You are on page 1of 16

Thin-Walled Structures 123 (2018) 467–482

Contents lists available at ScienceDirect

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

Full length article

Finite element analysis of large diameter high strength octagonal CFST short T
columns

M.F. Hassaneina, Vipulkumar Ishvarbhai Patelb, , Mohamed Elchalakanic, Huu-Tai Thaid
a
Department of Structural Engineering, Faculty of Engineering, Tanta University, Tanta, Egypt
b
School of Engineering and Mathematical Sciences, La Trobe University, Bendigo, VIC 3552, Australia
c
School of Civil, Environmental and Mining Engineering, Faculty of Engineering, Computing and Mathematics, The University of Western Australia, Australia
d
Institute of Research and Development, Duy Tan University, 03 Quang Trung, Da Nang, Viet Nam

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

Keywords: Octagonal masonry columns have been broadly used in outstanding architectural heritage due to their aesthe-
Axial compressive strength tical appearance and to support high loads. Nowadays, concrete-filled steel tubular (CFST) columns have been
Finite element analysis utilised to carry such high load. However, a recent trend in the construction of the CFST columns is to use the
Stub columns high strength concrete (HSC) to increase the usable floor spaces in different buildings due to the global limited
Design model
land areas. Therefore, this paper combines the architectural demand of the octagonal column shape with the
Concrete confinement
advantages of the CFST columns by investigating numerically, by means of finite element (FE) modelling, the
octagonal CFST short columns. Accordingly, validated FE models for the octagonal CFST short columns are
employed to perform parametric studies to widen the available knowledge about their behaviour. The paper is
mainly devoted for investigating large diameter columns using extensive series of diameter-to-thickness (D/t)
ratios ranging from 40 to 200, most of which filled with HSC up to 100 MPa. The ultimate strengths, based on
available experiments in literature and current FE investigation, are compared with the existing design models.
This comparison indicates that the existing design models are conservative and their accuracy is affected by the
D/t ratios. Therefore, a new design model is suggested based on the existing provisions by taking the D/t ratios of
the columns into account. This suggested design model is validated by using the existing experimental results
and is found to give excellent results. The important factors that affect the strength and behaviour of the oc-
tagonal CFST short columns filled with HSCs are additionally discussed in detail, with new conclusions added to
literature for the first time.

1. Introduction an outstanding example of Romanesque architecture which was built in


the late XIII century in Italy, are supported by two lines of seven oc-
The field of architectural engineering extensively contributes in the tagonal columns [4]. On the opposite, the columns of high-rise build-
choice of the main structural system in different projects. This is be- ings nowadays often are made of steel-concrete hybrid/composite sys-
cause the system to be built not only should have structural efficiency, tems, which is advantageous to provide wider free-column spaces with
material availability, low cost, ease of construction methods, and man lighter members [5–7]. Thus, larger column-free spaces are possible.
power, but also should guarantee that the architectural requirements of One of the most effective steel-concrete member is CFST column.
the building and its aesthetical appearance are taken into consideration This column composes of an empty steel tube filled with concrete, at
[1–3]. Hence, it is a common fact that the architecture of the building which both elements sustain together the applied axial load. A CFST
must suitably be integrated with its structural aspects to produce an column has been extensively used in civil engineering projects around
optimum economic balance. As the demand grew for lighter, taller and the world [1–3,5–7]. In this column, the outer steel tubular column
more rapidly constructed structures, the columns of the structure have confines the infilled concrete, while the infilled concrete prevents its
become of great importance. However, one of the columns that have inward buckling. The most used CFST columns in construction are those
been used extensively in ancient civilisations to sustain high loads (in of circular and rectangular (containing square) cross-sections. While the
high and huge temples for example) is the octagonal masonry columns. circular CFST columns are more efficient compared with the rectan-
Just as an example, the arches of the Basilica of Collemaggio, which is gular due to the enhanced confinement, the rectangular columns have


Corresponding author.
E-mail addresses: mostafa.fahmi@yahoo.com, mostafa.fahmi@f-eng.tanta.edu.eg (M.F. Hassanein), v.patel@latrobe.edu.au (V.I. Patel).

https://doi.org/10.1016/j.tws.2017.11.007
Received 21 July 2017; Received in revised form 21 October 2017; Accepted 7 November 2017
0263-8231/ © 2017 Elsevier Ltd. All rights reserved.
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

1. The paper provides wider range of D / t ratios for the current octa-
gonal CFST short columns than those currently available in litera-
ture [8,13]. The current FE models are generated from the concrete
constitutive model suggested by Patel et al. [14] and currently ca-
librated by using the test results of Tomii et al. [8] and Ding et al.
[13]. This is made by using the Abaqus computer package [18].
2. The ultimate strengths obtained numerically are, then, compared
with the three available design strengths by Yu et al. [10], Ding
et al. [13] and Patel et al. [14]. An emphasis is given to the effect of
D / t ratios on the accuracy of each mentioned design model
[10,13,14].
3. Based on the comparative results in the previous point, a new design
model is suggested to obtain better predictions for the octagonal
Fig. 1. Typical cross-section for octagonal CFST columns. large diameter CFST short columns over the full range of the in-
vestigated D / t ratios. Verification for this design model is made
through comparisons with international design specifications.
the advantage that they produce easier connections [1]. The advantages
4. The paper extends to discuss the main factors (i.e. the D / t ratios and
of both cross-sections may be accompanied by using the octagonal
the concrete compressive strength fc′) affecting the behaviour of the
cross-section (see Fig. 1), at which the confinement of the concrete core
large diameter octagonal CFST short columns filled with HSCs.
increases (compared with the rectangular) [8,9] with the ability to
5. An emphasis is given, at the end, to the combined effect of the D / t
generate easier connections in the orthogonal directions (relative to the
ratios and the fc′ values aiming at providing some tips that optimizes
circular) [6]. At the same time, it has an aesthetical appearance. Ad- the cross-sections of the octagonal steel tubes.
ditionally, these octagonal CFST columns may be used in modern
structures built in vicinity of historical architecture. By doing so, these
modern buildings may be in harmony with the existing ancient archi- 2. Finite element analysis
tectural features. However, it is a common fact that the steel-concrete
columns with slender steel tubular sections were examined in civil The novelty of the FE model is to develop the Python Script for the
engineering, whereas those columns utilising reduced section sizes were simulation of octagonal CFST columns. Although full-scale physical
initially investigated in architectural engineering [2]. This raises the tests provides an accurate insight into the structural characteristic of
importance of investigating large diameter columns which is the scope octagonal CFST columns, the experiments are quite expensive and time
of this paper. consuming, and they are less flexible to conduct the parametric study.
Experimental studies on octagonal CFST short columns have been Therefore, the systematic development of FE modelling can be utilised
initiated by Tomii et al. [8] in 1977, with D / t ratios varied between 37 for an accurate complementary investigation. This section provides the
and 75. After that, Susantha et al. [9] proposed a confined concrete details of the FE modelling utilised for replicating the experiments re-
model for the analysis of octagonal CFST columns in 2001. In 2013, Yu ported by Tomii et al. [8] and Ding et al. [13]. It is worth pointing out
et al. [10] suggested a unified compressive strength for both the circular that the test specimens by Ding et al. [13] were considered because they
and polygonal CFST columns. Yu et al. [10] verified this design strength are the only available tests on large diameter octagonal CFST short
by using the experimental results of Tomii et al. [8], beside the results columns, as discussed above.
of two experimental investigations provided in Chinese on small dia-
meter experiments [11,12]. However, it has been a long time (exactly in
2.1. Python scripts
2016) to another test series to take place by Ding et al. [13], with D / t
ratios in between 80 and 126. While Tomii et al. [8] tested eight small
The analysis of the CFST columns requires creating the individual
diameter experiments (D = 150 mm ), Ding et al. [13] examined those
components, assigning individual material properties, generating dif-
of relatively large diameter columns (four pairs of identical columns
ferent mesh sizes, assembling components, defining analysis steps and
with D = 480 mm ) and they suggested a new design model. More re-
also applying loads, constraints and boundary conditions based on
cently, a numerical investigation has been generated by Patel et al. [14]
geometric and material properties. The simulations are generally con-
to propose a material constitutive model for the confined concrete in
ducted in the same way multiple times for the verification as well as the
short circular, elliptical and octagonal CFST columns subjected to axial
parametric studies. The load levels, concrete stress-strain data, steel
compression. The material constitutive model for octagonal CFST short
stress-strain data, constraints and interactions change every time in the
columns given by Patel et al. [14] is a modified version of that given by
analysis. Accordingly, a python script, supported by Abaqus [18], was
Susantha et al. [9] based on FE analysis results. The FE models in Ref.
developed to perform the same task multiple times to generate the FE
[14] were validated by using the results of Tomii et al. [8] on the small-
simulations of the current octagonal short CFST columns. The geo-
diameter experiments, as shown earlier, which were filled by normal-
metric and material parameters were assigned as variables in the de-
strength concrete (NSC).
veloped Python script, at which the script runs in a loop for the given
From the above discussion, it can be recognised that several open
input variables. The script consists of two analysis steps. The first step is
points for the structural performance of the octagonal short CFST col-
the elastic buckling analysis which searches for the Eigenmode that
umns still call for additional research. Accordingly, this paper ad-
would be scaled to add the imperfect shape in the second step. The
dresses, by means of FE analyses, the axial compressive strength and
static general nonlinear analysis is, then, carried out in the consecutive
behaviour of large diameter octagonal short CFST columns filled with
step. The initial geometric imperfection value of L/1000 was considered
HSC. This is based on the two main facts that: (1) the above in-
in the FE simulation, as typically given in Refs. [14,19]. Despite the fact
vestigations based on small diameter columns cannot display the me-
that the effect of initial imperfection is not critical in short CFST col-
chanical behaviour of the large diameter ones used in practice [15,16]
umns [20], initial imperfections were included in this investigation to
and (2) the HSC is widely considered these days in different landmarks
allow the deformed shapes of the columns to be propagated. To the
around the world [7,17].
current authors, this would raise the confidence of the accuracy of the
The main objectives of the current paper are summarised as:
FE models and would accurately give the stress contours that enhance
the knowledge about the load transfer and stress concentrations.

468
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

tube and concrete core [22]. The inside surface of octagonal steel sec-
tions was defined as master surface and the outside surface of infilled
concrete was assigned as slave surface for the contact algorithm. The
interaction between two deformable surfaces in the normal direction
was specified by ‘Hard Contact’. This avoids the surface penetration in
compression and allows the separation in tension. The tangential be-
haviour of contacted surfaces was specified by using the Coulomb
friction model. As there is no slip up to the ultimate load between the
concrete and steel sections in octagonal CFST short columns when both
the steel tube and concrete core are loaded simultaneously [14], the
behaviour of octagonal short CFST columns under axial loading is not
affected by friction coefficient (μ) . The friction coefficient (μ) of 0.3 was
utilised in this study, as suggested by Patel et al. [14].

2.4. Load and boundary conditions

The loading end plates were not included in the current FE models.
Instead, the rigid body tie constraints with tie nodes were utilised for
applying the loads and boundary conditions. The reference points si-
tuated on the bottom and top levels of columns were assigned to control
the displacement-controlled loads and degrees of freedom. The degree
of freedom at the bottom reference point was fully restrained in
translations and rotations while the top reference point was allowed to
displace in the loading direction. This represents the clamped end
Fig. 2. A typical FE model for octagonal CFST short column showing the FE types and the conditions typical to that suggested previously by Tao el al. [19].
utilised mesh.
2.5. Design-oriented material law for confined concrete
2.2. Solid elements and element size
The confinement effect, which increases the compressive strength,
The eight-node reduced integration linear solid elements with three ultimate strain and ductility of concrete in octagonal CFST columns, is
translation degrees of freedom, C3D8R, were utilised for meshing oc- considered in the FE model. Patel et al. [14] developed a design-or-
tagonal concrete core. The thin-walled steel tube used in octagonal iented material model for confined concrete based on regression ana-
CFST short columns buckles locally outwards with the restraints pro- lyses and test results. The Drucker-Prager methodology is utilised for
vided by the filled concrete. The solid elements were effectively found simulating the infilled concrete in octagonal CFST short column. This
to capture the outer buckling of the thin-walled steel tubes subjected model is suitable for modelling materials having substantially greater
uniform axial loading [14]. Therefore, the octagonal steel tubes were compressive yield strengths relative to their tensile yield strengths
also discretised using the eight-node linear solid elements with three [18,22]. Additionally, such model well simulates the shear strength
translation degree of freedom (C3D8R). The sensitivity analysis was increase resulting from the increase in the hydrostatic pressure, which
carried out for examining the influence of the global mesh size of the characterises the behaviour of the core restrained concrete [24].
concrete core and steel tube on the accuracy of the predicted behaviour Therefore, this model has been widely used for FE modelling of con-
for octagonal short CFST columns. The sensitivity study indicates the fined concrete in many previous investigations; see for example Refs.
steel and concrete sections can be meshed with the 15 mm and 30 mm [24,25]. The angle of friction was taken as 20° while the flow-stress
element size, respectively. A typical FE model for octagonal CFST short ratio is 0.8 for the Drucker-Prager parameters based on the works of Hu,
columns contains over 20,000 solid elements. A typical FE mesh is et al. [26] and Wu [27]. The dilation angle for the Drucker-Prager
shown in Fig. 2. It should be noted that shell elements used for steel parameters is calculated from the equation given by Tao et al. [19]. The
tube causes convergence problems so the solid elements were utilised typical stress-strain curve of confined concrete in octagonal CFST
for discretising the steel tube [21]. Two layers of solid elements were column suggested by Patel et al. [14] is depicted in Fig. 3. The stress in
assigned through the thickness of the steel tube. It should be noted that concrete between Point O and A is determined using the formulas
the small finite element shown in Fig. 2 may cause the local stress proposed by Mander et al. [28], as:
concentration at the sharp corners. The mesh sensitivity study is con-
ducted to refine the mesh which avoids any stress concentration.

2.3. Surface-to-surface contact interaction

It is widely known that the bond between the inner surface of the
steel tube and the concrete core results from the chemical adhesion,
frictional resistance and physical interlocking of the concrete and steel
surfaces at the micro scale [22]. However, an accurate modelling for the
bond behaviour at the steel-concrete interfaces, to incorporate the
contributions of the above mentioned effects, is much challenging.
Therefore, numerous investigations have simplified the bond behaviour
by considering merely the frictional resistance in the FE modelling; see
for example [23].
On that basis, the surface-to-surface contact interactions were em- Fig. 3. Design-oriented material model of confined concrete for octagonal CFST short
columns [14].
ployed, in this investigation, on the deformable surfaces of the steel

469
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

fcc′ λ (εc / εcc′ ) 05


1.0
σc =
λ − 1 + (εc / εcc′ )λ (1)
1

Strength reduction factor γc


Ec
λ=
Ec − (fcc′ / εcc′ ) (2) 0.9
95

Ec = 3320 γc fc′ + 6900 (MPa) 0..9


(3)

in which σc represents the axial compressive concrete stress, εc denotes 0.8


85
the axial compressive concrete strain, fcc′ presents the axial compressive
confined strength of concrete, εcc′ shows the compressive strain of con- 0..8
fined concrete at fcc′ and Ec represents the Young's modulus of concrete,
ACI-363 [29]. 0.7
75
Liang and Fragomeni [30] in their investigation confirmed the
globally well-known fact that the uniaxial compressive strength (fcc′ ) 0..7
and the strain εcc′ of the confined concrete core are much higher than
0 100 200 300 400 500 0 600
those of the unconfined concrete. Hence, the formulas given by Mander Diameeter of concrette core Dc (mm
m)
et al. [28] in determining the concrete strength and strain of confined Fig. 4. Effect of column size on the strength reduction factor for concrete [36].
concrete were modified by Liang and Fragomeni [30]. The modified
formulas are given as:
⎧ (0.02508 − 0.0003977 D /t ) f y for 17 D /t 47

fcc′ = γc fc′ + k1 frp (4) frp = (0.004465 − 0.0000015D /t ) f y for 47 ≤ D /t < 150

⎪ 0.0 for D / t ≥ 150 (8)
εcc′ = εc′ (1 + k2 frp / γc fc′ ) (5) ⎩
in which D represents the diameter of octagonal section and t denotes
in which frp represents the confining pressure on the concrete, εc′ is the
the thickness of octagonal steel tube. As can be noted from Eq. (8), the
strain at the strength fc′ of the unconfined concrete, k1 and k2 are con-
current models assume that there is no confining pressure on the con-
stants. The values of k1 and k2 suggested by Richart et al. [31] as 4.1 and
crete core when D / t > 150 .
20.5, respectively, for the case of circular CFST columns were currently
Two linear segments AB and BC of the stress versus strain curve of
considered for the case of octagonal CFST columns.
confined concrete, depicted in Fig. 3, were assumed by Patel et al. [14]
On the other hand, the concrete compressive strength varies with
as:
the column size, concrete quality and loading rate. It can be recognised
from literature that the compressive strength of standard concrete cy- f ′ + (fe − fcc′ ) (εc − εcc′ )/(εe − εcc′ ) for εcc′ < εc ≤ εe
linders may not present the actual compressive strength of the concrete σc = ⎧ cc
⎨ fe + (fcu − fe ) (εc − εe )/(εcu − εe ) for εe < εc ≤ εcu (9)
in real CFST columns. Experimental results indicated that the com- ⎩
pressive strength of concrete varied from 0.85fc′ to 1.0fc′ [32–34]. As the in which εe and εcu are concrete compressive strains at points B and C,
concrete quality and loading rate are difficult to be estimated, a single respectively, as depicted in Fig. 3. The compressive strain εe is taken as
strength reduction factor is generally used to approximately account for 10εc′ corresponding to point B. The compressive strain εcu at point C is
the effects of these parameters, which is a function of the column size taken as 30εc′. The strains at Points B and C (Fig. 3) were suggested by
[35]. Accordingly, Liang [15] proposed a strength reduction factor (γc ) Dai and Lam [21]. In Eq. (9), fe is the stress corresponding to strain εe
for determining the effective compressive strength fcc′ of concrete, and fcu represents the compressive stress at strain εcu . The stresses fe and
which is determined as γc fc′ is expressed for circular sections by Eq. (6), fcu are given by:
which is currently used for the octagonal CFST short columns.
fe = α c fcc′ (10)
γc = 1.85Dc−0.135 (0.85 ≤ γc ≤ 1.0) (6)
fcu = βc fcc′ (11)
in which Dc represents the diameter of the inscribed concrete core and
is taken as (D − 2t ) for an octagonal cross-section. The parameters D where the factors α c and βc represent the confinement effect on the
and t are shown in Fig. 1. Although the ultimate strength of CFST ductility of the infilled concrete in the post-peak stage. These factors
columns is determined from the concentric tests, it varies with the depend on the strength of concrete [21]. The factors are proposed [14]
column size. This means that the ultimate strength of small scale spe- for octagonal CFST short columns as:
cimen may not represent the actual strength in real large scale CFST α c = 0.9729 − 0.0091 γc fc′ (12)
columns. The strength reduction factor as a function of the column size
determined using Eq. (6) is given in Fig. 4. The figure illustrates that the βc = 0.9987 − 0.0107 γc fc′ (13)
strength reduction factor decreases with an increase in the size of CFST
columns and its value varies between 0.85 and 1.0. For practical CFST
columns with Dc ≥ 300 mm , the strength reduction factor is limited to 2.6. Design-oriented material law for structural steels
0.85 [36].
The strain εc′ presents the strain at fc′ of unconfined concrete. The The structural steel is generally utilised in the construction of CFST
strain εc′ was given by Liang and Fragomeni [30]: columns. The stress versus strain response of steel is characterised by a
typical yield plateau. An idealised three-line stress versus strain beha-
⎧ 0.002 for γc fc′ ≤ 28 (MPa) viour was used in modelling of structural steel as presented in Fig. 5. In

εc′ = 0.002 + (γ f
c c′ − 28)/54000 for 28 < γc fc′ ≤ 82 (MPa) Fig. 5, σs represents the axial compressive steel stress, εs is the axial

⎪ 0.003 for γc fc′ > 82 (MPa) compressive steel strain, f y denotes the steel yield strength, εy stands for
⎩ (7)
the steel yield strain, εt is the hardening strain, fsu represents the steel
The equations for determining the lateral confining pressure on the tensile strength and εsu is the steel ultimate strain. The steel hardening
concrete core for octagonal CFST short column are proposed by Patel strain εt is considered as 10εy for structural steel. The steel ultimate
et al. [14] and it is given by: strain εsu is assumed to be 0.2 [15].

470
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

OST4 is measured as 321 MPa. The measured material and geometric


properties for all tested specimens are listed in Table 1, in which Pul . Exp
is the tested ultimate axial load and Pul . FE represents the ultimate axial
load obtained from the FE modelling. As shown in Table 1, the FE
model yields good estimations for the ultimate axial strengths of octa-
gonal short CFST columns under axial loading. The average ultimate
axial strength obtained from the FE modelling is 98% of the tested one.
The standard deviation (SD) of Pul . FE / Pul . Exp is 0.04, while its coefficient
of variation (COV) is 0.04. Hence, it is shown that the FE modelling
tends to give safe predictions for the ultimate axial strengths of octa-
gonal CFST short columns.
The axial load versus axial strain curves of the octagonal CFST
columns obtained from the FE modelling are compared with the ex-
Fig. 5. Design-oriented material model of structural steels for octagonal CFST short col- perimental results given by Tomii et al. [8] in Fig. 6. As shown in Fig. 6,
umns. the above described FE model reasonably predicts the experimentally
observed axial load versus axial strain response of the octagonal CFST
3. Evaluation of developed FE simulation columns. It is clearly seen that the predicted stiffness agrees well with
the experimental one except for specimens 3MN and 4MN. However, a
A FE model for simulating the structural performance of the axially- discrepancy can be observed between the predicted and the experi-
loaded octagonal short CFST columns reported by Tomii et al. [8] was mental curves in the post-yield range. However, this discrepancy falls
previously validated by Patel et al. [14]. The comparison of FE pre- within the acceptance range of ± 10%. This difference between the
diction and experiment given by Tomii et al. [8] is presented herein for predicted and experimental behaviours may be attributed, to the au-
completeness of the current paper. The D/t ratio of these tested octa- thors' knowledge, to the possible existence of initial imperfections in
gonal CFST short columns ranged from 37 to 75. The steel yield the tested specimens in the shape of outward deformation due to the
strength ranged from 294 MPa to 341 MPa while the concrete com- hydrostatic pressure when the wet concrete was poured into the octa-
pressive strength ranged between 17 MPa and 30 MPa. On the other gonal bare steel specimens. The effects of these outward deformations
hand, the latest test data on large diameter octagonal short CFST col- were not considered in the current paper. In addition, the minor stiff-
umns subjected to axial loading given by Ding et al. [13] are employed ness inaccuracies may result from the differences between the real
in this study to further investigate the evaluation of the FE models. Ding conditions within the specimens as the Young’ modulus of steel and
et al. [13] tested eight octagonal short CFST columns subjected to axial concrete materials, initial imperfections, boundary conditions and
compression with D/t ratios ranged between 80 and 126. The length of precision of the measurement devices with those assumed in the com-
these specimens is fixed to 1500 mm. The columns were filled with putational FE simulations. These real effects were not accurately re-
either concrete cube strength ( fcu ) of 39.3 MPa or 57.4 MPa. By using flected in the FE model as it is very difficult to fully simulate all the real
the equation found in Ref. [19] to convert fcu values to the corre- test conditions.
sponding cylinder strengths ( f ′c ), as given herein in Eq. (14), these test The computational and experimental axial load versus axial strain
specimens are filled by concrete with cylinder strength of 32.2 MPa and responses of the axially-loaded octagonal CFST short columns OST1-A,
49.0 MPa, respectively. OST1-B, OST2-A, OST2-B, OST3-A, OST3-B, OST4-A and OST4-B are
depicted in Fig. 7. It illustrates the elastic stiffness of the columns ob-
f tained from the FE modelling is marginally lower than that of the tested
fc′ = ⎡0.76 + 0.2 log10 ⎛ cu ⎞ ⎤ fcu
⎜ ⎟
⎢ ⎝ 19.6 ⎠ ⎥ (14) columns. The post-peak behaviour of the columns predicted by the FE
⎣ ⎦
modelling reasonably agrees with the measured post-peak behaviour by
The yield stress of the steel tubes for specimens OST1 and OST3 is Ding et al. [13]. Excellent agreement is achieved for the OST3-A
recorded as 311 MPa, while the steel yield stress of columns OST2 and

Table 1
Ultimate axial strengths of axially-loaded octagonal CFST short columns.

Specimens D (mm) t (mm) L (mm) D f y (MPa) fu (MPa) Es (GPa) fc′ (MPa) Pul, Exp (kN) Pul, FE (kN) Pul, FE Ref.
t Pul, Exp

2HN 150 2.00 300 75 341.3 430 206 30.1 989 946 0.96 [8]
3HN 150 3.20 300 47 300.2 430 206 30.1 1094 1095 1.00
4HN 150 4.00 300 38 294.3 430 203 30.1 1316 1263 0.96
2MN 150 2.00 300 75 341.3 430 206 21.9 771 813 1.05
3MN 150 3.20 300 47 300.2 430 206 21.9 916 966 1.05
4MN 150 4.00 300 38 294.3 430 203 21.9 1193 1136 0.95
3LN 148 3.20 300 46 300.2 430 206 16.7 856 885 1.03
4LN 149 4.00 300 37 294.3 430 203 16.7 1117 1074 0.96

OST1-A 485 3.85 1500 126 311 460 209 32.2 9297 9051 0.97 [13]
OST1-B 480 3.98 1500 121 311 460 209 32.2 9311 8942 0.96
OST2-A 483 6.02 1500 80 321 480 202 32.2 10,502 10,009 0.95
OST2-B 476 5.89 1500 81 321 480 202 32.2 10,713 9690 0.90
OST3-A 483 3.92 1500 123 311 460 209 49.0 12,362 12,066 0.98
OST3-B 480 4.02 1500 119 311 460 209 49.0 12,357 12,137 0.98
OST4-A 476 5.88 1500 81 321 480 202 49.0 12,992 12,643 0.97
OST4-B 478 5.98 1500 80 321 480 202 49.0 13,263 12,836 0.97
Mean 0.98
Standard deviation (SD) 0.04
Coefficient of variation (COV) 0.04

471
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

between the computational curve and the measured response as shown the HSCs in this paper is attributed to the fact that they are widely
in Fig. 7(e). Fig. 8 presents a deformed shape obtained from the FE utilised when reducing the weight of the structure is important or when
model for the octagonal CFST columns, where symmetric outward smaller load-carrying elements are required for architectural con-
buckles are formed. The predicted shape is the same as the experi- siderations. HSCs reduce the use of construction materials, which in
mentally observed shape [13]. It demonstrates the steel tube is failed by turn reduce water, energy and manpower in handling such materials
outwards local buckling which causes the separation between the steel [38]. The valuable work by Liew et al. [17] shows that high strength
tube and concrete core. As can be seen, the failure mode of the outer CFST columns were utilised for the construction of many high-rise
tube involved outward local buckling, involving separation of the tube buildings around the world.
from the concrete core. It should be noted that the current concrete constitutive law is ap-
plicable to HSC since the FE models were validated in this paper by
4. Parametric study using the experimental tests of the octagonal CFST columns [13] that
are filled with the HSC. This is because the concrete used by Ding et al.
The verified FE model, shown above, was used to conduct the [13] with fcu = 57.4 MPa is considered a HSC based on the Egyptian
parametric studies for the octagonal short CFST columns. It total, the design standard for concrete structures [39]. As there is no exact point
programme is composed of seventy two FE models. The lengths of the separating the NSC and HSC, it is approximately the lower bound
columns were assumed three times the cross-sectional depths of the ( fc′ = 49 MPa ) of the HSC according to the EN 1992-1-1 [37].
columns (D ), sufficiently to avoid the overall instability effects and end Within each group, three subgroups are considered with respect to
conditions. Table 2 shows the geometric and material properties em- the cross-sectional width (D ), which was taken as 400, 500 and
ployed in the parametric study, at which it can be noticed that the main 600 mm. Each subgroup consists of six columns with different thick-
studied parameters are the depth (D ) of the cross-sections, the depth-to- nesses (t ). The program is then formed from columns with a wide range
thickness ratios (D / t ) of the steel section, and the concrete compressive of depth-to-thickness (D / t ) ratios which vary between 40 and 200.
strengths ( fc′). As can be seen, the programme consists of four groups According to the current considered geometries, the range of the slen-
(G1, G2, G3 and G4), with the cylindrical strength ( fc′) of the infilled derness ratios (D / t ) of the current FE models is much wider than those
concrete varying between them. The considered fc′ values are 40, 65, 80 tested experimentally by Ding et al. [13]. On the other hand, matching
and 100 MPa, respectively, in groups G1, G2, G3 and G4. According to grades of steel and concrete materials should be used to form the CFST
the EN 1992-1-1 [37], the concrete is defined as HSC when columns [6,17], to avoid the crushing of the infilled concrete before
50 MPa < fc′ ≤ 90 MPa and ultra-high strength concrete (UHSC) when yielding of the steel tubular section. According to Liew et al. [17], steel
fc′ > 90 MPa . Therefore, group G1 is filled by NSC, groups G2 and G3 tubular sections having yield stresses up to 550 MPa may be used with
are filled with HSC and the UHSC is used in group G4. The emphasis on concrete strengths up to 190 MPa. Accordingly, the yield strength of the

1200 1200

1000 1000

800 800
Axial load (kN)

Axial load (kN)

600 600
Experimen
nt (2HN) Experiment (3HN)
400 400

Finite elem
ment model Finite element
e modeel
200 200

0 0
0 0.01 0.02 0.0
03 0.04 0.05 0 0.01 0.02 0.03 0.04
4 0.05
Axial straiin Axial strain
(aa) Specimen 2H
HN (b) Specim
men 3HN

1400 900
800
1200
700
1000
600
Axial load (kN)
Axial load (kN)

800 500

600 400
Experiment (4HN) Experiment (2MN)
300
400
Finite element model 200 Finite element model
200
100
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Axial strain Axial strain
(c) Specimen 4HN (d) Specimen 2MN

Fig. 6. Comparison of computational and tested [8] axial load versus strain responses.

472
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

1200 1400

1000 1200

1000
800
Axial load (kN)

Axial load (kN)


800
600
600
Experiment (3MN) Experiment (4MN)
400
400
Finite element model Finite element model
200
200

0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Axial strain Axial strain
(e) Specimen 3MN (f) Specimen 4MN
1000 1200
900
1000
800
700
Axial load (kN)

800

Axial load (kN)


600
500 600
400 Experiment (3LN) Experiment (4LN)
300 400

200 Finite element model Finite element model


200
100
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Axial strain Axial strain
(g) Specimen 3LN (h) Specimen 4LN

Fig. 6. (continued)

current steel tubes was constant at 350 MPa. The tensile strength of the ξ ⎞
Pul, Yu = ⎜⎛1 + 0.5kn ⎟ (f As + f
ck A c )
steel tube was taken as 430 MPa. The steel grades which would rea-
⎝ 1 + ξ⎠ y (15)
sonably be expected in Australian Construction, AS4100-1998 [40],
were selected. where ξ is the confinement factor, given as ξ = fs As / fck Ac , fck (calcu-
Table 2 also provides the FE ultimate strengths (Pul . FE ) of the octa- lated as 0.67fcu ) represents the characteristic strength value for the
gonal CFST short columns. As can be seen, the Pul . FE value increases concrete, while fcu stands for the concrete cube compressive strength.
with increasing the size of the octagonal steel section and with the use kn = 0.71 which is calculated from the expression
of HSC. Hence, by using the HSC, the weight of the column becomes kn = (n2 − 4)/(n2 + 20) , where n denotes the number of the sides in a
less, which is useful in earthquake design because the seismic response polygon (i.e. eight in case of octagon).
reduces when the dead load is reduced [41].

5.2. Design strength model by Ding et al. [13]


5. Design strength models and comparisons
According to Ding et al. [13], the strengths (Pul, Ding ) of the octagonal
To the authors' knowledge, there are no design strengths existing in CFST short columns are given as:
international design specifications for the octagonal CFST short col-
umns. Accordingly, in this section, the ultimate FE strengths (Pul, FE ) of Pul, Ding = fc′ Ac + 1.5f y As (16)
these columns are verified with the results of the design strengths
suggested by Yu et al. [10], Ding et al. [13] and Patel et al. [14]. It where fc′ is the unconfined concrete compressive cylinder strength.
should be noted that these design models were suggested based on
limited experimental variables, as shown in the Introduction.
5.3. Design strength model by Patel et al. [14]

The strengths (Pul, Patel ) of the octagonal CFST short columns, ac-
5.1. Design strength model by Yu et al. [10]
cording to Patel et al. [14], are suggested as:
Yu et al. [10] provided a unified strength of circular and polygonal Pul, Patel = fcc′ Ac + f y As (17)
CFST short columns subjected to axial loading. According to Yu et al.
[10], the axial load bearing strengths (Pul, Yu ) of the octagonal CFST
where fcc′ is the confined compressive strength of the concrete, given in
short columns are derived as:
Eq. (4).

473
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

5.4. Comparison with FE strengths Ding et al. [13] has the same philosophy used by the Japanese AIJ code
[43], which reflects the superior performance of the circular CFST by
It can be recognised by reviewing the design equations of Refs. enhancing the steel contribution in the resistance ( f y As ) by applying a
[10,13,14] that each one uses a different methodology to increase the factor equals to 1.27.
capacity of the columns due to the confinement of the infilled concrete The comparison between the experimental and FE strengths with
produced by the octagonal tubes. Yu et al. [10] amplifies the squash the above three design models is given in Fig. 9, in which the vertical
(
load of the column (f y As + fck Ac ) by using the term 1 + 0.5kn 1 + ξ . On
ξ
) axis represents the values of the design strengths [10,13,14] and the
horizontal gives the experimental (Pul, Exp ) or FE (Pul, FE ) strengths.
the opposite, Ding et al. [13] and Patel et al. [14] suggest to increase
Fig. 9(a) shows the experimental comparisons, while Fig. 9(b) illus-
the contribution of the steel section and the infilled concrete, respec-
trates the comparisons with the FE results. The dashed lines in Fig. 9
tively. In the former [13], the strength of the steel section is increased
represent the points where the design model and the ultimate strength
by 50%, while the latter [14] uses the confined concrete strength ( fcc′ )
given by Eq. (4). However, it is a common fact that the lateral expan- (Pul, Exp or Pul, FE ) are equal. The data points locating above the equality
sion of infilled concrete gradually becomes greater than that of steel in line represent design strengths that are unsafe compared with their
the inelastic loading range [41]. Hence, a radial pressure develops at ultimate strengths and vice-versa. As can be seen, Yu et al. [10] and
the interface between the inner steel surface and the concrete. This Patel et al. [14] conservatively predict the experimental (Pul, Exp ) and FE
pressure restrains the concrete core and develops a hoop tension in the (Pul, FE ) strengths of the octagonal CFST short columns, with the latter
steel tube. The concrete core, at this stage, is stressed tri-axially and the providing better estimates with an average value of about 0.87 of the
steel tube becomes under bi-axial stress state. So, the steel tube cannot (Pul, FE ) strengths compared with 0.82 for the former. Table 3 depicts the
resist the longitudinal yield stress with the presence of this hoop tensile average and standard deviation of relative design models, from which it
stresses. This well-known fact makes the proposed strength by Ding can be noticed that these available design models can safely be ex-
et al. [13] questionable, for the case of stocky octagonal cross-sections, tended to large diameter octagonal CFST short columns with D / t ratios
because the strength of the steel section is increased considerably be- up to 200 and filled with concrete with strengths as high as 100 MPa.
yond the ultimate strength of the used steel material. Conversely, oc- Additionally, it can be noticed that the results show that the design
tagonal cross-sections with large D / t ratios (characterised by a lower model by Ding et al. [13] is the most suitable on average. However, if
limit of 100, as shown latter in the paper) may benefit from the increase the effect of the D / t ratios is considered, an in-depth comparison could
in the steel contribution as a result of the delaying effect of the concrete be made.
core to the local buckling [42], while the increase in the concrete Table 4 is introduced to provide the relative design strengths of the
strength becomes very low. It is worth pointing out that the strength of octagonal CFST short columns for groups G1 and G3 as sample results.
This is because the results of the other two groups (G2 and G4) are

10000 10000
9000 9000
8000 8000
7000 7000
Axial load (kN)

Axial load (kN)

6000 6000
5000 5000
4000 4000
Experiment (OST1-B)
3000 Experiment (OST1-A) 3000
2000 2000
Finite element model Finite element model
1000 1000
0 0
0 0.005 0.01 0.015 0.02 0 0.01 0.02 0.03 0.04
Axial strain Axial strain
(a) Specimen OST-1A (b) Specimen OST1-B
12000 12000

10000 10000

8000
Axial load (kN)

8000
Axial load (kN)

6000 6000

4000 Expereiment (OST2-A) 4000 Expereiment (OST2-B)

2000 Finite element model Finite element model


2000

0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04 0.05
Axial strain Axial strain
(c) Specimen OST2-A (d) Specimen OST2-B

Fig. 7. Comparison of computational and tested [13] axial load versus strain responses.

474
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

14000 14000

12000 12000

10000 10000

Axial load (kN)


Axial load (kN) 8000 8000

6000 6000

Experiment (OST3-A) 4000 Experiment (OST3-B)


4000
Finite element model Finite element model
2000 2000

0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
Axial strain Axial strain
(e) Specimen OST3-A (f) Specimen OST3-B

14000 14000
1

12000 12000
1

10000 10000
1
Axial load (kN)

8000 Axial load (kN) 8000

6000 6000

4000 Experiment (OST4-A) Experim


ment (OST4-B
B)
4000
Finite elemeent model Finite element
e model
2000 2000

0 0
0 0.0
01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
Axial straain Axial strain
(g) Specimen OST
T4-A (h) Specimen
n OST4-B

Fig. 7. (continued)

qualitatively similar. As can be seen, the D / t ratio affects the accuracy Additional investigation is, therefore, needed for concrete strengths
of the design model given by Ding et al. [13]; i.e. the design model is higher than fc′ = 100MPa as those investigated previously by Liew et al.
unsafe for columns with relatively small D / t ratios (with similar con- [17].
clusion found from the tested program), while it becomes suitable for
columns constructed with steel tubes with large ratios. A limiting D / t ⎧1.1 × (fcc′ Ac + f y As ) for D / t ≤ 100
Pul,Prop =
ratio delineating between both behaviours is around 100. Hence, it is ⎨ fc′ Ac + 1.5f y As for D / t > 100 (18)

currently suggested to use the Ding et al. [13] only for columns with
D / t > 100 (given in grey highlight in the table). This confirms the
previous argument by the authors that this design model is suitable for 5.5. Additional verification of the current design model through
slender cross-sections, where the increase in the confined concrete comparisons with international design specifications
strength becomes very low while the infill concrete delays the local
buckling of the steel tubes [42]. On the other hand, for columns with 5.5.1. Eurocode 4 [44]
D / t ≤ 100 , the design model by Patel et al. [14] becomes better than Eurocode 4 [44] accounts for the concrete confinement for the
that given by Yu et al. [10], though it requires additional enhancement. prediction of ultimate capacity of circular CFST short columns sub-
Currently, it is suggested to increase the strength of Patel et al. [14] by jected to axial compression. The design equation given in Eurocode 4
10%, based on the best fit with the predicted strengths. The final pro- [44] is expressed by:
posed design strength (Pul,Prop ) is given in Eq. (18) and is graphically
t fy ⎞
represented in Fig. 10. When compared with Fig. 9, the current pro- Pul . EC 4 = ηa As f y + Ac fc ⎛⎜1 + ηc ⎟
⎝ D fc ⎠ (19)
posal seems to be the best design strength compared with the others. On
the other hand, an important issue that worth clarification is the con- in which
tinuity of the proposed design equation at the limit D / t = 100 . This is
checked herein in Fig. 11 which provides the relationship between the ηa = 0.25(3 + 2λ ) ≤ 1.0 (20)
proposed to FE strength ratios against the D / t ratios for the octagonal 2
CFST short columns. It can be seen that the current proposal provides ηc = 4.9 − 18.5λ + 17λ ≤ 1.0 (21)
continuous design predictions before and after the limit of D / t = 100 .
NPl . Rk
This increases the confidence of this proposal. However, this suggestion λ =
Ncr (22)
is limited for the large diameter octagonal CFST short columns within
the current studied upper limits of D / t = 200 and fc′ = 100MPa . NPl . Rk = As f y + Ac fc′ (23)

475
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

Pul . DBJ = (1.14 + 1.02ξ ) fc′ At (26)

As f y
ξ=
Ac fc′ (27)

in which At denotes the total cross-section area of octagonal CFST


columns and ξ is the confinement factor.

5.5.3. ACI-318-11 [46]


The design procedures reported in ACI-318-11 [46] do not consider
the concrete confinement provided by the steel tube in circular CFST
columns. The design equation is expressed as:
Pul . ACI = As f y + 0.85Ac fc (28)

The ultimate capacity of octagonal CFST columns obtained from the


design standards are compared with the large diameter experimental
data [13] in Table 5. As shown in Table 5, the DBJ/T13-51–2010 [45]
predictions agree with the experimental results, but for those filled with
concrete having fc′ = 49MPa . The code, ACI-318-11 [46], under-
estimates the ultimate capacity of octagonal CFST columns. This is due
to the fact that the code ACI-318-11 does not account for the increased
strength of concrete from the confinement effect in CFST columns. In
addition, Eurocode 4 [44] provides good estimates for the ultimate
capacity of octagonal CFST columns. This is because Eurocode 4 [44]
represents the equations for the design of circular CFST columns and
the confinement of circular composite section is considered. In Table 5,
the mean ratio for Eurocode 4-to-experiment [44], DBJ/T13-51–2010-
to-test [45] and ACI-318-11-to-experiment [46] is 0.96. 0.95 and 0.77,
respectively. For Eurocode 4 [44], the SD is 0.02, which is much better
than the others. On the other hand, the proposed design strength
Fig. 8. Deformed shape for Specimen OST3-A. (Pul,Prop ), given in Eq. (18), provides a mean ratio-to-experiments of 0.96
with a SD of 0.02. This means that this design model, besides that of
Eurocode 4 [44], is the best. However, the proposed design strength
π 2 (EI )eff
Ncr = (Pul,Prop ) is based on the confinement mechanism of the octagonal col-
L2 (24)
umns rather than the circular as that of the Eurocode 4 [44]. Hence, it is
currently recommended to use currently proposed design strength
(EI )eff = Es Is + 0.6Ecm Ic (25) (Pul,Prop ) in future estimations of the octagonal CFST short columns.

6. Structural behaviour
5.5.2. DBJ/T13-51-2010 [45]
The design equation given by DBJ/T13-51-2010 [45] also accounts This section presents the individual influence of the D / t ratios and
for the concrete confinement effects for the design of circular CFST concrete cylinder compressive strengths ( fc′) on the structural perfor-
columns. The equation reported in DBJ/T13-51-2010 [45] is given as: mance of the large diameter octagonal CFST short columns. Then, the

Table 2
Details and ultimate strengths of the FE models of octagonal CFST short columns.

Cross-sectional dimensions G1: fc′ = 40 MPa G2: fc′ = 65 MPa G3: fc′ = 80 MPa G4: fc′ = 100 MPa

D (mm) t (mm) D No. Pul . FE (kN) No. Pul . FE (kN) No. Pul . FE (kN) No. Pul . FE (kN)
t

400 2 200 C1 6848 C19 10,186 C37 11,929 C55 14,632


3 133 C2 7291 C20 10,577 C38 12,303 C56 14,946
4 100 C3 7720 C21 10,883 C39 12,680 C57 15,409
6 67 C4 8551 C22 11,698 C40 13,605 C58 16,093
8 50 C5 9354 C23 12,418 C41 14,298 C59 16,750
10 40 C6 11,020 C24 14,041 C42 15,854 C60 18,248
500 3 167 C7 11,038 C25 16,120 C43 19,179 C61 23,115
4 125 C8 11,585 C26 16,608 C44 19,655 C62 23,626
5 100 C9 12,119 C27 17,090 C45 20,120 C63 24,086
6 83 C10 12,640 C28 17,566 C46 20,578 C64 24,522
8 63 C11 13,649 C29 18,491 C47 21,472 C65 25,365
10 50 C12 14,624 C30 19,097 C48 22,334 C66 26,177
600 3 200 C13 15,502 C31 22,867 C49 27,205 C67 30,706
4 150 C14 16,169 C32 23,422 C50 27,831 C68 33,397
5 120 C15 16,821 C33 24,008 C51 28,395 C69 34,070
6 100 C16 17,460 C34 24,588 C52 28,956 C70 34,670
8 75 C17 18,698 C35 25,727 C53 30,055 C71 35,714
10 60 C18 19,897 C36 26,830 C54 31,120 C72 36,722

476
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

Fig. 9. Comparison of design and ultimate axial


loads for octagonal CFST short columns.

Table 3 The strain is currently computed as the end shortening of the column
Mean and standard deviation of the different design models. over its length. Overall, the axial load-strain curves show a typical re-
sponse with five different stages. Just as an example, five characteristic
Indices Pul, Yu Pul, Ding Pul, Patel Pul, Yu Pul, Ding Pul, Patel
Pul, Exp Pul, Exp Pul, Exp Pul, FE Pul, FE Pul, FE points are also highlighted by solid circles on the curve of C19 in
Fig. 12(a), while Fig. 13 shows the distributions of longitudinal axial
Mean 0.82 0.98 0.86 0.82 1.02 0.87 stress of the upper surface of the concrete core at these points. At the
Standard deviation 0.022 0.024 0.016 0.049 0.035 0.035 first stage (representing the elastic stage) from Point O to point A, the
steel tube and the concrete core support the axial load independently
because the steel tube has no restraining effect on the concrete core.
combined effect of both the D / t ratio and the value of fc′ is examined This is attributed to the Poisson's ratio of the concrete, which is in the
aiming at reaching an optimised cross-section for the thin-walled oc- range of 0.15–0.2 [21], is lower than that of the steel in this stage. The
tagonal steel tubes. In this section, the ductility index (DI) is calculated stress distribution of the concrete core at Point A, which is the limit of
as the vertical strain when the load falls to 85% of its ultimate load (in the elastic stage, is shown in Fig. 13(a), from which it can be seen that
the descending branch) over that strain corresponding to the ultimate the concrete is subjected to the same longitudinal stress. The column
load of the column [38]. begins to come into inelastic stage, as the longitudinal strain increases,
just after Point A, which extends to Point B where the yielding of the
6.1. Effects of depth-to-thickness (D / t ) ratio steel tube occurs. In this stage the curve is characterised by a low
stiffness compared with the elastic stage. With the increasing of the
The effect of the depth-to-thickness (D / t ) ratios on the ultimate applied load in this stage, the cracks initiate in the concrete and,
strength (Pul, FE ) and the load-strain curves of the axially-loaded octa- therefore, its volume begins to increase. The confinement of the con-
gonal CFST short columns is given herein. The external diameters of the crete core, provided by the steel tubular member, enhances due to the
steel sections (D ) and the concrete compressive strengths ( fc′) were kept increase in the transversal deformation of concrete as the Poisson's ratio
constant. The axial load versus strain responses of columns C19-C36 of concrete increases to 0.5 in the inelastic range [21]. However, the
(Table 2) with fc′ = 65 MPa , as sample results, are illustrated in Fig. 12. results, shown in Fig. 13(b), indicate that the distribution of the

Table 4
Comparison of relative design strengths of the FE models of octagonal CFST short columns for groups G1 and G3.

Cross-sectional dimensions G1: fc′ = 40 MPa G3: fc′ = 80 MPa

D (mm) t (mm) D No. Pul, Yu Pul, Ding Pul, Patel No. Pul, Yu Pul, Ding Pul, Patel
t Pul, FE Pul, FE Pul, FE Pul, FE Pul, FE Pul, FE

400 2 200 C1 0.79 0.96 0.78 C37 0.76 0.99 0.82


3 133 C2 0.82 0.99 0.90 C38 0.78 1.00 0.89
4 100 C3 0.85 1.02 0.90 C39 0.80 1.02 0.89
6 67 C4 0.90 1.06 0.91 C40 0.82 1.04 0.88
8 50 C5 0.95 1.11 0.92 C41 0.86 1.07 0.89
10 40 C6 0.90 1.05 0.92 C42 0.84 1.03 0.90
500 3 167 C7 0.80 0.97 0.78 C43 0.76 0.98 0.81
4 125 C8 0.82 0.99 0.89 C44 0.77 0.99 0.87
5 100 C9 0.85 1.01 0.90 C45 0.79 1.00 0.88
6 83 C10 0.87 1.03 0.90 C46 0.80 1.02 0.88
8 63 C11 0.91 1.07 0.91 C47 0.83 1.04 0.88
10 50 C12 0.95 1.10 0.91 C48 0.86 1.07 0.89
600 3 200 C13 0.79 0.96 0.78 C49 0.75 0.97 0.81
4 150 C14 0.81 0.97 0.89 C50 0.76 0.98 0.87
5 120 C15 0.83 0.99 0.89 C51 0.78 0.99 0.87
6 100 C16 0.85 1.01 0.90 C52 0.79 1.01 0.88
8 75 C17 0.88 1.05 0.90 C53 0.81 1.03 0.88
10 60 C18 0.92 1.08 0.91 C54 0.84 1.05 0.89

477
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

Fig. 10. Comparison of proposed design and ulti-


mate loads for octagonal CFST short columns.

steel tube is much larger than that of concrete in the centre. It can
additionally be seen that the tensile hoop stresses developed in the steel
tube becomes obvious in this stage. The fifth stage represents the stage
where the load stabilises from Points D to Point E, at which the long-
itudinal strain attains the value of 0.04. The strain softening behaviour
shown in Fig. 13(e) illustrates that the longitudinal stress distribution of
concrete core at Point E is the same as Point D. However, the long-
itudinal stress in the concrete core decreases with increasing the axial
strain from Point B to D. Obviously, it can be seen that increasing the
D / t ratio, for the columns with different diameters in Fig. 12(a), (b) and
(c), significantly reduces both the ultimate axial strength and stiffness
of the octagonal CFST short columns.
On the other hand, the Von-Mises stress distribution of the steel tube
is also shown in Fig. 13, from which it can be realised that the stress in
the tubes increases, as a result of the bi-axial stress state taking place in
Fig. 11. Proposed to FE strength ratios against the D/ t ratios for octagonal CFST short the CFST columns [41], at the corner areas of the octagonal cross-sec-
columns.
tion until the load reaches the column's strength (Point C). These corner
areas represent the parts of the active octagonal cross-section in con-
longitudinal stress of concrete still uniform over the cross-section. This fining the concrete core which was approximated by 0.17 of the side
stage is followed by the plastic stage which extends until the ultimate length of the octagon (a as shown in Fig. 1) by Ding et al. [13]. Ac-
load is reached, at which the confinement provided by the octagonal cordingly, this approximation remains valid for the case of large dia-
steel tube increases considerably allowing the longitudinal stress in the meter octagonal CFST columns filled with HSCs.
concrete to exceed the value of fc′. However, at the ultimate load, the
longitudinal stress of the concrete is distributed unevenly, with the
stress becomes greater near to the steel tube; see Fig. 13(c). A des- 6.2. Effects of concrete compressive strength
cending branch follows the ultimate load with a significant residual
strength at which the load stabilises after the Point D near a strain of The effect of the concrete compressive strengths ( fc′) on the axial
about 0.01. In this fourth stage (between Points C and D), the local compressive behaviour of the octagonal CFST short columns was ad-
buckling of steel tube considerably reduces the concrete confinement ditionally investigated. As shown previously in Table 2, four compres-
provided by the encased steel tube after steel yielding. The longitudinal sive concrete strengths were considered varying between 40 and
stress distribution in the concrete core at Point D is shown in Fig. 13(d). 100 MPa. The concrete compressive strength was selected based on the
It can be seen from the figure that the stress of concrete core near the guideline given in AS 3600-2009 (Clause 3.1.1.1) [47] for Australian
construction. Fig. 14 shows the comparison of the axial load-strain

Table 5
Comparison between experimental ultimate strengths [13] of octagonal CFST columns and strengths obtained from different international design codes.

Specimens D fc′ (MPa) Pul, Exp (kN) Eurocode 4 [44] DBJ/T13-51-2010 [45] ACI-318 [46] Current design model
t

Pul, EC 4 (kN) Pul, EC 4 Pul, DBJ (kN) Pul, DBJ Pul, ACI (kN) Pul, ACI Pul,Prop (kN) Pul,Prop
Pul .exp Pul .exp Pul .exp Pul .exp

OST1-A 126 32.2 9297 8836 0.95 8688 0.93 7075 0.76 8941 0.96
OST1-B 121 32.2 9311 8759 0.94 8595 0.92 7005 0.75 8873 0.95
OST2-A 80 32.2 10,502 10,328 0.98 9832 0.94 8084 0.77 10,048 0.96
OST2-B 81 32.2 10,713 9985 0.93 9527 0.89 7832 0.73 9737 0.91
OST3-A 123 49.0 12,362 11,847 0.96 12,169 0.98 9726 0.79 12,068 0.98
OST3-B 119 49.0 12,357 11,771 0.95 12,081 0.98 9658 0.78 12,000 0.97
OST4-A 81 49.0 12,992 12,826 0.99 12,930 1.00 10,377 0.80 12,537 0.96
OST4-B 80 49.0 13,263 12,987 0.98 13,079 0.99 10,497 0.79 12,678 0.96
Mean 0.96 0.95 0.77 0.96
Standard deviation (SD) 0.02 0.04 0.02 0.02

478
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

16000 Fig. 12. Axial load-strain curves for typical octa-


20000
C19 C20 C25 C26 gonal CFST short columns.
14000 C21 C22 C27 C28
C23 C24 16000
12000 C29 C30
Axial load (kN)

Axial load (kN)


10000
C 12000
8000 B
6000 8000
A
4000 D
E 4000
2000
O
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Axial strain Axial strain
(a) C19-C24 (b) C25-C30

30000
C31 C32
25000 C33 C34
C35 C36
Axial load (kN)

20000

15000

10000

5000

0
0 0.01 0.02 0.03 0.04 0.05
Axial strain
(c) C31-C36

curves for the columns with fc′ = 40 MPa (C1 and C6)) and 80 MPa (C37 which bears 27205 kN). This is however examined in detail in the fol-
and C42). The diameter of the columns C1 and C6 is 400 mm, while that lowing sub-section.
of C37 and C42 is 500 mm. Note that the results of the other concrete
compressive strengths are similar, so they were not included for the 6.3. Combined effects of D / t ratio and fc′ on the thin-walled octagonal tubes
clearness of the presentation. The curves also illustrate the columns
with the extreme considered D / t ratios; 200 and 40. It can be seen that In the above two sub-sections, the individual effect of D / t ratio and
increasing the concrete compressive strength significantly increases the fc′ was discussed. On the other hand, combined effects of both the D / t
ultimate axial strength of octagonal CFST short columns, while the ratio and fc′ are investigated with the aim of optimising the cross-sec-
initial stiffness increases slightly. It can as well be recognised that in- tion of the thin-walled octagonal steel tubes. This would affect the cost
creasing the value of fc′ for columns with different D / t ratios, decreases of the octagonal CFST short columns by providing more economical
the relative residual strength characterised by the portion DE; refer cross-section. Currently, the additional axial compression carried by the
back to Fig. 12(a). On the other hand, the ductility indices (DI) of the columns compared with the increased amount of steel is evaluated. The
columns were computed to further understand the effect of using HSC percentage of variation in the cross-sectional area of the steel tube,
on the ductility of the columns. They were 1.58 and 1.25 for C6 and concrete core and ultimate load (Pul, FE ) between two columns (Ci and Cj )
C42, respectively, while they were 2.69 and 1.26 for columns C1 and were calculated using the following expressions:
C37, respectively. Hence, a decrease of about 21% and 52% in the
ductility results from increasing the fc′ value from 40 MPa to 80 MPa . As, i − As, j
as, ij = for the variation in the cross
This confirms the widely known fact that increasing the value of fc′ As, i
reduces the ductility of different composite structures [42,48,49]. −sectional area of the steel tube (29)
Fig. 15 shows the ultimate axial load of octagonal CFST columns
with an outer diameter of 600 mm with different fc′ values, from which A c, i − A c, j
ac, ij = for the variation in the cross
four D / t ratios were only provided for the clearness of the presentation. A c, i
As can be seen, increasing the value of fc′ raises linearly the ultimate (30)
− sectional area of the concrete
axial strengths of the columns. On the other hand, this figure shed the
light into the applicability of reducing the octagonal steel cross-sec- Pul, FE , i − Pul, FE , j
tional area, while the ultimate load of the column increases through aP, ij = for the variation in the ultimate strength of
Pul, FE , i
using higher concrete strength. For example, the column C36
the column (31)
(t = 10 mm and fc′ = 65 MPa ) bears an ultimate load of 26830 kN,
while by using fc′ = 80 MPa higher strengths may be obtained even if Just for the sake of briefness, the values of as, ij , ac, ij and aP, ij were
small tube thicknesses are used (an example is C49 with t = 3 mm calculated for columns C1-C6 of group G1 ( fc′ = 40 MPa ) and C55-C60

479
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

Fig. 13. Longitudinal stress of concrete core cross-


section (with steel tube remains in place) at different
loading levels for Column C19 - (Units in MPa).

of group G4 ( fc′ = 100 MPa ). The values of these coefficients are pre- type of the concrete (i.e. NSC or HSC), it can be observed that the in-
sented in Table 6, noting that C1 and C55 are used as Cj for sub-groups crease in the ultimate load (by decreasing the D / t ratio) decreases when
C1-C6 and C55-C60, respectively. From this table, it can be realised that the HSC is used compared with the increase associated by using the
decreasing the D / t ratio of the steel tube (by increasing the thickness t ) NSC. As an example, the coefficient as, ij is 0.80 for both columns C6
increases both the cross-sectional area of the steel tube and the ultimate ( fc′ = 40MPa ) and C60 ( fc′ = 100 MPa ), while the coefficient aP, ij is 0.38
load of the column, while the cross-sectional area of the concrete core and 0.20 for columns C6 and C60, respectively. This indicates that using
remains almost very similar. However, the increase of the cross-sec- the HSC becomes economical by increasing the D / t ratio of the steel
tional area of the steel tube is much higher than the increase in the tube (i.e. decreasing the cross-sectional area of the steel tube). By doing
ultimate load of the column for both NSC and HSC. Considering the so, the total cost of the column reduces.

Fig. 14. Axial load-strain curves for typical octa-


gonal CFST short columns for different concrete
compressive strengths.

480
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

40000 25000
Octagonal hollow column
35000
20000 Octagonal CFST column
30000

Axial load (kN)


Axial load (kN)

25000 15000
20000
D/t = 150 10000
15000
D/t = 100
10000 5000
D/t = 75
5000 D/t = 60
0 0
0 30 60 90 120 0 0.01 0.02 0.03 0.04 0.05
Axial strain
Concrete compressive strength (MPa)
Fig. 16. Comparison of octagonal hollow steel tubular columns and CFST columns.
Fig. 15. Effects of concrete compressive strengths on the ultimate axial strengths of oc-
tagonal CFST columns for D = 600mm .
large diameter octagonal CFST columns filled with HSC.
6.4. Comparison between octagonal hollow tubes and CFST columns 2. The design models by Yu et al. [10] and Patel et al. [14] were found
to conservatively predict the experimental (Pul, Exp ) and FE (Pul, FE )
The FE analysis is conducted to compare the axial load-strain curves strengths of the octagonal CFST short columns, with the latter pro-
of octagonal steel tubular column and CFST column. Column C30 given viding better estimates with an average value of about 0.87 of the
in Table 2 is selected for this purpose. All properties are kept constant (Pul, FE ) strengths compared with 0.82 for the former.
including the mesh of steel tube in the analysis of hollow column and 3. The results showed that the design model by Ding et al. [13] is the
CFST column. The steel tubular column is modelled without the con- most suitable on average, despite giving many unsafe results for
crete core. Fig. 16 illustrates the comparison of axial load-strain curves columns characterised by relatively small D / t ratios.
for octagonal steel tubular column and CFST column. As shown in 4. A new design model, given herein by Eq. (18), was suggested. In this
Fig. 16, the stiffness of CFST column becomes much higher than that of model, it is proposed to use the expression of Ding et al. [13] for
steel tubular column because of the added concrete. Additionally, the columns with a diameter-to-thickness ratios (D / t ) greater than 100 .
concrete fill clearly increases the ultimate strength of octagonal CFST On the other hand, for columns with stockier tubes of D / t ≤ 100 , a
column as demonstrated in Fig. 16. This is attributed to the strength modified version of the design model by Patel et al. [14] was sug-
carried by concrete used to fill the void in the steel tube besides the gested.
confinement effect of the steel tube on the concrete core. Hence, the 5. Despite that design strength obtained from the Eurocode 4 [44]
current comparison shows the importance of using octagonal CFST obtained good results compared with the experimental large dia-
columns instead of thin-walled tubes in terms of the columns' strength meter octagonal columns, the proposed design strength Eq. (18) is
and stiffness. still recommended because it is based on the confinement me-
chanism of the octagonal columns rather than the circular used in
7. Conclusions the Eurocode 4 [44].
6. As confirmed by the current results, the widely known fact that
The current paper provides an extensive FE investigation on the increasing the value of fc′ reduces the ductility of the composite
large diameter octagonal CFST short columns, to address the lack of columns remains for the case of octagonal CFST short columns.
their behaviour in literature. The paper mainly investigates the effect of 7. The results indicated that using HSC becomes economical by in-
the HSC and the D / t ratios of the steel tubes on the behaviour of such creasing the D / t ratio of the octagonal steel tube (i.e. decreasing the
large diameter columns. Based on this investigation, the following cross-sectional area of the steel tube). By doing so, the total cost of
points could be summarised: the column is reduced.

1. The verification of the FE modelling showed that the confined The FE model presented in this paper provides the benchmark nu-
concrete model suggested by Patel et al. [14] may be extended to merical results for the design of octagonal CFST columns. Further

Table 6
Variation of cross-sectional area of the steel section, infilled concrete and ultimate load (Pul, FE ) .

Ci Cj fc′ (MPa) D As (mm2) Ac (mm2) Pul . FE (kN) as, ij ac, ij aP, ij


t

C1 C1 40 200 2638 129,911 6848 0.00 0.00 0.00


C2 133 3947 128,602 7291 0.33 −0.01 0.06
C3 100 5249 127,299 7720 0.50 −0.02 0.11
C4 67 7834 124,715 8551 0.66 −0.04 0.20
C5 50 10,392 122,157 9354 0.75 −0.06 0.27
C6 40 12,923 119,625 11,020 0.80 −0.09 0.38
C55 C55 100 200 2638 129,911 14,632 0.00 0.00 0.00
C56 133 3947 128,602 14,946 0.33 −0.01 0.02
C57 100 5249 127,299 15,409 0.50 −0.02 0.05
C58 67 7834 124,715 16,093 0.66 −0.04 0.09
C59 50 10,392 122,157 16,750 0.75 −0.06 0.13
C60 40 12,923 119,625 18,248 0.80 −0.09 0.20

481
M.F. Hassanein et al. Thin-Walled Structures 123 (2018) 467–482

experimental study on octagonal CFST columns with high strength [22] M.H. Mollazadeh, Load Introduction into Concrete-filled Steel Tubular Columns
concrete and steel is needed. (Ph.D. thesis), University of Manchester, 2015.
[23] M.F. Hassanein, O.F. Kharoob, L. Gardner, Behaviour and design of square concrete-
filled double skin tubular columns with inner circular tubes, Eng. Struct. 100 (2015)
Acknowledgements 410–424.
[24] T. Yu, J.G. Teng, Y.L. Wong, S.L. Dong, Finite element modeling of confined con-
crete-I: Drucker-Prager type plasticity model, Eng. Struct. 32 (2010) 665–679.
This research study is supported by both Tanta University and the [25] A. Mirmiran, K. Zagers, W. Yuan, Nonlinear finite element modelling of concrete
School of Engineering and Mathematical Sciences at La Trobe confined by fiber composites, Finite Elem. Anal. Des. 35 (2000) 79–96.
University with the start-up research fund. The financial support is [26] H.T. Hu, C.S. Huang, M.H. Wu, Y.M. Wu, Nonlinear analysis of axially loaded
concrete-filled tube columns with confinement effect, J. Struct. Eng., 129 (10)
gratefully acknowledged. (2003) 1322–1329.
[27] M.H. Wu, Numerical Analysis of Concrete Filled Steel Tubes Subjected to Axial
References Force (MS thesis), Dept. of Civil Engineering, National Cheng Kung Univ., Tainan,
Taiwan, R.O.C, 2000.
[28] J.B. Mander, M.N.J. Priestly, R. Park, Theoretical stress-strain model for confined
[1] M. Tomii, K. Sakino, A state of art report on concrete-filled steel tube structures, concrete, J. Struct. Eng. 114 (8) (1988) 1804–1826.
Part 1. elasto-plastic behavior of members, Concr. J. 2 (1974) 26–40. [29] ACI Committee 363, State-of-the-art report on high-strength concrete, J. Am. Concr.
[2] T. Kitada, Ultimate strength and ductility of state-of-the-art concrete-filled steel Inst. (1984).
bridge piers in Japan, Eng. Struct. 20 (1998) 347–354. [30] Q.Q. Liang, S. Fragomeni, Nonlinear analysis of circular concrete-filled steel tubular
[3] K. Roseline, D. Tensing, State of the art report on steel-concrete infilled composite short columns under axial loading, J. Constr. Steel Res. 65 (12) (2009) 2186–2196.
column, Int. J. Sci. Eng. Res. 4 (2013) 677–684. [31] F.E. Richart, A. Bradtzaeg, R.L. Brown, A study of the failure of concrete under
[4] P. Crespi, A. Franchi, N. Giordano, M. Scamardo, P. Ronca, Structural analysis of combined compressive stresses, University of Illionis, Engineering Experimental
stone masonry columns of the Basilica S. Maria di Collemaggio, Eng. Struct. 129 Station, Champaign (III), 1928 (Bull. 185).
(2016) 81–90. [32] S. Martinez, A.H. Nilson, F.O. Slate, Spirally reinforced high-strength concrete
[5] R. Chacón, Circular concrete-filled tubular columns: state of the art oriented to the columns, Acids Struct. J. 81 (5) (1984) 431–442.
vulnerability assessment, Open Civil. Eng. J. 9 (2015) 249–259. [33] Y.K. Yong, M.G. Nour, E.G. Nawy, Behavior of laterally confined high-strength
[6] L.-H. Han, L. Wei, R. Bjorhovde, Developments and advanced applications of con- concrete under axial loads, J. Struct. Eng. 114 (2) (1988) 332–351.
crete-filled steel tubular (CFST) structures: members, J. Constr. Steel Res. 100 [34] M.P. Collins, D. Mitchell, J.G. MacGregor, Structural design considerations for high-
(2014) 211–228. strength concrete, Concr. Int. Vo. 15 (5) (1993) 27–34.
[7] J.Y.R. Liew, M.X. Xiong, D.X. Xiong, Design of high strength concrete filled tubular [35] K. Sakino, H. Nakahara, S. Morino, I. Nishiyama, Behavior of centrally loaded
columns for tall buildings, J. High-Rise Build. 3 (2014) 1–7. concrete-filled steel-tube short columns, J. Struct. Eng. 130 (2) (2004) 180–188.
[8] M. Tomii, K. Yoshimura, Y. Morishita, , Experimental studies on concrete filled steel [36] V.I. Patel, Q.Q. Liang, M.N.S. Hadi, Nonlinear Analysis Of Concrete-filled Steel
tubular stub columns under concentric loading, in: International Colloquium on Tubular Columns, Scholar's Press, Germany, 2015.
Stability of Structures under Static and Dynamic Loads: Washington, DC, 1977, pp. [37] Eurocode 2, Design of concrete structures. Part 1-1. General rules and rules for,
718–741. buildings, 2004.
[9] K.A.S. Susantha, H.B. Ge, T. Usami, Uniaxial stress-strain relationship of concrete [38] H. Huang, L.-H. Han, Z. Tao, X.-L. Zhao, Analytical behaviour of concrete-filled
confined by various shaped steel tubes, Eng. Struct. 23 (2001) 1331–1347. double skin steel tubular (CFDST) stub columns, J. Constr. Steel Res. 66 (2010)
[10] M. Yu, X. Zha, J. Ye, Y. Li, A unified formulation for circle and polygon concrete- 542–555.
filled steel tube columns under axial compression, Eng. Struct. 49 (2013) 1–10. [39] ECP 203-2007, Egyptian code of practice for concrete design and construction,
[11] X. Zha, Hollow and Solid Concrete-filled Steel Tube Structures, Science Press, Cairo, Egypt, 2010.
Beijing, 2010 (in Chinese). [40] AS 4100–1998, Steel structures, Sydney, New South Wales, Australia: Standard
[12] Y.C. Zhang, Q.P. Wang, X.Y. Mao, B.Z. Cao, Research on mechanics behavior of stub Australia, 1998.
column of concrete-filled thin-walled steel tube under axial load, J. Build. Struct. 35 [41] N.E. Shanmugam, B. Lakshmi, State of the art report on steel-concrete composite
(1) (2005) 22–27 (in Chinese). columns, J. Constr. Steel Res. 57 (2001) 1041–1080.
[13] F.-X. Ding, Z. Li, S. Cheng, Z.-W. Yu, Composite action of octagonal concrete-filled [42] M. Elchalakani, A. Karrech, M.F. Hassanein, B. Yang, Plastic and yield slenderness
steel tubular stub columns under axial loading, Thin-Walled Struct. 107 (2016) limits for circular concrete filled tubes subjected to static pure bending, Thin-
453–461. Walled Struct. 109 (2016) 50–64.
[14] V.I. Patel, B. Uy, K.A. Prajwal, F. Aslani, Confined concrete model of circular, el- [43] AIJ-SRC, Standards for Structural Calculation of Tubular Steel Concrete Composite
liptical and octagonal CFST short columns, Steel Compos. Struct., Int. J. 22 (3) Structures, 5th ed., Architectural Institute of Japan, 2001 (in Japanese).
(2016) 497–520. [44] Eurocode 4, Design of composite steel and concrete structures, Part 1.1: General
[15] Q.Q. Liang, Performance-based analysis of concrete-filled steel tubular beam-col- rules and rules for building. BS EN 1994-1-1: 2004, British Standards Institution,
umns, Part I: theory and algorithms, J. Constr. Steel Res. 65 (2) (2009) 363–372. London (UK), 2004.
[16] L. Zhu, L. Ma, Y. Bai, S. Li, Q. Song, Y. Wei, L. Zhang, Z. Zhang, X. Sha, Large [45] DBJ/T 13-51-2010, Technical Specification for Concrete-filled Steel Tubular
diameter concrete-filled high strength steel tubular stub columns under compres- Structures, The Department of Housing and Urban-Rural Development of Fujian
sion, Thin-Walled Struct. 108 (2016) 12–19. Province, Fuzhou (China), 2010 (in Chinese).
[17] J.Y.R. Liew, M. Xiong, D. Xiong, Design of concrete filled tubular beam-columns [46] ACI-318-11, Building code requirements for structural concrete and commentary
with high strength steel and concrete, Structures 8 (2) (2016) 213–226. ACI Committee 318. Detroit (MI), 2011.
[18] ABAQUS Standard, User’s Manual The Abaqus Software is a product of Dassault [47] AS3600-2009, Concrete structures. Sydney, New South Wales, Australia: Standard
Systèmes Simulia Corp., Providence, RI, USA Dassault Systèmes, Version 6.8, USA, Australia, 2009.
2008. [48] M.F. Hassanein, O.F. Kharoob, Q.Q. Liang, Behaviour of circular concrete-filled lean
[19] Z. Tao, Z.-B. Wang, Q. Yu, Finite element modelling of concrete-filled steel stub duplex stainless steel tubular short columns, Thin-Walled Struct. 68 (2013)
columns under axial compression, J. Constr. Steel Res. 89 (2013) 121–131. 113–123.
[20] Z. Tao, B. Uy, L.H. Han, Z.B. Wang, Analysis and design of concrete-filled stiffened [49] M.F. Hassanein, M. Elchalakani, V.I. Patel, Overall buckling behaviour of circular
thin-walled steel tubular columns under axial compression, Thin-Walled Struct. 47 concrete-filled dual steel tubular columns with stainless steel external tubes, Thin-
(12) (2009) 1544–1556. Walled Struct. 115 (2017) 336–348.
[21] X. Dai, D. Lam, Numerical modelling of the axial compressive behaviour of short
concrete-filled elliptical steel columns, J. Constr. Steel Res. 66 (7) (2010) 931–942.

482

You might also like