Engineering Structures 275 (2023) 115316

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Flexural buckling behaviour of concrete-filled double skin aluminium

alloy columns
Shafayat Bin Ali a, b, *, George S. Kamaris a, Michaela Gkantou a
a
School of Civil Engineering and Built Environment, Liverpool John Moores University, Liverpool, L3 3AF, United Kingdom
b
Institute of Earthquake Engineering Research, Chittagong University of Engineering & Technology, Chattogram 4349, Bangladesh

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

Keywords: Concrete-filled double skin steel tubular columns consisting of two hollow steel tubes and concrete infill between
6082-T6 aluminium alloy their interspace, have become popular in modern construction owing to their high ultimate load, less self-weight
Concrete-filled double skin tubular sections and good ductility. The weight of these columns can be reduced further by using lightweight aluminium alloy
Columns
hollow tubes instead of steel ones. This paper experimentally and numerically studied the flexural buckling
Flexural buckling behaviour
Finite element analysis
behaviour of the concrete-filled double skin aluminium tubular (CFDSAT) columns subjected to axial
Design compression. A total of 8 CFDSAT columns were tested using a pin-ended set-up. The test results are presented in
terms of failure modes, ultimate load and load versus mid-height lateral displacement curves. Non-linear finite
element (FE) models of the specimens were developed and their accuracy was evaluated by comparing the FE
and test results. A numerical parametric investigation was conducted to study the influence of the hollow ratio,
the member slenderness, the cross-sectional slenderness of the hollow tubes and the concrete strength on the
structural behaviour of CFDSAT columns. The parametric study results revealed that the cross-sectional di­
mensions of outer section, the member slenderness and the concrete compressive strength have a significant
effect on the flexural buckling response of the columns, while the influence of the cross-sectional dimensions of
inner section is less prominent. In the absence of design standards for CFDSAT members, a design methodology is
proposed with a design buckling curve to predict the ultimate load of CFDSAT columns based on the Eurocode 4
framework. Moreover, a revised concrete correction factor is suggested to determine the effective flexural ri­
gidity of CFDSAT columns according to Eurocode 4.

sections can improve remarkably by incorporating the concrete infill

[7–9]. Therefore, by combining the effectiveness of double skin tech­
1. Introduction
nique and the profound features of aluminium alloys, a concrete-filled
double skin aluminium alloy tubular (CFDSAT) column can be an effi­
Concrete-filled double skin steel tubular (CFDSST) column is fabri­
cient structural member for structures located in aggressive environ­
cated using two hollow steel tubes and the space between them is filled
ments, such as offshore wind turbines, offshore oil platforms, seaside
with concrete. CFDSST columns have several benefits compared to the
piers, containment building of nuclear power plants, tall buildings and
conventional concrete-filled steel tubular (CFST) ones, such as, less self-
bridges in earthquake-prone regions.
weight because of the absence of inner concrete [1], good global and
A number of research studies are available on the structural behav­
local stability owing to the beneficial composite action [2] and better
iour of CFDSST columns. Experimental investigations were performed to
fire resistance due to thermal protection provided by the outer hollow
study the compressive behaviour of CFDSST short columns comprised
section to the inner hollow section and concrete [3]. Aluminium alloys
carbon steel square [10], rectangular [11] and circular [12–14] outer
are versatile metals with a combination of superior properties,
and inner profiles. In these studies, the failure pattern of CFDSST col­
including, high strength-to-weight ratio, excellent corrosion resistance,
umns was investigated and design guidelines were suggested to calcu­
good formability and attractive appearance [4–6]. Hence, the use of
late the ultimate capacity of the columns. Elchalakani et al. [15] and
aluminium alloy tubes in the concrete-filled double skin sections can
Han et al. [16] tested CFDSST stub columns produced by cold-formed
further reduce their weight significantly and enhance their resilience.
steel circular-square and square-circular outer and inner tubes and
On the other hand, the structural stability of aluminium alloy hollow

* Corresponding author.
E-mail addresses: [email protected], [email protected] (S. Bin Ali).

https://doi.org/10.1016/j.engstruct.2022.115316
Received 16 August 2022; Received in revised form 26 October 2022; Accepted 13 November 2022
Available online 24 November 2022
0141-0296/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Nomenclature Npl plastic resistance of column under axial compression

Nu ultimate load of column obtained from test and FE analysis
Ac area of concrete Nu,FE ultimate load of column obtained from FE analysis
Ao area of outer tube Nu,prop design ultimate load of column
Ai area of inner tube Nu,Test ultimate load of tested column
Bo width of outer tube n strain hardening exponent
Bi width of inner tube to thickness of outer tube
Do depth of outer tube ti thickness of inner tube
Di depth/diameter of inner tube α hollow ratio
E modulus of Elasticity α1 imperfection factor
EIe effective flexural rigidity of column δ lateral deflection at the mid-height of column
e0 initial loading eccentricity ε strain under tension
f tensile stress εf strain at fracture
fc concrete cylinder compressive strength εmax maximum strain
fcu concrete cube compressive strength εmin minimum strain
fu ultimate tensile stress εu strain at ultimate tensile stress
f0.1 0.1 % proof stress λ member slenderness
f0.2 0.2 % proof stress λ0 limit of horizontal plateau
I moment of inertia of cross-section ϕ a parameter
ke concrete correction factor χ reduction factor for determining flexural buckling capacity
L length of column of column
Le effective length of column ωg initial global imperfection amplitude
Ncr critical buckling load or Euler’s critical load of column

recommended design formulae for these columns. A series of compres­ conducted. In Section 5, a design methodology is proposed on the basis
sive tests were conducted on CFDSST stub columns made with square- of Eurocode 4 [34]. Finally, in Section 6, conclusions are provided.
rectangular [17], circular [18], square-circular [19] and round-up
rectangular-elliptical [20] carbon steel inner and stainless steel outer 2. Test programme
sections. These research studies suggested design rules for the afore­
mentioned compression members. Wang et al. [21] numerically inves­ 2.1. CFDSAT specimens
tigated the compressive response of square CFDSST cross-sections
constructed with stainless outer hollow sections and proposed modifi­ A total of 8 CFDSAT columns were tested under axial compression, in
cations of the current design standards to predict their design strength. which 6 specimens comprised square outer and inner tubes and 2
Tziavos et al. [22] conducted a numerical study to examine the specimens were made with square outer and circular inner tubes. The
compressive response of grout-filled double-skin tubular steel short hollow tubes were manufactured by 6082-T6 grade aluminium alloy,
columns and suggested design equations for these columns. Zhao et al. except two inner tubes, i.e., S19.2 × 1.6 and C19.2 × 3.2 which were
[23], Han et al. [24] and Zheng et al. [25] investigated the behaviour of produced by 6063-T6 aluminium alloy. Fig. 1 illustrates the cross-
CFDSST short columns subjected to large axial deformation, long-term sectional geometry of typical specimens and Table 1 presents the
sustained load and cyclic load. Research on the compressive behaviour measured geometric dimensions of all specimens, including depth/
of CFDSST slender columns was carried out by Hassanein and Kharoob diameter (D), width (B) and thickness (t), where the subscripts ‘o’ and ‘i’
[26], Sulthana and Jayachandran [27], Zhao et al. [28] and Ahmed et al. represent the outer and inner hollow sections, respectively. Moreover,
[29] considering different cross-sectional shapes and steel materials. the cross-section class of the outer and the inner tubes based on Euro­
They suggested a design equation to determine the ultimate capacity of code 9 [35] and column slenderness (λ) determined according to Euro­
these columns. code 4 [34] (see Equation (8) in Section 5.1) are also provided in
Few studies exist on the compressive response of CFDSAT stub col­ Table 1. In this study, 8 hollow sections were taken in total, where three
umns. Zhou and Young [30,31] experimentally and numerically studied large sections with measured depth/diameter to thickness ratio (Do/to)
the influence of geometry and concrete strength on the structural varying from 16.45 to 20.81 were utilised as the outer profiles and five
response of 6061-T6 alloy CFDSAT stub columns made with circular and small sections with measured depth/diameter to thickness ratio (Di/ti)
circular-square outer and inner profiles. They recommended design ranging from 6.09 to 16.74 were utilised as the inner profiles. All
guidelines to determine the compressive capacity of these cross-sections. specimens were 1000 mm long. The nominal effective length (Le) of the
Patel et al. [32] conducted a numerical study on the structural behaviour specimens, including pin-ended supports, was 1064 mm. The specimens
of CFDSAT short columns using a fiber element model and proposed a were labelled considering the geometric dimensions of the outer and the
formula to evaluate the lateral confining pressure developed on the inner hollow sections. For example, the label ‘S63.4 × 3.2-C19.2 × 3.2′
concrete infill. However, no research study is available on the buckling indicates a CFDSAT specimen consisting of ‘S63.4 × 3.2′ outer section
response of CFDSAT columns so far. Hence, this paper reports for the and ‘C19.2 × 3.2′ inner section. In the notation ‘S63.4 × 3.2′ , the letter
first time an experimental and numerical study on the flexural buckling ‘S’ signifies that the hollow section is square with nominal depth and
response of CFDSAT members under axial compression. The hollow width of 63.4 mm and thickness of 3.2 mm. In the notation ‘C19.2 × 3.2′ ,
tubes used in this study were manufactured by 6082-T6 aluminium alloy the letter ‘C’ represents a circular hollow section which has a nominal
which is a popular alloy owing to its high strength, excellent resistance diameter of 19.2 mm and thickness of 3.2 mm.
to corrosion and good weldability [7,33]. In this paper, the test pro­ Before concrete casting, the inner tubes of all specimens were pre­
gramme and the obtained test findings are presented in Sections 2 and 3, cisely located at the centroid of the outer hollow section using top and
respectively. In Section 4, non-linear finite element (FE) models of the bottom wooden plates. The wooden plates were prepared by attaching
specimens are developed and a numerical parametric study is two parts, i.e., an outer part and an inner part. The dimensions of the

2
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 1. Cross-sectional geometry of typical CFDSAT specimens.

Table 1
Measured cross-sectional dimensions, cross-section class of hollow tubes, member slenderness and initial global geometric imperfection of CFDSAT specimens.
Specimen Do Bo to Do/to Class of outer Di Bi ti Di/ti Class of inner λ ωg (mm)
(mm) (mm) (mm) section (mm) (mm) (mm) section

S50.8 × 3.2-S19.2 × 50.83 50.83 3.09 16.45 2 19.05 19.05 1.53 12.45 1 1.08 0.27
1.6
S63.4 × 3.2-S19.2 × 63.48 63.48 3.05 20.81 2 19.06 19.06 1.53 12.46 1 0.73 0.15
1.6
S63.4 × 3.2-S25.4 × 63.48 63.48 3.06 20.75 2 25.38 25.38 3.28 7.74 1 0.80 0.39
3.2
S76.2 × 4.8-S19.2 × 76.09 76.09 4.61 16.51 3 19.05 19.05 1.57 12.13 1 0.75 0.11
1.6
S76.2 × 4.8-S25.4 × 76.09 76.09 4.58 16.61 3 25.27 25.27 1.51 16.74 2 0.75 0.04
1.6
S76.2 × 4.8-S25.4 × 76.09 76.09 4.61 16.51 3 25.42 25.42 3.30 7.70 1 0.78 0.07
3.2
S63.4 × 3.2-C19.2 × 63.54 63.45 3.10 20.50 2 18.88 – 3.10 6.09 1 0.77 0.12
3.2
S63.4 × 3.2-C25.4 × 63.48 63.45 3.08 20.61 2 25.41 – 3.37 7.54 1 0.76 0.30
3.2

outer part were similar to the dimensions of the outer hollow section, 6892–1 [36]. A 50 kN hydraulic tensile machine was used for the tests.
whereas the dimensions of the inner part were the same as the di­ During the tests, flat grip faces were used for the flat coupons (Fig. 3(a))
mensions of the interspace between two tubes of the respective specimen and grip faces with a pitch serration were applied for the curved coupons
(Fig. 2(a)). The top end of the inner tubes was closed with plastic tape to for ensuring better contact between the specimen and the grip faces
avoid concrete falling inside them during casting. To avoid any concrete (Fig. 3(b)). A displacement-control load with a rate of 0.2 mm/min was
leakage the bottom wooden plates were sealed properly. The casting was applied during the tests. The longitudinal strains of the specimens were
conducted by filling concrete in layers. The concrete was compacted measured using an extensometer with a gauge length of 50 mm. The
using a vibrating table. After the cast, the columns were covered using a experimental stress–strain curves were replicated using Equations (1)
plastic sheet to avoid moisture loss from the concrete. The columns were and (2) recommended by Ramberg and Osgood [37] and updated by Hill
cured for 28 days and then the wooden plates were separated from both [38].
ends. Fig. 2 shows photos of the wooden plates and the typical cross- ( )n
f f
sections of the specimens. ε = + 0.002 (1)
E f0.2
The initial global geometric imperfection (ωg) has a strong effect on
the buckling behaviour and load-bearing capacity of slender columns
ln2
[29]. Therefore, prior to the test the values of ωg of all specimens were n= ( ) (2)
measured using a linear height gauge and considered in the numerical ln ff0.2
0.1

investigation reported in Section 4. The value of ωg corresponding to the

buckling axis was defined by determining the deviation at the mid- where ε represents the strain, f is the tensile stress, E denotes the
length of a face parallel to the buckling axis from the reference line modulus of Elasticity, f0.1 and f0.2 stand for the 0.1 % and 0.2 % proof
connecting both ends of a specimen [7]. The recorded values of ωg are stress, respectively and n is the strain hardening parameter. The material
listed in Table 1. properties of the aluminium alloy tubes determined from the tensile
tests of coupons are listed in Table 2, where fu represents the ultimate
tensile stress, εu is the strain at fu and εf denotes strain related to fracture.
2.2. Material properties The characteristic values of 0.2 % proof stress (f0.2,EC9), the ultimate
tensile strength (fu,EC9) and the strain hardening exponent (nEC9) sug­
2.2.1. Aluminium alloy hollow section gested by Eurocode 9 (EC9) [35] are also provided in Table 2. It can be
Tensile coupon tests were performed to determine the material seen that the values of f0.2 and fu obtained from the experiments are
properties of aluminium alloy hollow tubes. Two coupons were cut generally higher than the values suggested by Eurocode 9 for 6082-T6
longitudinally from the square (dogbone flat coupons) and the circular alloy. Fig. 3(c), 3(d) and 3(e) show the dimensions of a typical
(curved coupons) tubes by following the guidance given in BS EN ISO

3
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 2. Photographs of wooden plates and typical cross-sections of CFDSAT specimens.

coupon, typical failure modes of some coupons and typical experimental displacement and two were placed at the bottom channel to track the
stress–strain curves of S76.2 × 4.8 and C19.2 × 3.2 material coupons rotation of the specimens. Two strain gauges were attached at the mid-
along with the respective Ramberg-Osgood curves, respectively. height of two faces parallel to the bending axis of a specimen to record
the in-plane axial strains. A data logger with eight recording channels
2.2.2. Concrete was used to record all the data during the tests. At the beginning of the
For preparing the concrete ordinary (Type I) Portland cement, me­ tests, the actual initial loading eccentricity (e0) was determined based on
dium coarse silica sand, limestone with a maximum size of 10 mm and the observed lateral displacement and strain values to ensure that the
normal water were used. These ingredients were mixed by weight at a initial loading eccentricity was less than Le/1000 [7]. The e0 was
ratio of 1:1.5:2.5:0.5. Compressive tests of concrete cubes were carried calculated by Equation (3) [7] and listed in Table 4.
out to determine the material properties of the concrete infill. Four
EI(εmax − εmin )
standard cubes with 100 mm × 100 mm × 100 mm nominal size were e0 = − δ − ωg (3)
Do N
cast using the same concrete mix of the CFDSAT columns and cured for
28 days. The compressive tests were conducted according to the guid­ where I represents the second moment of area of the cross-section, N is
ance of BS EN 12390–3 [39] and the test results are reported in Table 3 the axial compressive load, δ is mid-height lateral displacement and εmax
including the density of the cubes. and εmin are the maximum and minimum strain values at recorded N,
respectively. Fig. 4 presents a photo and a schematic drawing of the test
setup.
2.3. Flexural buckling tests

3. Test results
Axial compression tests were carried out to study the flexural
buckling behaviour and determine the ultimate load of CFDSAT speci­
The failure pattern of all CFDSAT columns was examined and the
mens. Before the tests, both ends of all specimens were made flat by
failure pattern of a typical specimen is presented in Fig. 5. It was
milling and strengthened with carbon fibre-reinforced polymer (CFRP)
observed that the predominant failure pattern of all specimens was
to avoid any localised failure at any end [7,30]. Moreover, a plaster
flexural buckling (Fig. 5) as the pin-ended set-up allowed the specimens
layer was employed on the top surface of the infilled concrete of each
to rotate about the buckling axis. No inward or outward local buckling
end to ensure uniform distribution of the axial compressive load [7,30].
was identified at the outer tube of the CFDSAT columns. It attributes to
The pin-ended boundary condition was created using a steel plate con­
the concrete infill which effectively prevented the occurrence of inward
taining deep grooves and a wage plate with a knife edge and applied at
local buckling. Moreover, no fracture was noticed in the outer tube of
both ends of a specimen to allow rotation about the buckling axis. To
any specimen. To investigate the failure mode of the inner components,
ensure concentric alignment of the specimens along with the knife
the outer tube and concrete infill of typical specimens were partly cut
edges, a steel channel was loosely bolted using steel plates at each end. A
and detached after the tests. The failure pattern of the inner profile and
600 kN hydraulic testing machine was utilised for applying the
the concrete infill of a typical specimen is presented in Fig. 6(a) and 6(b),
displacement-control load at a rate of 0.2 mm/min. A load of 2 kN was
respectively. Fig. 6(a) illustrates that, same as the outer tube, the inner
employed initially to eliminate any gap at the loading point. During the
tube displayed obvious flexural buckling. However, no inward or out­
tests, four linear variable differential transducers (LVDTs) were used, in
ward bulging or fracture was found in the inner tube of the inspected
which two were positioned at the mid-length to measure the lateral

4
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 3. Coupon test set-up, dimensions of a typical coupon, typical failure modes of some coupons and typical experimental and corresponding Ramberg-Osgood
stress–strain curves.

specimens. It can be seen from Fig. 6(b) that some small horizontal the relationship between the axial load and the mid-height lateral
cracks in concrete appeared in the tension region near the mid-height of displacement is linear which indicates that the materials of the speci­
the specimen as tension stresses are higher in this region. mens were in the elastic stage. After the elastic stage, all curves diverge
Fig. 7 presents the applied axial load versus corresponding mid- from linear to nonlinear response as the growth of load decreases
height lateral displacement curves for all tested specimens. The curves compared to the lateral displacement. This is the evidence of the elas­
are plotted using the compressive load and lateral displacement data tic–plastic stage which exists up to the ultimate load. After reaching the
recorded from the load cell of the machine and the LVDT positioned at ultimate load, a descending part appears in the curves as the load started
the mid-length of the specimens, respectively. It is found that initially to fall gradually with the development of lateral displacement. It can be

5
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Table 2
Material properties of the aluminium hollow section.
Specimen E (GPa) f0.1 (MPa) f0.2 (MPa) f0.2,EC9 (MPa) fu (MPa) fu,EC9 (MPa) εu (%) (mm/mm) εf (%) (mm/mm) n nEC9

S19.2 × 1.6 72,500 185 189 160 215 195 4.57 5.45 28.83 24
S25.4 × 1.6 68,700 192 197 250 229 290 6.43 11.40 27.55 32
S25.4 × 3.2 70,000 276 283 250 328 290 7.66 16.00 28.06 32
S50.8 × 3.2 68,300 257 264 250 304 290 7.49 14.70 28.19 32
S63.5 × 3.2 69,900 157 164 250 212 290 6.69 15.60 21.94 32
S76.2 × 4.8 70,700 305 311 250 338 290 5.97 11.70 33.91 32
C19.2 × 3.2 72,800 303 306 160 320 195 5.18 8.94 36.53 24
C25.4 × 3.2 71,900 283 289 250 327 290 8.50 13.00 29.58 32

Table 3
Density and compressive strength of concrete cubes.
Concrete cube Density (kg/m3) Compressive strength (MPa)

1 2427 37.2
2 2392 35.7
3 2424 36.8
4 2432 36.5
Mean 36.5

Table 4
Initial loading eccentricity, failure mode, critical buckling load and ultimate
load of CFDSAT specimens.
Specimen e0 (mm) Failure mode Ncr (kN) Nu,Test (kN)

S50.8 × 3.2-S19.2 × 1.6 0.52 FB 192.17 143.30

S63.4 × 3.2-S19.2 × 1.6 0.89 FB 429.93 179.83
S63.4 × 3.2-S25.4 × 3.2 0.68 FB 437.68 216.85
S76.2 × 4.8-S19.2 × 1.6 0.81 FB 980.77 464.47
S76.2 × 4.8-S25.4 × 1.6 0.93 FB 978.74 455.64
S76.2 × 4.8-S25.4 × 3.2 0.66 FB 988.00 466.51
S63.4 × 3.2-C19.2 × 3.2 0.64 FB 434.63 210.67
S63.4 × 3.2-C25.4 × 3.2 0.60 FB 435.56 198.51
Note: FB = Flexural buckling

observed from the figure that a certain ductility is visible in all curves,
attributed to the beneficial composite action of the three elements of the
CFDSAT specimens. The ultimate load (Nu,Test) of all specimens deter­
mined from the tests is listed in Table 4 including their critical buckling
load (Ncr) (see Equation (9) in Section 5.1). It is found from the table that
the ultimate load of the CFDSAT specimens improved noticeably by
increasing the dimensions of outer tube when the dimensions of inner
tube are constant. For example, the ultimate load of the specimen S50.8
× 3.2-S19.2 × 1.6 improved by 25.5 % compared to the specimen S63.4
× 3.2-S19.2 × 1.6 as the outer dimension of outer tube of the later is
increased from 50.8 mm to 63.4 mm. Moreover, it is observed that the
ultimate load of the CFDSAT specimens decreased with the increase of
the depth/diameter of inner tube when the outer tube dimensions are
constant. This is because the area of concrete infill is decreased,
resulting in the reduction of cross-sectional resistance. For example,
when the outer diameter of the inner tube is increased from 19.2 mm to
25.4 mm, the ultimate load of the specimen S63.4 × 3.2-C19.2 × 3.2 is
reduced approximately by 5.5 % compared to the capacity of the spec­
imen S63.4 × 3.2-C25.4 × 3.2. The influence of the dimensions of outer
Fig. 4. Photograph and schematic diagram of the axial compression test setup.
and inner tubes on the ultimate load of the specimens is further studied
by the FE analysis and discussed in Section 4.3.
4.1. FE modelling

4. Finite element (FE) analysis

All CFDSAT specimens were modelled using the measured di­
mensions of the cross-sections and material properties listed in Tables 1,
In this study, ABAQUS software [40] was used to perform the FE
2 and 3. Eight-node hexahedral solid elements with reduced integration
analysis of the buckling behaviour of CFDSAT columns. Non-linear FE
(C3D8R) were applied to simulate the hollow tubes and the concrete
models of the CFDSAT specimens were developed and their accuracy
infill [7]. A mesh sensitivity investigation was carried out for choosing
was evaluated by comparing the FE results with those obtained from the
the optimum size of the C3D8R elements. It was found that 5 mm
experiments. A numerical parametric investigation was subsequently
element size is suitable for achieving accurate results in a reasonable
carried out.

6
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

strain data were inputted by converting them to true stress-

logarithmic plastic strain data for modelling the plastic behaviour. The
inbuilt concrete damage plasticity (CDP) model was used to replicate the
material properties of concrete. The modulus of Elasticity of concrete
was calculated using the formula provided in Eurocode 4 [34] and the
value of Poisson’s ratio was considered as 0.2. In the CDP model, the
dilation angle value of 40◦ was used [42], while the default values of 0.1
and 0 were applied for the flow potential eccentricity and the viscosity
parameter, respectively. The ratio of the compressive strength due to
biaxial loading over uniaxial compressive strength was defined based on
the formula recommended by Papanikolaou and Kappos [43] and the
compressive meridian was determined according to the suggestion given
by Yu et al. [44]. The confined concrete model proposed by Tao et al.
[42] (Fig. 8) is a well-known constitutive model to capture the
compressive behaviour of confined concrete. This model was developed
and calibrated for the concrete-filled steel tubular sections. Therefore,
its application is limited to the concrete confined by a single skin metal
section. To apply this model for the concrete confined by double skin
metal tubes a modified confinement factor is proposed in the previous
studies [18,19,21,45]. In this study, this model with the modified
confinement factor is adopted to simulate the compressive response of
the concrete of CFDSAT columns. The response of concrete infill sub­
jected to tensile force was considered by a linear relationship of
stress–strain up to 10 % of concrete compressive strength and a post-
peak branch based on fracture energy [46–48].
The built-in surface-to-surface contact option in ABAQUS was
applied to mimic the interaction between the concrete infill and the
aluminium hollow tubes. The outer faces of the concrete were set as the
Fig. 5. Flexural buckling of a typical CFDSAT specimen S76.2 × 4.8-S25.4 master surfaces, while the outer faces of the inner section and the inner
× 3.2. faces of the outer section were assigned as the slave surfaces. In the
vertical direction of the master and the slave surfaces, hard contact was
time. Thus, 5 mm element size was adopted for the concrete and the applied for permitting separation under tension and avoiding penetra­
hollow tubes by using the structured meshing approach. The thickness of tion under compression. Coulomb friction with a coefficient of 0.3 was
outer and inner hollow sections was discretised at least into three ele­ employed in the lateral direction of the two types of surfaces for
ments to obtain the nonlinear behaviour of the thin-walled sections allowing sliding [7].
[41]. The residual stresses generated during the production of aluminium
For simulating the properties of hollow sections, the elastic–plastic alloy are very low [49]. The impact of initial local geometric imper­
model based on the von Mises yield principle with isotropic hardening fection on the buckling behaviour of CFDSAT columns is not significant
was considered. The modulus of Elasticity of aluminium alloy reported as the sensitivity of the hollow sections to the local buckling is reduced
in Table 2 and Poisson’s ratio of 0.33 was used for the elastic response. due to the existence of concrete infill. Hence, the effects of residual
According to ABAQUS software’s guidelines, the nominal stress-plastic

Fig. 6. Failure pattern of (a) the inner tube of a typical CFDSAT specimen, (b) the concrete infill of a typical CFDSAT specimen.

7
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 7. Axial load versus mid-height lateral displacement curves of the CFDSAT specimens.

stresses and initial local geometric imperfections were not included in amplitude of the global imperfection of a specimen was calculated by
the FE modelling [7]. The effect of initial global imperfection was adding the value of ωg reported in Table 1 with the respective value of e0
considered in the non-linear FE analysis by integrating the lowest global mentioned in Table 4. The pin-ended supports were simulated by
buckling mode determined from an eigenvalue buckling analysis. The applying the boundary conditions on the reference points which were

8
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

results. An imperfection sensitivity investigation was carried out to

evaluate an appropriate global imperfection amplitude for using in the
numerical parametric study. In this study, four different global imper­
fection values were considered, i.e., the measured value (i.e., summation
of ωg and e0) and three values of fractions of the effective length, i.e., Le/
1000, Le/1500 and Le/2000. Table 5 reports the ratios of FE over test
ultimate load (Nu,FE/Nu,Test) for the four global imperfection amplitudes
and FE over test flexural rigidity (EIe,FE/EIe,Test) for Le/1000. It can be
found that the global imperfection value of Le/1000 offers an excellent
prediction of the ultimate load with a mean value of Nu,FE/Nu,Test 1.02
and the corresponding coefficient of variation (COV) of 0.04. Moreover,
the flexural rigidity values of the specimens obtained from the tests are

Table 5
Ratios of FE over test ultimate load and flexural rigidity.
Specimen Nu,FE/Nu,Test EIe,FE/EIe,
Fig. 8. Compressive behaviour of confined concrete [42].
Test

Global imperfection value

created on the centroid of the top and bottom ends of the specimens. The
distance between the two reference points was the same as the effective Measured Le/ Le/ Le/ Le/1000
1000 1500 2000
length of the specimens. It was assumed that the top reference point can
move along the longitudinal direction of the specimen (Z direction) and S50.8 × 3.2-S19.2 × 1.04 1.00 1.07 1.11 1.01
both reference points can rotate around the buckling axis (X direction), 1.6
S63.4 × 3.2-S19.2 × 1.09 1.10 1.14 1.15 1.06
as shown in Fig. 9(a). Due to simplicity, the CFRP warps were not 1.6
modelled in this study, however, potential local failures at the ends were S63.4 × 3.2-S25.4 × 0.98 0.99 1.03 1.05 0.97
prevented by applying the coupling constraints between each reference 3.2
point and each end of the specimens [7,50]. At the top reference point, S76.2 × 4.8-S19.2 × 1.02 1.02 1.04 1.05 0.98
1.6
the axial compressive load was employed using a displacement control
S76.2 × 4.8-S25.4 × 1.02 1.03 1.06 1.07 1.08
approach. Fig. 9 shows the geometry, loading direction, boundary 1.6
conditions and cross-section of a typical FE model. S76.2 × 4.8-S25.4 × 1.07 1.04 1.07 1.08 1.03
3.2
S63.4 × 3.2-C19.2 × 0.98 0.96 0.99 1.00 0.99
4.2. FE model validation 3.2
S63.4 × 3.2-C25.4 × 1.02 1.01 1.04 1.07 1.02
3.2
The accuracy of the developed FE model was determined by Mean 1.03 1.02 1.05 1.07 1.02
comparing the FE ultimate load, flexural rigidity, load versus mid-height COV 0.03 0.04 0.04 0.04 0.03
lateral displacement curve and failure mode with the corresponding test

Fig. 9. (a) Loading direction and boundary conditions, (b) cross-section of a typical FE model.

9
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

very close to the ones of the FE analyses suggesting a very good accuracy depth/diameter of the inner section (i.e., increase of α). These obser­
of the developed models. Fig. 10 presents the comparison between the vations can be further explained by Fig. 14, which shows the contribu­
FE (for Le/1000) and test axial load versus mid-height lateral displace­ tion of the outer tube, concrete infill and inner profile to the ultimate
ment graphs of two specimens. The figure illustrates that the developed load of typical CFDSAT specimens. It can be observed from Fig. 14(a)
model accurately simulates the axial load versus mid-height lateral and (b) that the outer profile has the largest contribution to the overall
displacement relationship of the specimens. Moreover, Fig. 11 demon­ capacity, followed by the concrete infill which provides the second-
strates a good agreement between the failure mode observed from the highest cross-sectional resistance to axial compression among the
test and the FE analysis of a typical specimen. From the above com­ three components, thereby justifying the increase in the ultimate load
parisons, it can be concluded that the developed FE model accurately noticed in Fig. 13(a). The contribution of the inner profile to the cross-
replicates the structural response of CFDSAT columns. sectional compressive resistance is the lowest compared to the other
components of CFDSAT columns. Therefore, the variation of the depth/
4.3. Parametric study diameter of the inner profile has negligible influence on the ultimate
load of CFDSAT columns.
Using the developed model, a numerical parametric investigation
was performed to study the influence of hollow ratio, member slender­ 4.3.2. Effect of depth to thickness ratio of the outer tube
ness, cross-section slenderness of hollow tubes and concrete strength on Fig. 15 presents the effect of depth to thickness ratio of the outer tube
the ultimate load of CFDSAT columns. The stress–strain curves of the (Do/to) on the axial load versus mid-length lateral displacement of
materials of CFDSAT cross-sections presented in Fig. 12 were used in the typical CFDSAT columns with the same length and concrete strength.
parametric study. In total 81 CFDSAT columns were modelled consid­ The Do/to ratio ranging from 6 to 150 was considered by altering the
ering different outer sections, i.e., 45 × 45, 60 × 60, 75 × 75, 100 × 100, outer section thickness. It is observed from the figure that the effect of
120 × 120 and 150 × 150 mm2 with wall thickness varied from 1 to 8 Do/to ratio is remarkable on the buckling behaviour of CFDSAT columns,
mm and inner sections, i.e., 20 × 20, 25 × 25, 30 × 30 and 35 × 35 mm2 as the ultimate load of the columns increased noticeably by decreasing
with wall thickness ranged from 1 to 5 mm. Two combinations of outer Do/to ratio (i.e., increase of thickness of outer tube). This is because the
and inner cross-section shapes, i.e., square-square and square-circular, larger thickness provides higher flexural rigidity of the outer tube,
three different member heights, i.e., 500, 1000 and 1500 mm and which results in an improved ultimate load of the specimens. Similar
three different concrete cube compressive strengths, i.e., 30, 40 and 50 observation was noticed for the other studied specimens having lower
MPa were chosen. Do/to ratio.

4.3.1. Effect of hollow ratio 4.3.3. Effect of depth/diameter to thickness ratio of the inner tube
To investigate the influence of hollow ratio on the ultimate load of Fig. 16 illustrates the influence of the depth/diameter to thickness
CFDSAT columns a wide range of depth/diameter of the outer and the ratio of the inner section (Di/ti) on the buckling behaviour of CFDSAT
inner cross-sections was considered in this study. The hollow ratio is columns having constant length and concrete strength. The Di/ti ratios
calculated using Equation (4). were considered from 4 to 35 by varying the thickness of square and
circular inner sections. Fig. 16 indicates that with the decrease of Di/ti
Di
α= (4) ratio the ultimate load of the CFDSAT columns generally improved, as
Do − 2to
the smaller Di/ti ratio means larger thickness of the inner profile which
offers an addition in the ultimate load of the specimens. However, the
where α is the hollow ratio which is varied from 0.14 to 0.86 by
strength increase of Di/ti is minimal, as the inner tube provides the
changing the depth/diameter of outer and inner hollow sections. The
lowest cross-sectional resistance among the three components of the
effect of α on the axial load versus mid-length lateral displacement
CFDSAT columns (see Fig. 14).
graphs of typical CFDSAT columns having constant length and concrete
strength is illustrated in Fig. 13. It can be found from Fig. 13(a) that the
ultimate load of CFDSAT columns significantly improved with the 4.3.4. Effect of member slenderness (λ)
decrease of α (i.e., increase of depth of the outer tube) when the inner To examine the effect of member slenderness (λ) on the capacity of
section’s dimensions remain the same. This is related to the fact that by CFDSAT columns, three different column lengths, i.e., 500, 1000 and
increasing the depth, the flexural rigidity of the outer profile increases 1500 mm were considered without changing the cross-sectional di­
along with the increase of concrete area, which contributes to the mensions of the specimens and concrete strength. The member slen­
improvement of the ultimate load of the specimens. Fig. 13(b) demon­ derness is determined according to Eurocode 4 [34] (see Equation (8) in
strates that when the dimensions of the outer tube are constant the ul­ Section 5.1). Fig. 17 illustrates the influence of λ on the axial load versus
timate load of CFDSAT columns decreased slightly by enlarging the mid-length lateral displacement graphs of a typical CFDSAT specimen.

Fig. 10. Comparison of test and FE axial load versus mid-height lateral displacement curves.

10
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 11. Comparison of test and FE failure modes of a typical CFDSAT specimen.

Fig. 12. Stress–strain curves of the materials of CFDSAT cross-sections used in the parametric study.

11
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

Fig. 13. Effect of hollow ratio on the axial load versus mid-height lateral displacement of typical CFDSAT specimens having constant length and concrete strength.

Fig. 14. Contribution of the outer tube, concrete infill and inner tube on the ultimate load of typical CFDSAT specimens.

Fig. 15. Effect of Do/to ratio on the axial load versus mid-height lateral Fig. 16. Effect of Di/ti ratio on the axial load versus mid-height lateral
displacement curves of typical CFDSAT specimens having constant length and displacement curves of typical CFDSAT specimens having constant length and
concrete strength. concrete strength.

The figure illustrates that by increasing λ (i.e., larger member length) the Similar observation is presented in a previous research article on the
ultimate load of CFDSAT columns reduced noticeably. This can be behaviour of short and slender CFDSST columns [29].
clarified by the fact that when the cross-sectional dimensions of the
specimens are constant, the critical buckling load of columns decreases 4.3.5. Effect of concrete compressive strength
with the increase of the member length, which makes the columns prone In the parametric study, three concrete cube compressive strengths
to global buckling failure, resulting in a reduced ultimate load. More­ (fcu), i.e., 30, 40 and 50 MPa were adopted to investigate the influence of
over, it can be observed from Fig. 17 that the curves of columns with concrete strengths on the ultimate load of CFDSAT columns. The effect
lower λ exhibit a sharp drop in load after reaching the ultimate capacity. of different concrete strengths on the axial load versus mid-height lateral

12
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

for the design of CFDSAT members [51]. Hence, in this study, a design
methodology with a design buckling curve is proposed based on the
Eurocode 4 [34] framework for determining the capacity of CFDSAT
columns. The accuracy of the proposed methodology is assessed by
comparing the design strengths with the corresponding test and FE data.
It should be noted that in the proposed methodology the material partial
safety factors were considered equal to unity.

5.1. Design methodology for CFDSAT columns based on Eurocode 4 [34]

In Eurocode 4 [34], a design formula for determining the ultimate

load of fully concrete-filled carbon steel structural columns is provided
by summing the plastic cross-sectional resistance of the outer tube and
the concrete infill. To determine the ultimate load of CFDSAT columns,
the formula is modified by substituting the material properties of carbon
steel for the outer tube with those of aluminium alloy and considering
the contribution of the inner tube by adding its plastic cross-sectional
Fig. 17. Effect of member slenderness (λ) on the axial load versus mid-height
resistance. Thus, the ultimate load of CFDSAT columns is calculated
lateral displacement curves of a typical CFDSAT specimen S100 × 1.5-C20 × 1.
using Equation (5).

displacement of typical CFDSAT columns having same length is shown Npl = Ao f0.2,o + αAc fc + Ai f0.2,i (5)
in Fig. 18. It can be observed from the figure that the concrete grade has
a profound influence on the buckling response of the CFDSAT columns, where Ao, Ac, and Ai are the cross-sectional areas of the outer section,
as the bearing capacity of the columns enhanced substantially with the concrete infill and inner section, respectively. f0.2,o and f0.2,i are the yield
increase of the compressive strength of concrete. This is related to the (0.2 % proof) stresses of the outer and the inner sections, respectively. fc
fact that concrete provides the second highest cross-sectional resistance is the compressive cylinder strength of concrete, which is considered 80
among the three components of the CFDSAT columns (see Fig. 14) and % of fcu. It is assumed that the concrete reaches its full fc due to the
improves the ultimate load of the outer tube by preventing the occur­ confinement provided by the outer and inner sections. Therefore, the
rence of inward buckling and slowing down the formation of outward concrete coefficient α is considered 1 instead of 0.85 [19].
buckling. Hence, the higher grade of concrete adds more cross-sectional The design strength of the CFDSAT section is converted to member
resistance and offers stronger support against the local buckling of the strength by employing a reduction factor χ recommended by Eurocode 4
outer profile. Moreover, it is also found that the improvement of ulti­ [34].
mate capacity due to higher concrete grade is higher for the columns Nu,prop = χ Npl (6)
with bigger cross-sections than the smaller ones. For example, the ulti­
mate load of the specimen S75 × 1-S20 × 1 increased by around 41 % for The expression of χ is given in Equation (7).
50 MPa concrete with respect to the 30 MPa ones (Fig. 18(a)), whereas [ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ]− 1
the value is more than 58 % for the specimen S150 × 1-C20 × 1 (Fig. 18 (7)
2
χ = φ + ϕ2 − λ ⩽1.0
(b)) which is the highest percentage ultimate load increase recorded
among the investigated specimens in this study. It indicates that the In the above equation, φ is a parameter and λ is the member slen­
bigger cross-section contains a larger concrete area which provides more derness. The λ is calculated by Equation (8).
cross-sectional resistance against the axial compression for higher con­ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
/
crete grades. λ = Npl Ncr (8)

5. Design recommendations The Ncr denotes the critical buckling load or Euler’s critical load,
which is defined by Equation (9).
The existing design code for composite structural members, i.e., /
Ncr = π2 EIe L2e (9)
Eurocode 4 [34] provides rules for the design of fully concrete-filled
carbon steel tubular columns. However, no design rules are available

Fig. 18. Effect of concrete cube compressive strength on the axial load versus mid-height lateral displacement curves of a typical CFDSAT specimen having con­
stant length.

13
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

where EIe represents the effective flexural rigidity of the composite Table 6
section, which is calculated by Equation (10). Comparison of the design capacity and effective flexural rigidity of the speci­
mens with the corresponding test and FE values.
EIe = Eo Io + ke Ec Ic + Ei Ii (10)
Specimen No Nu/Nu, EIe,Test/FE/ EIe,Test/FE/
EIe,EC4 EIe,EC4
where Eo, Ec, and Ei are the modulus of Elasticity of the outer tube, prop

concrete infill and inner tube, respectively. Io, Ic, and Ii are the moment (ke = 0.6) (ke = 1)
of inertia of the cross-section of the outer profile, concrete infill and Test S50.8 × 3.2-S19.2 1 1.22 1.31 1.10
inner profile, respectively. The symbol ke represents the correction fac­ × 1.6
tor which is used to reduce the gross flexural rigidity of the concrete. The S63.4 × 3.2-S19.2 1 1.20 1.30 1.06
× 1.6
value of ke is taken as 0.6 according to Eurocode 4 [34]. S63.4 × 3.2-S25.4 1 1.13 1.22 1.01
The φ is determined by Equation (11). × 3.2
[ 2]
S76.2 × 4.8-S19.2 1 1.16 1.21 1.03
φ = 0.5 1 + α1 (λ − λ0 ) + λ (11) × 1.6
S76.2 × 4.8-S25.4 1 1.14 1.20 1.02
in which, α1 is the imperfection factor and λ0 is the limit of the × 1.6
horizontal plateau. The value of α1 is taken equal to 0.34 which is S76.2 × 4.8-S25.4 1 1.17 1.28 1.08
× 3.2
suggested by Eurocode 4 [34] for composite columns having a ratio of
S63.4 × 3.2-C19.2 1 1.14 1.34 1.09
the cross-sectional area of the hollow tube over concrete infill greater × 3.2
than 3 %. The value of λ0 is considered as 0.1 according to the recom­ S63.4 × 3.2-C25.4 1 1.14 1.33 1.08
mendation provided by Eurocode 9 [35] for Class A materials. It should × 3.2
FE (mean) 81 1.05 1.41 1.06
be noted that in Eurocode 9 [35] the aluminium alloys are categorized
Mean 1.06 1.40 1.05
into two material classes, i.e., Class A and B and the 6082-T6 aluminium (all)
alloy used in this study is a Class A material. Fig. 19 shows the test and COV 0.08 0.13 0.07
FE ultimate load normalised by plastic capacity (i.e., Nu/Npl) against the (all)
member slendernessλ. Moreover, the proposed design buckling curve
considering the above-mentioned values of α1 and λ0 is also presented in
the same figure. From the figure, an anticipated decreasing trend of the
normalised load can be observed with the increase of the member
slendernessλ. It is found that most of the test and the FE data are above
the buckling curve, which implies that the proposed design curve is
suitable for the CFDSAT columns design and provides safe capacity
predictions.

5.2. Ultimate load prediction for CFDSAT columns

To assess the applicability of the proposed design methodology, the

predicted ultimate capacities are compared with the corresponding ones
obtained from the tests and FE analyses. Table 6 lists the ratios of test
and FE ultimate load over the predicted design capacity (Nu/Nu,prop)
along with the corresponding mean and COV values. The mean and COV
values of Nu/Nu,prop are 1.06 and 0.08, respectively, indicating that the
proposed methodology is appropriate to predict the ultimate load of
CFDSAT columns. In Fig. 20, the test and FE ultimate load are plotted Fig. 20. Comparison of the design capacity with the corresponding test and FE
against the predicted design capacity. It also illustrates the accuracy of ultimate load.
the proposed methodology as most of the data in the figure are close to
the diagonal line of Nu = Nu,prop. Overall, it can be summarized that the
of CFDSAT columns.
proposed design methodology can accurately predict the ultimate load

5.3. Assessment of Eurocode 4 [34] concrete correction factor

In order to evaluate the suitability of applying Eurocode 4 [34]

concrete correction factor (ke = 0.6) for determining the effective flex­
ural rigidity of CFDSAT columns, the effective flexural rigidities of the
specimens calculated from test and FE data (EIe,Test/FE) are compared
with the values predicted according to Eurocode 4 (EIe,EC4) (see Equation
(10) in Section 5.1). The ratios of EIe,Test/FE over EIe,EC4 are reported in
Table 6, including their mean and COV values, which are 1.40 and 0.13,
respectively. The high mean value can be related to the reduced flexural
rigidity of concrete, which is 60 % of concrete gross flexural rigidity and
leads to conservative design predictions for CFDSAT columns. Aiming to
improve the accuracy of this formula and make it applicable to CFDSAT
columns, the concrete correction factor ke is set equal to 1, assuming that
the concrete in the CFDSAT section can achieve its full flexural rigidity
because of the confinement provided by the outer and inner tubes. The
ratios of EIe,Test/FE/EIe,EC4 considering ke = 1 are also listed in Table 6. It
Fig. 19. Proposed design buckling curve for CFDSAT columns. can be seen that by considering the full flexural rigidity of concrete, the

14
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

accuracy of the flexural rigidity formula for CFDSAT columns is

improved by 35 %. Fig. 21 presents the test and FE effective flexural
rigidities (EIe,Test/FE) of the specimens against the predicted values (EIe,
EC4) considering ke equal to 0.6 (Fig. 21(a)) and 1 (Fig. 21(b)). It can be
seen from Fig. 21(b) that the proposed concrete correction factor of ke =
1 for the flexural rigidity formula provides accurate predictions of the
effective flexural rigidity of CFDSAT columns, as the points are closer to
the diagonal line.

6. Conclusions

This study reported an experimental and numerical investigation on

the flexural buckling behaviour of CFDSAT columns. A total of 8 col­
umns were tested and 81 columns were numerically modelled, consid­
ering two combinations of outer and inner cross-section shapes, i.e.,
square-square and square-circular, different cross-section slenderness,
member slenderness and concrete strengths. Based on the observed re­
sults, the following points can be summarised:

1) The predominant failure mode of all CFDSAT tested columns was

flexural buckling. No inward or outward local buckling was identi­
fied at the outer tube of any specimen. Overall, the concrete infill
effectively prevented the occurrence of inward local buckling.
2) The experimental axial load versus mid-height lateral displacement
curves demonstrated good ductility of all CFDSAT specimens, which
is attributed to the beneficial composite action of the three elements
of the CFDSAT cross-sections.
3) It was shown that the developed FE model with a global imperfection
amplitude of Le/1000 accurately replicated the flexural buckling
behaviour of CFDSAT columns. Hence, this model was adopted to
conduct a numerical parametric study which included 81 specimens,
to investigate the effect of hollow ratio, member slenderness, cross-
section slenderness of hollow tubes and concrete compressive
strength on the ultimate capacity of CFDSAT specimens.
4) The numerical parametric study results revealed that the cross-
sectional dimensions of the outer tube, member slenderness and
concrete compressive strength have significant influences on the
ultimate load of CFDSAT columns. It was demonstrated that the
larger cross-sectional dimensions of the outer tube and the higher
concrete compressive strength substantially improved the ultimate
capacity of the composite members. However, the variation of cross-
sectional dimensions of the inner tube has a negligible effect on the
Fig. 21. Comparison of test and FE effective flexural rigidities (EIe,Test/FE) of the
ultimate load of the columns, as the inner tube provides the least
specimens with the corresponding predicted values (EIe,EC4) considering ke
flexural buckling resistance among the three components of the
equal to 0.6 and 1.
composite cross-sections.
5) In the absence of design standards for CFDSAT members, a design
administration. Michaela Gkantou: Conceptualization, Supervision,
methodology was proposed with a design buckling curve to predict
Writing – review & editing, Funding acquisition.
the ultimate load of CFDSAT columns according to the Eurocode 4
framework. It was demonstrated that the proposed design method­
ology with imperfection factor and limit of the horizontal plateau Declaration of Competing Interest
equal to 0.34 and 0.1, respectively is appropriate for the design of
CFDSAT columns as it accurately predicts the ultimate load of the The authors declare that they have no known competing financial
composite columns. interests or personal relationships that could have appeared to influence
6) A revised concrete correction factor of ke = 1 was suggested based on the work reported in this paper.
test and FE data for the formula of Eurocode 4 to determine the
effective flexural rigidity of CFDSAT columns. It is concluded that by Data availability
using the proposed value for ke, the accuracy of the formula to pre­
dict the effective flexural rigidity of CFDSAT columns is improved by No data was used for the research described in the article.
35 %.
Acknowledgements
CRediT authorship contribution statement
The authors are grateful to the technicians of the Schools of Civil
Shafayat Bin Ali: Conceptualization, Methodology, Software, Vali­ Engineering and Built Environment and Engineering at Liverpool John
dation, Formal analysis, Data curation, Investigation, Visualization, Moores University for their valuable assistance. The financial support of
Writing – original draft. George S. Kamaris: Conceptualization, Su­ Faculty of Engineering and Technology of Liverpool John Moores Uni­
pervision, Writing – review & editing, Funding acquisition, Project versity is gratefully acknowledged.

15
S. Bin Ali et al. Engineering Structures 275 (2023) 115316

References [26] Hassanein MF, Kharoob OF. Analysis of circular concrete-filled double skin tubular
slender columns with external stainless steel tubes. Thin-Walled Struct 2014;79:
23–37.
[1] Wang F, Young B, Gardner L. Testing and numerical modelling of circular CFDST
[27] Sulthana UM, Jayachandran SA. Axial compression behaviour of long concrete
cross-sections with stainless steel outer tubes in bending. Eng Struct 2021;247:
filled double skinned steel tubular columns. Struct 2017;9:157–64.
113170.
[28] Zhao H, Wang R, Lam D, Hou CC, Zhang R. Behaviours of circular CFDST with
[2] Alqawzai S, Chen K, Shen L, Ding M, Yang B, Elchalakani M. Behavior of octagonal
stainless steel external tube: slender columns and beams. Thin-Walled Struct 2021;
concrete-filled double-skin steel tube stub columns under axial compression.
158:107172.
J Constr Steel Res 2020;170:106115.
[29] Ahmed M, Liang QQ, Hamoda A, Arashpour M. Behavior and design of thin-walled
[3] Li W, Ren QX, Han LH, Zhao XL. Behaviour of tapered concrete-filled double skin
double-skin concrete-filled rectangular steel tubular short and slender columns
steel tubular (CFDST) stub columns. Thin-Walled Struct 2012;57:37–48.
with external stainless-steel tube incorporating local buckling effects. Thin-Walled
[4] Mazzolani FM. Aluminium alloy structures. 2nd ed. London Chapman Hall; 1995.
Struct 2022;170:108552.
[5] Mazzolani FM. Competing issues for aluminium alloys in structural engineering.
[30] Zhou F, Young B. Concrete-filled double-skin aluminum circular hollow section
Prog Struct Eng Mater 2004;6(4):185–96.
stub columns. Thin-Walled Struct 2018;133:141–52.
[6] Georgantzia E, Gkantou M, Kamaris GS. Aluminium alloys as structural material: a
[31] Zhou F, Young B. Compressive strengths of concrete-filled double-skin (circular
review of research. Eng Struct 2021;227:111372.
hollow section outer and square hollow section inner) aluminium tubular sections.
[7] Georgantzia E, Ali SB, Gkantou M, Kamaris GS, Kansara KD, Atherton W. Flexural
Advances. Struct Eng 2019:1–17.
buckling performance of concrete-filled aluminium alloy tubular columns. Eng
[32] Patel VI, Liang QQ, Hadi MNS. Numerical study of circular double-skin concrete-
Struct 2021;242:112546.
filled aluminum tubular stub columns. Eng Struct 2019;197:109418.
[8] Ali SB, Kamaris GS, Gkantou M, Kansara KD, Hashim K. Numerical study of
[33] Kissell J, Ferry R. Aluminum structures: a guide to their specifications and design.
concrete-filled aluminium alloy tubular columns under eccentric compression. In
New York John Wiley Sons; 2002.
IOP Conference Series: Materials Science and Engineering 2021;1058:012010.
[34] BS EN 1994-1-1. Eurocode 4 (EC4): Design of composite steel and concrete
[9] Ali SB, Kamaris GS, Gkantou M, Kansara KD. Concrete-filled and bare 6082–T6
structures. Part 1-1: General rules and rules for buildings. European Committee for
aluminium alloy tubes under in-plane bending: experiments, finite element
Standardization (CEN), Brussels, 2004.
analysis and design recommendations. Thin-Walled Struct 2022;172:108907.
[35] BS EN 1999-1-1. Eurocode 9 (EC9): Design of aluminium structures. Part 1-1:
[10] Zhao XL, Grzebieta R. Strength and ductility of concrete-filled double skin (SHS
General structural rules - General structural rules and rules for buildings. European
inner and SHS outer) tubes. Thin-Walled Struct 2002;40:199–213.
Committee for Standardization (CEN), Brussels, 2007.
[11] Tao Z, Han LH. Behaviour of concrete-filled double skin rectangular steel tubular
[36] BS En ISO. 6892–1. metallic materials – tensile testing – part 1: Method of test at
beam–columns. J Constr Steel Res 2006;62:631–46.
room temperature. Brussels: European Committee for Standardization (CEN);
[12] Tao Z, Han LH, Zhao XL. Behaviour of concrete-filled double skin (CHS inner and
2009.
CHS outer) steel tubular stub columns and beam-columns. J Constr Steel Res 2004;
[37] Ramberg W, Osgood WR. Description of stress-strain curves by three parameters.
60:1129–58.
Technical Note No. 902. Washington, D.C., USA: National Advisory Committee for
[13] Uenaka K, Kitoh H, Sonoda K. Concrete filled double skin circular stub columns
Aeronautics, 1943.
under compression. Thin-Walled Struct 2010;48:19–24.
[38] Hill HN. Determination of stress-strain relations from “offset” yield strength values.
[14] Yan XF, Zhao YG. Compressive strength of axially loaded circular concrete-filled
Technical Note No. 927; 1944.
double-skin steel tubular short columns. J Constr Steel Res 2020;170:106114.
[39] En BS. 12390–3. Testing hardened concrete. Compressive strength of test
[15] Elchalakani M, Zhao XL, Grzebieta R. Tests on concrete filled double-skin (CHS
specimens. European Committee for Standardization (CEN), Brussels 2009.
outer and SHS inner) composite short columns under axial compression. Thin-
[40] Karlsson Hibbitt, Sorensen, Inc. ABAQUS. ABAQUS/Standard User’s Manual
Walled Struct 2002;40:415–41.
Volumes I-III and ABAQUS CAE Manual, 2016. Pawtucket (USA), Version 2016.
[16] Han LH, Tao Z, Huang H, Zhao XL. Concrete-filled double skin (SHS outer and CHS
[41] Milan CC, Albareda-Valls A, Carreras JM. Evaluation of structural performance
inner) steel tubular beam-columns. Thin-Walled Struct 2004;42(9):1329–55.
between active and passive preloading systems in circular concrete-filled steel
[17] Wang F, Young B, Gardner L. Experimental study of square and rectangular CFDST
tubes (CFST). Eng Struct 2019;194:207–19.
sections with stainless steel outer tubes under axial compression. J Struct Eng
[42] Tao Z, Wang ZB, Yu Q. Finite element modelling of concrete-filled steel stub
2019;145(11):04019139.
columns under axial compression. J Constr Steel Res 2013;89:121–31.
[18] Wang F, Young B, Gardner L. Compressive testing and numerical modelling of
[43] Papanikolaou V, Kappos A. Confinement-sensitive plasticity constitutive model for
concrete-filled double skin CHS with austenitic stainless steel outer tubes. Thin-
concrete in triaxial compression. Int J of Solids Struct 2007;44(21):7021–48.
Walled Struct 2019;141:345–59.
[44] Yu T, Teng J, Wong Y, Dong S. Finite element modeling of confined concrete-I:
[19] Wang F, Young B, Gardner L. CFDST sections with square stainless steel outer tubes
Drucker-Prager type plasticity model. Eng Struct 2010;32(3):665–79.
under axial compression: Experimental investigation, numerical modelling and
[45] Huang H, Han LH, Tao Z, Zhao XL. Analytical behaviour of concrete-filled double
design. Eng Struct 2020;207:110189.
skin steel tubular (CFDST) stub columns. J Constr Steel Res 2010;66(4):542–55.
[20] Han LH, Ren QX, Li W. Tests on stub stainless steel–concrete–carbon steel double
[46] Hordijk D. Local approach to fatigue of concrete (PhD Thesis). (Delft: Delft
skin tubular (DST) columns. J Constr Steel Res 2011;67(3):437–52.
University of Technology); 1991.
[21] Wang F, Young B, Gardner L. Compressive behaviour and design of CFDST cross-
[47] FIP. Ceb-Fip Model Code 1990. Ceb-Fip Model Code 1990. 1993.
sections with stainless steel outer tubes. J Constr Steel Res 2020;170:105942.
[48] Bažant Z, Becq-Giraudon E. Statistical prediction of fracture parameters of concrete
[22] Tziavos NI, Gkantou M, Theofanous M, Dirar S, Baniotopoulos C. Behaviour of
and implications for choice of testing standard. Cem Concr Res 2002;32:529–56.
grout-filled double-skin tubular steel stub-columns: numerical modelling and
[49] Mazzolani FM. Residual Stress Tests Alu-Alloy Austrian Profiles, ECCS Committee,
design considerations. Structures 2020;27:1623–36.
Brussels, 1975 Technical Report, Doc 16-75-1.
[23] Zhao XL, Tong LW, Wang XY. CFDST stub columns subjected to large deformation
[50] Zhou F, Young B. Numerical analysis and design of concrete-filled aluminum
axial loading. Eng Struct 2010;32:692–703.
circular hollow section columns. Thin-Walled Struct 2012;50(1):45–55.
[24] Han LH, Li YJ, Lao FY. Concrete-filled double skin steel tubular (CFDST) columns
[51] Ali SB, Kamaris GS, Gkantou M. Flexural behaviour of concrete-filled double skin
subjected to long-term sustained loading. Thin-Walled Struct 2011;49:1534–43.
aluminium alloy tubes. Eng Struct 2022;272:114972.
[25] Zheng Y, He C, Zheng L. Experimental and numerical investigation of circular
double-tube concrete-filled stainless steel tubular columns under cyclic loading.
Thin-Walled Struct 2018;132:151–66.

16