Fire Safety Journal 54 (2012) 113–120

Contents lists available at SciVerse ScienceDirect

Fire Safety Journal

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

A new temperature–time curve for fire-resistance analysis of structures

Yong Du a,n, Guo-qiang Li b
a
Nanjing University of Technology, College of Civil Engineering, 200# Zhongshan North Road, Nanjing 210009, China
b
Tongji University, Shanghai, China

a r t i c l e i n f o abstract

Article history: To evaluate the fire-resistance of large space buildings, smoke temperature distribution in large space
Received 27 January 2011 fires is one of the required conditions for determining the temperature elevation history in structures.
Received in revised form However, full scale fire experiments on large space building in State Key Laboratory of Fire Science in
25 December 2011
China revealed that the fire temperature distribution in large space fires are different with that in small
Accepted 26 July 2012
compartment fires. In order to standardize the hot smoke temperature in localized fire, a number of
Available online 8 September 2012
ventilated fire scenarios are designed with ceiling height ranging 4–20 m high, floor area ranging 500–
Keywords: 6000 m2 and the rate of heat release ranging 2–25 MW. A wide range of result from a series of large
Temperature–time curve space fire scenarios simulated by FDS (Fire Dynamics Simulator), revealed three important facts. Firstly,
Fire-resistance
temperature distributions throughout the large space fire are non-uniform and pole asymmetric from
Temperature
the fire source. Secondly, the key factors i.e. fire growth type, heat release rate, dimension of internal
Localized fire
space and fire area have significant influence on the temperature distribution. Thirdly, with the increase
of heat release rate, the maximum temperature is higher and with the increase of the floor area or the
internal space height, the maximum temperature is lower. The temperature decays from the plume
centerline in horizontal plane. Finally, a new parameter equation that represents the non-uniform fire
temperature distribution as a function of time and key factors has been developed by parametric
analysis method based on results from numerical tests by using FDS. The new parameter equation
agrees well with the numerical test and can be used as the temperature boundary conditions for the
lumped differential formulation, from which the transient temperature in elements can be derived. In
the meantime this equation can be available for global structural analysis exposed to localized fires.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction experiment was carried out in a hall with 22.4 m length, 11.9 m
width and 27.0 m ceiling height shown in Figs. 1 and 2. This fire
To evaluate the behavior of steel structures in fires, it is scenario is with 1.84 MW heat release rate source in the middle of
essential to know the temperature in steel members, which is the floor, and lasted 10 min while smoke diffused the whole
induced by thermal actions in fires. The thermal action of a fire is space. Data from the thermocouples, located at the corner from
usually represented by the standard time–temperature curve or each wall 2 m and at different heights, showed that the smoke
other realistic fire curves in related national codes and standards, temperature increased with the increase of fire height (see Fig. 3).
i.e. the standard curve given by prEN1991-1-2:2002 [1], ASTM Additionally, this experiment was simulated by the fire dynamics
E119 [2], NFPA 251 [3], UL263 [4], SFPE Hand Book [5], localized simulator, FDS, which is produced by NIST. In Fig. 3 results of the
fires in reference [1] and large space fires in NFPA 92B [6]. The FDS have been compared to data obtained from the experiment
thermal action mentioned above is based on the assumption of and to be generally favorable. Another large space fire scenario
zone models and uniform characteristics of hot smoke. with 15 MW heat release rate, 32  32 m2 floor area and 12.0 m
However, a significant amount of full scale large space fire height of internal space was simulated by FDS. Output informa-
experiments have been carried out by the State Key Laboratory tion describes that temperature distribution is non-uniform in the
of Fire Science in China. An abundance of database shows horizontal plane above the floor and the peak temperature is right
that there is different elevated temperature history and distribu- above the fire. Fig. 4 graphically shows that the transient
tion between fires in large volume space and in small volume temperature distribution in the horizontal plane at ceiling height
compartment. In reference [7], a full scale large space fire of 12 m after ignition 5 min. Fig. 5 graphically shows that the
spatial radial temperature distribution in the horizontal plane at
ceiling height of 12 m, which decreases from the initial point
n
Corresponding author. Tel./fax: þ 86 25 83587852. x¼1.5 m, where x is the distance from plume centerline. The
E-mail address: [email protected] (Y. Du). initial point corresponds to the edge of the fire source with

0379-7112/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.firesaf.2012.07.004
114 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120

Fig. 3. Temperatures from FDS versus the test data.

Fig. 1. Human view of hall for large space fire.

Fig. 4. FDS simulated temperature distribution in the horizontal plane.

Fig. 2. Internal view of hall for large space fire.

Fig. 5. Spatial radial temperature distribution in the ceiling plane.

3  3 m2 in the middle of the floor. Comparison of temperature
histories reveals that the peak temperature in ceiling right above
fire is lower than ISO834 curve shown as Fig. 6. of a grid steel roof covered a large space building under ISO834
curve versus a natural fire condition.

2. Structure fire analysis 2.1. Fire temperature

Do large space fires without flashover have much less impact As is graphically shown in Fig. 5, ISO834 curve is given by
on steel structures than small compartment fires with flashover?
A compelling result can be obtained after analysis of the behavior T g ¼ 20 þ 345log10 ð8t þ1Þ ð1Þ
 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120 115

Fig. 6. Peak temperature history versus ISO834 curve.

Fig. 8. Layout of the structural model. (a) Plan view of regular pyramidic space
truss and supports location. (b) Elevation of regular pyramidic space.

Fig. 7. Virtual space for fire simulation.

where Tg is the hot smoke temperature (in 1C), t is the time (in
minutes).
To compare with ISO834, a natural fire scenario is designed for
numerical simulation using FDS. In the middle of the ground,
15 MW fire was carried out in a prismatic structure with the Fig. 9. Deformation of FE model regular for pyramidic space truss in large
dimension of 24  24  12 m3 (see Fig. 7). The walls are made of space fire.
brick and the roof is made of concrete supported by a regular
pyramidic grid steel structure. At the bottom of the structure, Non-linear temperature dependent material properties, such
there are four vents at the bottom of each wall to provide as stress-strain relationship, specific heat, thermal conductivity
combustion oxygen. The fire is fuel restricted and follows t-square and elongation for steel are taken from China code [6].
fast growth. The non-uniform temperature distribution obtained
from FDS, and the peak temperature in ceiling reached 540 1C. 2.3. The effects of different temperature conditions

The finite element model was developed using commercial

software ANSYS for analysis the behavior of the regular pyramidic
2.2. Structural model space truss (see Fig. 9). The results of a fully non-linear thermal
analysis (see Fig. 10) show that the displacements in the middle
The finite element model of global structure is established for of the grid structure increase rapidly while the global structure
analysis of fire-resistance of the regular pyramidic grid steel losing load-bearing capacity. However, the structure failed at
structure. The grid steel structure is simply supported by per- 340 1C in the natural fire, and still had good load-bearing capacity
fectly straight concrete columns at its top chord node and has a before 570 1C exposure to standard fire ISO834. The reason of
depth of 1.5 m, as is shown in Fig. 8. It is assumed that the natural fire with lower temperature cause structure failure is that
supports are restrained at four corners, and others are fixed only in large space fire, the non-uniform smoke temperature distribu-
in the vertical direction, but all rotations remain free. a series of tion induces the non-uniform temperature elevated in steel
pin-ended link the chords. Loading from the roof slab to the space structure. Then the cooler parts can provide restraints to the
truss is taken as a series of point loads positioned at the joints heated area and induces thermal stresses. In the meantime the
between top chords and web members. Cross sections of the top, additional thermal stresses increase the overall compression
bottom and web members are chosen from a range of hollow stress in the top chords and cause instability. Following the first
bar such as +60  3.5, +76  3.75 and +89  4 according to its compression chords failure, subsequent redistribution of load and
axial internal force. The maximum loading ratio among the material properties decaying with elevated temperature were
chords is dominated by 0.6. sufficient to cause the collapse of the system. The non-uniform
116 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120

various factors mention above, a series of transient temperature

distributions will result from each fire scenario.

4. New parametric curve for large space fire

Fig. 4 shows the typical non-uniform temperature distribution,

which may be suggested to be pole asymmetric from the fire
origin. So the non-uniform temperature distribution can be
described by T(x,z,t) (see Fig. 7), where x is the distance from
the plume centerline; z is the height above the fire origin; t is the
Fig. 10. Displacement curve of regular pyramidic space truss.
time after ignition. The temperature decreases from the plume
centerline to the point in the horizontal plane (see Fig. 5). Fig. 6
distribution of temperature in large space fires decays the fire shows the typical temperature history in large space building
resistance of steel structures. Hence it is important to describe fires. During fire growth period the smoke temperature increases
this non-uniform distribution for large space fire with new rapidly. Then, the temperature grows slowly and keeps constant
temperature–time curve. during steady period.
The feature of temperature in large space fires is different from
those based on zone model, CFD model and localized fire model 4.1. Basic equation
are given by Eurocode, North American standards and other
national stands. However, zone models are based on the hypoth- As the characteristics of smoke temperature in large space fires
esis that temperature is uniform in the upper and lower zones. It mentioned above, the basic equation that describes the transient
does not agree well with the real large space fire. At present, field non-uniform temperature distribution, T(x,z,t), can be expressed
model is the most accurate model for fire simulation with by
numerical techniques from Computational Fluid Dynamics
Tðx,z,tÞ ¼ T g ð0Þ þT max
g f ðtÞksm ð2Þ
(CFD). A software FDS developed by NIST applies field model to
simulate fire scenarios, which is employed to fulfill numerical where Tg(0) is the ambient temperature; T max
is the maximum
g
tests for large space building fires. The availability for FDS soft- temperature; f(t) is the function of time for temperature history;
ware is verified by the full scale large space fire mentioned above. ksm is the regressing function of distance, and x is the distance
However, empirically-derived constants need to be provided to from the plume centerline to points in the horizontal plane.
FDS to describe turbulence and dynamics of the flame plume that Eq. (2) describes the smoke temperature which is one of the
is difficult for civil engineers without knowledge of fluid boundary conditions, and induces the temperature elevated in
dynamics, combustion and heat transfer. It maybe a feasible structural element. Flame radiation effects have been illustrated
way to develop an empirical function to describe the temperature in reference [9].
in large space fire with parametrical analysis based on abundant
test databases.
4.2. Maximum temperature, T max
g

In Fig. 4 there is a peak in the non-uniform temperature field,

3. Numerical experiments
which increases and keeps constant with the development of the
fire (see Fig. 5). T max
g is the maximum peak value through the
Numerical tests are economy and more easily than full scale
whole fire. Usually, some key parameters are needed to create the
experiments to get wide range of database for parametrical
FDS input data file to describe the scenario, i.e. fire growth type;
analysis. In this paper FDS software is employed to discover the
heat release rate, Q; fire area, Aq; floor area, Asp; and ceiling
characteristics of temperature in large space fires. Since FDS
height, H. So the effect of key parameters on the maximum
requires a relatively fast CPU and a substantial amount of
temperature T max
g can be developed by parameter study.
random-access memory (RAM), all simulations are operated by
Fig. 11 shows the maximum temperature history T max g with a
the PC with the processor as fast as 1.8 GHz Pentium Dual Core,
range of fire growth styles as slow, medium, fast and ultra fast.
with at least 1 GB RAM and with hard drivers of at least 240 GB to
The heat release rate, fire area and space size are constants. Fig. 11
store the output files. A Smokeview Tool can be employed to
illustrates that the peak temperature increases in the fire growth
visualize numerical predictions generated by FDS.
phase with different trend and keeps the same constant value
These numerical tests have been conducted in virtual spaces
which is defined as maximum temperature T max g . The fire growths
with ceiling heights ranging from 6 m to 20 m and floor areas
effect the time after ignition for the maximum temperature to
ranging from 500 m2 to 6000 m2. In all cases there are only four
grow, and the variation of the maximum temperature is slight.
ventilators with 2 m height and 10 m width at the bottom of each
wall to provide combustion air. In reference [8], fire size is
expressed in terms of the heat release rate with 2 MW, 5 MW
and 25 MW, and unsteady fire growth is as t-squared curve with
slow, medium, fast and ultra-fast type. For this paper, a heat
release rate per floor area of 225 kW/m2 will be used for a
restricted fuel, and 500 kW/m2 will be used for large fire. Then,
the fire occurring over 9.3 m2 of floor space will result a fire of
2.1 MW. If a fire occurred over 9.3 m will result a fire of 4.6 MW. A
large fire of 25 MW will involve an area of 50 m2 at 500 kW/m2. It
is suggested that the location of the fire source is in the middle of
the floor. There is no energy exchange between internal and
external space through wall. The whole fire lasts 2 h. Under Fig. 11. Peak temperature histories with different fire growth constants.
 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120 117

Fig. 12. Peak temperature histories with different heat release rates.

Fig. 15. Temperature histories at two points in large space with 2  2 m2 fire
source.

Fig. 13. Peak temperature histories with different floor areas.

Fig. 16. Temperature histories at two points in large space with 3  3 m2 fire
Fig. 14. Peak temperature histories with different heights of space. source.

Fig. 12 shows the peak temperature histories with a range of heat The maximum temperatures appear the similar tendency with
release rates between 2 MW and 25 MW. The fire growths, heat different fire size shown as Figs. 17–19. At first, the function of the
release rate per floor and space size are constants. The maximum floor area and ceiling height is established as T max g ¼ A þBH þ
temperatures increase with the larger heat release rates. Fig. 13 C=Asp , and fitting the coefficient A, B and C in the range of heat
shows the peak temperature histories with a range of floor areas release rate by the least squares method. Then, the function of the
between 500 m2 and 6000 m2. The heat release rate, ceiling heat release rate is established as T max
g ¼ ða1 þb1 Q Þ þ ða2 þ b2 Q ÞH þ
height and fire growth type are constants. The maximum tem- ða3 þ b3 Q Þ=Asp and fitting the coefficient a1 and b1 by the least
peratures decrease with the larger floor area. Fig. 14 shows the squares method once more. Through curve fitting of 72 simula-
peak temperature histories with a range of ceiling heights tion results listed in Table 1 T max
g can be expressed as
between 4 m and 20 m. The fire growths, heat release rate and
floor area are constants. The maximum temperatures decrease T max
g ¼ ð20Q þ 80Þð0:4Q þ 3ÞH þ ð52Q þ 598Þ  102 =Asp ð3Þ
with the higher ceiling height. Fig. 15 and Fig. 16 show the
temperature elevated history at two points caused by the square
of 2  2 m2 and 3  3 m2 fire source with 2 MW heat release rate 4.3. Accretion function, f(t)
respectively. There are the similar temperature histories at the
same point in large space with 1000 m2 floor area and 6 m height. Fig. 11 shows that accelerations of the peak temperature are
Based on a series of fire scenarios the area of the fire source has different among the fire growth types. All temperature–time
slight influence on the maximum smoke temperature, but enlarge curves above illustrate that temperatures accelerate sharply in
the peak temperature distribution as Eq. (5). So the maximum fire growth phase and tend to keep constants value, which is
temperature can be assumed to be the function of Q, Asp and H. defined as the maximum temperature in steady phase. This
The large space building fires with a range of heat release rate, temperature history can be described by two terms exponential
floor area and ceiling height are simulated by FDS software, and function of fire time as
the maximum temperatures with different key parameters listed
in Table 1. f ðtÞ ¼ 10:8eðbtÞ 0:2eð0:1btÞ ð4Þ
118 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120

Table 1 Table 2
Maximum temperatures T max
g (1C) with different key parameters. Factor b with fire growth tupes.

Floor area (m2) Space height (m) Q ¼2 MW Q ¼5 MW Q ¼25 MW Fire growth type

500 4 180 330 880 Slow Medium Fast Ulter

6 170 300 790
9 160 300 780 b 0.001 0.002 0.003 0.004
12 140 300 780
15 120 280 780
20 90 190 640

1000 4 180 230 830

6 130 230 700
9 110 230 660
12 100 210 630
15 90 170 610
20 80 150 580

3000 4 150 230 660

6 110 180 630
9 90 140 530
12 80 140 480
15 70 130 450
20 60 120 350

6000 4 140 160 560

6 100 140 540
9 80 120 480 Fig. 20. Effects of b on the peak temperature history with a range of fire
12 70 110 410 growth types.
15 60 100 380
20 50 80 340
Table 3
Factor Z with volume of large space building.

Asp (m2) H (m)

6 9 12 15 20

500 0.60 0.65 0.70 0.80 0.85

1000 0.50 0.55 0.60 0.70 0.75
3000 0.40 0.45 0.50 0.55 0.60
6000 0.25 0.30 0.40 0.45 0.50

where b is regression parameter dependent on the fire growth types

and fitting of simulation results listed in Table 2(dimensionless), t is
Fig. 17. Maximum temperatures induced by fire 2 MW with different ceiling the time from fire ignition (s).
heights and floor areas. Fig. 20 shows the acceleration of the peak temperature with
the variation of b.

4.4. Regression ratio of maximum temperatures, ksm

Fig. 4 illustrates the non-uniform temperature distribution of

hot smoke in the horizontal plane above the fire source. Fig. 5
shows that the maximum temperature decays along the distance
from the plume centerline in ceiling level. The ratio of the
maximum temperature regression varies with the floor area, Asp,
and the height of the ceiling level, H. Such tendency can be
described by exponential function mathematically as
Fig. 18. Maximum temperatures induced by fire 5 MW with different ceiling
heights and floor areas. ksm ¼ Z þ ð1ZÞeðD=2xÞ=7 , if x r D=2, then x ¼ D=2 ð5Þ
qffiffiffiffiffiffiffiffiffiffiffi
D ¼ 2 Aq =p

where D is the effective diameter, Aq is the area of the fire, Z is the

shape factor dependent on the floor area and ceiling height, and
fitting of simulation results listed in Table 3.
Fig. 21 illustrates that factor Z has effects on the ratio of the
regression parameter ksm (dimensionless).

4.5. Example

Fig. 19. Maximum temperatures induced by fire 25 MW with different ceiling To illustrate application of new calculation method given by
heights and floor areas. Eq. (2), an example for giving fire scenario is provided. Dimension
 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120 119

Then, introduce T max

g , f(t) and ksm into Eq.(2)
Tðx,z,tÞ ¼ 20 þ 206½10:8expð0:003tÞ
0:2expð0:1  0:003tÞ½0:6 þð10:6Þ eð1:5xÞ=7  ð6Þ
Point A is in the ceiling right above the fire and point B with
distance 16 m from the plume centerline in ceiling plane.
The temperature histories at two points fit to FDS results well
(see Fig. 22). The regression of the maximum temperature from
the plume centerline in the ceiling level of the fire is shown in
Fig. 23. The temperature history of each point in ceiling plane can
be calculated by Eqs. (2)–(5) and fits to results from FDS well.
To conservatively apply Eqs. (2)–(5), it is suggested that the
location of the fire source be conservatively selected. For example,
Fig. 21. Effects of Z on the ratio of the maximum temperature regression along the
horizontal direction. if fires may be possible anywhere from the floor to 3 m above the
floor, conservation selection of ceiling height would be from fire
source.

5. Conclusions

There is different behavior of a typical regular pyramidic space

truss exposure to fire between ISO834 curve and a large space
building fire. So a new temperature–time relation is developed to
describe the transient non-uniform temperature distribution.
By inducing key factors i.e. fire size, fire area, dimension of large
spaces and fire growth types, the proposed new temperature–time
curve is able to address the transient temperature distribution
within the ceiling level mathematically. The maximum tempera-
ture history can be obtained by Eq. (4), and the temperature
decreased from plume centerline can be obtained by Eq. (5). This
method is suitable to a wide range of fire situations with fire
Fig. 22. Temperature histories fitting to data from FDS.
growth rates from slow to ulter-fast, heat release rates from 2 MW
to 25 MW, floor areas from 500 m2 to 6000 m2, ceiling heights
from 6 m to 20 m, fuel areas from 9 m2 to 50 m2, and the fire
location is in the middle of the floor. If fires are above the floor,
conservation selection of ceiling height would be from fire source.
In the mean time, the effect of key factors on the temperature
characteristic are as follows:

 with the increase of heat release rate, the maximum tempera-

ture is higher;
 with the increase of the floor area, the maximum temperature
is lower;
 with the increase of the internal space height, the maximum
temperature is lower; and
 the temperature decays from the plume centerline in
horizontal plane.

This new temperature–time curve can be readily applied to

Fig. 23. Regression ratio of maximum temperatures from FDS versus calculation. boundary conditions for temperature elevated in structural ele-
ments and global structure analysis.
of the large space is 15 m height, 32  32 m2 area respectively.
At the bottom of each wall there is ventilator with 2 m height
and 10 m width to provide combustion air. A fire with 5 MW Acknowledgments
heat release rate and x ¼1.5 m was selected along with a fast
growth type in t-square fire and last 2 h. Introduce parameters The authors would like to thank State Key Laboratory of
into Eqs. (3)–(5), where b and Z are from Table 2 and Table 3 Fire Science in China for providing information of full scale
respectively. experiment.

T max
g ¼ ð20  5 þ80ð0:4  5 þ 3Þ  12 þð52  5 þ 598Þ
References
102 =1024 ¼ 206 1C
[1] prEN 1991-1-2 (1992), Eurocode 1—Actions on Structures, Part 1–2: General
Actions—Actions on Structures Exposed to Fire, Final Draft Stage 49, European
f ðtÞ ¼ 10:8expð0:003tÞ0:2expð0:1  0:003tÞ
Committee for Standardization, Brussels, January 10, 2002.
[2] ASTM, Standard methods of fire tests of building construction and materials
(ASTM Standard E119-05), American Society for Testing and Materials, West
ksm ¼ 0:6 þ ð10:6Þeð1:5x=7 Conshohocken., PA, 2005.
120 Y. Du, G. Li / Fire Safety Journal 54 (2012) 113–120

[3] NFPA, Standard Methods of Tests of Fire Endurance of Building Construction [7] Y.Z. Li, Studies of Smoke Movement and Control in Atrium Building Fire, Ph.D.
and Materials. NFPA 251, National Fire Protection Association, Quincy, MA, Thesis, University of Science and Technology of China, 2001.
2006. [8] NISTIR 5516, Method of Predicting Smoke Movement in Atria With Application
[4] UL, Fire Tests of Building Construction and Materials, UL 263, Underwriters to Smoke Management, NIST, November 1994.
Laboratories Inc, Northbrook, Illinois, 2003. [9] Y. Du, G.Q. Li, Effects of flame radiation on temperature elevation of steel
[5] SFPE, SFPE Handbook of Fire Protection Engineering, Society of Fire Protection members in large space building fire, in: Proceedings of the International
Engineers, SFPE, 2004. Conference on Application of Structural Fire Engineering, Prague, 2009,
[6] CECS 200, Technical Code for Fire Safety of Steel Structures in Buildings, pp. 356–361.
National Institute of Standards for Engineering, China, 2006.