Journal of Constructional Steel Research 71 (2012) 182–188

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Creep effect on buckling of axially restrained steel columns in real fires

Guo-Qiang Li a, b, Chao Zhang b,⁎
a
Sate Key Laboratory for Disaster Reduction in Civil Engineering, 1239 Siping Road, Shanghai, China
b
College of Civil Engineering, Tongji University, 1239 Siping Road, Shanghai 200092, China

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

Article history: At present, Harmathy creep model is used in most fire resistance analysis, which explicitly consider creep.
Received 21 April 2011 Harmathy creep model only predicts creep strains with acceptable accuracy for the case of constant stresses,
Accepted 5 September 2011 but becomes invalid for the case of variable stresses. For the case of axially restrained steel columns subjected
Available online 28 September 2011
to fire, the fire induced stresses vary considerably and rapidly with time and temperature. In this paper, the
effect of creep on the buckling behavior of axially restrained steel columns in real fires has been investigated.
Keywords:
Axially restrained steel columns
A creep model in ANSYS, which is capable of predicting creep strain regardless of any coupling between time
Buckling temperatures and either stress or temperature of steel, is used to predict creep strains. The results for buckling tempera-
Real fires tures and axial deformations, predicted with and without considering creep in numerical simulations, are
Creep model compared. Both fast and slow fires are considered. The study found that for axially restrained steel columns
Numerical study in slow fires, considering creep gives higher buckling temperatures than those not considering creep; and for
axially restrained steel columns in fast fires, considering creep might give higher or lower buckling temper-
atures than those not considering creep.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction In steel-framed buildings, the columns are restrained by the sur-

rounding structures. So far, the behaviors of restrained steel columns
Traditionally, the fire resistance of a structural component is de- in fire conditions have been investigated experimentally by Simms et
termined by prescriptive methods which are based on standard fire al. [3], Rodrigues et al. [4], Ali and O'Connor [5], Tan et al. [6] and Li
tests conducted on simply isolated members subjected to the stan- et al. [7], and numerically by Neves [8], Wang [9] and Franssen [10].
dard fire exposure such as ISO834 [1]. The standard fire tests have Both axial and rotational restraints have been separately or simulta-
been generally recognized to have many shortcomings, the two neously considered. It was found that axial restraint significantly re-
main aspects of which are on one hand the standard fire bears little duces the buckling temperature of heated columns, whereas
resemblance to a real fire and on the other hand the behavior of iso- rotational restraint increases the critical temperature (which is defined
lated members cannot represent the global behavior of structures in as the temperature at which the column internal axial force decreases to
fire condition. Thus, prescriptive methods cannot always predict the its initial value at ambient temperature). Besides, the column slender-
realistic behavior of the structural component coupled in a global ness ratio and load utilization factor are also critical factors. In those in-
structure and yield the suitable fire protection design. vestigations, however, only bare steel columns are used and the effect of
As an alternative, performance based method (PBM) has been heating rate has not been considered.
prompted by many countries for scientific fire safety design. PBM, In real buildings, the steel members are usually protected, and the
using state-of-the-art knowledge in structural fire engineering, has heating condition in potential real fires are different from that in labora-
its capacity of considering fire loads, ventilation and robust mechani- tory. Huang et al. [11] conducted a series of numerical studies to inves-
cal behavior of global structures in real fires [2]. In a performance- tigate the effect of heating rate on fire resistance of steel columns. A self-
based fire protection design, the thermal and mechanical behaviors developed FE program FEMFAN2D was applied for the analysis and
of building components exposed to design fire scenarios should be creep strain had been explicitly considered. Harmathy model [12] was
reasonably predicted. In most situations, post-flashover fires are used as the creep model in FEMFAN2D. Two constant heating rates,
used to represent design fires. namely 10 °C/min and 50 °C/min were considered. The study found
that increasing the heating rate will increase the column critical tem-
peratures. This is due to the fact that during the same temperature in-
⁎ Corresponding author at: College of Civil Engineering, Tongji University, Shanghai,
crement, the shorter the heating time, the smaller is the creep strain
200092, China. Tel.: + 86 21 65985318; fax: + 86 21 65983431. developed. The study also found that creep governs the behavior of
E-mail address: [email protected] (C. Zhang). steel columns beyond about 400 °C under general fire conditions.

0143-974X/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2011.09.006
 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188 183

Kodur and Dwaikat [13] studied the effect of high temperature creep on
the fire response of restrained steel beams. A validated model was ap-
plied to investigate the effect of load level, heating rate, fire scenario
and fire induced axial restraint on the extent of creep deformations.
Current high temperature creep models are compared with test data.
Harmathy creep model was found to be generally sufficient for reason-
ably predicting the creep deformations for unprotected steel members
but becomes inaccurate when high stress and temperature are main-
tained in steel members for long time durations. The results of the
study shows that neglecting high temperature creep in fire resistance
analysis of steel structures can lead to unconservative predictions.
This paper investigates the behavior of axially restrained steel col-
umns, which is the common case for building components, exposed to Fig. 1. Illustration of the HRR history in a NFSC fire.
real fires. Four different fire scenarios including two arbitrary fires and
two post-flashover fires are considered. The temperatures of steel The heat generation rate of the source is input as HRR specified by
members in post-flashover fires are obtained by solving the one zone NFSC. The compartment boundaries, and the insulation coatings of
compartment fire model with considering the heat sink effect of steel the steel members are modeled using LINK32. Convection and radia-
members. The structural behavior is modeled using a validated FE tion at fire or environment exposed surfaces are modeled using
model. The effect of creep on the buckling behavior of axially restrained LINK34 and LINK31, respectively. Gas and steel volumes are modeled
steel columns in real fires has been investigated. using MASS71. Radiation at opening is modeled using LINK31. Con-
vection at opening by mass exchange is modeled using LINK34.
2. Methodology LINK32, LINK31 and LINK34 are 2D conduction bar, 2D radiation
link and convection link elements, respectively; and MASS71 is a
2.1. Steel temperature in post-flashover fires point thermal mass element [26]. The FEM model has been verified
by program OZone [19], as shown in Fig. 3.
Post-flashover fires are commonly considered in fire resistance de-
sign. In literature, the temperature of steel members exposed to a 2.2. Creep model
post-flashover fire is usually determined by first modeling the fire phe-
nomenon by a empirical correlation (e.g. parameter fire model [14]) or Creep is the progressive time-dependent inelastic deformation
advanced computer simulation (e.g. Fire Dynamic Simulation [15]) to under constant load and temperature [21]. Generally, under constant
obtain a temperature–time curve to represent the fire environment, temperature and stress level, creep strain passes through three
then substituting the fire curve into a 1D condensed heat transfer phases, namely primary creep, secondary creep and tertiary creep.
model to obtain the steel temperatures [16,17]. The temperature of The tertiary creep stage is usually not analyzed since it implies
steel members in a fire can also be determined by advanced computer impending failure. Thus, in high temperature creep models, the pri-
simulations [18]. mary and secondary phases of high temperature creep strain are gen-
It is obvious that the steel members in a fire compartment will ab- erally used in the fire resistance analysis.
sorb a portion of the heat released by combustion. That portion of Creep strains are derived from steady state tests during which
heat will heat the steel members on one hand and cool the compart- stress is kept constant. There are usually two cases for calculating
ment on the other hand. As a result, the temperature of a steel member creep strains. The first case is when the stress state is close to a steady
within a fire compartment is dependent on the heating mechanism of state and the second case is when stress is changing with time. As for
the compartment. However, in current model as mentioned above, the first case, a “time hardening rule” is used where creep strain rate
the temperature of a steel member within a fire compartment is related is written as a function of time and stress in steel and is integrated
to the fire curve which is determined without considering the heat sink with respect to time; and for the second case, a “strain hardening
effect of the steel member. Here, the heat sink effect of steel members in rule” is used where creep strain rate is written as a function of
fire compartments has been considered by adding a quantity to the tra- creep strain and stress in steel and is integrated with respect to
ditional heat balance equation for one zone compartment model, thus time [13]. In calculation, creep can be considered implicitly and
explicitly.
HRR ¼q̇g þq̇w þq̇o;c þ q̇o;r þ q̇s ð1Þ At present, Harmathy creep model is used in most fire resistance
analyzes [11,22]. Harmathy creep model is mainly based on Dorn's
where HRR is heat release rate due to combustion; q̇ g is rate of theory, which assumes constant stress. However, in the case of axially
heat storage in the gas volume; q̇ w is rate of heat loss through the restrained steel columns subjected to fire, the fire induced stresses
walls, ceiling and floor; q̇ o;c is rate of heat loss due to replacement
of hot gases by cold; q̇ o;r is rate of heat loss by radiation through
the openings; and q̇ s is rate of heat storage in steel members. The nat-
ural fire safety concept (NFSC) [14,19] is used to represent the fire
conditions. The NFSC fire is assumed to be t-square in the growth
stage and decay stage begins at the time when 70% of design fire
load is consumed. Fig. 1 shows the HRR history in a NSFC fire. The
detail explanation and calculation of the remaining terms in Eq.(1)
can be found in [20].
Solve Eq.(1) by using technologies like finite differential method
(FDM) and finite element method (FEM), we can obtain both gas
and steel temperatures in fire compartments. Here, the FEM software
package ANSYS is employed to simulate the model. Fig. 2 shows the
FEM thermal model. In the FEM model, the heat source is modeled
by a perfect conductor, which is represented by one LINK32 element. Fig. 2. Illustration of the FEM thermal model for heat transfer analysis.
184 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188

theory on plasticity and creep strains, can model both the primary
and secondary creep, and is capable of predicting creep strain regard-
less of any coupling between time and either stress or temperature of
steel. The model is given in a simplified format, as

c1 σsc2 t c3 þ1 e−c4 =Ts c −c =T

εcr ¼ þ c5 σs 6 te 7 s ð2Þ
c3 þ 1

where, εcr is equivalent creep strain; σs is steel stress in MPa; Ts is steel

temperature in °C; and c1 through c7 are coefficients obtained through
nonlinear regression analysis to high temperature creep test data.
In [13], c1 = 6 × 10− 6 per min, c2 =6.95, c3 = − 0.4, c4 = 16500°C,
Fig. 3. Verification of the FEM model by Ozone.
c5 = 6.0 × 10− 6 per min, c6 = 6 × 10− 5 and c7 = 5 × 10− 3°C, here, t is
time at end of sub step in min.

vary considerably and rapidly with time and temperature, and these 2.3. Temperature-stress–strain models
conditions are not captured by Dorn's theory. Kodur and Dwaikat
[13] compared creep strains predicted using Harmathy model with Temperature-stress–strain curves, for structural steel, given by
test data for different cases of constant and variable stresses, as EC3 [23], Lie [24] and Poh [25] are widely used in fire resistance sim-
shown in Fig. 4. For the case of constant stress and temperature, ulations. In EC3 constitutive model, creep is implicitly included, while
Harmathy creep model predicts creep strains within acceptable accu- in Lie and Poh models, creep is excluded.
racy. However, for the case of varying stress with time, Harmathy High temperature creep is highly nonlinear transient phenome-
creep model diverges from the test data. The creep strains predicted non and creep characteristics change with time as stress and temper-
using ANSYS are also plotted in Fig. 4. The ANSYS creep model used ature vary significantly with time too. Therefore, it is impossible to
in [13] predicts creep strains within acceptable accuracy for both fully include the effect of creep in static-strain curves that do not in-
cases. corporate the time factor (whether rates or absolute times). In [13],
In our investigation, the ANSYS creep model used in [13] is temperature-stress–strain relations recommended by Poh was used
adopted. The model, namely implicit creep model 11 in ANSYS for the analysis.
which is based on work of Zienkiewicz and Cormeau on unified
2.4. FEM structural model

In this paper, the FEM program ANSYS [26] is employed as the nu-
merical tool for both thermal and structural analysis. The FEM model
for structural analysis is given below.
Fig. 5 shows the FEM structural model of axially restrained col-
umns. The steel column is modeled using 3D linear finite strain
beam element, BEAM188. BEAM188 is based on Timoshenko beam
theory and is suitable for analyzing slender to moderately stubby/
thick beam structures. The axial restraint is modeled by an axial
spring using spring-damper element, COMBIN14.
The initial column crookedness is assumed to be a half sine wave.
The initial deflection amplitude at mid-height, if not specified, is
taken as 0.1% of the column length. The load utilization ratio, βL in
Fig. 5, is defined as

P0
βL ¼ ð3Þ
Pcr

where, P0 is the applied axial compression force at room temperature;

and Pcr is the column buckling resistance determined from Chinese
structural design code GB50017 [27]. The axial restraint ratio is
given by

ka
βa ¼ ð4Þ
kc0

Fig. 4. Creep strains predicted using ANSYS and Harmathy models compared to test
data, given by Kodur and Dwaikat [13]. Fig. 5. Illustration of the FEM model of axially restrained steel column.
 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188 185

where, ka is the stiffness of axial restraint; and kco = E0A/lc is the Table 1
column axial stiffness at room temperature, in which E0 is elastic Comparison between the predicted and measured results for TP112 test in [3].

modulus of steel at room temperature, and lc is column length. Name Failure temperature Restraint force Maximum reaction force

°C kN kN

Test data 233 12.36 57.19

3. Case studies
Lie-creep 233 13.26 55.96
Lie 238 14.44 57.14
3.1. Validation of the structural model EC3 204 13.07 55.77

The structural model used in the following studies is validated

using test data from Lie and Macaulay [24] and Simms et al. [3].
Lie and Macauly reported test on fire resistance of protected steel 3.2. Different fires
columns. Test on column No.1 in [24] is used for validation. The tested
column was of cross section W10×60 and had length of 3810 mm. As given in Table 2, two arbitrary fires and two real fires are con-
The column was uniformly insulated by 55 mm sprayed mineral sidered in case studies. In arbitrary fires, the steel temperatures are
fiber. End condition was fixed-fixed. The applied axial load was linearly increased with specified constant rates. The steel tempera-
1760 kN. Yield stress and elastic modulus of the steel at room temper- ture rising rates (eTs/△ t) in arbitrary fast fire (Fire1) and arbitrary
ature was 300 MPa and 2 × 10 5 MPa, respectively. The furnace heating slow fire (Fire2) are taken as 100 and 1 °C/min, respectively. The
condition followed ASTM-E119 and the temperature rising rate of real fires include a fast and slow NFSC fire in a compartment of 3 m
steel changed from 0.57 to 5.15. Fig. 6 shows the results for the width, 4 m depth and 2.7 m height. For the fast NFSC fire (Fire3),
axial deformations predicted by FEM using different constitutive the compartment has a 1 m 2 square vent and the maximum heat re-
models. EC3 model [23] gives better prediction than Lie model [24]. lease rate of the fire is 1.5 MW. For the slow NFSC fire (Fire4), the
However, when Lie model combined with ANSYS implicit creep compartment has a rectangle vent of 0.5 m width and 1 m height
model, the combined model gives better prediction than EC3 model. and the maximum heat release rate of the fire is 0.75 MW. The floor
Constitutive models given by Lie [24] and Poh [25] give similar pre- fire load densities are taken as 1000 MJ/m 2 and 1500 MJ/m 2 for the
dictions. Considering the expression for Poh model is complex, in fast and slow NFSC fires, respectively. The gas temperature–time
the following studies Lie model is used. curves for Fire3 and Fire4 are shown in Fig. 7.
Simms et al. [3] reported test on structural performance of axially The cross section of the steel columns in Fire1 to Fire3 is
restrained steel columns subjected to elevated temperatures. Test H300 ×300× 10 ×15, and that of the steel column in Fire4 is
case TP112 in [3] is used for validation. The tested column was of H200 ×200× 8× 12. The steel columns in different fires have the same
cross section IPE80imes46 and had length of 1600 mm. End condition height of 2.7 m, which is the height of the compartment for Fire3 and
about the minor axis was pinned-pinned. The load utilization ratio Fire4. The steel columns in Fire3 and Fire4 are insulated with 10 mm
was 0.61 and the axial restraint ratio was 0.042. The measured yield and 20 mm SFRMs, respectively. The properties of SFRM are
stress and elastic modulus of the steel at room temperature was ki = 0.12 W/mK, ρi =250 kg/m3, and ci =800 J/kgK. The temperature of
320 MPa and 2.05 × 10 5 MPa, respectively. The furnace heating rate the steel columns in different fires are shown in Fig. 7. Fig. 8 shows the
was 10 °C/min. The steel temperatures was not recorded in the test. development of the steel temperature rising rate in Fire3 and Fire4. The
In FEM predictions, the temperature rising rate of the steel column maximum rate in Fire3 and Fire4 are 7.5 and 3.3 °C/min, respectively.
is taken as equal to the heating rate of the furnace. Table 1 gives The yield stress and elastic modulus of the steel at room temperature is
some results predicted by FEM using different constitutive models in- 235 MPa and 2 ×105 MPa, respectively.
clude EC3 model, Lie model and combined Lie-creep model. The test
data are also presented for comparison. Lie-creep model gives perfect 3.3. βL and βa
prediction, while Lie model alone gives very good but slightly higher
prediction of the failure temperature. The over-prediction of the fail- The load utilization ratio of 0.3, 0.5 and 0.8 are considered. The in-
ure temperature by Lie model is very small (only 5 °C), because the vestigated axial restraint ratio include 0, 0.02, 0.06 and 0.1.
creep effect for the investigated case with fast heating rate (10 °C/
min), short time duration (about 25 min) is small. EC3 model gives 4. Results and discussions
under-prediction of failure temperature, which might due to EC3
model considers “proportionality limit” as an end of the linear phase. 4.1. Results

For steel columns in Fire1, the results predicted by using Lie-creep

and Lie model are consistent, as shown in Fig. 9. In Fire1, the tempera-
ture rising rate of the steel columns is very fast (up to 100 °C/min) that
the columns reach their buckling temperatures in a very short time
(only several minutes). As a result, the creep strain developed in steel
columns in Fire1 is very small, and the effect of creep on the behavior
of steel columns is ignorable.

Table 2
Different fires in case studies.

Name Description Compartment Vent Maximum heat release rate

m× m × m m×m MW

Fire1 Arbitrary fast fire – – –

Fire2 Arbitrary slow fire – – –
Fire3 Real fast fire 3 × 4 × 2.7 1×1 1.5
Fig. 6. Measured and predicted results for axial deformation for column test No. 1 in Lie
Fire4 Real slow fire 3 × 4 × 2.7 0.5 × 1 0.75
and Macaulay [24].
186 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188

Table 3
Predicted failure temperatures in Fire2.

βL σ−ε βa

0 0.02 0.06 0.1

0.3 Lie-creep 635.0 586.1 516.5 467.3

Lie 635.8 571.6 500.1 379.3
0.5 Lie-creep 509.6 474.7 418.8 370.5
Lie 510.3 449.1 394.3 340.2
0.8 Lie-creep 294.9 264.8 216.5 185.9
Lie 296.9 238.4 188.4 151.6

Fig. 7. Steel and gas temperatures in different fires.

Fig. 10. Axial deformations predicted by using Lie-creep and Lie model for βL =0.5 in Fire2.

cases, Lie-creep model gives prediction of smaller axial deformations

than Lie model, as shown in Fig. 11.
Table 5 gives the results for the buckling temperatures of steel col-
umns in Fire4, predicted by using Lie-creep and Lie model. Fig. 12
Fig. 8. Steel temperature rising rates in real fires. gives some results for axial deformations. For free steel columns
(βa = 0), the buckling temperatures predicted by using different
models are nearly consistent. For axially restrained steel columns,
Table 3 gives the results for the buckling temperatures of steel col- Lie-creep model gives higher buckling temperatures than Lie model.
umns in Fire2, predicted by using Lie-creep and Lie model. Fig. 10 For all columns, Lie-creep model gives prediction of smaller axial de-
gives some results for axial deformations. For free steel columns formations than Lie model, as shown in Fig. 12.
(βa = 0), the buckling temperatures predicted by using different
models are nearly consistent. For axially restrained steel columns, 4.2. Discussions
Lie-creep model gives higher buckling temperatures than Lie model.
For all columns, Lie-creep model gives prediction of smaller axial de- At high temperature, the total strain of steel is composed of [28]
formations than Lie model, as shown in Fig. 10.
Table 4 gives the results for the buckling temperatures of steel col- εtot ¼ εmec ðσ; T Þ þ εth ðT Þ þ εcr ðσ; T; t Þ ð5Þ
umns in Fire3, predicted by using Lie-creep and Lie model. The values
in bracket are time at buckling. Fig. 11 gives some results for axial de- where, εmec is mechanical strain; and εth is thermal strain. For
formations. For free steel columns (βa = 0), the buckling tempera- compressed steel columns in fire, the sign of εmec is consistent with
tures predicted by using different models are nearly consistent. For that of εcr, and opposite to that of nth.
steel columns with βa = 0.02, Lie-creep model gives higher buckling For compressed members in fire, creep has effect of reducing the
temperatures than Lie model; however, for steel columns with elongation strain or increasing the compressive strain. For free steel col-
βa = 0.1, Lie-creep model gives lower buckling temperatures. For all umns on which the axial forces are constant during the whole fire expo-
sure time, due to creep induced compressive strain, considering creep
will give slightly lower buckling temperatures, as shown in Tables 3–5.
However, for axially restrained steel columns, the restrained elongation
strain will induce additional compressive force at the restraint. The

Table 4
Predicted failure temperatures in Fire3.

βL σ−ε βa

0 0.02 0.06 0.1

0.3 Lie-creep 654.2 (132.9) 577.4 (105.6) 481.2 (80.7) 412.4 (65.4)
Lie 655.9 (133.9) 570.4 (103.5) 520.9 (90.4) 466.6 (77.4)
0.5 Lie-creep 529.0 (92.5) 473.2 (78.9) 382.2 (59.4) 323.4 (48.1)
Lie 530.2 (92.8) 469.2 (78.0) 373.1 (57.6) 360.5 (55.2)
0.8 Lie-creep 314.3 (46.5) 261.6 (37.4) 197.4 (27.2) 162.9 (22.1)
Fig. 9. Axial deformations predicted by using Lie-creep and Lie model for βL = 0.5,
Lie 316.9 (47.0) 258.0 (36.8) 213.0 (29.6) 176.8 (24.1)
βa = 0.1 in Fire1.
 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188 187

Fig. 11. Axial deformations predicted by using Lie-creep and Lie model for βL =0.5 in Fire3. Fig. 13. Calculated creep and thermal strain for case βL = 0.3, βa = 0 in Fire2.

Overall, for weakly restrained steel columns (β = 0.02), considering

creep always gives higher bucking temperatures.
Table 5
Predicted failure temperatures in Fire4.
5. Conclusions
βL σ−ε βa

0 0.02 0.06 0.1 Using a creep model which can predict creep strains with acceptable
0.3 Lie-creep 653.3 (374.4) 590.9 (302.1) 500.8 (223.0) 436.6 (178.2)
accuracy for conditions with constant and variable stresses, the buckling
Lie 661.8 (389.4) 574.3 (285.9) 473.9 (203.3) 431.7 (175.2) behavior of axially restrained steel columns in different fire scenarios has
0.5 Lie-creep 536.1 (251.5) 486.3 (212.0) 402.8 (157.9) 343.6 (126.2) been investigated. Based on the results of this study, the following con-
Lie 537.6 (252.7) 475.6 (204.5) 379.7 (144.9) 319.2 (114.5) clusions can be drawn:
0.8 Lie-creep 330.6 (120.0) 285.8 (99.6) 219.2 (72.6) 182.4 (59.1)
Lie 333.2 (121.2) 277.8 (96.0) 207.4 (68.2) 169.7 (54.8)
• Creep has effect on the buckling temperature of axially restrained
steel columns in real fires. For axially restrained steel columns in
slow fires, considering creep will give higher buckling temperatures
more the restrained elongation, the larger is the additional compressive than those not considering creep. For steel columns in fast fires, con-
force. As a result, considering creep will give smaller additional com- sidering creep might give higher or lower buckling temperatures.
pressive force which might lead to prediction of higher buckling tem- • For steel columns in real fires, considering creep will give smaller
peratures. It should be noted that the converse effect of creep, namely axial deformations than those not considering creep.
“stress relaxation” [26], will reduce the yield stress of steel which also • For weakly restrained steel columns, considering creep will give
leads to prediction of lower buckling temperatures. higher bucking temperatures than those not considering creep;
For axially restrained steel columns in slow fires like Fire2 and and for free steel columns, considering creep will give slightly
Fire4, the steel columns reach buckling after a long time that the de- lower buckling temperatures than those not considering creep.
veloped creep strains in columns are comparatively large, as shown in
Fig. 13. The creep strain in Fig. 13 is calculated by Eq. (2). Correspond- References
ingly, considering creep will reduce the additional compressive force
considerably which gives higher buckling temperatures, as shown in [1] BS 476–20. British Standard BS 476, Part 20: Method for determination of the fire
resistance of load bearing elements of construction (General Principles). London:
Tables 3–5. For axially restrained steel columns in fast fires like British Standard Institution; 1987.
Fire3, at buckling the developed creep strains in columns are compar- [2] Zhang C, Li GQ, Yin YZ, Luo MC. Fire resistance design of large space grid structures by
atively small. If the positive effect of creep of reducing the additional performance-based approach - a case study of the fire resistance design of the roof
structure of Kunming International Airport. Proceedings of the Sixth International
compressive force is greater than the negative effect of creep of in- Conference on Advances in Steel Structures, HongKong; 2009. p. 776–85.
creasing the compressive strain, considering creep will give higher [3] Simms WI, O'Connor DJ, Ai F, Randall M. An experimental investigation on the
buckling temperatures, as founded in the cases with βa = 0.02 in structural performance of steel columns subjected to eleveated temperatures.
Journal of Applied Fire Science 1996;5:269–84.
Table 4. Conversely, if the positive effect is less than the negative ef- [4] Rodrigues JP, Neves IC, Valente JC. Experimental research on the critical temperature
fect, considering creep will give lower buckling temperatures, as of compressed steel elements with restrained thermal elongation. Fire Safety Journal
founded in the cases with βa = 0.1 in Table 4. 2000;35:77–98.
[5] Ali F, O'Connor D. Structural performance of rotationally restrained steel columns
in fire. Fire Safety Journal 2001;36:679–91.
[6] Tan KH, Toh WS, Huang ZF, Phng CH. Structural responses of restrained steel columns
at elevated temperatures. Engineering Structures 2007;29:1641–52.
[7] Li GQ, Wang PJ, Wang YC. Behaviour and design of restrained steel column in fire,
Part 1: Fire test. Journal of Constructional Steel Research 2010;66:1138–47.
[8] Neves IC. The critical temperature of steel columns with restrained thermal elongation.
Fire Safety Journal 1995;24:211–27.
[9] Wang YC. Postbuckling behavior of axially restrained and axially loaded steel columns
under fire conditions. Journal of Structural Engineering 2004;130:371–80.
[10] Franssen JM. Failure temperature of a system comprising a restrained column
submitted to fire. Fire Safety Journal 2000;34:191–207.
[11] Huang ZF, Tan KH, Ting SK. Heating rate and boundary restraint effects on fire resistance
of steel columns with creep. Engineering Structures 2006;28:805–17.
[12] Harmathy TZ. A comprehensive creep model. Journal of Basic Engineering
1967;89:469–502.
[13] Kodur VK, Dwaikat MM. Effect of high temperature creep on the fire response of
restrained steel columns. Materials and Structures 2010;43:1327–41.
[14] EC1-1-2. Eurocode 1: Actions on structures - Part 1–2: General actions - Actions
Fig. 12. Axial deformations predicted by using Lie-creep and Lie model for βL =0.5 in Fire4. on structures exposed to fire. British Standard Institution (BSI); 2002.
188 G.-Q. Li, C. Zhang / Journal of Constructional Steel Research 71 (2012) 182–188

[15] McGrattan K, McDermott R, Hostikka S, Floyd J. Fire Dynamics Simulator (Version [21] Naumenko K, Altenbach H. Modeling of creep for structural analysis. Springer;
5) User's Guide. NIST Special Publication 1019–5. National Institute of Standards 2007.
and Technology (NIST); 2010. [22] Zeng JL, Tan KH, Huang ZF. Primary creep buckling of steel columns in fire. Journal
[16] Li GQ, Zhang C. Thermal response to fire of uniformly insulated steel members: of Constructional Steel Research 2003;59:951–70.
background and verification of the formulation recommended by Chinese code [23] Eurocode 3: Design of steel structures - Part 1–2: General rules - Structural fire
CECS200. Advanced Steel Construction 2010;6:788–802. design. BSI; 2005.
[17] Zhang C, Li GQ, Wang YC. Sensitivity study on using different formulae for calculating [24] Lie TT, Macaulay BA. Evaluation of the fire resistance of protected steel columns.
the temperature of insulated steel members in natural fires. Fire Technology 2011, doi: Internal Report No. 583, National Research Concil Canada; 1989.
10.1007/s10694-011-0225-x. [25] Poh KW. Stress–strain-temperature relationship for structural steel. Journal of
[18] Li GQ, Zhang C. Thermal response of steel columns exposed to localized fires - Materials in Civil Engineering 2000;13:371–9.
numerical simulation and comparison with experimental results. Proceedings [26] ANSYS user manual book. 2005
of the Sixth International Conference on Structures in Fire, MI, USA; 2010. [27] GB50017. Code for design of steel structures. China Planning Press; 2003.
p. 35–42. [28] Anderberg Y. The impact of various material models on structural fire behaviour
[19] Cadorin JF, Franssen JM. A tool to design steel elements submitted to compartment prediction. Proceedings of the Fifth International Conference on Structures in
fires - OZone V2. Part 1: pre- and post-flashover compartment fire model. Fire Safety Fire, Singapore; 2008. p. 253–65.
Journal 2003;38:395–427.
[20] Drysdale D. An introduction to fire dynamics. 2nd edition. John Wiley and Sons;
1999.