Journal of Constructional Steel Research 65 (2009) 17761786

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Nonlinear behavior of steel structures considering the cooling phase of a fire

K.H. Lien a,b , Y.J. Chiou a,c, , R.Z. Wang c , P.A. Hsiao d
a

Department of Civil Engineering, National Cheng Kung University, Tainan, 701, Taiwan, ROC

Department of Civil Engineering, Kao Yuan University, Kaohsiung, 821, Taiwan, ROC

National Center for Research on Earthquake Engineering, Taipei, 106, Taiwan, Roc

Department of Asset Management Science, Hsing-Kuo University of Management, Tainan, 709, Taiwan, ROC

article

info

Article history:
Received 25 July 2008
Accepted 24 March 2009
Keywords:
Vector Form Intrinsic Finite Element
(VFIFE)
Nonlinear analysis
Steel structures
Fire
Cooling effect

abstract
The nonlinear behavior of steel structures considering the heating and the cooling phases of a fire is
investigated by the Vector Form Intrinsic Finite Element (VFIFE) method. The temperature dependent
constitutive relations of steel which include strain reversal effects are adopted. The numerical model is
first verified by comparing the results with the published analytical and experimental results for steel
structures in the cooling phase. Several numerical examples are then fully studied to investigate the
cooling behavior of steel structures. The proposed numerical model can effectively predict the nonlinear
behavior of the steel structure in both the heating and cooling phases.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Steel structures have been widely used in high-rise buildings,
although their temperature sensitivity is a weakness. Since the
properties of steel, such as the Youngs modulus and yielding
strength, drop rapidly with increasing temperature, the loading
capacity of such structures in fires will reduce dramatically. The
large deformation induced by the thermal strain will result in
structural damage or collapse. A number of researchers [17]
adopted the conventional finite element method to investigate
the fire response of steel structures, studying the inelastic
large deformations of beams, columns and frames due to the
degradation of strength and thermal expansion. There are many
experimental results that have been adopted as design codes, such
as ECCS [8], Eurocode 3 [9], BS 5950-8 [10] and AISC [11]. However,
these studies focused on the structural behavior during the heating
phase, and ignoring the cooling process.
The previous researches predicted the maximum allowable
temperature of the structure under constant loading, and these
results are important in preventing the collapse of buildings in
a fire, as well as ensuring the safety of rescue workers in such
situations. However, it is also important to evaluate structural

Corresponding address: No.1, University Road, Tainan City 701, Taiwan, ROC.
Tel.: +886 6 2757575x63159; fax: +886 6 2358542.
E-mail addresses: [email protected] (K.H. Lien), [email protected]
(Y.J. Chiou), [email protected] (R.Z. Wang), [email protected]
(P.A. Hsiao).
0143-974X/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2009.03.015

damage after the fire and to draw up adequate repair plans. Steel
structures that experience a high temperature and then cool after
the fire will lose their original structural performance due to
residual stress and strain in some elements. A numerical model
for more accurate assessment is thus required to investigate the
structural behavior in the cooling phase.
The famous British Cardington test provides much experimental information about composite steel frames during the cooling
phase, such as the frame configuration, temperature distribution,
and structural behavior [1216]. In addition, Cong et al. [17] and Li
et al. [18] provided time histories of the cooling temperatures and
structural behaviors for steel beams and steel columns in fire resistant experiments, respectively. More recently, Li and Guo [19] performed experiments that examined the fire resistance of restrained
steel beams during the heating and cooling phases. With respect
to the numerical analysis, because strain reversal happens during
the cooling phase, cooling can be treated as an unloading behavior when considering thermal effects as the equivalent loading.
El-Rimawi et al. [20], Bailey et al. [21], Bailey [22], Iu et al. [23], Li
and Guo [24], and Guo and Li [25] studied the cooling behavior on
the basis of the model proposed by Franssen [26], which assumed
that the plastic strain of the material is irrelevant to step-wise temperature change.
The analysis of the fire response of steel structures frequently
requires the modeling of progressive failure and collapse, which
needs the consideration of large rigid body motions, large
deformations, and unbalanced applied forces. If the continuum is
subjected to a set of non-equilibrium external forces, the rigid body
motion is not zero and global equilibrium cannot be reached. In

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

time = t 2'

addition, if the order of magnitude of the rigid body component

is much larger than the deformation, errors in the calculation of
the internal forces may lead to numerical instability. There are
fundamental difficulties in using conventional finite elements to
treat large rigid body motion. Recently, Ting et al. [27,28] and Shih
et al. [29] proposed a vector form intrinsic finite element (VFIFE)
procedure. This procedure is designed to calculate the motions of
a system of rigid and deformable bodies, which may include large
rigid body motions and large geometrical changes.
This study adopts the VFIFE method to investigate the nonlinear
behavior of steel structures during the cooling as well as the
heating phases of fire. A two-dimensional frame element is
employed. The numerical model is first verified by comparing the
results with the published analytical and experimental results of
the steel structure during the cooling phase. Several numerical
examples are then fully studied to investigate the cooling behavior
of steel structures.
2. Fundamentals of Vector Form Intrinsic Finite Element
(VFIFE)
The VFIFE is a vector mechanics based mathematical calculation
method for structures with a large deformation. It is based on an
intrinsic finite element modeling approach, an explicit algorithm,
and a co-rotational formulation of kinematics [2730]. The VFIFE
maintains the intrinsic nature of the finite element method and
makes a strong form of equilibrium at the connection nodes of
members. All the forces balanced at each node are obtained from
the principle of virtual work. The VFIFE method adopts an explicit
solution procedure to avoid the difficulties that are caused by
iterations of material non-linearity and an incremental theory. All
the material properties, stress distribution, particle velocities and
geometry are defined from the calculation results of the previous
time step. The primary objective of this method is to handle the
motion and deformation of a system of multiple continuous bodies
and their interactions.
The VFIFE approach introduces a general description of particle
motion to handle large rigid body motion by incorporating
an adaptive convected material frame in the basic definition
of the displacement vector and strain tensor. The convected
material frame is a modification of the co-rotational approach. The
conventional co-rotational approach provides a simple kinematic
description. Its constitutive equation is expressed by total stress
and total strain, and it is well-suited to considering intermediate
large rotation [5,29,30]. For the large displacement study of frame
structures, small deformations superposed on large rotations are
commonly assumed.
In this study, the proposed method models the structure as a
system of discrete mass points, and these positions of mass points
characterize the shape and motion of the structure. The motion
of each mass point satisfies Newtons laws. The internal forces
induced by deformation and the external load are applied at each
mass point. The internal forces of the elements are then calculated
by using the convected material reference frame and fictitious
reversed rigid body motion [31]. A simple central difference
scheme is adopted to solve a set of equations of motions for each
particle. The rigid body motion and deformation displacement are
decoupled for each increment. By assuming a lumped mass matrix
of diagonal form, the explicit finite element analysis involves only
vector assemblage and vector storage.
Fig. 1 shows the schematic configuration of deformation of the
frame elements. The initial position of an element at time t0 is
assumed to be (10 , 20 ), the position at time ta = t is (1, 2),
and the current position of an element at time t is (10 , 20 ). The
displacement increment vectors 1d1 and 1d2 can be calculated
from the equations of motion with a reference coordinate at (1, 2).

1777

e
( d1)

1'

e
d1

d1

d2

(d2 d1)

1
2
time = ta = t -

10
y

d2

20 time = t0

Fig. 1. Schematic configuration of deformation of the frame elements.

The component of rigid body motion is decomposed from the

displacement increment vectors by the fictitious reverse rigid body
motion. First, the element (10 , 20 ) translates to make points 1 and 10
coincide, and then rotates an angle 1 to the direction of element
(1, 2). The deformation displacement increment vector 1dd is thus
expressed as

1dd = 1d 1dr

(1)

where 1d and 1d are the total displacement increment vector

and displacement increment vector induced by rigid body motion,
respectively. A two-node plane frame element has only three
independent variables of deformation displacement, and it is
expressed as
r

dTe = 1e

(2)

where 1e is the axial deformation increment, and 1 and 2 are

the changes of slopes of the nodes. The displacement field of the
BernoulliEuler frame element can be written as
ud = s1e [(1 4s + 3s2 )1 + (2s + 3s2 )2 ]y

(3a)

v = (s 2s + s )l1 + (s + s )l2

(3b)

where s = x/l, and l is the element length. The axial strain

increment in the frame element is

1 = Bde =

1
l

[1

(4 6s)y

( )
1e
(2 6s)y] 1 .
2

(4)

With regard to the thermal effect, the axial strain increment is

written as

1 = 1 + 1th

(5)

where 1 is the strain increment induced by stress, and 1th is

thermal strain increment.

1th = (Tt Ta )

(6)

where = (T ) is the thermal expansion coefficient, and Tt =

Tt (s, y) and Ta = Ta (s, y) represent the temperature distribution
at time t and ta , respectively.
The virtual internal energy increment of the planar frame is

Ue =

(1)T dV = (de )T fint

e

(7)

where and fint

e are the total stresses at time t and the internal
force of the element. The incremental internal force 1fint
e in each
time increment is written as

1fint
e =

BT 1 dV =

BT E 1 dV =

BT E (1 1th )dV (8)

1778

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

1.2
Stress-Strain curve at time ta,
Temperature = Ta < Tt

E
Equilibrium
position at
time t(point E)

ETa

'

ETa

Stress-Strain curve at time t,

Temperature = Tt > Ta

ETt

ETt

pl

C'

ETt

ETt

Stress-Strain curve at time ta,

Temperature = Ta > Tt

C'

(12)

where M is the lumped mass matrix of diagonal form, d is the

displacement vector, Fext is the external force vector, and Fint is
the internal force vector. The explicit time integration method
is adopted to solve these equations and a second order central
difference formulation which involves the vector operation only is
used as the time integration technique. The current displacement
is thus obtained as follows.

(without temperature change)

ETt
ETt

1200

= Fext Fint
Md

Equilibrium position
at time ta (point A)

ETa

1000

Using the principle of virtual work, the equation of motion for

the plane frame element is derived as [30]

(with temperature change)

E
B

400
600
800
Temperature (C)

Fig. 4. Reduction of elastic modulus and yield stress.

Stress-Strain curve at time t,

Temperature = Tt < Ta

'

pl

200

Equilibrium position
at time t (point E)
D

0.4

0.6

Fig. 2. Equilibrium positions for the heating phase.

ETa

Q235-sy(T)

0.8

0.2

> 0

EC3-E(T)
Q235-E(T)
EC3-sy(T)

(without temperature
change)
(with temperature
change)

Reduction factor

Equilibrium position
at time ta (point A)

di+1 = 1t 2 M1 (Fext
Fint
i
i ) + 2di di1 .

(13)

< 0

3. Mechanical properties of steel at elevated temperature

Fig. 3. Equilibrium positions for the cooling phase.

where E = E (j ) = (d /d)j = f 0 (j ) is the tangent modulus,

f () is a generalized nonlinear function used to derive the inelastic
constitutive relation.
The stressstrain behaviors for the cases of the heating and
cooling phases are shown in Figs. 2 and 3, respectively. On the
calculation of incremental internal force, the plastic strain is
assumed to be irrelevant to step-wise temperature change [26].
The final strain at the previous time ta is taken as the initial strain
for the equilibrium of the current time t.

(ta , Tt ) = (ta , Ta ).

(9)

Thus, the strain at time t with current temperature Tt is

(t , Tt ) = (ta , Tt ) + 1

(10)

where 1 is strain increment.

Referring to Figs. 2 and 3, point A is the final balance position
at previous time ta with temperature Ta . The stressstrain curve
of temperature Tt with plastic strain pl is thus represented by the
CDE curve in Figs. 2 and 3. The strain at point D is equal to the
strain at point A. Thus the stress increment induced by the strain
increment 1 is

1 = E A = D + 1 0 A .

(11)

The stress increment during heating with 1 < 0 and during

cooling with 1 > 0 can be calculated by a similar approach.

The variation of the mechanical properties of steel with temperature has been verified through extensive testing programs by
a number of researchers [32,33]. The focus of these investigations
was mainly on the stressstrain behavior of steel at elevated temperatures, the modulus of elasticity, yield stress, and coefficient of
thermal expansion. Numerous experimental studies on the variation of modulus of elasticity with temperature have been carried
out and a number of formulas published [34,35]. In general, the
value of the modulus of elasticity decreases gradually up to about
400 C and then drops rapidly to zero at approximately 1000 C.
The stressstrain relationships of steel at elevated temperature of
Eurocode 3 [9] and Q235 steel, the latter of which is commonly
used in Mainland China [36], are adopted in this study. The reduction factor for Youngs modulus E (T ) and yield stress of steel
y (T ) at different temperatures are presented in Fig. 4. However,
the available data for Q235 steel are only within 600 C, and thus
the material properties of this steel are assumed to be linear from
600 C to 1000 C. The modulus of elasticity and yielding strength
are both zero at temperature 1000 C. In calculating the thermal
strain, the coefficient of thermal expansion of steel is according
to EC3 Pt. 1.2 [9], which is taken as a constant at 1.4 105 / C.
The bilinear stressstrain relationship at elevated temperature
[4,37] is employed in this study, and the strain hardening parameter is assumed to be 1% unless otherwise specified.

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

1779

Load & Temperature history :

w = 30.6 kN/m

load

6m
F
E = 210 kN/mm2
time(sec)

y = 308 N/mm2

1
temperature

Section :35617151UB
Temperature distribution :

T1

T/2
T

T2
20

20

t1

t1+15

t2

time(sec)

Fig. 5. A simply supported beam subjected to thermal gradient.

4. Results and discussion

4.1. Cooling behavior of a simply supported beam

A simply supported beam with a heatingcooling phase is first
adopted to verify the proposed VFIFE approach, as shown in Fig. 5.
This beam consists of a 356 171 51 UB section with a span
of 6 m subjected to a uniform load of 30.6 kN/m. The beam is
first heated to 650 C, and then cooled to room temperature. The
temperature along the axial direction is assumed to be uniform,
while a non-uniform temperature distribution in the cross-section
is adopted, with the bottom and web at the assigned temperature
T and the top flange at half of this value. At ambient temperature,
the elastic modulus and yield stress are 210 GPa and 308 MPa,
respectively. El-Rimawi et al. [20] investigated this beam by using
the RambergOsgood and bilinear material models to describe
the steel behavior in the heating and cooling stages. Later, Bailey
et al. [21] also adopted the RambergOsgood model to investigate
the heating and cooling behavior, while Iu et al. [23] used the
material properties of Eurocode 3 Part 1.2 [9] and linear unloading
with constant plastic strain in the cooling phase to study this beam.
This study adopts the material properties of Eurocode 3 Part
1.2 and the bilinear material model, and the implementation of
applied load and temperature are shown in Fig. 5. In the heating
stage, the temperature is monotonically increased from ambient
temperature to temperature T1 , and then kept at a constant
value for time duration t1 to obtain a steady-state result. In the
cooling stage, the temperature is monotonically decreased from
temperature T1 to temperature T2 , and then kept at a constant
value for time duration t2 . The mid-span deflections of this simply
supported beam are presented in Fig. 6, in which it is seen that
as the temperature rises above about 500 C, the VFIFE numerical
results are close to earlier analyses. In the heating stage, the
deflections between the current analysis and the Iu et al. [23] are

-50
Mid-span deflection (mm)

This study concentrates on studying the nonlinear structural

behavior of steel structures during the cooling as well as the
heating phases of a fire. The numerical results are first compared
with the published analytical and experimental results for steel
structures during the cooling phase to verify the proposed
numerical model. Several examples are then thoroughly studied
to investigate the parameters which affect the cooling behavior of
such structures.

-100
El-Rimawi et al.(1996)
Bailey et al.(1996)
Iu et al.(2005)
VFIFE(present fire analysis)

-150
-200
-250
-300
-350
-400
-450
0

100

200

300
400
Temperature (C)

500

600

700

Fig. 6. Mid-span deflections of simply supported beam subjected to thermal

gradient.

consistent until 650 C. However, there are marked differences

between the VFIFE numerical result and the results of El-Rimawi
et al. [20] and Bailey et al. [21] until 600 C, since the material
properties are different at high temperatures. In the cooling stage,
the numerical results of this study show the same tendency as
those in earlier studies. The permanent deformation obtained by
this study agrees well with that of El-Rimawi et al. [20].
4.2. Fire resistant test of a simply supported beam
The simply supported beam consists of a H250 125 6 9
section with a span of 4.2 m subjected to four concentrated loads
of 10 kN at 1/8, 3/8, 5/8 and 7/8 of the span length, respectively,
as seen in Fig. 7. Cong et al. [17] carried out the fire resistant test
of this beam, and the temperature distribution and deformation
were measured, including the heating phase and the early stage of
the cooling phase, but the material properties were not presented.
Lateral buckling occurred at the end of their test. Although this
test was not designed to investigate the entire heating and cooling
effects, the information obtained in the early stage of the cooling
phase can be employed to verify the accuracy of the numerical
model proposed in this work.

1780

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

10 kN

10 kN

10 kN

10 kN
800
700

4.2 m

600
Temperature (C)

E = 210 kN/mm2
y = 330 N/mm2
: H25012569
Load & Temperature history :

500
400
300
Top flange

load

200

web
Bottom flange

100

0
1

time(sec)
Time (min)

Fig. 7. Geometric configuration and implementation of loading and heating for an axially unrestrained simply supported beam.

At ambient temperature, the tested yield stress of steel is

y = 330 MPa, and the elastic Youngs modulus is assumed to

4.3. Fire resistant test of a simply supported column

The simply supported column consists of a H244 175 7
11 section with a height of 3300 mm subjected to axial loads of

Mid-span deflection (mm)

-40
Fire test (Cong et al. , 2005)
E2=2,100 N/mm2 =0.01E1

-80
-120

E2=1,500 N/mm2
E2=1,000 N/mm2

-160
-200
-240
-280
-320
0

100

200
300
400
500
600
Temperature of bottom flange (C)

700

800

Fig. 8. Mid-span deflection histories for an axially unrestrained simply supported

beam during the heating and the early cooling stages.

45
Fire test (Cong et. al , 2005)
E2=2,100 N/mm2 =0.01E1

40
Axial displacement (mm)

be 210 GPa. The stressstrain relationships of steel at elevated

temperature of Q235 steel [36] are adopted in this example.
This study carried out a structural fire analysis according to the
measured temperatures at the top flange, the bottom flange, and
the web in the real fire test. During the analysis, the applied
loads on the beam were maintained constant after 10 s, while the
temperatures were increased according to the recorded data, as
shown in Fig. 7. The mid-span deflection and the axial deformation
at the roller end against temperature are presented in Figs. 8 and
9, respectively, in which it is found that the deflections obtained
by VFIFE are generally close to the test results. When the bottom
flange temperature was increased to above roughly 600 C, the rate
of deflection increased significantly. Alternatively, when the beam
was cooled, the deflections reversed gradually in the early stage
of the cooling phase. The steel frequently reached a plastic state
during the fire tests, and thus the material properties at elevated
temperatures have a considerable influence on the structural
behavior in the heating and cooling phases. The bilinear model was
assumed to model the Q235 steel material properties, and various
strain hardening parameters were adopted. In the heating stage
before 600 C, the strain hardening parameter seems to cause little
difference in the mid-span deflection and the axial deformation,
and the numerical results are almost consistent with the test
results of Cong et al. [17]. However, in the early cooling stage, Figs. 8
and 9 show significant differences in deflections between various
strain hardening parameters. As the strain hardening parameters
increased, the final deflections or the permanent deflections
decreased. In addition, the axially unrestrained simply supported
beam elongated in the heating stage, and the thermal expansion
effect might balance the shortening effect caused by the transverse
deflections. When the bottom flange temperature exceeded 650 C,
the axial shortening caused by the beam deflections was greater
than the elongation due to the thermal expansion effect. The strain
hardening parameters thus appeared to have significant influences
on the structural responses of the beam at the critical temperature
and in the cooling stage after the maximum temperature was
reached.

35

E2=1,500 N/mm2

30

E2=1,000 N/mm2

25
20
15
10
5
0
-5
0

100

200
300
400
500
600
Temperature of bottom flange (C)

700

800

Fig. 9. Axial deformation histories at the roller end for an axially unrestrained
simply supported beam during the heating and the early cooling stages.

308 kN, as seen in Fig. 10. Li et al. [18] carried out the fire resistance
test of this column, and lateral buckling occurred in the strong axis
as well as in the weak axis at the end of the test. The temperature
distribution and deformation of this simply supported column

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

1781

700
P=308kN
H244175711
Ttop

600
Temperature (C)

E = 210 kN/mm2
y = 285 N/mm2

3.3m

0
Tmiddle
Tbottom

Load & Temperature history :

load

500
400
300
200

Average Temperature
Top zone Temperature (Ttop )

100

Middle zone Temperature (Tmiddle )

Bottom zone Temperature (Tbottom )

Temperature distribution &

geometrical imperfection

0
1

time(sec)

10

15 20 25
Time (min)

30

35

Fig. 10. Geometric configuration and implementation of loading and heating for an axially unrestrained column.

4.4. Axial restraining effect on beams

The beam consists of a 356 171 51 UB section with a
span of 6 m subjected to a uniform load of 30.6 kN/m, as seen

205
225
Lateral deflection (mm)

200
175
150
125

Fire test (Li et al. , 2005)

E2=10,000 N/mm2

100

E2=10,100 N/mm2

75

E2=10,500 N/mm2=0.05E1

50
25
0
-25
0

100

200
300
400
500
Average temperature (C)

600

700

Fig. 11. Lateral deflection against temperature for simply supported column.

25
20
Axial displacement (mm)

were measured, including the heating phase and the early stage of
the cooling phase. The exposed column height was 3000 mm in the
furnace, and the temperature of the column was measured at three
cross-sections using a total of fifteen K -type thermo-couples. For
each cross-section, five thermo-couples were located on the inside
or outside of the flanges along the specimen at heights 610 mm,
1410 mm, and 2800 mm from the specimen base. The temperature
profile over the four-face heated cross-section was uniform in
each zone. However, along the axial direction the temperature
distribution appeared to have the highest temperature in the
middle zone, the next highest in the upper zone and the lowest
in the bottom zone of the specimen.
At ambient temperature, the tested yield stress of steel is
285 MPa, and the elastic Youngs modulus is assumed to be
210 GPa. The stressstrain relationships of steel at elevated
temperature of Q235 steel [36] are adopted in this example. To
simulate the buckling behavior of the steel column, the initial
geometry was prescribed as a sine curve with a geometrical
imperfection at the mid-length of the column taken as 1/1500
of the column length. A structural fire analysis was carried out
by adopting the measured temperatures in the real fire test.
During the analysis, the applied load was linearly increased to
the experimental load and maintained constant with a rise time
of 1 s and a total duration of 10 s, while the temperatures were
increased according to the recorded data, as shown in Fig. 10.
The deflection at height 1510 mm and the axial deformation at
the roller end against temperature are presented in Figs. 11 and
12, respectively, and the numerical results show an acceptable
agreement compared with the test results.
Referring to Fig. 11, the lateral deflections of the column can
be divided into three stages. There was nearly no lateral deflection
in the first pre-buckling stage, and then the lateral deflection
increased significantly in the second buckling stage. Finally, there
was permanent deformation in the third cooling stage. In the later
stage of buckling followed by the early cooling stage, Fig. 11 shows
that various strain hardening parameters can result in significant
discrepancies in permanent deflections. Alternatively, referring
to Fig. 12, axial deformation of the column can be divided into
expansion and contraction stages. The contraction stage occurred
almost simultaneously with the column buckling. With the same
strain hardening parameter, the thermal expansion coefficients
seemed to have a significant influence on the axial deformation.

15
10
5
0
Fire test (Li et al. , 2005)
Thermal expansion=1.4e-5 (E2=0.05E1)

-5

Thermal expansion=1.2e-5 (E2=0.05E1)

-10
0

100

200

300
400
500
Average temperature (C)

600

700

Fig. 12. Axial deformation against temperature for simply supported column.

in Fig. 13. At ambient temperature, the elastic Youngs modulus

is 210 GPa, and the yield stress is 308 MPa, the strain hardening
parameter is assumed to be 0.01, and the temperature distribution
is taken to be uniform. The material properties are assumed to
follow Eurocode 3 Part 1.2, and the thermal expansion coefficient is
taken as 1.4 105 / C [9]. Due to the extra axial reaction induced
by the restraint, the temperature at which the axially restrained

1782

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

6m

6m
Section :35617151UB

w = 30.6 kN/m
Temperature distribution :

E = 210 kN/mm2

y = 308 N/mm2

T
Fig. 13. Geometric configuration and implementation of loading and heating for axially restrained and unrestrained beams.
Temperature history : 20C - Tmax - 20C
Tmax=200C : Permanent deformation=61.6 (mm)

Temperature history : 20C - Tmax - 20C

Tmax=550C : Permanent deformation=0 (mm)
Tmax=575C : Permanent deformation=0.8 (mm)

Tmax=400C : Permanent deformation=69.6 (mm)

Tmax=600C : Permanent deformation=151.3 (mm)

Tmax=600C : Permanent deformation=130.4 (mm)

Tmax=610C : Permanent deformation=311.3 (mm)

Tmax=700C : Permanent deformation=228.4 (mm)

Tmax=800C : Permanent deformation=407.4 (mm)

Tmax=620C : Permanent deformation=487.3 (mm)

-100

-100

-200

Mid-span deflection (mm)

Mid-span deflection (mm)

-300
-400

-200
-300
-400
-500

-500

-600

-600
0

100

200

300
400
Temperature (C)

500

600

700

Fig. 14. Mid-span deflections of the simply supported beam cooling from different
maximum temperatures during the heating and cooling stages.

beam reached the plastic state was much lower than that of the
simply supported condition. This beam reached the plastic state
when the temperatures were about 568 C and 100 C for simply
supported and axially restrained conditions, respectively.
The mid-span deflections of the simply supported and the
axially restrained beams during the heating and cooling stages
are depicted in Figs. 14 and 15, respectively. In the former, it
is found that the residual deformation of the simply supported
beam increased with the maximum temperature and showed
great sensitivity in the rate of deflection when the maximum
temperature exceeded 575 C. In Fig. 15, the residual deformation
for the axially restrained beam also increased with the maximum
temperature. However, Fig. 15 also shows that the maximum
deflection and the residual deformation were lower than those for
the simple beam.
The axial reaction versus temperature during various cooling
phases in the axially restrained beam is presented in Fig. 16,
which shows that even in the case of the temperature history
cooling from 100 C, there was a significant residual axial force.
The variations of the axial reactions were closely related to the
combined effects of the temperature changes, vertical deflection,
and the material strength at the heating and cooling phases.
In the early heating stage, since the vertical deflection and the

100

200

300 400 500 600

Temperature (C)

700

800

900

Fig. 15. Mid-span deflections of the axially restrained beam cooling from different
maximum temperatures during the heating and cooling stages.

material degradation were not significant, the restraint mainly

resisted thermal elongation and the compressive axial force then
increased rapidly, reaching a maximum at about 100 C in this
study. When the temperature rose, although the axial elongation
still increased, the compressive axial force decreased. In addition,
the axial compressive reaction induced a second order p effect
and an increase in vertical deflection. The vertical deflection
decreased the axial compressive force and eventually the axial
reaction transferred to the tension state as the temperature
increased to 600 C. In the cooling phase, the axial reactions
were in a tension trend, since the restraints resisted the shorting
effect due to the decrease in temperature. The analytical results
show that once the beam reached a plastic state at the heating
phase, the residual reaction became tension when the beam
cooled to room temperature. The development of a residual
reaction is highly associated with the maximum temperature
that the beam has suffered. Referring to Fig. 16, it is found
that when the maximum temperature was 400 C, the axial
reaction resulted in the maximum residual tension forces. When
the maximum temperature was lower than 400 C, the residual
tension forces increased with increasing temperature. However,
when the maximum temperature was higher than 400 C, the
residual tension forces decreased with increasing temperature.
This was because that the beam lost little strength and the vertical

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

Temperature history : 20C - Tmax - 20C
Tmax =100C

1783

Temperature history :
Laterally unrestrained beam : 20C - 620C - 20C
Laterally restrained beam : 20C - 800C - 20C

Tmax =200C

Tmax =400C
Tmax =600C

Tmax =700C

-100
Mid-span deflection (mm)

Tmax =800C

2000

1750

1500

1250
Axial reaction (kN)

1000
750
500

-200
w

-300

-400
-500

250
0

-600

-250

-500

100

200

300 400 500 600

Temperature (C)

700

800

900

-750
-1000
0

100

200

300 400 500 600

Temperature (C)

700

800

900

Fig. 16. Axial reaction forces versus temperature during various cooling phases
from different maximum temperatures in the axially restrained beam.

deflection was not too large at temperatures below 400 C. The

residual tension forces at the supports were mainly caused by
the shortening effect of the members, and the vertical deflection
effect can be neglected. The structural behavior of the members
below 400 C remained as for the beam type. Once the maximum
heating temperature was higher than 400 C, the beam lost more
strength and the vertical deflection increased significantly. The
structural behavior of the members was like cables, and their
horizontal tension at the supports were inversely proportional
to the vertical deflection. As the members cooled from a higher
maximum temperature, the residual tension at the supports thus
became smaller. The shorting effect was not apparent, and can be
neglected compared to the vertical deflection effect. It should be
noted that the maximum temperature of 400 C was the result
of performing an analysis every 100 C in this study. When the
1 C temperature increment near 400 C was used to perform
the analysis, the maximum residual tension forces produced by
the axial reaction force after cooling happened at a maximum
temperature of 405 C.
Fig. 17 shows the comparison of temperaturedeflection curves
of axially restrained and unrestrained beams during the heating
and cooling stages. The first case is a simply supported beam
subjected to a maximum temperature Tmax = 620 C, and the
second one is the axially restrained beam subjected to a maximum
temperature Tmax = 800 C. Referring to Fig. 17, it is found that the
vertical deflection increased very slowly in the simply supported
beam before 500 C during the heating stage. However, when
the temperature was higher than 600 C, this beam reached a
plastic state and its material properties were severely degraded,
with the deflection becoming much larger. In the axially restrained
beam, the vertical deflection also increased very slowly before
100 C during the heating stage. As the temperature increased,
the axial reaction of the axially restrained beam will change from
compression to tension. When the axial reaction is compression,
the second order p effect caused a much larger deflection than
that in the simply supported beam. When the temperature was
higher than 600 C during the heating stage, the axial reaction of
the axially restrained beam changed from compression to tension,

Fig. 17. Temperaturedeflection curves of axially restrained and unrestrained

beams during the heating and cooling stages.

and the deflections became smaller than those in the simply

supported beam due to the catenary action.
The catenary action can prevent an axially restrained beam
from premature failure. The axially restrained beam can thus
undergo a much higher temperature and show an early-warning
deflection before collapse. However, the axially restrained beam
will reach the plastic stage at a lower temperature, and will
show substantial residual deformation or reaction when it is
cooled to ambient temperature. The magnitude of the residual
deflection or reaction will dictate whether the structure needs
repairing, strengthening, or dismantling for reconstruction. These
findings indicate that the members may reach a plastic stage
and cause great damage in the early stage of fires, even if the
fire is extinguished promptly. In engineering practice, new design
approaches can be suggested in order to adopt the optimum
stiffness of the restraints and to develop compromise approaches
so that the structure can avoid both collapse at high temperatures
and reaching the plastic stage in the early stage of fires.
4.5. Five-storey three-span frame subjected to localized fire
Fig. 18 shows a five-storey, three-span frame subjected to
localized fire. The material properties of Eurocode 3 Part 1.2 [9]
were adopted. It was assumed that only the elements near the fire
area were influenced by temperature, and that the temperature
was constant. The numerical results considering heating and
cooling are shown in Figs. 1921.
The relationship between joint displacement and temperature
is shown in Figs. 19 and 20. In Fig. 19, it is found that during
the temperature rising phase, the displacement of column top
U1 rose first and then shrank. Fig. 20 shows that the horizontal
displacement of the beam-to-column connection U5 was extended
out first, and then pulled inwards under the integrated effects
of element expansion due to temperature, the decrease in steel
strength, and the increase in vertical deflection of the beam. During
the cooling phase, Figs. 19 and 20 show that the curves of joint
displacements are nearly parallel with the heating response curve
and the permanent deformation increased with the increase of the
maximum historical temperature. Fig. 21 shows the deformation of
the frame under different temperatures in the fire. The shifts were
quite small compared to the size of the frame, so the shifts in the
figure were enlarged 10 times. The figure shows that the horizontal

1784

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

U1

U2

80
Heating
Cooling , Tmax=600C
Cooling , Tmax=700C
Cooling , Tmax=800C

Zone 2

Zone 1

Displacement of U5 (mm)

U3

4m

U5

U4

4m

4m

60
40
20
0
-20
-40
-60
4m

100

200

300 400 500 600

Temperature (C)

700

800

900

4m

Fig. 20. Column top horizontal displacement U5 against temperature.

T = 600C

Orginal Shape

T = 750C
T = 700C
T = 20 C (Cooling ,Tmax=750)

6m

w = 30.6 kN/m

6m

All beams : 35617151 UB

4m

6m

E = 210 kN/mm2
All columns : 305305137 UC

Fig. 18. Five-storey three-span frame subjected to fires.

56.9 mm
A

224 mm

C
4m

40

4m

y = 308 N/mm2

4m

0
Heating
Cooling , Tmax=800C
Cooling , Tmax=850C

-20

4m

Displacement of U1 (mm)

20

-40
-60

6m

-80
0

100

200

300 400 500 600

Temperature (C)

700

800

900

Fig. 19. Column top axial displacement U1 against temperature.

displacement at position A expanded out first and then pulled

inwards during the heating phase. In addition, the displacement at
the mid-point of beam BC was obviously larger than that of beam
AB at 700 C. Because the right side of beam BC was not affected
by temperature, its beam axial bonding stiffness was larger than
that of beam AB, resulting more obviously in the second p effect.
As the temperature continued rising, the strength of the steel
decreased and the vertical deflection increased. The behavior of the
beam was gradually closer to the behavior of the cable, and their
vertical deflections became very close at 750 C. Fig. 21 also shows
that the obvious permanent deformation still existed after cooling
to room temperature.
The analyses of structures in a fire at different times were
carried to consider the failure behaviors of different steel structure
members in a fire. In Fig. 18, the fire started at Zone 1 and spread

6m

6m

Fig. 21. Final deformed geometry of steel frame under different temperatures in
the fire (All deformation are in 10 times enlargement.).

to the left due to the fireproof area, and then spread to Zone 2 after
10 min. The temperature of the structural members in the fire and
the temperature of the fire scene were assumed to be in accordance
with the ISO 834 heating curve. In addition, the structure began
cooling 55 min after the fire started, and all members cooled to
20 C in 20 min. The temperature time histories of each area are
shown in Fig. 22, while the VFIFE analysis results are shown in
Figs. 2325.
Fig. 23 shows the overall displacement-time history curves of
each node of U1 U4 during the heating and cooling stages. The
analysis result shows the structure was permanently deformed
after cooling, with very little deformation recovery. Fig. 24 shows
the first 3000 s displacement-time history curves of each point
of U1 U4 . The figure clearly demonstrates the gradual structural
failure of each member in a fire. The analysis of the results
of this study show that beam failure due to large deformation
happened much earlier than column buckling. Since the fire started
in Zone 1 approximately 10 min earlier than in Zone 2, the

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

Progressive collapse of frame

Orginal Shape

Time (min)
0

10

20

30

40

50

60

70

1785

80

1000

T i me = 1500 sec

T i me = 2500 sec

900

T i me = 3000 sec

T i me = 3500 sec

Temperature (C)

800
700
600
500
400

Zone 1

300

Zone 2

200
100
0
0

500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Time (sec)

Fig. 22. Timetemperature relationship of each district for fire spread scene.

400
0
Fig. 25. Progressive deformation of steel frame subjected to fire spread.

Displacement (mm)

-400

rate increased slightly, but without losing control. Obviously, the

failure mechanism was not yet formed, and the structure kept
stable and loaded. Finally, the left column also started buckling
at 2400 s, and the failure mechanism was formed at this time,
with both U1 and U2 deforming rapidly. The structure would thus
partially collapse if it were not cooled. Fig. 25 shows the structural
deformation process under the fire effect, and also the sequence
of large beam and column deformation and the later rapid column
deformation due to buckling.

-800
-1200
-1600

U1
U2
U3
U4

-2000
-2400
-2800
-3200
0

500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Time (sec)

Fig. 23. Timedisplacement relationship for fire spread scene (05000 s).

Displacement (mm)

-100
-200
-300
-400

U1
U2
U3
U4

-500
-600
0

500

1000

1500
Time (sec)

2000

2500

3000

Fig. 24. Timedisplacement relationship for fire spread scene (03000 s).

beam in Zone 1 failed earlier than that in Zone 2 due to larger

deformations. However, the structure kept stable and was able to
bear loading. Under the temperature effect, the middle column
started buckling at 1500 s. At this time, the displacement increase

5. Conclusions
This study applied the Vector Form Intrinsic Finite Element
(VFIFE) method to investigate the nonlinear behavior of steel
structures considering the cooling as well as the heating phase
of fire. The numerical model was first verified by comparing the
results with the published analytical and experimental results
for steel structures during the cooling phase. Several numerical
examples were then fully studied to investigate the cooling
behavior of steel structures, and the proposed numerical model can
effectively predict the nonlinear behavior of such structures during
both heating and cooling phases.
The numerical results show that the boundary conditions,
temperature distributions, and properties of the high-temperature
steel will significantly affect the fire response of steel structures.
The steel beam without axial constraint will seriously distort
due to large displacement when the temperature reaches the
critical value. In contrast, the steel beam with axial constraint will
suppress the displacement effectively due to a catenary action.
However, structures with constraints in fires will induce thermal
stresses, and the members of those structures will reach the
plastic stage at lower temperatures when they are cooled to room
temperature. This phenomenon indicates that these structures will
probably reach the plastic phase in the early stage of a fire, and
induce a considerable permanent deformation that causes great
damage even after the fire is put out. Therefore, in future work it is
worth considering the feasibility of using adequate beam stiffness
design or other methods to avoid the structure reaching the plastic
phase in the early stage of a fire and large structural deformations
at high temperatures, which may affect escape and rescue.

1786

K.H. Lien et al. / Journal of Constructional Steel Research 65 (2009) 17761786

Acknowledgment
This study is supported by the National Science Council of
Republic of China under grant NSC93-2211-E-006-016.
References
[1] Liew JYR, Tang LK, Holmaas T, Choo YS. Advanced analysis for the assessment
of steel frames in fire. Journal of Constructional Steel Research 1998;47:1945.
[2] El-Rimawi JA, Burgess IW, Plank RJ. Studies of the behavior of steel sub-frames
with semi-rigid connections in fire. Journal of Constructional Steel Research
1999;49:8398.
[3] Zhao JC. Application of the direct iteration method for non-linear analysis of
steel frames in fire. Fire Safety Journal 2000;35:24155.
[4] Song L, Izzuddin BA, Elnashani AS, Dowling PJ. An integrated adaptive
environment for fire and explosion analysis of steel frames-part I: Analytical
models. Journal of Constructional Steel Research 2000;53(1):6385.
[5] Tan KH, Ting SK, Huang ZF. Visco-elasto-plastic analysis of steel frames in fire.
Journal of the Structural Engineering ASCE 2002;128(1):10514.
[6] Iu CK, Chan SL. A simulation-based large deflection and inelastic analysis of
steel frames under fire. Journal of Constructional Steel Research 2004;60:
1495524.
[7] Landesmann A, Batista EM, Alves JLD. Implementation of advanced analysis
method for steel-framed structures under fire conditions. Fire Safety Journal
2005;40:33966.
[8] ECCS. European recommendations for the fire safety of steel structures.
Technical committee 3; 1983.
[9] Eurocode 3. Design of steel structures. Part 1.2: General rules structural fire
design, ENV 1993-1-2. British Standard Institution; 2001.
[10] BS5950. Structural use of steelwork in building. Part 8. Code of practice for fire
resistant design. British Standard Institution; 2003.
[11] AISC. Specification for structural steel buildings. American Institute of Steel
Construction; 2005.
[12] Bailey C. Computer modeling of the corner compartment fire test on the largescale Cardington test frame. Journal of Constructional Steel Research 1998;48:
2745.
[13] Wang YC. An analysis of the global structural behaviour of the Cardington
steel-framed building during the two BRE fire tests. Engineering Structures
2000;22:40112.
[14] Elghazouli AY, Izzuddin BA. Analytical assessment of the structural performance of composite floors subject to compartment fires. Fire Safety Journal
2001;36:76993.
[15] Lennon T, Moore D. The natural fire safety conceptFull-scale tests at
Cardington. Fire Safety Journal 2003;38:62343.
[16] Gillie M, Usmani AS, Rotter JM. A structural analysis of the Cardington British
steel corner test. Journal of Constructional Steel Research 2002;58:42742.
[17] Cong SP, Liang ST, Dong YL. Experimental investigation of behavior of simple
supported steel beams under fire. Journal of Southeast University (Natural
Science Edition) 2005;35(Sup I):668.

[18] Li XD, Dong YL, Lv JL. Experimental investigation of the behaviors of axially
loaded H-section steel column under fire. Journal of Experimental Mechanics
2005;20(3):32834.
[19] Li GQ, Guo SX. Experiment on restrained steel beams subjected to heating and
cooling. Journal of Constructional Steel Research 2008;64:26874.
[20] El-Rimawi JA, Burgess IW, Plank RJ. The treatment of strain reversal in
structural members durig the cooling phase of a fire. Journal of Constructional
Steel Research 1996;37(2):11535.
[21] Bailey CG, Burgess IW, Plank RJ. Analyses of the effects of cooling and fire
spread on steel-framed buildings. Fire Safety Journal 1996;26:27393.
[22] Bailey CG. Development of computer software to simulate the structural
behaviour of steel-framed buildings in fire. Computers & Structures 1998;67:
42138.
[23] Iu CK, Chan SL, Zha XX. Nonlinear pre-fire and post-fire analysis of steel frames.
Engineering Structures 2005;27:1689702.
[24] Li GQ, Guo SX. Analysis of restrained steel beams subjected to temperature
increasing and descending Part I: Theory. Journal of Disaster Prevention and
Mitigation Engineering 2006;26(3):24150.
[25] Guo SX, Li GQ. Analysis of restrained steel beams subjected to temperature
increasing and descending Part II: Validation and parametrical analyses.
Journal of Disaster Prevention and Mitigation Engineering 2006;26(4):35968.
[26] Franssen JM. The unloading of building materials submitted to fire. Fire Safety
Journal 1990;16:21327.
[27] Ting EC, Shih C, Wang YK. Fundamentals of a vector form intrinsic finite
element: Part I Basic procedre and a plane frame element. Journal of Mechanics
2004;20(2):11322.
[28] Ting EC, Shih C, Wang YK. Fundamentals of a vector form intrinsic finite
element: Part II Plane solid elements. Journal of Mechanics 2004;20(2):
12332.
[29] Shih C, Wang YK, Ting EC. Fundamentals of a vector form intrinsic finite
element: Part III Convected material frame and examples. Journal of Mechanics
2004;20(2):13343.
[30] Wang YK. A general curved element for very flexible beam. Ph. D. dissertation.
Purdue University; 1996.
[31] Wang RZ. Vector form motion analysis of structure. Ph. D. dissertation.
Department of Civil Engineering, National Central University, Taiwan, ROC;
2005.
[32] Kirby BR, Preston RR. High temperature properties of hot-rolled structural
steels for use in fire engineering design studies. Fire Safety Journal 1987;13:
2737.
[33] Lie TT. Structural fire protection. ASCE manuals and reports on engineering
practice No.78; 1992.
[34] Brockenbrough RL. Theoretical stress and strains from heating curving. Journal
of Structural Division ASCE 1970;96:142144.
[35] Proe DJ, Bennetts ID, Thomas IR, Szeto WT. Handbook of fire protection
materials for structure steel. Australian Institute of Steel Construction; 1989.
[36] Zhang H. Studies of steel structure material mechanical performance under
the fire condition. Journal of the Chinese Peoples Armed Police Force Academy
2004;20(3):335.
[37] Izzuddin BA, Song L, Elnashani AS, Dowling PJ. An integrated adaptive
environment for fire and explosion analysis of steel frames- part II: Verification
and application. Journal of Constructional Steel Research 2000;53(1):87111.