Fig. 16(b1),
iours of columns (i) containing the most significant initial imper-
which also shows a yield line crossing the whole mid-span
fections, i.e., (i1) the pure global (h = 90° – the choice of this
cross-section bottom flange). Along the equilibrium path
particular initial imperfection shape is due to the fact that the elas-
descending branch, yielding spreads progressively along
tic column was found to exhibit one of the lowest limit loads17), (i2)
the lower web and bottom flange-lip central regions, while
all other column areas remain elastic (see diagram III in
17
Recall that the h = 105° column has the lowest limit load, differing 0.5% from the Fig. 16(b1)).
h = 90° column one.
 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434 433

Table 1 Initially, one addressed the lipped channel column (i) buckling
Variation of the ultimate load ratio Pu/Pcr with h and fy/rcr. behaviour and (ii) uncoupled distortional and global post-buckling
h fy/rcr h fy/rcr behaviours, a task that (i) made it possible to select the most
(°)
1.1 1.7 2.5 1
(°)
1.1 1.7 2.5 1 appropriate column length (i.e., that maximising the D/G interac-
tion effects) and (ii) ended up disclosing a few unexpected (and
0 0.850 0.931 0.945 0.945 90 0.721 0.797 0.807 0.807
15 0.826 0.921 0.937 0.937 105 0.711 0.792 0.803 0.803
surprising) features. Then, one presented and discussed several
30 0.807 0.912 0.909 0.909 120 0.716 0.797 0.807 0.807 numerical results concerning the elastic post-buckling behaviour
45 0.797 0.863 0.864 0.864 135 0.726 0.807 0.819 0.819 of lipped channel columns with the selected length (i.e., experienc-
60 0.764 0.831 0.836 0.836 150 0.754 0.831 0.839 0.839 ing strong D/G mode interaction effects) and containing critical-
75 0.740 0.811 0.817 0.817 165 0.792 0.864 0.870 0.870
mode initial imperfections with various configurations and the
90 0.721 0.797 0.807 0.807 180 0.854 0.940 0.960 0.961
same amplitude (linear combinations of the two competing
buckling mode shapes). These results consisted of (i) non-linear
(iii) In all h = 1–26° columns the plastic strain evolution is qual- equilibrium paths and (ii) figures providing the evolution, along
itatively quite similar to the one described in the previous those paths, of column and cross-section deformed configurations.
item – see Fig. 16(b2) for the h = 26° column. There is one Finally, the paper closed with a few post-buckling results concern-
important difference, though: yielding starts and spreads ing elastic–plastic columns with (i) the same geometry and initial
around the column regions located near the outer half-wave imperfection shapes and (ii) three yield stress values. Besides
crests, where the largest distortional deformations occur. the non-linear equilibrium paths and deformed configuration evo-
(iv) In columns containing pure distortional imperfections lution, one addressed also (i) issues related to the onset and spread
(h = 0° or 180°), the plastic strain evolution is qualitatively of plasticity, as well as (ii) the variation of the column ultimate
different from those described in the previous two items. strength with the initial imperfection shape and yield stress value.
Plasticity first appears at both lip free ends in the vicinity Among the various conclusions drawn from this investigation,
of the outer distortional half-wave crests. Collapse follows the following ones deserve to be specially mentioned:
shortly after and is due to the full yielding of both web-
flange corners (see diagram II in Fig. 16(b3)). (i) A GBT analysis revealed that a large portion of the column
(v) The ultimate load ratios given in Table 1 show that the var- critical buckling curve descending branch corresponds to
iation of Pu/Pcr with h is qualitatively similar for all column distortional–flexural–torsional buckling modes with relevant
sets. Indeed, one observes that (v1) the h = 105° column anti-symmetric distortional components – this contradicts
always exhibits the lowest value and that (v2) there is very the general belief that such column lengths are associated
little variation within the 90° 6 h 6 120° interval (the maxi- with global buckling. Thus, the columns analysed in this
mum and minimum Pu/Pcr are never more than 1.4% apart) – work are affected by the interaction between (i1) three
this confirms that, for practical purposes, one may view the half-wave symmetric distortional and (i2) single half-wave
pure global initial imperfection (h = 90°) as the most detri- (anti-symmetric) distortional–flexural–torsional modes.
mental one.18 (ii) The participation of the anti-symmetric distortional mode
(vi) The strength erosion stemming from the distortional/global just mentioned was shown to reduce the column post-criti-
mode interaction effects is quite considerable. For the cal strength and provided the explanation for the surprising
h = 90° column, the ultimate strength corresponds to 29% unstable ‘‘global’’ post-buckling behaviour exhibited by the
(fy = 235 MPa), 21% (fy = 355 MPa) and 20% (fy > 355 MPa) intermediate-to-long lipped channel columns.
drops with respect to the critical buckling load – it is interest- (iii) The equilibrium paths describing the post-buckling behav-
ing to notice there is no benefit in having a yield stress much iours of the columns affected by distortional/global (‘‘global’’
larger than fy = 355 MPa, since the column collapse is totally means ‘‘distortional–flexural–torsional’’) mode interaction
governed by elastic distortional/global interaction effects. exhibit features that vary considerably with the initial
(vii) The features addressed in the previous two items are bound imperfection shape. Those equilibrium paths can be grouped
to have far-reaching implications in the design of cold- in three categories, depending on whether the initial imper-
formed steel columns experiencing D/G interaction, mostly fection shape is (iii1) pure distortional, (iii2) predominantly
because the (uncoupled) distortional and global post-buck- distortional with outward outer half-waves or (iii3) none of
ling behaviours are looked upon as stable (even if only mar- the above – this last category comprises the vast majority
ginally) – to the authors’ best knowledge, no adverse mode of the post-buckling paths and led to the identification of a
interaction effects have ever been reported. ‘‘coupled buckling mode’’, defined by CG  0.45CD.
(iv) A key aspect concerning the distortional/global mode inter-
action is the influence exerted by the outer half-wave flange-
5. Concluding remarks lip motions (distortional feature) on the major axis flexure
(global feature). Indeed, (iv1) while inward flange-lip
This work dealt with a numerical (shell finite element) investiga- motions reduce the cross-section major moment of inertia,
tion on the elastic and elastic–plastic post-buckling behaviour and thus facilitating the occurrence of the corresponding flexure,
strength of simply supported cold-formed steel lipped channel col- (iv2) the outward flange-lip motions have precisely the
umns affected by distortional/global mode interaction. The analy- opposite effect.19
ses, performed in the code ABAQUS, involved columns containing (v) The columns containing pure distortional initial imperfec-
initial imperfections with shapes obtained by combining differently tions are affected by a different type of distortional/global
the two competing distortional (three half-waves) and global (one mode interaction phenomenon, which (v1) involves only
half-wave) buckling modes with amplitudes equal to (i) 10% of the minor axis flexure and (v2) is not caused by the closeness
wall thickness t (distortional mode) and (ii) L/1000 (global mode).
19
This aspect is particularly relevant because the (symmetric) distortional buckling
18
It is worth recalling that a pure outward distortional initial imperfection was mode exhibits three half-waves (2 outward and 1 inward or vice-versa) - if the mode
found to be the most detrimental in the context of simply supported lipped channel has an even or larger odd half-wave number, it should play a much lesser role (this
columns affected by local-plate/distortional mode interaction [11]. issue is currently under investigation).
434 P.B. Dinis, D. Camotim / Computers and Structures 89 (2011) 422–434
of two buckling loads – it stems from the occurrence of hor-
izontal shifts of the column cross-section effective centroids,
due to the ‘‘weakening’’ (axial stiffness drop) of the flange-
lip assemblies brought about by high compressive normal
stresses.
(vi) Regardless of the initial imperfection shape, all elastic equi-
librium paths exhibit limit points that take place below the
critical applied load level. Concerning the elastic–plastic
post-buckling behaviour, it was found that (vi1) the strength
erosion due to distortional/global modal interaction effects is
considerable, (vi2) there is virtually no elastic–plastic
strength reserve and/or ductility (the onset of yielding often
triggers the column failure) and, for yield stresses a bit larger
than fy = 355 MPa, (vi3) the column collapse is fully governed
by elastic distortional/global interaction.
(vii) For practical purposes, pure global initial imperfections may
be taken as the most detrimental ones, in the sense that they
lead to the lowest ultimate loads, both in the elastic and
elastic–plastic columns.
(viii) Finally, a minor imperfection-sensitivity study carried out
showed that the initial imperfection global amplitude has
a relatively small impact on the column elastic post-buck-
ling behaviour and ultimate strength – nevertheless, it is
far more relevant than its distortional counterpart.
Some of the features just described are bound to have far-reach-
ing implications in the design of cold-formed steel columns expe-
riencing distortional/global interaction, mostly because (i) the
uncoupled distortional and global post-buckling behaviours are
now looked upon as stable (even if only marginally) and (ii) no ad-
verse mode interaction effects between them had ever been re-
ported. The authors are currently investigating the influence of
this mode interaction phenomenon in the post-buckling and ulti-
mate strength behaviours of lipped channel columns with other
cross-section dimensions and support conditions – the completion
of this task should pave the way towards the development and val-
idation of a direct strength approach to design this type of
columns.
References
[1] Kwon Y, Hancock GJ. Postbuckling analysis of thin-walled channel sections
undergoing local and distortional buckling. Comput Struct 1993;49(3):507–16.
[2] Prola L, Camotim D. On the distortional post-buckling behavior of cold-formed
lipped channel steel columns. In: Proceedings of structural stability research
council annual stability conference (Seattle, 24–26/4); 2002. p. 571–90.
[3] Camotim D, Silvestre N. GBT distortional post-buckling analysis of cold-formed
steel lipped channel columns and beams. In: Program and book of abstracts of
17th asce engineering mechanics conference (Newark, 13–16/6); 2004. p. 38–
39 [full paper in Cd-Rom proceedings].
[4] Silvestre N, Camotim D. Local-plate and distortional post-buckling behavior of
cold-formed steel lipped channel columns with intermediate stiffeners. J Struct
Eng (ASCE) 2006;132(4):529–40.
[5] Batista EM. Local-global buckling interaction procedures for the design of cold-
formed columns: effective width and direct method integrated approach. Thin-
Wall Struct 2009;47(11):1218–31.
[6] Schafer BW. Direct strength method design guide. Washington, DC: AISI
(American Iron & Steel Institute); 2005.
[7] Schafer BW. Review: the direct strength method of cold-formed steel member
design. J Constr Steel Res 2008;64(7–8):766–78.
[8] Schafer BW, Peköz T. Laterally braced cold-formed steel members with edge
stiffened flanges. J Struct Eng (ASCE) 1999;125(2):118–27.
[9] Yang D, Hancock GJ. Compression tests of high strength steel columns with
interaction between local and distortional buckling. J Struct Eng (ASCE)
2004;130(12):1954–63.
[10] Ungureanu V, Dubina D. Recent research advances on ECBL approach (Part I:
plastic–elastic interactive buckling of cold-formed steel sections. Thin-Wall
Struct 2004;42(2):177–94.
[11] Dinis PB, Camotim D, Silvestre N. FEM-based analysis of the local-plate/
distortional mode interaction in cold-formed steel lipped channel columns.
Comput Struct 2007;85(19–20):1461–74.
[12] Camotim D, Dinis PB, Silvestre N. Local/distortional mode interaction in lipped
channel steel columns: post-buckling behaviour, strength and DSM design. In:
Proceedings of fifth international conference on thin-walled structures
(Brisbane, 18–20/6), vol. 1; 2008. p. 99–114.
[13] Kwon YB, Kim BS, Hancock GJ. Compression tests of high strength cold-formed
steel channels with buckling interaction. J Constr Steel Res
2009;65(2):278–89.
[14] Silvestre N, Camotim D, Dinis PB. Direct strength prediction of lipped channel
columns experiencing local-plate/distortional interaction. Adv Steel Constr –
Int J 2009;5(1):45–67.
[15] Dinis PB, Camotim D. Post-buckling analysis of cold-formed steel lipped
channel columns affected by distortional/global mode interaction. In:
Proceedings of SSRC annual stability conference (Nashville, 2–5/4); 2008. p.
405–31.
[16] Rossi B. Mechanical properties, residual stresses and structural behavior of
thin-walled stainless steel profiles. Ph.D. thesis, University of Liège; 2009.
[17] Thomasson PO. Thin-walled C-shaped panels in axial compression. Swedish
Council for Building Research, D1:1978.
[18] Höglund T. Design of trapezoidal sheeting provided with stiffeners in the
flanges and webs. Swedish Council for Building Research, D28:1980.
[19] Desmond TP, Peköz T, Winter G. Edge stiffeners for thin-walled members. J
Struct Eng (ASCE) 1981;107(2):329–53.
[20] European Convention for Constructional Steelwork (ECCS). Manual on the
stability of steel structures. 2nd ed. Brussels; 1976.
[21] Schafer BW, Peköz T. Computational modelling of cold-formed steel:
characterising geometric imperfections and residual stresses. J Constr Steel
Res 1998;47(3):193–210.
[22] Dubina D, Ungureanu V. Effect of imperfections on numerical simulation of
instability behaviour of cold-formed steel members. Thin-Wall Struct
2002;40(3):239–62.
[23] Maiorana E, Pellegrino C, Modena C. Imperfections in steel girder webs with or
without perforations under patch loading. J Constr Steel Res
2009;65(5):1121–9.
[24] Simulia Inc. Abaqus standard (vrs. 6.7-5); 2008.
[25] Dinis PB, Camotim D. On the use of shell finite element analysis to assess the
local buckling and post-buckling behaviour of cold-formed steel thin-walled
members. In: Soares CAM et al., editors. Book of abstracts of III European
conference on computational mechanics: solids, structures and coupled
problems in engineering (Lisboa, 5–9/6); 2006. p. 689 [full paper in Cd-Rom
proceedings].
[26] Schafer BW. Cufsm (version 2.6); 2003. <www.ce.jhu.edu/bschafer/cufsm>.
[27] Camotim D, Silvestre N, Dinis PB. Numerical analysis of cold-formed steel
members. Int J Steel Struct 2005;5(1):63–78.
[28] Camotim D, Silvestre N, Gonçalves R, Dinis PB. GBT analysis of thin-walled
members: new formulations and applications. In: Loughlan J, editor. Thin-
walled structures: recent advances and future trends in thin-walled structures
technology (International workshop – Loughborough, 25/6); 2004. p. 137–68.
[29] Camotim D, Silvestre N, Basaglia C, Bebiano R. GBT-based buckling analysis of
thin-walled members with non-standard support conditions. Thin-Wall Struct
2008;46(7–9):800–15.
[30] Bebiano R, Silvestre N, Camotim D. GBTUL 1.0b – code for buckling and
vibration analysis of thin-walled members; 2008. <http://www.civil.ist.utl.pt/
gbt>.
[31] Bebiano R, Silvestre N, Camotim D. GBTUL – a code for the buckling analysis of
cold-formed steel members. In: LaBoube R, Yu W-W, editors. Proceedings of
19th international specialty conference on recent research and developments
in cold-formed steel design and construction (St. Louis, 14–15/10); 2008. p.
61–79.
[32] Young B, Rasmussen KJ. Shift of effective centroid in channel columns. J Struct
Eng (ASCE) 1999;125(5):524–31.