Engineering Structures 56 (2013) 1986–2004

Contents lists available at ScienceDirect

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

Performance of concrete-filled RHS columns exposed to fire on 3 sides

Hua Yang a,b,⇑, Faqi Liu a,b,c, Leroy Gardner c
a
Key Lab of Structures Dynamic Behavior and Control (Harbin Institute of Technology), Ministry of Education, Heilongjiang, Harbin, 150090, PR China
b
School of Civil Engineering, Harbin Institute of Technology, Heilongjiang, Harbin, 150090, PR China
c
Department of Civil and Environmental Engineering, South Kensington Campus, Imperial College London, London SW7 2AZ, UK

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

Article history: The behaviour of concrete-filled steel tubular (CFST) columns subjected to uniform fires has been well
Received 30 January 2013 studied over the past few decades. However, knowledge of the performance of concrete-filled rectangular
Revised 10 August 2013 hollow section (RHS) columns exposed to fire on 3 sides, which is a common scenario in practice, is lim-
Accepted 13 August 2013
ited. Hence, this paper presents an experimental and numerical study into the response of concrete-filled
Available online 21 September 2013
RHS columns subjected to 3-sided fire exposure. Three full-scale concrete-filled RHS columns, two of
which were exposed to fire on 3 sides and the other on 4 sides, were tested to failure. The temperature
Keywords:
distributions, axial displacements, lateral displacements and failure modes were all recorded and dis-
Concrete-filled RHS columns
Finite element analysis
cussed. Following the experiments, a sequentially coupled thermal–stress numerical model was devel-
Fire resistance oped, featuring heat transfer analysis and stress analysis. The FE model was validated against the test
Fire testing results, and used to assist in the understanding of the observed failure mechanisms and to extend the
3-Sided exposure investigated range of key parameters. These parameters included load ratio, load eccentricity, cross-sec-
tional dimensions, slenderness ratio, steel ratio and strengths of the constituent materials. It was found
that the load ratio, cross-sectional dimensions and load eccentricity have a significant influence on the
fire resistance. Based on the results of the parametric studies, a simplified design formula was developed
for predicting the fire resistance of concrete-filled RHS columns subjected to 3-sided exposure. Further-
more, a reduction factor method was proposed for the design of concrete-filled RHS columns exposed to
3-sided fire, based on fire safety design methods for concrete-filled RHS columns in uniform fire, which
may be more convenient for engineering applications.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction et al. [22] studied the fire resistance of steel tubular columns filled
with unreinforced high strength concrete and reinforced high
Concrete-filled steel tubular (CFST) columns are gaining strength concrete with bars or steel fibres. Lu et al. [23] tested fire
increasing usage in practice owing to their excellent structural per- performance of self-consolidating concrete filled double skin steel
formance and ease of construction. Extensive research has been tubular columns. Han et al. [24] studied fire performance of self-
conducted on behaviour of circular [1–5], elliptical [6–8], octagonal consolidating concrete filled stainless steel tubular columns. Ding
[9], square [10–12] and rectangular [13,14] CFST columns at ambi- and Wang [25] and Han et al. [26] carried out tests on fire perfor-
ent temperature. Since fire safety is one of the key aspects of struc- mance of CFST columns in structures. Huo et al. [27] tested impact
tural design, it is essential to develop a full understanding of the behaviour of CFST columns at elevated temperatures. Although fire
fire performance of CFST columns. testing is expensive and time consuming, it remains the most di-
Considerable experimental and numerical research has been rect and reliable method of evaluating the fire resistance of mem-
carried out to investigate the fire performance of CFST columns bers and developing an understanding of their behaviour.
in the past few decades. Hass [15], Lie and Chabot [16,17], Kodur Meanwhile, numerical simulations are also gaining acceptance
and Lie [18] conducted experimental studies of the fire resistance as a means of examining fire behaviour. Lie and Chabot [28], Kodur
of concrete-filled circular and square hollow section columns filled and Lie [29] and Han [30] developed fibre models to predict the fire
with plain concrete, reinforced concrete and steel–fibre-reinforced resistance of CFST columns. Such models can provide accurate re-
concrete. Han et al. [19] examined experimentally the behaviour of sults, though local buckling and the interaction between the steel
concrete-filled rectangular and square hollow section columns ex- tube and core concrete cannot be precisely considered. Ding and
posed to standard fire conditions. Kodur et al. [20,21] and Romero Wang [31], Hong and Varma [32], Espinos et al. [33] and Lu et al.
[34] developed three-dimensional finite element models for pre-
dicting the fire performance of CFST columns. The three-dimen-
⇑ Corresponding author. Tel.: +86 451 86282079; fax: +86 451 86282083.
sional finite element method can account for various
E-mail address: [email protected] (H. Yang).

0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.08.019
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1987

Nomenclature

Ac cross-sectional area of concrete core n load ratio

As cross-sectional area of steel tube Nf constant compressive load applied in fire
A cross-sectional area of composite section, A = As + Ac Nu ultimate strength of composite columns at ambient
cc specific heat of concrete temperature
cs specific heat of steel t heating time
cw specific heat of water tR fire resistance time
D outer depth of RHS ts wall thickness of the steel tube
B outer width of RHS T temperature
e load eccentricity a steel ratio, a = As/Ac
e/r load eccentricity ratio, r = D/2 ac coefficient of thermal expansion of concrete
Ec(T) Young’s modulus of concrete at temperature T as coefficient of thermal expansion of steel
Es Young’s modulus of steel d elongation percentage at fracture of steel
fc concrete prismatic strength, 0.95 times the concrete cyl- ec strain in concrete
inder strength es strain in steel
fc0 concrete cylinder strength etr transient strain in concrete
fc0 (T) concrete cylinder strength at temperature T ls Poisson’s ratio of steel
fcu,28 concrete cube strength at 28 days qc density of concrete
fcu,test concrete cube strength at the test day of specimens qs density of steel
fc0 (T) concrete tensile strength at temperature T qw density of water
fu ultimate tensile strength of steel rc stress in concrete in compression
fy yield strength of steel rs stress in steel
H height of columns rt stress in concrete in tensionpffiffiffi
k reduction factor k slenderness
pffiffiffi ratio, k ¼ 2 3L=D (major axis) or
kc thermal conductivity of concrete k ¼ 2 3L=B (minor axis)
ks thermal conductivity of steel
L unsupported length of columns

complexities of the thermal and mechanical behaviour of CFST col- In this paper, the performance of concrete-filled RHS columns
umns, for example, the non-uniformity of temperature along the exposed to fire on 3 sides is investigated experimentally and
longitudinal direction of the member, local buckling and the inter- numerically. Three full-scale concrete-filled RHS columns, two of
action between the steel tube and core concrete. Previous authors which were subjected to 3-sided fire exposure and the other one
have also conducted sensitivity analyses to study the influence of a to 4-sided fire exposure, were tested to study their fundamental
range of modelling factors, including the interaction behaviour be- fire behaviour and obtain data for the validation of finite element
tween the steel and concrete, geometric imperfections and con- (FE) models. In the experiments, temperature distributions, axial
crete tensile properties. These studies have shown that the finite displacements, lateral deflections and failure modes were all re-
element method can accurately simulate the observed physical corded and discussed. The FE models employed sequentially cou-
behaviour of CFST columns in fire, and can provide good predic- pled thermal–stress analyses, with a heat transfer analysis
tions of fire resistance. Nevertheless, discrepancies between followed by a stress analysis, using the program ABAQUS. Paramet-
numerical simulations and the experimental response of members ric studies were conducted to identify the influences of key param-
under fire loading tend to be larger than those at ambient temper- eters on the fire resistance of the column, including load ratio,
ature, due to the greater number of variables and uncertainties. cross-sectional dimensions, slenderness ratio and load eccentricity.
Previous research has focused largely on the fire performance of Based on the parametric studies, a simplified design method was
CFST columns in uniform fire. However, in real structures, columns proposed for predicting the fire resistance of concrete-filled RHS
may be subjected to non-uniform exposure due to localised fires or columns subjected to 3-sided exposure. Furthermore, a reduction
the barrier effect of adjacent walls. For concrete-filled rectangular factor method, which is considered more convenient for practical
hollow section (RHS) columns, this means that, in different situa- applications, was devised to extend existing fire safety design
tions, only some sides will be exposed to fire, i.e., one side, two methods for concrete-filled RHS columns in uniform fire to those
sides or three sides. It has been shown that the number of sides ex- columns in 3-sided fires.
posed to fire (hereafter referred to as the fire boundary) has a con-
siderable influence on the thermal distribution and fire resistance 2. Experimental study
of structural members [35–38]. In previous work [38,39], it was
found that slender square shaped CFST columns (slenderness ratio Following previous experiments on square shaped CFST col-
>30) subjected to a 3-sided fire have lower fire resistance than umns exposed to non-uniform fire [38], experiments are con-
those in uniform fire; this can be attributed to the shift of the cen- ducted herein to study the fire behaviour of concrete-filled RHS
tre of stiffness of the column, which effectively induces eccentric columns under 3-sided exposure, in order to develop an under-
loading, and the additional curvature induced by the asymmetric standing of the fire response of these columns and to obtain test
thermal distribution. Distinct from square shaped cross-sections, results to verify FE models, which are described in Section 3.
rectangular cross-sections have a major axis and minor axis, result-
ing in more complicated responses when exposed to non-uniform 2.1. Specimens
fire. To date, no research has been reported on the fire performance
of concrete-filled RHS columns subjected to 3-sided fire exposure, Three full-scale concrete-filled RHS columns were tested to fail-
and this is the subject of the present paper. ure, two of which (R1 and R2) were subjected to 3-sided fire expo-
1988 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

30
End plate
Stiffener

675
Φ20 vent hole

No.2, Φ40 hole

for thermocouples

1230
ts

Exposed length=3m Steel tube

3810

D
No.1, Φ40 hole
for thermocouples
Concrete
B
1230

No.3, Φ40 hole

for thermocouples

Φ20 vent hole

30

(a) (b)
Fig. 1. Column specimen details: (a) elevation; and (b) cross-section (unit: mm).

sure and the third (R3) was in uniform fire. The edge width, edge elongation percentage at fracture d, were determined by means
depth and total length of the specimens are 200 mm, 300 mm of a tensile coupon test. The coupons were cut from the steel tube
and 3810 mm respectively. Details of a typical column are shown in the longitudinal direction of the member. The measured steel
in Fig. 1. Two enlarged end plates with a cross-section of properties are summarised in Table 2. The carbonate aggregate
660 mm  460 mm and a thickness of 30 mm were welded to the concrete mix proportions, together with the average compressive
top and bottom of the column. The centre of the end plates coin- cube (150 mm) strengths at 28 days after pouring fcu,28 and the test
cided with the geometric centre of the steel tubes to attain concen- day of the first specimen fcu,test (at 147 days after pouring) are gi-
tric loading. Further details of the test specimens are given in ven in Table 3.
Table 1, in which L is the unsupported column length, B, D and ts Two semi-circular holes of 20 mm diameter were drilled at each
are the breadth,
pffiffiffi depth and thickness, respectively, of the steel sec- end of the steel tube to provide vent holes for water vapour pro-
tion, k ¼ 2 3L=D is the major axis column slenderness ratio and tR duced during the test. Holes of 40 mm diameter were located at
is the fire resistance time. the mid-height and 675 mm away from both ends of the columns
The effect of load ratio on the fire resistance of concrete-filled for the installation of thermocouples. Type K chromel–alumel ther-
RHS columns was investigated in addition to the fire boundary. mocouples, with a diameter of 1 mm, were installed to measure
The load ratio is defined as follows: temperatures at different locations through the cross-section
(Fig. 2). Any gaps were filled with high-strength non-shrinkage
Nf
n¼ ð1Þ grout. The temperatures at different positions along the longitudi-
Nu
nal axis of the columns were measured to verify the longitudinal
where Nf is the constant compressive load applied in the fire test uniformity of temperature.
and Nu is the ultimate strength of the composite column at ambient In order to simulate the unexposed (protected) sides, two layers
temperature. of 50 mm thick ceramic fibre blankets were attached to the column
The key material properties of the steel tube – elastic modulus surface with high-temperature glue that can endure 1200 °C. To
Es, yield strength fy, ultimate strength fu, Poisson’s ratio ls and further secure the blankets, thermocouples were used to tie the

Table 1
Specimen details.

Column no. B  D  ts (mm) k fy (N/mm2) fcu,test (N/mm2) Nf (kN) n Fire boundary Boundary conditions tR (min) Failure mode
R1 200  300  5.75 44.0 352.3 59.3 1520 0.4 3-sided Pinned–rollered 34.3 Global buckling
R2 200  300  5.75 44.0 352.3 59.3 2280 0.6 3-sided Pinned–rollered 14.3 Global buckling
R3 200  300  5.75 44.0 352.3 59.3 2280 0.6 4-sided Pinned–rollered 13.0 Local buckling
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1989

Table 2
Key material properties from tensile coupon tests.

Column no. ts (mm) Es (N/mm2) fy (N/mm2) fu (N/mm2) ls d (%)

5
R1, R2, R3 5.75 2.047  10 352.3 450.3 0.277 27.4

Table 3
Concrete mix proportions and compressive strengths.

Cement (kg/m3) Fines (kg/m3) Coarse (kg/m3) Water (kg/m3) Silica fume (kg/m3) Super-plasticiser (kg/m3) fcu,28 (N/mm2) fcu,test (N/mm2)
385 705 1074 175 53 7.67 52.6 59.3

Ceramic fibre blanket Ceramic fibre blanket Ceramic fibre blanket

1 8
2 9
367 10 13 14 15

4 11
5 12

(a) (b) (c)

Fig. 2. Layout of thermocouples of specimens R1–R3: (a) no. 1 hole; (b) no. 2 hole; and (c) no. 3 hole (unit: mm).

Table 4 effectiveness of the adopted covering method. The fibre ceramic

Thermal properties of the ceramic fibre blanket. blankets were modelled using solid elements to simulate the real
Properties T (°C) (experimentally-achieved) level of protection on the unexposed
side. Fig. 3 shows the temperatures of the specimen with the real
<600 800 1000
and ideal (i.e. fully protected) exposure conditions. It may be seen
Density (kg/m3) 128
that the adopted experimental approach of covering the surface
Thermal conductivity (W/(m °C)) 0.12 0.16 0.23
Specific heat (J/(kg °C)) 878.85 with a ceramic fibre blanket provides close to ideal conditions.
Allowable temperature (°C) 1260
2.2. Test conditions and procedure

blanket to the specimens at about 300 mm intervals. The insulative The fire tests were conducted in a purpose-built furnace
properties of the blankets are given in Table 4. The heat transfer FE specially built for testing structural columns, which can provide
model described in the next section was used to verify the combined actions of elevated temperature and structural load.

900 300
5(5) 270
800
Fire Fire
700 240 6(6)
7(7)
C

1 210
o

600 2
Fire

Fire
Fire

Fire

7
Temperature

Temperature

180 36
4(4)
500
1 150 4 2
400 5
Fire 120 Fire
300
1 90
200 60 3(3)
100 30
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time Time
(a) (b)
Fig. 3. Comparisons of temperatures of specimen with real and ideal protection: (a) steel tube; and (b) concrete core.
1990 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

(a) (b)
Fig. 4. End boundary condition: (a) general view; and (b) details (unit: mm).

No.b, L No.b, R

Draw wires

Furnace
No.b, L insulation No.b, R
Concrete-filled
No.a, L RHS column No.a, R
Displacement No.a, L No.a, R
transducer Anchor point

(a) (b)
Fig. 5. Layout of displacement transducers for measuring lateral deflection: (a) elevation; and (b) plan.

The furnace is a rectangular box with a plan area of 3 m  2 m and Since the draw wires on either side of the specimens were of similar
a height of 3.3 m. The maximum load capacity is 400 tonnes. The length and temperature, their thermal expansion would be almost
interior surfaces of the furnace chamber are lined with ceramic fi- equal. Hence, the effect of thermal expansion can be eliminated
bre materials that efficiently transfer heat to the test specimens. by taking the true lateral displacement of the column u as:
The furnace temperature is measured with five platinum–rhodium
thermocouples and controlled by programmed fire curves, and air utest;L þ utest;R
u¼ ð2Þ
pressure can be also adjusted. Details of the furnace are given in 2
Yang et al. [38].
where utest,L and utest,R are the measured lateral displacements of
Specimens were installed in the furnace by bolting their end
the left side and right side of the column, respectively.
plates to cast steel support plates (Fig. 4). The upper cast steel plate
was fully restrained except for movement in the vertical direction
and rotation about the major axis of the cross-section of the
column. The lower cast steel plate was fully restrained except for
rotation about the major axis. The boundary conditions of the col-
umns were considered to be pinned–rollered.
Axial load was applied concentrically approximately 30 min be-
fore the start of the test, until no further increase of transient axial
displacement was measured. The load was held constant with a
servo controlled hydraulic jack throughout the test.
The ambient temperature at the start of each test was about
10 °C. During the test, the furnace temperature was increased fol-
lowing the ISO-834 standard fire curve [40]. The central 3000 mm
length of the column was subjected to fire, leaving 560 mm at the
top and 250 mm at the bottom unexposed. The furnace, steel and
concrete temperatures and the axial and lateral displacements of
the specimens were measured during the test. A high temperature
camera also monitored the specimens throughout the testing
process.
In order to measure the lateral displacement of the columns,
draw wires were bound to the specimens at the mid-height of the
chamber (i.e., 1.65 m away from the bottom of the specimen) and
were attached to four displacement transducers outside the furnace
(Fig. 5). The four displacement transducers were placed in pairs on
the two opposite sides of the furnace to monitor the torsional
Fig. 6. Specimens after testing: (a) R1 (n = 0.4, 3-sided fire); (b) R2 (n = 0.6, 3-sided
deflection, denoted as Noa,L, Noa,R and Nob,L, Nob,R, respectively. fire); and (c) R3 (n = 0.6, 4-sided fire).
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1991

Local Buckling

Local Buckling Local Buckling

(a) (b) (c)

Fig. 7. Local buckling of specimens: (a) R1 (n = 0.4, 3-sided fire); (b) R2 (n = 0.6, 3-sided fire); and (c) R3 (n = 0.6, 4-sided fire).

1000 mid-height is considered to be the temperature of the specimen.

Since thermocouples 3 and 4 of specimen R1 failed during the fire
test, the corresponding temperatures measured near the top end of
800 the specimens were used in their place, though in the case of ther-
mocouple 4, the corresponding thermocouple (thermocouple 11)
C

also failed, so no results are available for this point. Fig. 10 shows
o

600
Temperature

the measured temperature curves of the three test specimens; as

expected, the temperature of the outer steel tube is much higher
ISO-834
R1 than that of the concrete core. For the specimens exposed on fire
400
R2 on 3 sides, the temperature of the unexposed side is much lower
R3 than that of the exposed sides, with the difference reaching around
200 200–300 °C at the time of failure. However, the temperatures on
the exposed sides are very similar, regardless of whether the fire
is on 3 or 4 sides (Fig. 10d).
0 Fig. 11 presents the measured time–displacement curves of the
0 5 10 15 20 25 30 35
test specimens, in which a positive value means expansion and
Time
bending towards the exposed side for the axial and lateral
Fig. 8. Comparisons of furnace temperature with ISO-834 standard fire curve. displacements, respectively. Time–displacement velocity curves
were also calculated and are presented in Fig. 11. The time–axial
displacement curves have three similar stages for columns with
The failure criteria were set according to ISO-834 [40]. Speci-
both 3-sided and 4-sided fire exposure: (i) expansion, (ii) stable
mens were considered to have failed and the tests were terminated
contraction and (iii) rapid contraction. The time–lateral displace-
when either of the following conditions was met: (a) the axial
ment also experiences three corresponding stages. However, spec-
deformation exceeded 0.01H mm; (b) the rate of the axial deforma-
imens subjected to 3-sided exposure firstly bend towards the
tion exceeded 0.003H mm/min, where H is the height of the
exposed side then to the reverse direction, due to the countering
specimens in millimetres.
effects of the asymmetric expansion and the shift of the centre of
stiffness. The phenomenon is also observed for square shaped CFST
2.3. Test results columns exposed to 1-sided and 3-sided fire [38]. For the specimen
under uniform fire loading, lateral displacement is minimum as no
It was observed that the specimens under 3-sided fire exposure significant global buckling occurred. The two measurements of
failed by global buckling accompanied with local buckling on the lateral displacement shown in Fig. 11b, d and f were calculated
exposed surfaces, which is similar to the failure modes observed from the data of the pairs of displacement transducers (Noa,L and
for square shaped CFST columns with 1-sided or 3-sided fire expo- Noa,R; Nob,L and Nob,R) according to Eq. (2). Since the difference be-
sure [38]. For the specimen in uniform fire, failure was by material tween the two curves is very small, the average result will be used
yielding and local buckling with no significant lateral deformation hereafter.
of the column, similar to the failure mode described in [16,19]. A
general view and close-up view showing local buckling of the spec-
imens after testing are shown in Figs. 6 and 7, respectively. 3. Finite element modelling
The measured time–temperature curves of the furnace in the
tests were in good agreement with the standard ISO-834 time– There are two types of coupled thermal–stress analysis meth-
temperature curve (Fig. 8). Note that the control system had to ods: sequentially coupled thermal–stress analysis and fully
be restarted during the test of specimen R1, causing a temporary coupled thermal–stress analysis. Fully coupled thermal–stress
drop in furnace temperature, which is evident in Fig. 8. The tem- analysis solves for the stress/strain and the temperature fields
peratures along the longitudinal direction of the R2 specimen, simultaneously, and is used when thermal and mechanical
which was typical of all specimens, are shown in Fig. 9, and solutions affect one another strongly. For most conventional struc-
confirm the uniformity of the temperature along the member tures, it is reasonable to assume their structural behaviour in fire
length. Hereafter, the temperature of the cross-section at depends on the thermal field but that there is no interdependency
1992 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

[41]. Romero et al. [22] have shown that load and corresponding For the stress analyses, the concrete compression model pro-
deformation typically have insignificant influence on the thermal posed by Lie [43], and outlined in the Appendix, was adopted,
distribution. Therefore the sequentially coupled approach was which had been shown by Hong and Varma [32] and Romero
used in this study. It was performed by firstly solving the pure heat et al. [33] to be suitable for the prediction of the fire behaviour
transfer problem, then reading the temperature solution into the of CFST columns. The concrete tension model proposed by Hong
stress analysis as a predefined field [42]. [32] was used to define the tensile stress–strain relationship of
the concrete (see Appendix). The total strain in the concrete under
3.1. Material properties fire condition has four components, namely the mechanical strain
caused by externally applied stress, the thermal strain caused by
For the heat transfer analyses, the density, thermal conductivity the change of temperature, the creep strain caused by dislocations
and specific heat of the materials are required. The thermal prop- in the microstructure of the material and the transient strain
erty models recommended by Lie [43] for steel and concrete were caused by the change of chemical composition [44]. The concrete
adopted and are presented in the Appendix. compressive model proposed by Lie [43] accounts for the creep

350 8 350
9
Fire
Fire

300 300 7

Fire
12

Fire
Fire 3 6
Temperature oC

Temperature oC
250 5 250 Fire
1 12
200 2 200 7
Fire
Fire

1 15
5
150 150 15

Fire
Fire
Fire 13 14
8
100 100 Fire
6
50 50 14
9 2 13
0 0
0 3 6 9 12 15 0 3 6 9 3 12 15
Time Time
(a) (b)
Fig. 9. Comparisons of cross-section temperatures at different heights in R2 specimen: (a) comparison between mid-height and top cross-section; and (b) comparison
between mid-height and bottom cross-section.

600 350
Fire Fire
500 (fully protected) 300 (fully protected)
1 1
5
Temperature / oC
Temperature / oC

2 2
250
Fire

Fire

Fire

Fire

400 3 7 3 7
6 6
4 4 7
5
7 200 5 5
300
Fire Fire
150
1 1
200
100
100 6 50 6 4
2 2
3 3
0 0
0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16
Time /min Time /min
(a) (b)
350 600 R1
R2
300 Fire R3
500
1 1 Fire
Temperature / oC
Temperature / oC

2 (fully protected for R1, R2)

250
Fire

Fire

3 7 7 400
6
4
200
Fire

Fire

5
5 300
Fire
150 5
200 Fire
100

50 2 100
6 4
3
0 0
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35
Time /min Time /min
(c) (d)
Fig. 10. Measured time–temperature curves: (a) R1; (b) R2; (c) R3; and (d) temperature of thermocouple 5 in R1, R2 and R3.
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1993

strain implicitly. Previous research [44] has indicated that the tran- developed by Lie [43] for steel at elevated temperatures was
sient strain can have a significant influence on the deformation of employed (see Appendix), implicitly including the effect of
columns and it may be unsafe to ignore it. Hence the stress–total creep.
strain of concrete was incorporated into ABAQUS, implicitly The thermal expansion models for concrete and steel were
including the transient strain. The transient strain model proposed taken from Lie [43], as detailed in the Appendix. The Poisson’s ratio
by Guo and Shi [37] was employed, which is expressed as follows: was set to be 0.2 and 0.3 for concrete and steel respectively, inde-
" pendent of the temperature. The elastic modulus of concrete and
 2  #
rc T T steel at elevated temperatures were taken as the tangent modulus
etr ¼ 72   103 ð3Þ at the zero stress point of the stress–strain curve.
fc 1000 1000

where etr is the transient strain, rc is the corresponding stress, fc is 3.2. Heat transfer analysis
the concrete prismatic strength (0.95 times the concrete cylinder
strength) and T is the temperature in °C. Heat transfer analyses were conducted to simulate the heat
For the structural steel, the total strain includes the mechan- transferred from the fire to the structural members. Heat is
ical strain, the thermal strain and the creep strain. The model transferred from the fire to the outer surfaces of the specimens

20

Displacement velocity / mm/min

5 3

Lateral displacement / mm
iii
0
Axial displacement / mm

0 0
-3 i ii iii
-5
i ii
-10 Test displacement
-6 -20
-15 Test velocity -9
Axial displacement velocity limit
-20 -12 -40
Failure time=34.3min Tested (No. a,L , No.a,R)
-25 -15
Tested (No. b,L , No.b,R)
-30 -18 -60 Averaged
-35 -21
Axial displacement limit
-40 -24 -80
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Time / min Time / min
(a) (b)
20
Displacement velocity / mm/min

5 3
Lateral displacement / mm
Axial displacement / mm

0 0
0
-5 i ii iii -3 iii
i ii
-10 Test displacement -6 -20
-15 Test velocity -9
Axial displacement velocity limit
-20 -12 -40
Failure time=14.3min
-25 -15 Tested (No. a,L, No.a,R )
-30 -18 Tested (No. b,L, No.b,R )
-60
Averaged
-35 Axial displacement limit -21
-40 -24 -80
0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
Time / min
Time / min
(c) (d)
20
Displacement velocity / mm/min

5 3
Lateral displacement / mm

0
Axial displacement / mm

0
0
i ii iii -3
-5
Tested (No. a,L , No.a,R)
-10 Test displacement -6 Tested (No. b,L , No.b,R)
Test velocity
-20
-15 -9 Averaged
Axial displacement velocity limit
-20 -12
-40
Failure time=13min -15
-25
-30 -18
-60
-35 -21
Axial displacement limit
-40 -24
0 2 4 6 8 10 12 14 16 -80
0 2 4 6 8 10 12 14 16
Time / min
Time / min
(e) (f)
Fig. 11. Displacement of specimens: (a) R1: axial displacement; (b) R1: lateral displacement; (c) R2: axial displacement; (d) R2: lateral displacement; (e) R3: axial
displacement; and (f) R3: lateral displacement.
1994 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

via radiation and convection. In this study, the convective heat fire curve [45], depending on the specific conditions of the test
transfer coefficient and resultant emissivity were taken as 25 W/ members. Unless otherwise specified, the ISO-834 standard fire
(m2 K) and 0.5 respectively for the exposed sides. Heat is trans- [40] was employed.
ferred within the specimens through conduction. It was assumed For the unexposed side, the convective heat transfer coefficient
that the steel tube and the concrete core were in perfect contact was set to be 9 W/(m2 K), which also makes an allowance for heat
such that there was no heat loss at the interface. Fire conditions transferred by radiation [46]. The ambient temperature was
followed the ISO-834 standard fire curve [40] or the ASTM-E119 defined to be 10 °C for the simulation of the tests, and 20 °C for

800 1-tested 200 2-tested 200 3-tested 800 5-tested

Temperature C

Temperature C

Temperature C

Temperature C
1-predicted 160 2-predicted 160 3-predicted 5-predicted
o

o
600 600
120 120
400 400
80 80
200 40 40 200

0 0 0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Time Time Time Time
(a) (b) (c) (d)
200 6-tested 800 7-tested 200 1-tested 200 2-tested
Temperature C

Temperature C

Temperature C

Temperature C
o

o
160 6-predicted 7-predicted 160 1-predicted 160 2-predicted
600
120 120 120
400
80 80 80
40 200 40 40

0 0 0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time Time
(e) (f) (g) (h)
200 3-tested 200 4-tested 400 5-tested 200 6-tested
160 3-predicted 160 4-predicted 5-predicted 160 6-predicted
Temperature C
Temperature C

Temperature C

Temperature C

300
o

o
120 120 120
200
80 80 80
40 40 100 40

0 0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time Time

(i) (j) (k) (l)

400 7-tested 400 1-tested 200 2-tested 200 3-tested
7-predicted 1-predicted 160 2-predicted 160 3-predicted
Temperature C

Temperature C

Temperature C

Temperature C

300 300
o

120 120
200 200
80 80
100 100 40 40

0 0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time Time

(m) (n) (o) (p)

200 4-tested 400 5-tested 200 6-tested 400 7-tested
Temperature C

Temperature C

Temperature C

160 4-predicted 5-predicted 160 6-predicted 7-predicted

o

o
Temperature C

300 300
o

120 120
200 200
80 80
40 100 40 100

0 0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time Time
(q) (r) (s) (t)
Fig. 12. Comparisons of predicted and measured time–temperature curves: R1 (a–f); R2 (g–m); and R3 (n–t).
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1995

290
530
Symmetric surface

Boundary
Bottom plate: U1=U2=U3=0
4330

top plate: U1=U2=0

3000

Symmetric surface

Solid block
including end plate and
220

pinned support
290

Fig. 13. FE model of the tested specimens.

Table 5
Details of square shaped CFST columns subjected to non-uniform exposure [38].

Column no. D  ts (mm) fy (N/mm2) fcu,test (N/mm2) Nf (kN) n e (mm) Fire boundary Boundary conditions tR,test (min) tR, cal (min)
S1 300  5.80 361.7 59.3 3186.0 0.6 0 3-sided Pinned–rollered 14 15.8
S2 300  5.80 361.7 59.3 2121.0 0.4 0 3-sided Pinned–rollered 33 30.1
S3 300  5.80 361.7 59.3 2312.0 0.6 40 3-sided Pinned–rollered 13 10.9
S4 300  5.74 299.1 59.3 1693.0 0.6 80 3-sided Pinned–rollered 11 13.1
S5 300  5.74 299.1 59.3 2976.6 0.6 0 1-sided Pinned–rollered 43 40.4
S6 300  5.74 299.1 59.3 1986.6 0.4 0 1-sided Pinned–rollered 99 126.9

Table 6
Details of square and rectangular shaped CFST columns in uniform fire [16,19].

Column no. D  B  ts (mm) fy (N/mm2) fc0 (N/mm2) e (mm) Fire protection (mm) Nf (kN) Boundary conditions tR,test (min) tR, cal (min)

SQ-01 [16] 152.4  152.4  6.35 350 58.3 0 0 376 Fixed–fixed 66 55.4
SQ-02 [16] 152.4  152.4  6.35 350 46.5 0 0 286 Fixed–fixed 86 60.6
SQ-07 [16] 177.8  177.8  6.35 350 57.0 0 0 549 Fixed–fixed 80 61.5
R-3 [19] 300  150  7.96 341 39.2 0 0 1906 Pinned–rollered 16 14.8
RP-1 [19] 300  200  7.96 341 39.2 0 13 2486 Pinned–rollered 104 104.1
RP-2 [19] 300  200  7.96 341 39.2 0 20 2486 Pinned–rollered 146 156.5

the latter parametric studies. The concrete core and the steel tube Moisture migration and evaporation in concrete occur between
were modelled with 8-node continuum solid elements (DC3D8) about 100 °C and 200 °C, absorbing heat and affecting the temper-
and 4-node shell elements (DS4), respectively. ature distribution in the concrete. Moisture migration is a complex
1996 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

5 5 5
0 0 0
-5 -5 -5

Axial displacement
Axial displacement
Axial displacement
Tested Tested Tested
-10 Predicted -10 Predicted -10 Predicted
-15 -15 -15
Fire
-20 -20 -20

Fire

Fire

Fire
Fire

Fire

Fire
-25 -25 -25
-30 -30 -30
Fire Fire Fire
-35 -35 -35
-40 -40 -40
0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time
(a) (b) (c)
5 5 5
0 0 0

Axial displacement
Axial displacement

Axial displacement
-5 Tested -5 Tested -5 Tested
Predicted Predicted -10
Predicted
-10 -10
-15 -15 -15
Fire

Fire

Fire
Fire

Fire

Fire
-20 -20 -20
Fire Fire Fire
-25 -25 -25
-30 -30 -30
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16
Time Time Time
(d) (e) (f)
5 5 5
0 0 0
Axial displacement

Axial displacement

Axial displacement
-5 Tested -5 Tested -5 Tested
Predicted Predicted Predicted
-10 -10 -10
-15 -15 -15
Fire
Fire

-20 -20 -20

Fire Fire Fire
-25 -25 -25
-30 -30 -30
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35 40 45 0 20 40 60 80 100 120 140
Time Time Time
(g) (h) (i)
20 30 20
Tested
10 20 Predicted 10
Axial displacement

Axial displacement
Axial displacement

10
0 0
Tested 0 Tested
-10 -10 Predicted
Predicted -10
-20 Fire Fire -20 Fire
-20
Fire

Fire

Fire
Fire

Fire

Fire

-30 -30 -30

Fire Fire Fire
-40 -40 -40
0 10 20 30 40 50 60 70 0 20 40 60 80 100 0 20 40 60 80
Time Time Time
(j) (k) (m)
5 5 5
0 0 0
-5 -5 -5
Axial displacement

Tested Tested Tested

Axial displacement
Axial displacement

-10 Predicted -10 Predicted -10 Predicted

-15 -15 -15
Fire Fire Fire
-20 -20 -20
Fire

Fire

Fire
Fire

Fire

Fire

-25 -25 -25

-30 -30 -30
Fire Fire Fire
-35 -35 -35
-40 -40 -40
0 2 4 6 8 10 12 14 16 18 0 20 40 60 80 100 120 0 20 40 60 80 100 120 140 160
Time Time Time
(n) (p) (q)
Fig. 14. Comparisons of predicted and measured time–axial displacement curves: (a) R1; (b) R2; (c) R3; (d) S1 [38]; (e) S2 [38]; (f) S3 [38]; (g) S4 [38]; (h) S5 [38]; (i) S6 [38];
(j) SQ-01 [16]; (k) SQ-02 [16]; (m) SQ-07 [16]; (n) R-3 [19]; (p) RP-1 [19]; and (q) RP-2 [19].

process and is seldom accounted for in FE models. Reasonable water content was assumed to be 5% by weight and all water was
accuracy can usually be obtained by including moisture evapora- assumed to be vaporised at about 100 °C.
tion but ignoring moisture migration. Therefore moisture migra-
tion is not modelled in this paper. The influence of water 3.3. Stress analysis
vaporisation in concrete at elevated temperature was incorporated
into the concrete thermal properties by transforming the heat Nonlinear stress analyses were carried out subsequently to
required for water vaporisation into the specific heat [47]. The free investigate the structural response of the specimens under
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1997

20 20 20

Lateral displacement 0 0 0

Lateral displacement

Lateral displacement
-20 -20 -20
Tested Tested Tested
-40 Predicted -40 Predicted -40 Predicted
-60 -60 -60 Fire

Fire

Fire

Fire
Fire

Fire

Fire
-80 -80 -80
-100 Fire -100 Fire -100 Fire
-120 -120 -120
0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time Time
(a) (b) (c)
20 20 20
0 0 0
Lateral displacement

Lateral displacement

Lateral displacement
-20 -20 -20
Tested Tested Tested
-40 Predicted -40 Predicted -40 Predicted
-60 -60 -60
Fire

Fire

Fire
Fire

Fire

Fire
-80 -80 -80
Fire Fire Fire
-100 -100 -100
-120 -120 -120
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16
Time Time Time
(d) (e) (f)
20 20 20
0 0 0
Lateral displacement

Lateral displacement

Lateral displacement
-20 -20 -20
Tested Tested Tested
-40 Predicted -40 Predicted -40 Predicted
-60 -60 -60
Fire
Fire

-80 -80 -80

Fire Fire Fire
-100 -100 -100
-120 -120 -120
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35 40 45 0 20 40 60 80 100 120 140
Time Time Time
(g) (h) (i)
Fig. 15. Comparisons of predicted and measured time–lateral displacement curves: (a) R1; (b) R2; (c) R3; (d) S1 [38]; (e) S2 [38]; (f) S3 [38]; (g) S4 [38]; (h) S5 [38]; and (i) S6
[38].

combined constant axial load and non-uniform fire loading. A behaviour was defined as ‘hard contact’ and the tangential
three-dimensional finite element model was developed. The con- direction behaviour was taken as penalty friction, with a coefficient
crete core and the steel tube were modelled with linear reduced- of 0.3.
integration 8-node continuum element (C3D8R) and linear Global geometric imperfections were specified as the first
reduced-integration 4-node shell elements (S4R) respectively. In (lowest) elastic buckling model shape (deflecting towards the
order to read the temperature data for each node efficiently, the unexposed side) with an amplitude of L/1000, where L is the
finite element meshes of the stress analysis model were the same unsupported length of the column. The first buckling mode shape
as those of the corresponding heat transfer analysis model. was attained by means of eigenvalue analysis.
The concrete damage plasticity model given in ABAQUS was The axial load was applied up to the design level and then held
used to represent the concrete material. The model assumes that constant, after which thermal loading was applied by reading
the two main failure mechanisms of the concrete are tensile crack- nodal temperatures obtained from the heat transfer analysis. The
ing and compressive crushing and requires the definition of the boundary conditions of columns were generally defined as pin-
stress–strain relationships in compression and tension. For the ned–pinned, unless otherwise stated.
steel, the isotropic multiaxial plasticity model with Von Mises
yielding and associated plastic flow with temperature dependency
was used. 4. Validation of the FE model
The interaction behaviour between the steel tube and the con-
crete core was simulated with a contact interaction algorithm in The measured temperatures and displacements from the tests
ABAQUS. The surfaces of the steel and concrete in contact were on the concrete-filled RHS columns subjected to 3-sided fire
defined as a contact pair, with the concrete outer surface as the performed in this study, as well as square shaped CFST columns ex-
master surface and the steel tube inner surface as the slave surface. posed to non-uniform fire [38] and uniform fire [16] and rectangu-
The mechanical properties of the contact interaction consist of lar shaped CFST columns exposed to uniform fire [19] were
normal and tangential direction behaviour. The normal direction employed to verify the FE model.
1998 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

Table 7
Values of studied parameters.

Parameter Values Fixed

value
n 0.4, 0.5, 0.6, 0.7 0.5
A (104 mm2) 16, 36,
64, 100,
144,
196,
Local buckling 256
64
D/B 1.0, 1.5, 2.0 1.0, 1.5,
2.0
k 20, 30, 40, 50, 60 30
a 0.08, 0.12, 0.16, 0.20 0.12
fy (MPa) 235, 345, 390, 420 345
fc0 (MPa) 20, 30, 40, 50 40
e/r 1.0, 0.8, 0.6, 0.4, 0.3, 0.2, 0.1, 0, 0.1, 0.2, 0
0.3, 0.4, 0.6, 0.8, 1.0
(a) (b)
Fig. 16. Failure modes of specimen R1: (a) predicted; and (b) test.
of location or imperfect insulation of the thermocouples. The FE
models replicated the test end boundary conditions (Fig. 4), with
Comparisons between the predicted and measured tempera- the total length of the specimens being 4330 mm between the
tures are shown in Fig. 12, in which the unexposed side was pin ends (H + 2Hpin). For simplicity, the end plates of the columns
modelled using the real exposure conditions. Layouts of thermo- and the pinned supports were modelled together as a solid block.
couples are shown in Fig. 2. The predicted temperatures generally A typical FE model of one of the test specimens is depicted in
accord well with the temperatures measured in the fire tests. There Fig. 13, in which the solid block was considered to be perfectly
are some discrepancies between predicted and tested tempera- elastic. Only half of the columns were modelled due to their geo-
tures (e.g., Fig. 12e–g), which may have been caused by deviation metric symmetry.

1.0 5 1.0 5
0 0
0.8 -5 0.8 -5
i ii iii i ii
Axial force ratio

Axial force ratio

iii
Axial displacement

Axial displacement
-10 -10
0.6 0.6
-15 -15
-20 -20
0.4 0.4
-25 -25

0.2 -30 0.2 -30

-35 -35
0.0 -40 0.0 -40
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time Time
(a) (b)
Fig. 17. Time–axial force ratio curves: (a) R2; and (b) R3.

0.4 0.4 Fire

1 1
Tension Tension
Fire
Fire

Fire
Fire

0.2 0.2
Longitudinal stress / fy
Longitudinal stress / fy

2 2
0.0 Fire 0.0 Fire

-0.2 Compression
-0.2 Compression

-0.4 -0.4

-0.6 -0.6

-0.8 -0.8

-1.0 -1.0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time Time

(a) (b)
Fig. 18. Time–longitudinal stress curves: (a) R2; and (b) R3.
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 1999

Details of the specimens modelled herein but tested in other subjected to 4-sided exposure, increases more than that of speci-
studies [16,19,38] are summarised in Tables 5 and 6. Lie and Cha- men R2, exposed to 3-sided fire, due to its higher temperature.
bot [16] employed the ASTM-E119 standard temperature–time With increasing temperature, the axial contraction induced by
curve [45]. Comparisons between the predicted and measured deterioration of the material properties and subsequent buckling
time–axial displacement curves and time–lateral displacement becomes more significant than the thermal expansion, and the col-
curves of these columns are presented in Figs. 14 and 15. umn starts to shorten (stage ii). The axial force is then gradually
The predicted and test failure modes of the specimen R1 are transferred from the steel tube to the concrete core. The heating
illustrated in Fig. 16. The specimen failed by global buckling, rate of the steel tube subjected to 4-sided exposure is much higher
though local buckling was also observed on the exposed faces of than that subjected to 3-sided fire; therefore its axial force ratio
the steel tube. Clearly the FE failure mode matches well with the decreases more dramatically. Approaching the fire resistance
test failure mode. Similar observations were made for other (stage iii), the axial force ratio of the concrete core increases rap-
specimens. idly and the concrete cannot bear the axial load, resulting in failure
of the columns.
The time–longitudinal stress curves for the steel tube on the
5. Failure mechanisms
exposed and unexposed sides, normalised by the yield strength
of the steel at ambient temperature fy, are shown in Fig. 18. For
In order to develop a greater understanding of the failure mech-
specimen R2 exposed to 3-sided fire, the discrepancy in longitudi-
anisms of concrete-filled RHS columns subjected to 3-sided fire
nal stress between the exposed and unexposed sides is much high-
exposure, the axial force ratio in the steel tube and concrete core
er than that of specimen R3 under 4-sided fire loading. This is
of columns exposed to 3-sided fire and 4-sided fire were analysed
attributed to the shift of the centre of stiffness that occurs due to
with the numerical model. The axial force ratio is defined as the
the asymmetric thermal distribution. The shift of the centre of
ratio of the load carried by the steel or concrete to the entire load
stiffness is a particular phenomenon associated with columns
applied to the column. Fig. 17 shows the time–axial force ratio
subjected to asymmetric fire conditions. Determination of the
curves together with time–axial displacement curves for speci-
effective centroid of the column’s cross-section can help to under-
mens R2 and R3.
stand the mechanical behaviour of columns subjected to 3-sided
Initially (stage i), the heating rate of the steel tube is much high-
fire exposure. This problem will be discussed in later sections.
er than that of the concrete core, inducing greater expansion in the
steel tube; hence, the axial force ratio of the steel tube increases
while the axial force ratio of the concrete decreases. The delay of
6. Parametric studies
local buckling of the steel tube due to the concrete infill allows
the steel tube to bear much of the applied load for a significant per-
The validated FE model was used to study the influence of key
iod of time. The axial force ratio in the steel tube of specimen R3,
parameters on the fire resistance of concrete-filled RHS columns
exposed to 3-sided fire, including load ratio, cross-sectional area,
Unexposed Unexposed slenderness ratio, steel ratio, strength of steel and concrete, load
y y
eccentricity ratio and depth to width ratio. Values of these studied
parameters are listed in Table 7. The key parameters were varied
individually while the other parameters were held constant. A total
Mx of 780 parametric results were generated.
An RHS section has a major and minor axis; hence, the location
Fire

Fire
Fire

Fire

of the unexposed side and the direction of applied moment create

x x x x
four possible conditions, shown in Fig. 19. Compared to condition
My B, condition A is much more detrimental because the effect of addi-
tional eccentricity induced by the asymmetric thermal field is
superimposed onto that of the moment. A similar situation occurs
for conditions C and D. Therefore conditions A and D were studied,
y y
Fire Fire and the more adverse of the two defined the fire resistance of the
column.
(a) (b) The calculated results reveal that very similar fire resistances
were achieved for both conditions A and D; therefore only results
Fire Fire
y y for condition A are presented in Fig. 20. Similar to CFST columns
in uniform fire, the load ratio and cross-sectional area have signif-
icant influences on the fire resistance of columns in 3-sided fires,
Mx while steel ratio, strength of materials and depth to width ratio
have far less influence. However, unlike in the case of uniform fire,
Unexposed

Unexposed

slenderness ratio (between 20 and 60) has only a relatively small

Fire

Fire

x x x x influence on the fire resistance of CFST columns in 3-sided fires.

Load eccentricity ratio was also found to have an interesting
My
influence on the fire resistance of columns in 3-sided fires. Since
the centre of stiffness would shift towards the unexposed side
due to the asymmetric thermal distribution, an additional eccen-
y y tricity consequently occurs. The additional eccentricity intensifies
Fire Fire the influence of any load eccentricity. The parametric results show
(c) (d) that the longest fire resistances occur when the load is applied
nearer to the unexposed side instead of at the geometric centre
Fig. 19. Fire boundary conditions considered in parametric study: (a) A; (b) B; (c) C; of the cross-section (Fig. 20g). This may be explained by the fact
and (d) D. that the effective centroid of the section shifts with time due to
2000 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

120 A=64×104 mm2

120 n =0.5 D/B=1
D/B=1
α =0.12 D/B=1.5 α =0.12 D/B=1.5
λ =30 λ =30 D/B=2
100 D/B=2 100
fy =345MPa fy =345MPa

Fire resistance / min

Fire resistance / min

fc '=40MPa fc '=40MPa
80 80

60 60

40 40

20 20

0 0
0.3 0.4 0.5 0.6 0.7 0.8 50 100 150 200 250 300
Load ratio n) Cross-sectional area A × 104mm2
(a) (b)
120 A=64×104mm2 D/B=1 120 A=64×104mm2 D/B=1
α =0.12 D/B=1.5 n =0.5 D/B=1.5
100 n =0.5 D/B=2
100 λ =30 D/B=2
fy =345MPa fy =345MPa
Fire resistance / min

Fire resistance / min

fc '=40MPa fc '=40MPa
80 80

60 60

40 40

20 20

0 0
10 20 30 40 50 60 70 0.05 0.10 0.15 0.20 0.25
Slenderness ratio λ ) Steel ratio α )
(c) (d)
120 120 D/B=1
A=64×104 mm2 D/B=1 A =64 ×104mm2
α =0.12 D/B=1.5 α =0.12 D/B=1.5
λ =30 D/B=2 λ =30 D/B=2
100 100
n =0.5 fy =345MPa
Fire resistance / min

Fire resistance / min

fc '=40MPa n =0.5
80 80

60 60

40 40

20 20

0 0
200 250 300 350 400 450 10 20 30 40 50 60
Yield strength of steel fy Strength of concrete fc
(e) (f)
120 120 A=16 ×104 mm2
Fire resistance / min

D/B=1 n =0.5
A =64×104mm2 D/B=1.5 α =0.12 A=64 ×104 mm2
n =0.5 D/B=2
α =0.12 100 100
λ =30 A=144×104 mm2
f y =345MPa
λ =30
Fire resistance / min

e,positive f c '=40MPa
fy =345MPa
80
Fire

80
Fire

fc '=40MPa

60 Fire e,negative 60

40 40

20 20

0 0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.0 1.2 1.4 1.6 1.8 2.0
Load eccentricity ratio (e/r) Depth to width ratio D/B)
(g) (h)
Fig. 20. Influences of parameters on fire resistance: (a) load ratio; (b) cross-sectional area; (c) slenderness ratio; (d) steel ratio; (e) yield strength of steel; (f) strength of
concrete; (g) load eccentricity; and (h) depth to width ratio.

the decaying of the mechanical material properties with increasing in Fig. 21. The shift tends to increase with a decrease in load ratio
temperature, which means that an initial load eccentricity at room and an increase in cross-sectional area. A study of extended ranges
temperature may well reduce in fire as the position of the applied of the factors (Fig. 22) shows that the shift in effective centre of
load aligns more closely with the effective centroid in fire. The stiffness is between 0.05r and 0.25r, where r = D/2. Since the
principal factors that affect fire resistance, e.g. load ratio and beneficial effect of load eccentricity occurs within a narrow range
cross-sectional area, may also affect the shift of the centre of stiff- of eccentricity, it is recommended that this beneficial effect be
ness. The influence of these factors on the centroidal shift is shown ignored.
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 2001

Fire resistance / min

Fire resistance / min

200 n=0.4 n=0.5 200 D=400mm
A=64×104mm2 D=B
n=0.6 n=0.7 D=800mm
D =B n =0.5
α =0.12 α =0.12 D=1200mm
160 e,positive 160
λ =30 λ =30
e,positive

Fire
Fire
fy =345MPa fy =345MPa

Fire
Fire
fc '=40MPa fc '=40MPa
120 120
Fire e,negative
Fire e,negative
80 80

40 40

0 0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Load eccentricity ratio (e/r) Load eccentricity ratio (e/r)
(a) (b)
Fig. 21. Influences of load ratio and cross-sectional area on effective centroid.

suitable for both conditions A and D, though it should be noted that

2.4
n=0.4 the load applied on a given column is constant in real structures,
D =B n=0.5 resulting in different load ratios for the two conditions A and D;
n =0.5 n=0.6 hence, its fire resistance should be determined for both conditions
α =0.12 n=0.7 A and D, the shorter of which will be considered critical. The limits
2.0
λ =30 of validity of the simplified design formula are: n = 0.4–0.7,
R, e=0

fy =345MPa A = 16  104 mm2 to 256  104 mm2, D/B = 1–2, k = 20–60,

fc '=40MPa a = 0.08–0.2, fy = 235–420 MPa, fc0 ¼ 20—50 MPa.
/t

1.6
R, e

7.2. Reduction factor method

t

1.2
Differences in fire resistance between concrete-filled RHS col-
umns with 3-sided exposure and 4-sided exposure were studied
to extend fire safety design methods for those columns in 4-sided
fires to those in 3-sided fires. The parameters influencing the fire
0.8 resistance design of CFST columns in uniform fire (load ratio,
cross-sectional dimensions and slenderness ratio), were consid-
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
ered. The fire resistances of concrete-filled RHS columns in 3-sided
and 4-sided fires, as derived from the FE models, are shown in
Load eccentricity ratio (e/r )
Fig. 24. It may be seen that concrete-filled RHS columns subjected
Fig. 22. Locations of effective centroid. to 3-sided fires possess lower fire resistance than those columns
subjected to 4-sided fires when the slenderness ratio is less than
It should be noted that the same load ratio was maintained with about 30, whereas the reverse situation occurs at higher slender-
different load eccentricities, which means smaller load will be ness ratios. The results show that slenderness ratio (between 20
applied on the column with the increasing of load eccentricity. This and 60) has a small influence on the fire resistance of columns in
could be the reason why an increase of load eccentricity leads to a 3-sided fires, whereas it has significant influence on fire resistance
greater fire resistance when the load eccentricities exceed the of columns in 4-sided fires.
influence range of the centroidal shift. If current fire safety design methods for concrete-filled RHS col-
umns in uniform fires are to be applied to columns exposed to
3-sided fires, a reduction factor, depending on the slenderness ra-
7. Methods for predicting fire resistance tio, should be adopted. The relation can be expressed as following:

7.1. Simplified method tR;3 ¼ kt R;4 ð5Þ

where tR,3 is the fire resistance of concrete-filled RHS columns ex-
Based on the results of the parametric studies, it was found that posed to 3-sided fires, k is a reduction factor and tR,4 is the fire resis-
the load ratio and cross-sectional dimensions have pronounced tance of concrete-filled RHS columns exposed to 4-sided fires.
influences on the fire resistance of concrete-filled RHS columns If a column may be subjected to either 3 or 4-sided exposure,
subjected to 3-sided fire exposure. The following formula was the reduction factor given by Eq. (6) may be applied, covering both
established to relate fire resistance time and the two principal the 3-sided and 4-sided fire exposure conditions.
parameters determining it. 
pffiffiffi 0:85 k 6 30
k¼
B 1:0 k > 30
t R ¼ 0:2 2 ð4Þ
n
If a column can only be exposed to a 3-sided fire, the reduction
where tR is the fire resistance in hours, B is the shorter edge width of factor of Eq. (7) can be adopted, with better accuracy.
the cross-section in metres and n is the load ratio. Comparisons be-  2
tween the predicted fire resistances form Eq. (4) and the FE model k
k ¼ 1:9 þ 0:8 ð7Þ
are shown in Fig. 23, which indicates good agreement. Eq. (4) is 100
2002 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

3.0 3.0
Fire
2.5 2.5
+20% +20%

Fire

Fire
Fire
t R, formula / hour

t R, formula / hour
2.0 2.0
-20% -20%
1.5 Fire 1.5 Fire

1.0 1.0

0.5 0.5

0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
tR, FE / hour tR, FE / hour

(a) (b)
Fig. 23. Comparisons of predicted fire resistances from Eq. (4) and the FE models: (a) condition A; and (b) condition D.

120 3-sided D =B
120 3-sided D =800mm
4-sided α =0.12 4-sided D =B
n=0.5 α =0.12
100 100 fy =345MPa
fy =345MPa
Fire resistance / min

Fire resistance / min fc '=40MPa

fc'=40MPa
80 80
n=0.4
D =1200mm
60 60 n=0.5

40 D =800mm 40
n=0.6
D =400mm
20 20

0 0
10 20 30 40 50 60 10 20 30 40 50 60
Slenderness ratio λ) Slenderness ratio λ)
(a) (b)
Fig. 24. Differences of fire resistance between the columns exposed to 3-sided fires and 4-sided fires.

2.0 1.0
Eq.(4)
Eq.(7)
1.8
0.8

1.6
/ hour

0.6
1.4
/t

1.2 0.4
t

1.0
0.2

0.8

0.0
0.6 0.0 0.2 0.4 0.6 0.8 1.0
20 30 40 50 60
slenderness ratio (λ) t R,test / hour

Fig. 25. Comparisons of predicted fire resistances between the reduction method Fig. 26. Comparisons between predicted and tested fire resistances.
and the FE model results.

Comparisons of predicted fire resistances between the reduction shaped CFST columns tested in [38] were also predicted with
factor method and the FE model are shown in Fig. 25. The reduc- Eqs. (4) and (7), as shown in Fig. 26, where the FE model was
tion factor method is convenient for engineering application and employed to predict fire resistance under 4-sided fire exposure
can attain acceptable results. The fire resistances of the for the reduction method. Acceptable results can be obtained with
concrete-filled RHS columns tested in this study and square the proposed design methods.
 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004 2003

8
Although the proposed methods have been validated with both >
> ð0:004T þ 3:3Þ  106 0  C 6 T 6 650  C
>
>
the FE results and a few test data and are found to be reasonably < ð0:068T  38:3Þ  106 650  C < T 6 725  C
accurate, extensive validation is still required using further test qs c s ¼
>
> ð0:086T þ 73:35Þ  106 725  C < T 6 800  C
data. These equations may be used as an initial proposal at the >
>
:
stage of lacking sufficient experimental information. 4:55  106 T > 800  C
Coefficient of thermal expansion [m/(m °C)]:
8. Conclusions (
ð0:004T þ 12Þ  106 T < 1000  C
as
An experimental and numerical investigation into the fire per- 16  10 6
T  1000  C
formance of concrete-filled RHS columns subjected to 3-sided
exposure was conducted. Three-full scale concrete-filled RHS col- Stress–strain relations:
umns were tested to failure to explore the fundamental behaviour ( f ðT;0:001Þ
0:001 se es 6 ep
of these columns and provide data for the validation of FE models. rs ¼ f ðT;0:001Þ
A three-dimensional finite element model was developed with the 0:001 pe þ f ½T; ðes  ep þ 0:001Þ  f ðT; 0:001Þ es > ep
program ABAQUS to perform heat transfer and stress analyses. Fol- where ep = 4  106fy.
lowing validation of the models, parametric studies were carried pffiffiffiffiffiffiffiffiffiffiffiffiffi
out and two design methods were proposed for predicting the fire f ðT; 0:001Þ ¼ ð50  0:04TÞ  ½1  expð30 þ 0:03TÞ 0:001  6:9
resistance of concrete-filled RHS columns exposed to 3-sided fires.
The following conclusions can be drawn based on the studies
f ½T; ðes  ep þ 0:001Þ ¼ ð50  0:04TÞ  ½1  expð30 þ 0:03TÞ
performed: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 es  ep þ 0:001  6:9
(1) The shift of the centre of stiffness, together with asymmetric
thermal expansion (thermal bowing), which are particular
phenomena associated with columns exposed to asymmet- A.2. Concrete properties
ric fire conditions, promote global buckling. The collated
numerical analyses revealed that the effective centroid of Density:
the cross-section under 3-sided fire exposure typically
qc ¼ 2350 kg=m3
shifted between 0.025 times and 0.125 times the depth
towards the unexposed side. Thermal conductivity [W/(m °C)]:
(2) Parametric studies revealed that load ratio, cross-sectional 
1:355 0  C < T 6 293  C
dimensions and load eccentricity ratio have significant influ- kc ¼
ences on the fire resistance of concrete-filled RHS columns 0:001241T þ 1:7162 T > 293  C
exposed to 3-sided fires, while slenderness ratio, steel ratio, Specific heat [J/(kg °C)]:
material strength and depth to width ratio are less 8
important. >
> 2:566  106 0  C 6 T 6 400  C
>
>
(3) A simplified design method and a reduction factor method >
> ð0:1765T  68:034Þ  106 400  C < T 6 410  C
>
>
>
>
for predicting the fire resistance of concrete-filled RHS col- >
> ð0:05043T þ 25:00671Þ  106 410  C < T 6 445  C
>
>
umns exposed to 3-sided fires were proposed, the latter of >
< 2:566  106 445  C < T 6 500  C
which extends current fire safety design methods for con- qc c c ¼
crete-filled RHS columns in uniform fire to columns exposed > ð0:01603T  5:44881Þ  106
> 500  C < T 6 635  C
>
>
>
> ð0:16635T  100:90225Þ  106
to 3-sided fires. These methods may be used as an initial >
> 635  C < T 6 715  C
>
>
proposal at the stage of lacking sufficient experimental >
> ð0:22103T þ 176:07343Þ  106 715  C < T 6 785  C
>
>
information. >
:
2:566  106 T > 785  C

Density of water:
Acknowledgements
qw ¼ 1000 kg=m3
The research presented in this paper was sponsored by the Na-
Specific heat of water [J/(kg °C)]:
tional Natural Science Foundation (No. 51278153), a Special Fund
of Harbin Technological Innovation (No. 2012RFLXG025) and China qw cw ¼ 4:2  106
Scholarship Council (CSC); their financial support is highly
appreciated. Modified specific heat of concrete to account for moisture
evaporation:

Appendix A. Thermal and mechanical properties of materials 0:95qc cc þ 0:05qw cw T < 100  C
q0c c0c ¼
qc c c T P 100  C
A.1. Steel properties
Coefficient of thermal expansion [m/(m °C)]:
Density: ac ¼ ð0:008T þ 6Þ  106
qs ¼ 7850 kg=m3 Uniaxial stress–strain relations in compression:
8   2
Thermal conductivity [W/(m °C)]: >
>
> 0 emax ec
ec 6 emax
 < fc ðTÞ 1  emax
0:022T þ 48 0  C 6 T 6 900  C rc ¼ 
ks ¼ >  2
28:2 T > 900  C >
> e emax
ec > emax
: fc0 ðTÞ 1  c3emax
Specific heat [J/(kg °C)]:
2004 H. Yang et al. / Engineering Structures 56 (2013) 1986–2004

where emax ¼ 0:0025 þ ð6T þ 0:04T 2 Þ  106 . [19] Han LH, Yang YF, Xu L. An experimental study and calculation on the fire
resistance of concrete-filled SHS and RHS columns. J Construct Steel Res
8 0 2003;59(4):427–52.
< fc
> 0  C < T < 450  C
[20] Kodur VKR. Performance of high strength concrete-filled steel columns
0 T20
fc ðTÞ ¼ fc0 2:011  2:353 1000
450  C 6 T 6 874  C exposed to fire. Can J Civ Eng 1998;25(6):975–81.
>
: [21] Schaumann P, Kodur VKR, Bahr O. Fire behaviour of hollow structural section
0 T > 874  C steel columns filled with high strength concrete. J Construct Steel Res
2009;65(8):1794–802.
Uniaxial stress–strain relations in tension: [22] Romero ML, Moliner V, Espinos A, Ibañez C, Hospitaler A. Fire behavior of
8 axially loaded slender high strength concrete-filled tubular columns. J
> E ðTÞet et 6 etk Construct Steel Res 2011;67(12):1953–65.
< c
0:1f 0 ðTÞ [23] Lu H, Han LH, Zhao XL. Fire performance of self-consolidating concrete filled
rt ¼ ft0 ðTÞ  e t ðet  etk Þ etk < et 6 2etk double skin steel tubular columns: experiments. Fire Safety J
>
: tk
2010;45(2):106–15.
0:9ft0 ðTÞ et > 2etk [24] Han LH, Chen F, Liao FY, Tao Z, Uy B. Fire performance of concrete filled
stainless steel tubular columns. Eng Struct 2013;56:165–81.
where ft0 ðTÞ ¼ 0:09fc0 ðTÞ. [25] Ding J, Wang YC. Experimental study of structural fire behaviour of steel beam
to concrete filled tubular assemblies with different types of joints. Eng Struct
ft0 ðTÞ 2007;29(12):3485–502.
etk ¼ [26] Han LH, Wang WH, Yu HX. Experimental behaviour of reinforced concrete (RC)
Ec ðTÞ
beam to concrete-filled steel tubular (CFST) column frames subjected to ISO-
834 standard fire. Eng Struct 2010;32(10):3130–44.
[27] Huo JS, He YM, Chen BS. Experimental study on impact behaviour of concrete-
filled steel tubes at elevated temperatures up to 800 °C. Mater Struct 2013.
References
http://dx.doi.org/10.1617/s11527-013-0059-8.
[28] Lie TT, Chabot M. A method to predict the fire resistance of circular concrete
[1] Han LH, Yao GH, Tao Z. Behaviors of concrete-filled steel tubular filled hollow steel columns. J Fire Prot Eng 1990;2(4):111–28.
members subjected to combined loading. Thin-Wall Struct 2007;45(6): [29] Kodur VKR, Lie TT. Evaluation of fire resistance of rectangular steel columns
600–19. filled with fibre-reinforced concrete. Can J Civ Eng 1996;24(3):339–49.
[2] Ellobody E, Young B, Lam D. Behaviour of normal and high strength concrete- [30] Han LH. Fire performance of concrete filled steel tubular beam–columns. J
filled compact steel tube circular stub columns. J Construct Steel Res Construct Steel Res 2001;57(6):697–711.
2006;62(7):706–15. [31] Ding J, Wang YC. Realistic modelling of thermal and structural behaviour of
[3] Moon J, Roeder CW, Lehman DE, Lee HE. Analytical modeling of bending of unprotected concrete filled tubular columns in fire. J Construct Steel Res
circular concrete-filled steel tubes. Eng Struct 2012;42:349–61. 2008;64(10):1086–102.
[4] Pi YL, Liu CY, Bradford MA, Zhang SM. In-plane strength of concrete-filled steel [32] Hong S, Varma AH. Analytical modeling of the standard fire behavior of loaded
tubular circular arches. J Construct Steel Res 2012;69(1):77–94. CFT columns. J Construct Steel Res 2009;65(1):54–69.
[5] Wang YY, Geng Y, Ranzi G, Zhang SM. Time-dependent behaviour of expansive [33] Espinos A, Romero ML, Hospitaler A. Advanced model for predicting the fire
concrete-filled steel tubular columns. J Construct Steel Res response of concrete filled tubular columns. J Construct Steel Res
2011;67(3):471–83. 2010;66(8):1030–46.
[6] Yang H, Lam D, Gardner L. Testing and analysis of concrete-filled elliptical [34] Lu H, Zhao XL, Han LH. FE modeling and fire resistance design of concrete filled
hollow sections. Eng Struct 2008;30(12):3771–81. double skin tubular columns. J Construct Steel Res 2011;67(11):1733–48.
[7] Zhao XL, Packer JA. Tests and design of concrete-filled elliptical hollow section [35] Xu YY, Wu B. Fire resistance of reinforced concrete columns with L-, T-, and
stub columns. Thin-Wall Struct 2009;47(6):617–28. +-shaped cross sections. Fire Safety J 2009;44(6):869–80.
[8] Sheehan T, Dai XH, Chan TM, Lam D. Structural response of concrete-filled [36] Mao XY, Kodur VKR. Fire resistance of concrete encased columns under 3- and
elliptical steel hollow sections under eccentric compression. Eng Struct 4-side standard heating. J Construct Steel Res 2011;67(3):270–80.
2012;45:314–23. [37] Guo ZH, Shi XD. Behaviour of reinforced concrete at elevated temperature and
[9] Yu M, Zha XX, Ye JQ, Li YT. A unified formulation for circle and polygon its calculation. Beijing: Tsinghua University Press; 2003 [in Chinese].
concrete-filled steel tube columns under axial compression. Eng Struct [38] Yang H, Liu FQ, Zhang SM, Lv XT. Experimental investigation of concrete-filled
2013;49:1–10. square hollow section columns subjected to non-uniform exposure. Eng Struct
[10] Han LH, Wang WD, Zhao XL. Behaviour of steel beam to concrete-filled SHS 2013;48:292–312.
column frames: finite element model and verifications. Eng Struct [39] Liu FQ, Yang H, Zhang SM. Comparison of fire resistance of concrete-filled SHS
2008;30(6):1647–58. columns subjected to 3-sided and 4-sided exposure. In: 10th international
[11] Guo LH, Wang YY, Zhang SM. Experimental study of concrete-filled rectangular conference on advances in steel concrete composite and hybrid structures;
HSS columns subjected to biaxial bending. Adv Struct Eng 2012. p. 889–896.
2012;15(8):1329–44. [40] ISO 834-1. Fire resistance tests – elements of building construction – Part 1:
[12] Wang YY, Yang YL, Zhang SM. Static behaviors of reinforcement-stiffened general requirements. Switzerland: International Organization for
square concrete-filled steel tubular columns. Thin-Wall Struct 2012;58: Standardization; 1999.
18–31. [41] Iding RH. Calculating structural response to fire. Fire Protect Eng Soc Fire
[13] Ellobody E, Young B. Design and behaviour of concrete-filled cold-formed Protect Eng SP 2003:18–23.
stainless steel tube columns. Eng Struct 2006;28(5):726–8. [42] SIMULIA. ABAQUS/standard version 6.9 analysis user’s manual. Pawtucket,
[14] Yang YF, Han LH. Experiments on rectangular concrete-filled steel tubes Rhode Island: Hibbit, Karlsson & Sorenson, Inc.; 2009.
loaded axially on a partially stressed cross-sectional area. J Construct Steel Res [43] Lie TT. Fire resistance of circular steel columns filled with bar-reinforced
2009;65(8):1617–30. Concrete. J Struct Eng ASCE 1994;120(5):1489–509.
[15] Hass R. On realistic testing of the fire protection technology of steel and [44] Li LY, Purkiss J. Stress–strain constitutive equations of concrete material at
cement supports. Translations of BHPR/NL/T/1444, Melbourne, Australia; elevated temperatures. Fire Safety J 2005;40(7):669–86.
1991. [45] ASTM E119-88. Standard methods of fire tests on building construction and
[16] Lie TT, Chabot M. Experimental studies on the fire resistance of hollow steel materials. Philadelphia: American Society for Testing and Materials; 1990.
columns filled with plain concrete. NRC-CNRC internal report, No. 611, Ottawa, [46] EN1991-1-2. Eurocode 1: actions on structures – Part 1–2: general actions –
Canada; 1992. actions on structures exposed to fire. Brussels: European Committee for
[17] Chabot M, Lie TT. Experimental studies on the fire resistance of hollow steel Standardization; 2002.
columns filled with bar-reinforced concrete. NRC-CNRC internal report, No. [47] Lu H, Zhao XL, Han LH. Finite element analysis of temperatures in concrete
628, Ottawa, Canada; 1992. filled double skin steel tubes exposed to fires. In: 4th International structural
[18] Kodur VKR, Lie TT. Experimental studies on the fire resistance of circular engineering and construction conference; 2007. p. 1151–1156.
hollow steel columns filled with steel–fibre-reinforced concrete. NRC-CNRC
internal report, No. 691, Ottawa, Canada; 1995.