Thin–Walled Structures 159 (2021) 107224

Contents lists available at ScienceDirect

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

Built-up back-to-back cold-formed steel compression members failing by

local and distortional buckling
Son Tung Vy a, b, Mahen Mahendran a, *, Thananjayan Sivaprakasam a
a
Queensland University of Technology (QUT), Brisbane, Australia
b
National University of Civil Engineering (NUCE), Hanoi, Viet Nam

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

Keywords: The current design methods for built-up back-to-back cold-formed steel (CFS) channel compression members are
Cold-formed steel considered inadequate despite several research studies in recent times. These research studies have proposed
Built-up members improved DSM based design equations, however, they have in general led to predictions of varying accuracy.
Back-to-back sections
Further, there were some contradicting observations regarding the effects of screw connections on their
Local buckling
Distortional buckling
compression behaviour and capacities. Therefore, this research focused on the compression behaviour and ca­
Screw connections pacity of built-up back-to-back CFS channel sections subject to local and/or distortional buckling. It included
compression tests of short single and back-to-back channel members failing by local-distortional interaction
buckling, development and validation of finite element models and finally a parametric study to examine the
effects of screw fastener characteristics and arrangements on the compression behaviour and capacities of back-
to-back channel members failing in local and/or distortional buckling. Using the results, the accuracy of
currently available design equations in the CFS design standards and recently published research papers was
investigated and a simple DSM based design method is recommended. This paper presents the details of this
research and the results including its findings.

1. Introduction Strength Method (DSM) for the calculations of CFS member compression
capacities. They state that the compression capacity of built-up members
Recently, cold-formed steel (CFS) applications are increasing in the can be determined using the same methods as single members, but with
construction of various residential and industrial buildings. In conven­ a modified slenderness ratio (le/r)m defined by Eq. (1) instead of using
tional applications, CFS structural members are normally made of single that of single/individual members (le/ri).
channel or zed sections, which have limited load-bearing capacities due ( ) √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )2 ( )2
to their low torsional and flexural rigidities. For applications with higher le le s
= + (1)
load-bearing capacities, single channel or zed sections are inadequate r m r o ri
and CFS built-up screw-connected back-to-back channel (BC) members
shown in Fig. 1 are often used by engineers. This is because these BC where (le/r)o is the overall slenderness ratio of the entire section about
members can be easily fabricated from commercially available single the built-up member axis, s is the intermediate fastener or spot weld
channel members, and they may have greater compression (or bending) spacing, and ri is the minimum radius of gyration of the full unreduced
capacities than the sum of their individual member capacities. In other cross-section of an individual member in the built-up section. This
words, they can provide an economical solution in many applications replacement results in the improvement of the flexural buckling stress,
where higher load-bearing capacities are needed. which may lead to an increase in the compression capacity of the built-
The BC members under compression loading can fail in one of the up member (to more than twice of its individual member capacity) if the
three failure modes: global (G), local (L) and distortional (D) buckling, failure mode is global buckling or global-local interaction buckling.
or an interaction among these three failure modes. The two major cold- The CFS design standards [1,2] are important for the practical design
formed steel design standards in North America and Australia/New of BC members, however, they still have many limitations. Firstly, Stone
Zealand [1,2] use the Effective Width Method (EWM) and the Direct and LaBoube [3] showed that using the modified slenderness ratio in

* Corresponding author.
E-mail address: m.mahendran@qut.edu.au (M. Mahendran).

https://doi.org/10.1016/j.tws.2020.107224
Received 26 April 2020; Received in revised form 30 August 2020; Accepted 14 October 2020
Available online 14 November 2020
0263-8231/© 2020 Elsevier Ltd. All rights reserved.
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

and global buckling modes. Thus, they proposed a novel DSM-based

design method for BC members that may fail due to these interaction
buckling modes. They also showed that the effect of screw spacing on the
local and distortional buckling capacities of BC members is insignificant.
Additionally, recent research studies [9,10] have developed compound
strip method based models, which can include the effects of discrete
screw fasteners and provide the elastic buckling loads of BC members.
Abbasi et al. [11] used these models and DSM to predict the compression
capacities of BC members.
Fratamico et al. [12] conducted a compression test series of short BC
members failing by local or local-distortional buckling mode. They used
the test results to (1) investigate the effects of screw spacing on the web
buckling compatibility, and (2) show that the screw fasteners have a
marginal effect on those members’ capacities, which agreed with the
Fig. 1. Single channel and built-up back-to-back channel sections.
findings of [5,8]. Then, Fratamico et al. [13] used an experimental study
to (1) show the considerable effects of the end fastener groups on the
predicting the compression capacity of BC members is over-conservative compression capacities of long BC members that fail mainly by
with differences between test results and design standard predictions local-global interaction (or global) buckling, and (2) propose a
varying in the range of 16–65%. Roy et al. [4] showed that the current DSM-based design method for these members. Besides, they stated that
design standards are unconservative by about 12% for short BC composite action can develop in BC members when global (flexural)
compression members failing mainly by local buckling. Secondly, while buckling controls. Meanwhile, Fratamico et al.’s [14] test results showed
the modified slenderness ratio is only useful to calculate the minor axis that (1) BC compression members lined with oriented strand boards
flexural buckling stress of BC members, no guidelines are given to (OSB) on both sides failed by local buckling, and (2) the steel-to-steel
determine their torsional or flexural-torsional buckling stresses. Thirdly, screw fasteners did not influence considerably the member failure
the DSM design guidelines for predicting the compression capacities of modes and capacities. Based on these results, they developed suitable
BC members failing in local or distortional buckling mode are not given design approaches using an extension of DSM.
in these standards, which suggest the EWM design method of single Recently, Kechidi et al. [15] developed advanced FE analyses of
channel members without any justification. Finally, no guidelines are unlined and OSB-lined BC members, verified the experimental findings
available for the required number and arrangement of screws along each in Refs. [13,14] and showed that the current standard requirements for
row in BC members, which may result in different assumptions as shown the shear capacity of screws in BC compression members are
in Fig. 2. over-conservative. Some other research studies, such as [16,17], pro­
To overcome the limitations of the current CFS design standards [1, posed improved formulae to predict the flexural stiffness and the elastic
2] and enhance the knowledge and understanding of the compression flexural buckling stress of BC members. These improved formulae pro­
behaviour and capacity of BC members, many research studies have vided more accurate predictions than the standard method based on the
been conducted in recent years. Li et al. [5] developed finite element modified slenderness ratio. Furthermore, improved EWM-based design
(FE) models of BC members and showed that (1) the compression ca­ methods were proposed by research studies [4,18]. Ting et al. [19]
pacities of short BC members, which fail in local buckling, cannot be investigated the effect of screw spacing on the compression behaviour
greater than the sum of their individual member capacities, (2) for BC and capacities of BC members with varying slenderness ratios. They
members buckling about the symmetric axis, the torsional or found that the compression capacities of short BC members, failing in
flexural-torsional buckling does not have to be considered, and (3) the local buckling, might reduce by about 5–10% when the screw spacing is
effect of screw spacing on their compression capacities is not evident. doubled, which contradicts the finding in Refs. [5,8,12].
They also proposed new design methods of BC compression members This paper focuses on short back-to-back channel (BC) compression
based on the EWM and DSM. Zhang and Young [6] conducted members failing in local and/or distortional buckling mode. These
compression tests of BC members made of single channel sections with failure modes can govern the compression capacity of BC members,
edge and web stiffeners. Using these test results, Zhang and Young [7] when they are used as short truss members, or studs in sheathing
developed and validated FE models and proposed DSM-based design restrained light gauge steel frame (LSF) wall panels [14,15]. Some past
methods for these members. Lu et al. [8] conducted a compressions test research studies [3,4] have shown that the current design standards are
series of BC members made of plain lipped single channel members, inadequate while some others [5,7,8] have proposed improved DSM
where many specimens failed by the interaction of local, distortional design rules based on different assumptions and predicted capacities
with varying levels of accuracy. Besides, there were contradicting ob­
servations regarding the effects of screw fasteners on the compression
capacity of short BC compression members [5,8,12,19]. Therefore, the
predictions of all these new design guidelines and observations will be
investigated by using experimental tests and FE analyses in this
research, aimed at developing suitable design guidelines for short BC
compression members failing in local and/or distortional buckling
mode.
In this research, compression tests were first conducted on short
single channel and BC members failing in local-distortional interaction
buckling mode. Finite element (FE) models of tested columns were then
developed and validated to provide a base for conducting parametric
studies of BC members with different section shapes, sizes and screw
arrangements, and failing by local, distortional or local-distortional
interaction buckling. The results from tests and FE analyses are pre­
sented in this paper, based on which, the effects of screw sizes, screw
Fig. 2. Different number of screws and arrangements in BC members. spacing and number of screws along each row are discussed. More

2
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

importantly, the DSM design guidelines in the current standards [1,2] Table 1
and past research studies [5,7,8] are investigated before recommending Dimensions of SC87D section.
a reliable and simplified DSM based approach. Dimension Value (mm) Dimension Value (mm)

h1 13.42 b1 35.71
2. Experimental investigation h2 2.09 b2 0.80
h3 56.24 b3 39.30
2.1. Test specimens h4 2.06 b4 0.89
h5 3.48 d1 7.01
h6 1.62 d2 8.63
A commercially available channel section (SC87D) that is commonly h7 8.31
used as LSF wall studs in Australia was selected in this experimental
investigation. It was used as single channel and back-to-back channel
(BC) members in this investigation. This section shown in Fig. 3 is made
of a high strength steel grade, G550 (minimum yield strength of 550
MPa), and its base metal thickness without zinc-coating is 0.95 mm.
Table 1 provides the average measured dimensions of the channel sec­
tions used in this investigation. They are designed as asymmetric sec­
tions to enable them to be nested as closed sections for use in
applications with higher compression capacities. This also minimizes
any unwanted imperfections during transportation.
The specimen length was chosen as 800 mm, which was 200 mm
greater than the actual length (600 mm) used in the capacity calcula­
tions and FE models. This was because in the test set-up, 100 mm seg­
ments of specimens were clamped inside a specially designed connector
at each end. To limit the length tolerance and any non-uniform stresses
applied to the specimen, each specimen was cut to the specified length of
800 mm and milled flat at both ends. Fig. 4. BC87D section with one and two screws along each screw row.
To fabricate a BC member as shown in Fig. 4, two single channel
members were connected together by using 10 g self-drilling hex-head • for single channel members, the label “SC87D-600” or “SC87L-600”
screws. Three different screw arrangements were used as shown in indicates that channel depth is 87 mm and length is 600 mm, and the
Fig. 5. The relative positions of channel sections were adjusted carefully letter “D” or “L” means that distortional or local buckling mode is
to ensure that the built-up member was doubly symmetric and its ends dominant, however, “D” also includes local-distortional interaction
were flat. The fabricated BC members formed two groups based on the buckling mode.
number of screws along each row, i.e., one or two screws at the mid- • for BC members, the label “BC87D-600-1S-150” or “BC87L-600-2S-
depth of the web element. The screw spacing (s) used was 150 or 300 150” indicates that individual channel depth is 87 mm, length is 600
mm with the first row of screws from each end located at 15 mm. mm and screw spacing is 150 mm, while the letter “D” or “L” means
For ease of identification, all specimens were labelled based on their distortional or local buckling mode is dominant in its individual
cross-sectional parameters, lengths, screw spacing and other specific section with “D” also including local-distortional interaction buck­
features. For example. ling mode, and “1S” or “2S” means there are one or two screws per
row.

2.2. Test set-up and procedure

Fig. 6 shows the test set-up used in this investigation, where a single
or BC channel specimen was vertically loaded in compression under a
stable and strong test frame fixed to a strong reinforced concrete slab.
The 100 mm long bottom end segment of the specimen was fixed inside a
specially designed steel connector, attached to a heavy beam located at
the bottom. Similarly, the 100 mm long top end segment of the specimen
was fixed inside another connector, located under a load cell. A 200 kN
hydraulic jack placed above the load cell was used to apply the
compression load to the specimen. The top end of the hydraulic jack was
fixed to the heavy beam of the loading frame, which ensured that it was
kept stable during the test. The applied compression load and associated
axial shortening of the specimen were recorded by the load cell and a
draw-wire displacement sensor located between the hydraulic jack and
the top connector, respectively.
The use of specially designed top and bottom connectors shown in
Figs. 7 and 8 ensured that end boundary conditions of each single
channel or BC specimen are fixed. Each connector included four 20 mm
thick steel side plates welded together, and then welded to a 10 mm
thick steel end plate to form a box. The end plate was used to stop the
axial displacement of the specimen’s end-segment. Besides, there were
four M12 bolts (restraining bolts) attached on each steel side plate by
nuts welded to the side plate. They supported steel plates F1 and F2, and
Fig. 3. SC87D section. steel blocks B1 and B2 in preventing the lateral movements, bending and

3
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 5. Screw fastener locations in BC members.

between the specimen and the connectors’ end plates, an axial

compression load of about 3 kN was applied to the specimen using the
hydraulic jack, and was then released. The test then commenced when
the compression load was applied to the specimen using the displace­
ment control method at a rate of 1 mm/min. During the test, a speci­
alised computer program, named UDAQ, was used for collecting the data
of the applied load and shortening of the specimen at 1 s intervals. The
test was terminated following a reasonable unloading phase.

2.3. Test results

Fig. 9 shows the deformed shapes of specimens after the tests, and
their failures are dominated by distortional buckling as expected.
However, the deformed shapes exhibiting flange rotations along the
flange-web junction and local buckling deformations at some positions
of the web indicate that SC87D specimens have failed by the interaction
of local and distortional buckling (LD) modes. BC87D specimens have
also failed by LD interaction buckling mode, but distortional buckling is
more dominant since flange rotations could be clearly seen than local
web buckling deformations. Fig. 10 shows the compression load versus
shortening curves for the tested specimens, while Table 2 presents their
ultimate loads (Nc,Test). These results show that the compression capac­
ities of BC87D-600-1S-150 (average 91.2 kN), BC87D-600-2S-150
(average 94.9 kN) and BC87D-600-2S-300 (average 90.7 kN) are
nearly similar (the differences are around 5%), and are almost equal to
twice the capacity of SC87D-600 (average 47.6 kN). This shows that the
effects of using two screws instead of one screw or changing the screw
spacing on the compression capacity are marginal, and the compression
capacity of BC members can be taken as twice that of single channel
members.
Fig. 6. Test set-up.
3. Finite element modelling

twisting of the specimen’s end-segment. The thickness of these steel

3.1. General
plates was designed as 10 mm while the width of the steel blocks was
0.5–1 mm smaller than the inner depth of the specimen’s cross-section to
The finite element (FE) program Abaqus/CAE version 2019 was
eliminate the effect of the gaps between the steel blocks and the flanges
utilized to develop the FE models of tested specimens and to predict the
of specimens. Since the stiffness of the flanges and the webs of the
behaviour and compression capacities of CFS single channel and BC
specimen’s end-segment were ignorable compared with that of the steel
members. Following the guidelines in Ref. [20], the FE models included
plates, the steel blocks and the restraining bolts, the flange and web
the accurate section geometry (Fig. 11), material properties, initial
elements were fully restrained against the lateral movements, bending
geometric imperfections and end boundary conditions of tested speci­
and twisting. Based on the same mechanism, the lips of the specimen’s
mens. For the case of the BC members, the nonlinear contact between
end-segments were also restrained by steel plates S1 to S4, attached to
elements and the nonlinear shear load versus slip displacement curve of
steel plates F1 and F2.
screw fasteners were also incorporated. The FE analyses included two
Before each test, the end plates of top and bottom connectors were
steps. Eigenvalue buckling analyses of single channel and BC members
levelled, and their central points and the hydraulic jack axis were
were performed first. Based on their first eigenvalue buckling modes,
ensured to be on a vertical line. When the specimen was clamped inside
nonlinear analyses were conducted by incorporating the initial geo­
the top and bottom connectors, the length and the compression load of
metric imperfections to obtain the ultimate loads and deformed shapes.
each restraining bolt were adjusted carefully to ensure the specimen was
Finally, because of the complexity of the section geometry and contacts
vertical and placed at the central points of the connectors’ end plates.
in each BC member, the Static/General solver was applied where the
The verticality of the test specimen was checked by two laser levels
specific damping factor was 0.0005 and the maximum ratio of stabili­
located on the front and left sides of the test frame. To minimize the gaps
zation was 0.005.

4
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 7. Specimens fixed inside the bottom connector: (a) single channel and (b) BC.

Fig. 8. Sketches of specimens fixed inside the bottom and top connectors: (a) single channel and (b) BC.

Fig. 9. Deformed shapes of SC87D and BC87D specimens from Tests and FE analyses.

5
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 10. Load versus axial shortening curves for SC87D and BC87D specimens.

3.3. Material properties

Table 2
Comparison of test and FE analysis results of SC87D and BC87D specimens.
The material properties were obtained from the tensile coupon tests
Specimen Test Failure Nc,Test Nc,FEA Nc,FEA/Nc, of 0.95 mm thick G550 steel, conducted by Rokilan and Mahendran
mode (kN) (kN) Test
[21]. The measured yield and ultimate strengths were 615 MPa and 625
SC87D-600 1 LD 46.32 47.20 1.02 MPa, respectively, while the elastic modulus was 205,000 MPa. The
2 LD 48.89 47.20 0.97
Poisson’s ratio was taken as 0.3. Fig. 13 shows the stress versus strain
BC87D-600-1S- 1 LD 91.85 91.35 0.99
150 2 LD 87.87 91.35 1.04 curve of 0.95 mm G550 steel obtained from the tensile coupon tests and
3 LD 93.86 91.35 0.97 the corresponding true stress versus strain curve used in the FE models.
BC87D-600-2S- 1 LD 95.71 91.39 0.95 Using the recommendations of Schafer and Peköz [22], the higher yield
150 2 LD 94.04 91.39 0.97 stress in the corner regions was ignored as the effect of residual stresses
3 LD 95.08 91.39 0.96
BC87D-600-2S- 1 LD 93.19 91.14 0.98
was excluded in this study (Section 3.6).
300 2 LD 88.21 91.14 1.03
3.4. Boundary conditions and loading method
Average 0.99
COV 0.03
According to the test set-up used, fixed end support boundary con­
ditions were used in all FE models of single channel and BC members
3.2. Element type and mesh size considered in this study (Fig. 14). Firstly, each end cross section was
constrained by using MPC type TIE to a reference point (RP-1 or RP-2)
The S4R element was used to model the elements of each member. located at its centroid. Then, all displacements (Ux, Uy and Uz) and ro­
This is a four-node (with six degrees of freedom at each node), quadri­ tations (URx, URy and URz) of the bottom reference point (RP-1) were
lateral, stress/displacement shell element with reduced integration and restrained. The top reference point (RP-2) was set similarly except
a large-strain formulation. The S4R element is suitable for a wide range releasing the axial displacement (Uz). Displacement control was used to
of applications including thin-walled structures and is computationally apply the axial compression loading via RP-2.
efficient. Using a convergence study, a mesh of 5 mm × 5 mm was
selected (Fig. 12).

Fig. 11. FE models of test specimens: (a) single channel, (b) BC.

6
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 12. FE mesh of single channel and BC members.

3.5. Interaction and constraint in BC members

Due to the complex shape of the BC87D section, the interactions

between elements in each member were complicated. Hence, the “gen­
eral contact” with “hard contact” normal behaviour (ensuring no surface
overlapping) was used in the FE models of BC members. The friction
coefficient between individual members was taken as 0.19 [15]. Besides,
the Abaqus “point-based” fasteners, as shown in Fig. 14, were used to
model screw connections in BC specimens. The property of the
“point-based” fasteners was defined as using “basic” connector section
type “Cartesian” translational. This connector section can include the
shear behaviour of each screw connection in FE models of BC members
by incorporating the shear load versus slip displacement curves of the
screw connection along the transverse and longitudinal directions of the
webs of BC members. These shear load versus slip displacement curves
Fig. 13. Stress-strain curves of 0.95 mm thick G550 steel [21]. can be determined by using tests or FE analyses. Recently, Phan and

Fig. 14. Boundary conditions and loading in FE models of specimens: (a) SC87D, (b) BC87D-1S-150, (c) BC87D-2S-150.

7
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 15. Shear load versus slip displacement curves for screw fasteners and 0.95 mm steel plates.

Fig. 16. Sections used in Group 1 parametric study.

8
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Rasmussen [23] proposed equations to determine these curves, which 3.7. Validation of the developed finite element models
are based on the elastic modulus (E) and thickness (t) of steel plates, and
the screw diameter (d). Using their proposal, the required shear load The developed FE models were validated based on failure modes,
versus slip displacement curves were derived as shown in Fig. 15, where load-displacement curves and ultimate loads. Fig. 9 exhibits good sim­
the screw sizes vary from 6 g to 14 g for 0.95 mm thick steel plates with ilarity between the deformed shapes of specimens observed in tests and
elasticity modulus of 205,000 MPa. FE analyses. It also shows that the failure mode of specimens predicted
by FE analyses is local-distortional interaction buckling as observed in
3.6. Initial geometric imperfections and residual stresses the tests. As shown in Fig. 10 and Table 2, the ultimate loads predicted
by FE analyses agree well with those from the tests. In Fig. 10, the
The single channel and BC members considered in this study are patterns of the load versus shortening curves from the tests and FE an­
subjected to distortional or local buckling failure. Hence, their FE alyses are similar. However, the slopes of the load versus shortening
models included the initial geometric imperfections based on these curves from FE analyses were higher than those from test results. The
√̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅
modes, with their magnitudes as δD = 0.3t fy /foD or δL = 0.3t fy /foL reasons for this difference are considered to be (1) modelling only the
for distortional or local buckling mode, respectively, where t is the clear length of 600 mm, (2) additional local bearing deformation of the
section thickness, fy is the yield stress, foD and foL are the distortional and clamped segments and (3) localized imperfections and variations along
local buckling stresses of single channels, respectively [2]. CFS members the specimen length. Overall, the developed FE models are considered
can also be subjected to residual stresses caused by cold-forming pro­ reliable for predicting the compression capacity and failure mode of
cess. However, the residual stresses were not included in FE models, single channel and BC members.
because some research studies [24–26] have shown that residual stresses
only have a small effect on the ultimate loads of CFS members.

Fig. 17. Sections used in Group 2 parametric study.

9
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

4. Parametric study BC87L-600-2S-150 members remained almost unchanged, despite the

varying shear load-slip displacement characteristics (Fig. 15). Hence it is
4.1. General concluded that screw fastener characteristics do not affect the
compression capacities of BC members if their failures are dominated by
Based on the validated FE models, parametric studies were con­ distortional and/or local buckling.
ducted to investigate the effects of important parameters, such as the As shown in Fig. 19, when the screw spacing was reduced, the level
screw size, screw spacing and number of screws along each row, on the of the web buckling compatibility in BC87D members increased. For
compression behaviour and capacity of short BC members. Twelve CFS BC87D members with a screw spacing of 60 mm, which is less than the
sections shown in Figs. 16 and 17 were used, which are made of single local and distortional buckling half-wavelengths (69.1 and 283.7 mm,
channel sections subject to either distortional buckling (Group 1 except respectively) of SC87D section, the full web buckling compatibility was
SC87D sections used in the experimental study, which fail by local- not achieved since two webs of the individual members still buckled
distortional interaction buckling) or local buckling (Group 2). These separately between two rows of screws. The web of BC87D-600-2S-30
sections had three different depths of 75, 87 and 120 mm and thick­ member, using very small screw spacing, achieved the full buckling
nesses of 0.75, 0.95, 1.00 and 2.20 mm. The member lengths (L) were compatibility, where both webs of the individual members exhibited the
selected in the range of 600–800 mm. The selected SC75D, SC87D and same buckling deformed shape. The same observations were also noted
SC120D sections are based on the channel sections used commonly in for most of the other BC members.
LSF walls in Australia, while the selected plain channel sections have Table 5 summarizes the failure modes and compression capacities
similar ratios of section height to flange width and flange width to lip (Nc,FEA) of the investigated BC members, and the ratios of their capac­
width as the commonly used commercial products. The thicknesses were ities to the sum of their individual member capacities (Nc,FEA/ΣNc,FEA-i).
chosen so that the sections failed in either distortional or local buckling Despite being connected by screw fasteners, these BC members failed by
modes. The elastic local and distortional buckling loads (Nol and Nod), local or distortional buckling if the behaviour of individual members
half-wavelengths, dominant failure modes and compression capacities was dominated by local or distortional buckling mode, respectively.
(Nc,FEA-i) of the single channel sections were determined by using the Further, when the screw spacing was reduced from L/2 to L/20, the
finite strip program CUFSM [27] and FE analyses, and are summarized compression capacity of BC members increased slightly, while the
in Table 3. The shear load versus slip displacement curves of the screw highest and lowest value differed by 1–8%. Similarly, when the number
fasteners shown in Fig. 15 were modified for varying channel section of screws along each row was increased from one to two, the compres­
thicknesses using the proposal given in Ref. [23]. The material proper­ sion capacity of BC members remained almost the same.
ties of all investigated members were assumed to be the same as those of Table 5 also shows that the compression capacities of BC members
the tested specimens (Fig. 13). To investigate the effects of screw size, FE subject to local buckling, such as BC75L, BC87L and BC120L members,
analyses of BC75D-600-2S-150, BC87D-600-2S-150 and are equal to about the sum of their individual member capacities, despite
BC87L-600-2S-150 members were conducted with screw sizes in the the presence of screw fasteners. The same observation is also made for
range of 6 g–14 g, as well as with simplified rigid or flexible connectors. BC87D members, which fail by local-distortional interaction buckling.
The shear behaviour of the flexible connector was assumed to follow According to Fratamico and Schafer [28], the composite action in a BC
that of 6 g screws, however, its stiffness was reduced by 95% (Fig. 15). member can increase the compression capacity to be more than twice
To investigate the effects of the remaining parameters, the screw spacing the capacity of its individual members, but the full composite action
of investigated BC members was varied from L/20 to L/2 while the happens in a BC member only when its web elements are fully connected
number of screws per row was either one or two. along the member length. Thus, the composite action was not achieved
in BC75L, BC87L, BC120L and BC87D members. However, the
compression capacities of BC75D and BC120D members subject to
4.2. Results and discussion
distortional buckling are higher than the sum of their individual member
capacities by up to 16%. Hence, in these members, a partial composite
Fig. 18 shows the typical deformed shapes of investigated BC mem­
action must have been achieved. Besides, the relationship between their
bers at their ultimate compression loads. This confirms that failures of
capacities and screw spacing is not evident. This study assumes that the
BC75D and BC120D members are dominated by distortional (D) buck­
combined effects of screw fasteners and interaction between the webs of
ling while members in Group 2 and BC87D members fail by local (L) and
individual members can increase the distortional buckling stress of BC
local-distortional (LD) interaction buckling, respectively.
members since they can partly restrain the minor web deformation in
As shown in Table 4, when the screw size was varied from 6 g to 14 g,
the distortional buckling mode with longer half-wavelength.
or the screw fasteners were modelled by rigid or flexible connectors, the
The current design standards [1,2] state that the ratio of the screw
compression capacities of BC75D-600-2S-150, BC87D-600-2S-150 and

Table 3
Additional information of single channel sections used in parametric study.
Section Elastic buckling loads Half-wavelengths (mm) FE analyses

Nol (kN) Nod (kN) Local buckling Distortional buckling Failure mode Nc,FEA-i (kN)

SC75D-600 50.95 39.76 83.90 287.70 D 50.68

SC75D-800 50.95 39.76 83.90 287.70 D 49.44
SC87D-600 25.02 32.48 69.10 283.70 LD 47.20
SC87D-800 25.02 32.48 69.10 283.70 LD 43.78
SC120D-600 341.71 245.82 167.70 435.80 D 243.61
SC120D-800 341.71 245.82 167.70 435.80 D 234.22
SC75L-600 13.78 33.98 58.50 536.40 L 37.81
SC75L-800 13.78 33.98 58.50 536.40 L 37.79
SC87L-600 25.45 53.67 70.10 669.40 L 59.44
SC87L-800 25.45 53.67 70.10 669.40 L 59.97
SC120L-600 19.60 51.54 93.70 790.30 L 68.59
SC120L-800 19.60 51.54 93.70 790.30 L 67.71

10
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 18. Deformed shapes of BC members at their ultimate compression loads: (a) BC75D, (b) BC87D, (c) BC120D, (d) BC75L, (e) BC87L and (f) BC120L (Note that
the deformation magnitudes are scaled by a factor of three).

Table 4
Effect of screw size on the compression capacity.
Specimen Screw size Screw diameter (mm) Nc,FEA Specimen Screw size Nc,FEA Specimen Screw size Nc,FEA
(kN) (kN) (kN)

BC75D-600-2S-150 Flexible – 105.19 BC87D-600-2S-150 Flexible 91.07 BC87L-600-2S-150 Flexible 115.29

6g 3.5 107.67 6g 91.38 6g 115.23
8g 4.2 107.84 8g 91.46 8g 115.26
10 g 4.7 107.97 10 g 91.39 10 g 115.28
12 g 5.4 108.21 12 g 91.48 12 g 115.29
14 g 6.3 108.32 14 g 91.19 14 g 115.18
Rigid – 108.64 Rigid 93.36 Rigid 115.28

spacing to the minimum radius of gyration of an individual member

If λl ≤ 0.776: Ncl = Nce (5)
(s/ri) in a built-up member should not exceed one-half of the governing
slenderness ratio of the built-up member. Using this, the screw spacing [ ( )0.4 ]( )0.4
Nol Nol
of the investigated BC members in Table 4 should not be greater than If λl > 0.776: Ncl = 1 − 0.15 Nce (6)
L/3 to L/5. As the above discussions on the effects of screw spacing Nce Nce
indicate, this design rule is conservative for short BC members. √̅̅̅̅̅̅̅
Nce
λl = (7)
5. Direct strength method based design guidelines Nol

If λd ≤ 0.561: Ncd = Ny (8)

5.1. DSM based design guidelines in the current design standards [1,2]
[ ( )0.6 ]( )0.6
Nod Nod
The compression capacity of a CFS member is the least of the global If λd > 0.561: Ncd = 1 − 0.25 Ny (9)
Ny Ny
buckling capacity (Nce), local buckling capacity (Ncl) and distortional
buckling capacity (Ncd), where Nce is determined by: √̅̅̅̅̅̅̅
Ny
( ) λd = (10)
(2)
2
For λc ≤ 1.5: Nce = 0.658 λc Ny Nod

( ) where Nod and Nol are the elastic buckling loads in distortional and local
0.877
For λc > 1.5: Nce = 2
Ny (3) buckling mode, respectively.
λc For CFS built-up members, including BC members, the DSM design
√̅̅̅̅̅̅̅ guidelines are not available in the CFS standards [1,2]. In this study,
Ny three methods are proposed for predicting the compression capacities of
λc = (4)
Noc BC members failing in local or distortional buckling mode.
where Ny is the yield capacity, Noc is the least of the elastic compression
• Method 1: Nc,Prop-1 = min (2Ncd, 2Ncl) where Ncd and Ncl are deter­
member buckling load in flexural, torsional and flexural-torsional
mined using Eqs. (5)–(10), while Noc is the least of the elastic flex­
buckling.
ural, torsional and flexural-torsional buckling loads. Note that the
Ncl and Ncd are determined by:
elastic flexural buckling load is determined by using the modified
slenderness ratio of the BC member (Eq. (1)), Noc = Ag π 2 E/[(le/r)m]2,

11
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 19. Web buckling compatibility in BC87D members with varying screw spacing (Note: the longitudinal cut plane is located at mid-height of the cross section).

where Ag is the gross area of BC section, while the elastic torsional η = [λ/(w/t) − 1.4] ​ /0.6 if 1.4 < λ/(w/t) ≤ 2.0 (13)
and flexural-torsional buckling loads are calculated as twice those of
its individual member. Section 4.2 showed that the compression η = 1 if λ/(w/t) > 2.0 (14)
capacity of BC members subject to distortional buckling might be up
/
to 16% higher than that of its individual members. However, this is λ = Lyi ryi (15)
ignored and Ncd is calculated by using the distortional buckling stress
of the individual member.
• Method 2: Nc,Prop-2 is determined as for Method 1, but Noc is the
elastic flexural buckling load, determined by using the modified where Lyi, Iyi and ri are the effective length, second moment of area and
slenderness ratio of the BC member. radius of gyration about the weak axis of an individual member,
• Method 3: Nc,Prop-3 is determined as for Method 1, but Noc is twice the respectively; A is the individual member’s cross-sectional area; d is the
elastic flexural buckling load of the individual member (single distance between the individual section’s and built-up section’s cen­
channel member). troids; w and t are the flange width and thickness, respectively.

5.2. DSM based design guidelines proposed by past research studies • For members buckling about the symmetry axis, the torsional or
flexural-torsional buckling does not have to be considered.
Li et al. [5] proposed design guidelines to determine the compression • The member distortional capacity shall be determined in accordance
capacity of BC members as follows: with the related provisions in the current design standards [1,2].

• The built-up member capacity is the sum of its individual member Zhang and Young [7] proposed two design methods for BC members
capacities. For built-up members of length less than 400 mm, a with longitudinal web stiffeners (Fig. 20 (a)). The first method used Eqs.
strength reduction factor of 0.9 is recommended to take into account (5)–(10) in the current standards [1,2] while the second method used
the effect of installation error. modified Eqs. (16)–(19) that consider the local instability in multiple
• To determine the elastic flexural buckling load of BC members, the compression elements. Besides, the elastic local and distortional buck­
second moment of area about the weak axis (Iy) of an individual ling loads are determined by a rational buckling analysis of the sections
member shall be calculated as follows: as shown in Fig. 20 (a) while the flexural buckling load is determined by
using the modified slenderness ratio of the BC member, (le/r)m.
Iy = Iyi + ηAd2 (11)
If λl ≤ 0.673: Ncl = Nce (16)
η = 0 if λ/(w/t) ≤ 1.4 (12)

12
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Fig. 20. Back-to-Back sections investigated by: (a) Zhang and Young [7], (b) Lu et al. [8].

[ ( )0.5 ]( )0.5
Nol Nol DSM design guidelines in the current design standards [1,2] and past
If λl > 0.673: Ncl = 1 − 0.22 Nce (17) research studies [5,7,8] are presented. The finite strip program CUFSM
Nce Nce
was used to determine the elastic local and distortional buckling loads
If λd ≤ 0.761: Ncd = Ny (18) required for each method (Fig. 21). To conduct the buckling analyses for
the proposals of Zhang and Young [7], the solid stiffeners were replaced
[ ( )0.6 ]( )0.6 by flat elements at the screw locations with the thickness modified to t +
Nod Nod
If λd > 0.761:Ncd = 1 − 0.2 Ny (19) 6(t)0.5. Meanwhile, the least of the capacity predictions, using their two
Ny Ny
design methods, was considered as the BC member capacity. For the
Lu et al.’s [8] design guidelines are based on determining the elastic buckling analyses of BC members in other design methods, individual
local and distortional buckling loads by the rational buckling analysis of channel member was modelled in CUFSM.
the section shown in Fig. 20 (b), and the elastic flexural buckling load by Table 5 summarizes the ratios of the BC members’ compression ca­
using the overall slenderness ratio of the entire BC section, (le/r)o. The pacities predicted by the above design methods to those by FE analyses.
compression capacity (Nc) of BC members was determined by: Note that the effective length factor was 0.5 since the support conditions
( *
Nc = min Nce , Ncle , Ncd
) *
if Nol > Nod or Ncd > Ncle (20) were chosen as fixed-ended. Table 5 also compares the dominant failure
modes of these members predicted by the design methods and FE ana­
( ) lyses. Since the design method proposed by Li et al. [5] provided the
*
Nc = min Nce , Ncle *
, Ncd , Nclde *
if Nol < Nod and Ncd < Ncle (21)
same predictions as the third method proposed in this paper, Table 5
* * does not include Li et. al.’s [5] predictions. Eqs. (11)–(15) in Li et al.’s
where Ncle and Nclde are the member compression capacities for the local-
[5] design method are likely to be more suitable for BC members subject
global interaction buckling and local-distortional-global interaction
*
to global buckling failures. These equations are unnecessary and
buckling modes, respectively. The equations to determine Ncle and N*clde
cumbersome for short BC members failing by local or distortional
are based on the proposals of Kumar and Kalyanaraman [29], who buckling, and thus were not considered.
developed an improved local buckling capacity curve that considers the
effect of web to flange widths ratios, and Schafer [30], who suggested
methods to determine the compression capacities of CFS members 5.3. Discussion
failing in local-distortional interaction mode. These equations are not
presented herein because of their complexity. The definitions of other As shown in Table 5, the design guidelines proposed by Zhang and
parameters are as defined in Section 5.1. Young [7] may not be suitable for the investigated BC members in this
In this section, the results of investigations conducted based on the study. The average ratio of their predictions to FE analysis results is
1.062, however, the COV is 0.107. Especially, for the case of BC members,
which fail by local buckling (L), their design guidelines are unsafe by up
to 24%, and gives incorrect member failure mode (as distortional buck­
ling, D). This is because their buckling analysis method based on Fig. 20
(a) gives higher local buckling loads of BC members than the real value as
shown in Fig. 22. In contrast, the design guidelines proposed by Lu et al.
[8] are very conservative, by 22% on average (overall average of 0.778
with COV of 0.123). For the BC members in Group 2, their design
guidelines predict their failure mode as local-distortional-buckling
(LDG), which was not observed in the test and FE analyses.
The current design standards [1,2] and Li et al.’s [5] proposal
(Method 3 in this paper) are able to predict the compression capacities
and failure modes of BC members, which can fail in either local or
distortional buckling mode. The ratios of their predictions to FE analysis
results vary in the range of 0.80–1.07 (average of 0.953 and COV of
0.064). Such variation can be justified considering the tolerance of the
DSM equations developed for different single channel sections, the ap­
Fig. 21. Buckling analyses used for design as proposed in: (a) this paper, Lu proximations in FE analysis results and the effect due to ignoring the
et al. [8] and Li et al. [5], (b) and (c) Zhang and Young [7]. higher capacity of BC members subject to distortional buckling. The

13
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Table 5
Effects of screw fastener arrangement on the compression capacities of short BC members.
Specimen FE Analysis by Zhang and Young [7] by Lu et al. [8] by DSM in design standards [1,2]

Nc,FEA Failure Nc,FEA/ Nc, [7]/Nc, Failure Nc, [8]/Nc, Failure Nc,prop-1/ Nc,prop-2/ Nc,prop-3/ Failure
(kN) mode-FEA ΣNc,FEA-i FEA mode [7] FEA mode [8] Nc,FEA Nc,FEA Nc,FEA mode-p

BC75D-600- 106.08 D 1.05 0.92 D 0.93 D 0.93 0.93 0.93 D

1S-300
BC75D-600- 106.78 D 1.05 0.91 D 0.92 D 0.92 0.92 0.92 D
1S-150
BC75D-600- 107.90 D 1.06 0.90 D 0.91 D 0.91 0.91 0.91 D
1S-100
BC75D-600- 108.62 D 1.07 0.90 D 0.91 D 0.91 0.91 0.91 D
1S-60
BC75D-600- 109.47 D 1.08 0.89 D 0.90 D 0.90 0.90 0.90 D
1S-30
BC87D-600- 91.14 LD 0.97 0.95 D 0.78 LDG 0.97 0.97 0.97 D
1S-300
BC87D-600- 91.35 LD 0.97 0.94 D 0.78 LDG 0.97 0.97 0.97 D
1S-150
BC87D-600- 91.53 LD 0.97 0.94 D 0.78 LDG 0.97 0.97 0.97 D
1S-100
BC87D-600- 91.36 LD 0.97 0.94 D 0.78 LDG 0.97 0.97 0.97 D
1S-60
BC87D-600- 92.04 LD 0.98 0.94 D 0.77 LDG 0.96 0.96 0.96 D
1S-30
BC120D-600- 530.35 D 1.09 0.84 D 0.85 D 0.85 0.85 0.85 D
1S-300
BC120D-600- 524.90 D 1.08 0.85 D 0.86 D 0.86 0.86 0.86 D
1S-150
BC120D-600- 533.64 D 1.10 0.83 D 0.85 D 0.85 0.85 0.85 D
1S-100
BC120D-600- 546.76 D 1.12 0.81 D 0.83 D 0.83 0.83 0.83 D
1S-60
BC120D-600- 562.01 D 1.15 0.79 D 0.80 D 0.80 0.80 0.80 D
1S-30
BC75L-600-1S- 75.74 L 1.00 1.04 D 0.71 LDG 0.94 0.93 0.95 L
300
BC75L-600-1S- 75.83 L 1.00 1.04 D 0.71 LDG 0.94 0.95 0.95 L
150
BC75L-600-1S- 75.39 L 1.00 1.04 D 0.72 LDG 0.95 0.96 0.95 L
100
BC75L-600-1S- 75.94 L 1.00 1.03 D 0.71 LDG 0.94 0.96 0.94 L
60
BC75L-600-1S- 76.05 L 1.01 1.03 D 0.71 LDG 0.94 0.96 0.94 L
30
BC87L-600-1S- 115.24 L 0.97 1.11 D 0.73 LDG 1.05 1.05 1.06 L
300
BC87L-600-1S- 115.39 L 0.97 1.11 D 0.73 LDG 1.05 1.06 1.06 L
150
BC87L-600-1S- 115.09 L 0.97 1.11 D 0.73 LDG 1.05 1.07 1.06 L
100
BC87L-600-1S- 115.46 L 0.97 1.11 D 0.73 LDG 1.05 1.07 1.06 L
60
BC87L-600-1S- 116.27 L 0.98 1.10 D 0.73 LDG 1.04 1.06 1.05 L
30
BC120L-600- 137.19 L 1.00 1.03 D 0.65 LDG 0.96 0.95 0.96 L
1S-300
BC120L-600- 137.52 L 1.00 1.02 D 0.65 LDG 0.95 0.96 0.96 L
1S-150
BC120L-600- 137.11 L 1.00 1.03 D 0.65 LDG 0.96 0.96 0.96 L
1S-100
BC120L-600- 137.54 L 1.00 1.02 D 0.65 LDG 0.95 0.96 0.96 L
1S-60
BC120L-600- 137.59 L 1.00 1.02 D 0.65 LDG 0.95 0.96 0.96 L
1S-30
BC75D-600- 106.37 D 1.05 1.04 D 0.93 D 0.93 0.93 0.93 D
2S-300
BC75D-600- 107.97 D 1.07 1.02 D 0.91 D 0.91 0.91 0.91 D
2S-150
BC75D-600- 109.85 D 1.08 1.00 D 0.90 D 0.90 0.90 0.90 D
2S-100
BC75D-600- 111.36 D 1.10 0.99 D 0.89 D 0.89 0.89 0.89 D
2S-60
BC75D-600- 113.82 D 1.12 0.97 D 0.87 D 0.87 0.87 0.87 D
2S-30
BC87D-600- 91.14 LD 0.97 1.12 D 0.78 LDG 0.97 0.97 0.97 D
2S-300
(continued on next page)

14
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Table 5 (continued )
Specimen FE Analysis by Zhang and Young [7] by Lu et al. [8] by DSM in design standards [1,2]

Nc,FEA Failure Nc,FEA/ Nc, [7]/Nc, Failure Nc, [8]/Nc, Failure Nc,prop-1/ Nc,prop-2/ Nc,prop-3/ Failure
(kN) mode-FEA ΣNc,FEA-i FEA mode [7] FEA mode [8] Nc,FEA Nc,FEA Nc,FEA mode-p

BC87D-600- 91.39 LD 0.97 1.12 D 0.78 LDG 0.97 0.97 0.97 D

2S-150
BC87D-600- 91.20 LD 0.97 1.12 D 0.78 LDG 0.97 0.97 0.97 D
2S-100
BC87D-600- 91.05 LD 0.96 1.12 D 0.78 LDG 0.97 0.97 0.97 D
2S-60
BC87D-600- 91.91 LD 0.97 1.11 D 0.78 LDG 0.96 0.96 0.96 D
2S-30
BC120D-600- 536.69 D 1.10 1.05 D 0.84 D 0.84 0.84 0.84 D
2S-300
BC120D-600- 546.31 D 1.12 1.03 D 0.83 D 0.83 0.83 0.83 D
2S-150
BC120D-600- 535.79 D 1.10 1.05 D 0.84 D 0.84 0.84 0.84 D
2S-100
BC120D-600- 541.59 D 1.11 1.04 D 0.83 D 0.83 0.83 0.83 D
2S-60
BC120D-600- 566.44 D 1.16 0.99 D 0.80 D 0.80 0.80 0.80 D
2S-30
BC75L-600-2S- 76.01 L 1.01 1.16 D 0.71 LDG 0.94 0.93 0.94 L
300
BC75L-600-2S- 75.75 L 1.00 1.16 D 0.71 LDG 0.94 0.95 0.95 L
150
BC75L-600-2S- 75.43 L 1.00 1.17 D 0.72 LDG 0.95 0.96 0.95 L
100
BC75L-600-2S- 75.87 L 1.00 1.16 D 0.71 LDG 0.94 0.96 0.95 L
60
BC75L-600-2S- 76.27 L 1.01 1.15 D 0.71 LDG 0.94 0.95 0.94 L
30
BC87L-600-2S- 115.42 L 0.97 1.23 D 0.73 LDG 1.05 1.05 1.06 L
300
BC87L-600-2S- 115.28 L 0.97 1.23 D 0.73 LDG 1.05 1.06 1.06 L
150
BC87L-600-2S- 115.31 L 0.97 1.23 D 0.73 LDG 1.05 1.07 1.06 L
100
BC87L-600-2S- 115.52 L 0.97 1.23 D 0.73 LDG 1.05 1.07 1.06 L
60
BC87L-600-2S- 116.64 L 0.98 1.22 D 0.72 LDG 1.04 1.06 1.05 L
30
BC120L-600- 137.01 L 1.00 1.23 D 0.65 LDG 0.96 0.95 0.96 L
2S-300
BC120L-600- 137.49 L 1.00 1.22 D 0.65 LDG 0.95 0.96 0.96 L
2S-150
BC120L-600- 137.29 L 1.00 1.23 D 0.65 LDG 0.96 0.96 0.96 L
2S-100
BC120L-600- 137.68 L 1.00 1.22 D 0.65 LDG 0.95 0.96 0.96 L
2S-60
BC120L-600- 137.55 L 1.00 1.22 D 0.65 LDG 0.95 0.96 0.96 L
2S-30
BC75D-800- 103.41 D 1.05 0.94 D 0.95 D 0.95 0.95 0.95 D
1S-400
BC75D-800- 103.54 D 1.05 0.94 D 0.95 D 0.95 0.95 0.95 D
1S-200
BC75D-800- 104.94 D 1.06 0.93 D 0.94 D 0.94 0.94 0.94 D
1S-160
BC75D-800- 105.00 D 1.06 0.93 D 0.94 D 0.94 0.94 0.94 D
1S-100
BC75D-800- 105.74 D 1.07 0.92 D 0.93 D 0.93 0.93 0.93 D
1S-40
BC87D-800- 89.28 LD 1.02 0.97 D 0.79 LDG 0.99 0.99 0.99 D
1S-400
BC87D-800- 89.04 LD 1.02 0.97 D 0.79 LDG 1.00 1.00 1.00 D
1S-200
BC87D-800- 89.33 LD 1.02 0.97 D 0.79 LDG 0.99 0.99 0.99 D
1S-160
BC87D-800- 89.91 LD 1.03 0.96 D 0.78 LDG 0.99 0.99 0.99 D
1S-100
BC87D-800- 89.69 LD 1.02 0.96 D 0.79 LDG 0.99 0.99 0.99 D
1S-40
BC120D-800- 485.81 D 1.04 0.91 D 0.93 D 0.93 0.93 0.93 D
1S-400
BC120D-800- 491.69 D 1.05 0.90 D 0.92 D 0.92 0.92 0.92 D
1S-200
484.99 D 1.04 0.92 D 0.93 D 0.93 0.93 0.93 D
(continued on next page)

15
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Table 5 (continued )
Specimen FE Analysis by Zhang and Young [7] by Lu et al. [8] by DSM in design standards [1,2]

Nc,FEA Failure Nc,FEA/ Nc, [7]/Nc, Failure Nc, [8]/Nc, Failure Nc,prop-1/ Nc,prop-2/ Nc,prop-3/ Failure
(kN) mode-FEA ΣNc,FEA-i FEA mode [7] FEA mode [8] Nc,FEA Nc,FEA Nc,FEA mode-p

BC120D-800-
1S-160
BC120D-800- 495.41 D 1.06 0.90 D 0.91 D 0.91 0.91 0.91 D
1S-100
BC120D-800- 490.68 D 1.05 0.91 D 0.92 D 0.92 0.92 0.92 D
1S-40
BC75L-800-1S- 75.39 L 1.00 1.04 D 0.71 LDG 0.92 0.90 0.93 L
400
BC75L-800-1S- 75.45 L 1.00 1.04 D 0.71 LDG 0.92 0.94 0.93 L
200
BC75L-800-1S- 75.47 L 1.00 1.04 D 0.71 LDG 0.92 0.94 0.93 L
160
BC75L-800-1S- 75.47 L 1.00 1.04 D 0.71 LDG 0.92 0.95 0.93 L
100
BC75L-800-1S- 75.74 L 1.00 1.04 D 0.71 LDG 0.92 0.95 0.92 L
40
BC87L-800-1S- 115.54 L 0.96 1.10 D 0.73 LDG 1.03 1.03 1.04 L
400
BC87L-800-1S- 115.36 L 0.96 1.11 D 0.73 LDG 1.03 1.05 1.05 L
200
BC87L-800-1S- 115.66 L 0.96 1.10 D 0.73 LDG 1.03 1.05 1.04 L
160
BC87L-800-1S- 115.27 L 0.96 1.11 D 0.73 LDG 1.03 1.06 1.05 L
100
BC87L-800-1S- 116.46 L 0.97 1.10 D 0.72 LDG 1.02 1.05 1.04 L
40
BC120L-800- 136.44 L 1.01 1.03 D 0.65 LDG 0.95 0.94 0.96 L
1S-400
BC120L-800- 136.37 L 1.01 1.03 D 0.65 LDG 0.95 0.96 0.96 L
1S-200
BC120L-800- 136.67 L 1.01 1.03 D 0.65 LDG 0.95 0.96 0.95 L
1S-160
BC120L-800- 136.29 L 1.01 1.03 D 0.65 LDG 0.95 0.97 0.96 L
1S-100
BC120L-800- 136.41 L 1.01 1.03 D 0.65 LDG 0.95 0.97 0.96 L
1S-40
BC75D-800- 105.15 D 1.06 1.05 D 0.94 D 0.94 0.94 0.94 D
2S-400
BC75D-800- 105.31 D 1.07 1.05 D 0.94 D 0.94 0.94 0.94 D
2S-200
BC75D-800- 105.97 D 1.07 1.04 D 0.93 D 0.93 0.93 0.93 D
2S-160
BC75D-800- 107.72 D 1.09 1.02 D 0.92 D 0.92 0.92 0.92 D
2S-100
BC75D-800- 107.55 D 1.09 1.02 D 0.92 D 0.92 0.92 0.92 D
2S-40
BC87D-800- 88.72 LD 1.01 1.15 D 0.79 LDG 1.00 1.00 1.00 D
2S-400
BC87D-800- 88.45 LD 1.01 1.16 D 0.80 LDG 1.00 1.00 1.00 D
2S-200
BC87D-800- 89.04 LD 1.02 1.15 D 0.79 LDG 1.00 1.00 1.00 D
2S-160
BC87D-800- 89.84 LD 1.03 1.14 D 0.78 LDG 0.99 0.99 0.99 D
2S-100
BC87D-800- 89.44 LD 1.02 1.14 D 0.79 LDG 0.99 0.99 0.99 D
2S-40
BC120D-800- 483.31 D 1.03 1.16 D 0.94 D 0.94 0.94 0.94 D
2S-400
BC120D-800- 505.76 D 1.08 1.11 D 0.89 D 0.89 0.89 0.89 D
2S-200
BC120D-800- 513.59 D 1.10 1.09 D 0.88 D 0.88 0.88 0.88 D
2S-160
BC120D-800- 522.98 D 1.12 1.07 D 0.86 D 0.86 0.86 0.86 D
2S-100
BC120D-800- 518.36 D 1.11 1.08 D 0.87 D 0.87 0.87 0.87 D
2S-40
BC75L-800-2S- 75.21 L 1.00 1.17 D 0.71 LDG 0.92 0.90 0.93 L
400
BC75L-800-2S- 75.43 L 1.00 1.17 D 0.71 LDG 0.92 0.94 0.93 L
200
BC75L-800-2S- 75.35 L 1.00 1.17 D 0.71 LDG 0.92 0.95 0.93 L
160
75.47 L 1.00 1.17 D 0.71 LDG 0.92 0.95 0.93 L
(continued on next page)

16
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

Table 5 (continued )
Specimen FE Analysis by Zhang and Young [7] by Lu et al. [8] by DSM in design standards [1,2]

Nc,FEA Failure Nc,FEA/ Nc, [7]/Nc, Failure Nc, [8]/Nc, Failure Nc,prop-1/ Nc,prop-2/ Nc,prop-3/ Failure
(kN) mode-FEA ΣNc,FEA-i FEA mode [7] FEA mode [8] Nc,FEA Nc,FEA Nc,FEA mode-p

BC75L-800-2S-
100
BC75L-800-2S- 75.89 L 1.00 1.16 D 0.71 LDG 0.91 0.95 0.92 L
40
BC87L-800-2S- 115.71 L 0.96 1.23 D 0.73 LDG 1.03 1.03 1.04 L
400
BC87L-800-2S- 115.65 L 0.96 1.23 D 0.73 LDG 1.03 1.05 1.04 L
200
BC87L-800-2S- 115.45 L 0.96 1.23 D 0.73 LDG 1.03 1.06 1.04 L
160
BC87L-800-2S- 115.63 L 0.96 1.23 D 0.73 LDG 1.03 1.06 1.04 L
100
BC87L-800-2S- 116.45 L 0.97 1.22 D 0.72 LDG 1.02 1.05 1.04 L
40
BC120L-800- 136.05 L 1.00 1.24 D 0.65 LDG 0.95 0.95 0.96 L
2S-400
BC120L-800- 136.46 L 1.01 1.23 D 0.65 LDG 0.95 0.96 0.96 L
2S-200
BC120L-800- 136.61 L 1.01 1.23 D 0.65 LDG 0.95 0.96 0.95 L
2S-160
BC120L-800- 136.29 L 1.01 1.24 D 0.65 LDG 0.95 0.97 0.96 L
2S-100
BC120L-800- 136.69 L 1.01 1.23 D 0.65 LDG 0.95 0.96 0.95 L
2S-40

Average 1.062 0.778 0.949 0.955 0.953

COV 0.107 0.123 0.060 0.064 0.062

Note: “Failure mode-FEA”, “Failure mode [7]” and “Failure mode [8]” mean the failure modes of BC members given by FE analyses, the design methods of Zhang and
Young [7] and Lu et al. [8], respectively, while “Failure mode-p” means those given by the three proposed design methods.

Fig. 22. Buckling analysis results of BC75L-1S and BC75L-2S sections.

three proposed DSM design methods based on the current design stan­ 6. Summary and conclusions
dards, provide nearly similar predictions. The second design method is
the most accurate (average of 0.955 and COV of 0.064), however, this This paper has presented the details and results of an investigation of
paper recommends the third method for practical use since it is the most short CFS single channel and BC members failing in distortional and
simplified, but is still reliable (average of 0.953 and COV of 0.062). local buckling mode. Compression tests were undertaken first on short
Although BC87D members exhibited local-distortional interaction CFS single channel and BC members failing in local-distortional inter­
buckling (Table 5), the proposed design methods were able to predict action buckling. FE models of the tested members were developed and
their capacities reasonably well. Based on Martins et al. [31], the indi­ validated using the test results. Using the validated FE models, a para­
vidual channel section (SC87D) may fail by “true local-distortional metric study was undertaken for 120 cases of BC members with different
interaction” buckling since its ratio of distortional to local buckling sections, screw sizes, screw spacing and number of screws per rows. The
loads is 1.298, which is in the range of 0.8–1.3. However, SC87D section BC members in the parametric study failed mostly by local or distor­
cannot represent all the channel sections subject to “true tional buckling but in some cases by local-distortional interaction
local-distortional interaction” buckling. Therefore, further research buckling.
studies on BC members made of this type of channel sections are needed. This research has provided a good understanding of the compression
behaviour of short BC members. The results showed that the screw
characteristics and arrangement have only a small effect on the
compression behaviour and capacities of BC members if their failures are

17
S.T. Vy et al. Thin-Walled Structures 159 (2021) 107224

dominated by local or distortional buckling modes. Although the screw [6] J.H. Zhang, B. Young, Compression tests of cold-formed steel I-shaped open
sections with edge and web stiffeners, Thin-Walled Struct. 52 (2012) 1–11.
spacing can affect the level of web buckling compatibility in these BC
[7] J.H. Zhang, B. Young, Numerical investigation and design of cold-formed steel
members, a screw spacing much less than the local and distortional built-up open section columns with longitudinal stiffeners, Thin-Walled Struct. 89
buckling half-wavelengths may be needed to achieve full web buckling (2015) 178–191.
compatibility. Further, a partial composite action is present in BC [8] Y. Lu, T.H. Zhou, W.C. Li, H.H. Wu, Experimental investigation and a novel direct
strength method for cold-formed built-up I-section columns, Thin-Walled Struct.
members subject to distortional buckling, which may give slightly 112 (2017) 125–139.
higher compression capacities than the sum of their individual member [9] M. Abbasi, M. Khezri, K.J.R. Rasmussen, B.W. Schafer, Elastic buckling analysis of
capacities. For BC members subject to local or local-distortional buck­ cold-formed steel built-up sections with discrete fasteners using the compound strip
method, Thin-Walled Struct. 124 (2018) 58–71.
ling, such composite action may not be achieved and its capacity is about [10] M. Khezri, M. Abbasi, K.J.R. Rasmussen, Application of the compound strip method
twice that of their individual members. in buckling analysis of cold-formed steel built-up sections, ce/papers 1 (2–3)
Using the FE analysis results, the accuracy of the design guidelines in (2017) 1667–1676.
[11] M. Abbasi, M. Khezri, K.J.R. Rasmussen, On extending the direct strength method
the current design standards and past research studies was investigated to the design of cold-formed steel built-up columns, ce/papers 1 (2–3) (2017)
in terms of the predicted compression capacity and the dominant failure 1600–1608.
modes of short BC members. It was found that the DSM design guidelines [12] D.C. Fratamico, S. Torabian, K.J.R. Rasmussen, B.W. Schafer, Experimental
Investigation of the Effect of Screw Fastener Spacing on the Local and Distortional
proposed by some past research studies are not suitable for these Buckling Behavior of Built-Up Cold-Formed Steel Columns, International Specialty
members, while those in the current design standards are conservative. Conference on Cold-Formed Steel Structures, Baltimore, Maryland, U.S.A, 2016.
This paper has proposed a reliable and simplified method based on the [13] D.C. Fratamico, S. Torabian, X. Zhao, K.J.R. Rasmussen, B.W. Schafer, Experiments
on the global buckling and collapse of built-up cold-formed steel columns,
current CFS design standards.
J. Constr. Steel Res. 144 (2018) 65–80.
[14] D.C. Fratamico, S. Torabian, X. Zhao, K.J.R. Rasmussen, B.W. Schafer,
Credit author statement Experimental study on the composite action in sheathed and bare built-up cold-
formed steel columns, Thin-Walled Struct. 127 (2018) 290–305.
[15] S. Kechidi, D.C. Fratamico, B.W. Schafer, J.M. Castro, N. Bourahla, Simulation of
Son Tung Vy: Investigation, Methodology, Formal analysis, Valida­ screw connected built-up cold-formed steel back-to-back lipped channels under
tion, Data Curation, Visualization, Writing - Original Draft. Mahen axial compression, Eng. Struct. 206 (2020).
Mahendran: Conceptualisation, Methodology, Writing - Review & [16] D.K. Phan, K.J.R. Rasmussen, Flexural rigidity of cold-formed steel built-up
members, Thin-Walled Struct. 140 (2019) 438–449.
Editing, Supervision, Resources, Project administration, Funding [17] T.H. Zhou, Y.C. Li, H.H. Wu, Y. Lu, L. Ren, Analysis to determine flexural buckling
acquisition. Thananjayan Sivaprakasam: Investigation, Formal analysis, of cold-formed steel built-up back-to-back section columns, J. Constr. Steel Res.
Supervision. 166 (2020).
[18] K. Piyawat, C. Ramseyer, T.H.K. Kang, Development of an axial load capacity
equation for doubly symmetric built-up cold-formed sections, J. Struct. Eng. 139
Declaration of competing interest (12) (2013).
[19] T.C.H. Ting, K. Roy, H.H. Lau, J.B.P. Lim, Effect of screw spacing on behavior of
axially loaded back-to-back cold-formed steel built-up channel sections, Adv.
The authors declare that they have no known competing financial Struct. Eng. 21 (3) (2018) 474–487.
interests or personal relationships that could have appeared to influence [20] ABAQUS, Analysis User’s Manual-Version 6.14–2, ABAQUS Inc, USA, 2014.
the work reported in this paper. [21] M. Rokilan, M. Mahendran, Elevated temperature mechanical properties of cold-
rolled steel sheets and cold-formed steel sections, J. Constr. Steel Res. 167 (2020).
[22] B.W. Schafer, T. Peköz, Computational modeling of cold-formed steel:
Acknowledgments characterizing geometric imperfections and residual stresses, J. Constr. Steel Res.
47 (3) (1998) 193–210.
The authors would like to thank QUT, Australian Research Council [23] D.K. Phan, K.J.R. Rasmussen, Cold-formed Steel Bolted and Screw-Fastened
Connections in Shear, in: Eighth International Conference on THIN-WALLED
(Grant Number LP170100951), National Association of Steel Framed STRUCTURES, Lisbon, Portugal, 2018.
Housing (NASH) and Stoddart Group for providing financial support, [24] E. Ellobody, B. Young, Behavior of cold-formed steel plain angle columns, J. Struct.
research facilities and test materials, and Banyo Laboratory Technical Eng. 131 (3) (2005) 457–466.
[25] L. Gardner, D.A. Nethercot, Numerical modeling of stainless steel structural
staff for their assistance during the tests. The first author appreciates the components-A consistent approach, J. Struct. Eng. 130 (10) (2004) 1586–1601.
assistance of Mr David Fratamico from Johns Hopkins University, Mr [26] S. Narayanan, M. Mahendran, Ultimate capacity of innovative cold-formed steel
Ngoc Hieu Pham and Mr Dang Khoa Phan from the University of Sydney, columns, J. Constr. Steel Res. 59 (4) (2003) 489–508.
[27] Z. Li, B.W. Schafer, Buckling analysis of cold-formed steel members with general
and Assoc. Prof Hung Cuong Bui from NUCE Vietnam in relation to boundary conditions using CUFSM: conventional and constrained finite strip
screw shear tests, DSM and CUFSM. methods, in: Proceedings of the 20th Int;l. Spec. Conf. on Cold-Formed Steel
Structures, St. Louis, MO, 2010.
[28] D.C. Fratamico, B.W. Schafer, Numerical studies on the composite action and
References buckling behavior of built-up cold-formed steel columns, in: 22nd International
Specialty Conference on Cold-Formed Steel Structures, St. Louis, Missouri, USA,
[1] AISI S100, North American Specification for the Design of Cold-Formed Steel 2014.
Structural Members, American Iron and Steel Institute, Washington, D.C., 2016. [29] M.V.A. Kumar, V. Kalyanaraman, Design strength of locally buckling stub-lipped
[2] AS/NZS 4600, Cold-Formed Steel Structures, Australian/New Zealand Standard, channel columns, J. Struct. Eng. 138 (11) (2012) 1291–1299.
Sydney, Australia, 2018. [30] B.W. Schafer, Local, distortional, and euler buckling of thin-walled columns,
[3] T.A. Stone, R.A. LaBoube, Behavior of cold-formed steel built-up I-sections, Thin- J. Struct. Eng. 128 (3) (2002) 289–299.
Walled Struct. 43 (12) (2005) 1805–1817. [31] A.D. Martins, D. Camotim, P.B. Dinis, On the direct strength design of cold-formed
[4] K. Roy, T.C.H. Ting, H.H. Lau, J.B.P. Lim, Effect of thickness on the behaviour of steel columns failing in local-distortional interactive modes, Thin-Walled Struct.
axially loaded back-to-back cold-formed steel built-up channel sections - 120 (2017) 432–445.
experimental and numerical investigation, Structure 16 (2018) 327–346.
[5] Y.Q. Li, Y.L. Li, S.K. Wang, Z.Y. Shen, Ultimate load-carrying capacity of cold-
formed thin-walled columns with built-up box and I section under axial
compression, Thin-Walled Struct. 79 (2014) 202–217.

18