You are on page 1of 10


Stochastic Analysis of Structures in

Fire by Monte Carlo Simulation
Kaihang Shi1, Qianru Guo2 and Ann E. Jeffers3,*
1Undergraduate Student Researcher 2Graduate Student Research Assistant
3Assistant Professor

Department of Civil and Environmental Engineering, University of Michigan, Ann

Arbor, MI 48109, USA

This paper describes a preliminary study to explore the use of Monte Carlo simulation to
assess the reliability of structures in fire given uncertainty in the fire, thermal, and
structural model parameters. The methodology requires (1) the probabilistic
characterization of the uncertain parameters in the system, (2) a stochastic model for the
thermo-structural response, and (3) a limit state function that describes the failure of the
system. The study focuses on assessing the failure probability of a protected steel beam
under natural fire exposure. The system was modeled stochastically using a series of
sequentially coupled thermo-structural finite element analyses that were embedded
within a Monte Carlo simulation. Although the example considered here is relatively
simplistic in that it focuses on member level performance, it effectively demonstrates the
application of the proposed reliability method and provides insight into the practicalities
of extending the approach to more complex structural systems.

Keywords: Reliability, Monte Carlo Simulation, Finite Element Analysis, Probability,


In the broad context of fire protection engineering, a significant amount of research has focused on the
performance-based design of fire safety systems. Risk assessment is a key component of performance-
based design, as it enables the fire protection engineer to study tradeoffs in cost and utility to determine
the best engineered solution for a given target level of performance [1]. For the design of fire safety
systems, probabilistic methods, such as those based on event tree or fault tree analyses, are often the most
useful because such methods provide a quantitative measure of the risk associated with the design [2].
While the fire protection engineering discipline regularly uses risk-based methods in the
performance-based design of fire safety systems, the fire resistant design of building structures is
carried out in a purely deterministic manner, without direct consideration for uncertainty in the
governing parameters. The current methodology has resulted in a practice in which the reliability of
structures in fire is indeterminate [3-4] and inconsistent with the design for other types of hazards [5].
Furthermore, the current approach is in disagreement with empirical evidence suggesting that there are
a large number of uncertainties in the nature of the problem and that the response is highly sensitive to
a number of significant parameters [6-9]. These factors suggest that a probabilistic assessment of
structural fire resistance could provide key insight into the safe, economical design of structures
threatened by catastrophic fire. Furthermore, the quantification of structural reliability is a necessary
advance in the field of structural fire engineering because risk analysis is essential to performance-
based design.
A review of literature reveals that the topic of structural reliability in fire dates back to the early
works of Magnusson and Pettersson [10], who provided an initial framework for the reliability

*Corresponding author:

Volume 4 · Number 1 · 2013

38 Stochastic Analysis of Structures in Fire by Monte Carlo Simulation

assessment of structures in fire. More recently, reliability theory has been applied in the field of
structural fire engineering to derive probability-based load and resistance factors [5, 11-12], to account
for uncertainty observed in the analysis of fire resistance tests [6-9], and to support risk-based structural
fire design [3-4, 13-14]. Research to date has resulted in important advances in the reliability-based fire
resistant design of structures but has not fully explored the use of simulation-based stochastic methods
that are commonly used in other engineering disciplines.
The present study seeks to advance the field of reliability-based structural fire design by studying
the use of Monte Carlo simulation [15] as a means to quantify the reliability of structures in fire given
uncertainty in key model parameters. The approach specifically involves (i) characterizing the sources
of uncertainty, (ii) quantifying the probabilistic characteristics of each uncertain parameter, (iii)
defining performance criteria for the structure based on strength, stability, and/or serviceability
requirements, (iv) evaluating the structural response stochastically (e.g., by Monte Carlo simulation),
and (v) calculating the probability of failure. Once the probability of failure is determined, the adequacy
of the design can then be evaluated in terms of an acceptable level of risk. The methodology is
illustrated here by an example in which the fire resistance of a steel beam is evaluated based on
uncertainties in fire load and structural resistance parameters. The paper focuses specifically on the
implementation of the Monte Carlo simulation in the finite element analysis software Abaqus [16]
using commands that are contained in a separate Python script file. The approach extends the standard
modeling capabilities of Abaqus for performing parametric analyses. Although the example considered
here is relatively simplistic and does not include all possible uncertain parameters, it effectively
demonstrates the application of the proposed reliability method and provides insight into the
practicalities of extending the approach to more complex structural systems.

Monte Carlo simulation (MCS) is an iterative method in which random samples of each uncertain
parameter are generated based on their probabilistic characteristics, and a series of deterministic
analyses are subsequently conducted based on each possible combination of random parameters. Once
all deterministic simulations have been carried out, probabilistic information about the system can be
synthesized from the results [15]. For the purposes of evaluating the reliability of a system, the
probability of failure pf can be determined by observing the number of simulations for which the
response exceeds a given failure criterion, i.e.,

p =
, (1)

where Nf is the number of simulations for which the system failed and N is the total number of
simulations. As the total number of simulations N is increased, the failure probability converges to the
true failure probability of the system.
MCS has been used extensively to evaluate uncertainty in a wide range of engineering applications.
The technique has been broadly used in the fire protection engineering discipline to conduct
performance-based assessments of fire safety systems [17-18] and has been the primary method used
to explore uncertainty in structural fire engineering applications [6-9]. Despite its versatility, MCS
tends to be very computationally expensive (i.e., often impractical for a personal computer) because a
large number of iterations are often required to achieve a desired level of accuracy, especially in regions
with low probabilities of failure. Efficiency can be improved with special sampling techniques such as
Latin hypercube sampling [19], although such techniques do not completely overcome the limitations
in regions with exceptionally low failure probabilities. Nevertheless, MCS is an extremely versatile
technique that can provide insight into the probabilistic characteristics of complex, multidisciplinary
systems and therefore is the focus of this paper.
A sequentially coupled analysis of fire-structure interaction involves three components: an analysis

Journal of Structural Fire Engineering

Kaihang Shi Qianru Guo and Ann E. Jeffers 39

of the fire behavior to determine the thermal boundary conditions at the surface of the structure, a heat
transfer analysis to determine the temperature propagation within the structure, and a structural analysis
to evaluate the forces and deformations associated with the prescribed heating. As shown in Fig. 1, a
stochastic model for fire-structure interaction involves a propagation of uncertainty that affects each
stage of the response. In the fire model, uncertainty in the compartment geometry, type and distribution
of fuel, and ventilation conditions result in uncertainty in the thermal boundary conditions associated
with the given fire event. This in turn introduces uncertainty in the predicted temperatures in the
structure and ultimately affects the mechanical response of the structure. Additional uncertainty
associated with the material properties of the structure, the thermal and structural boundary conditions,
and magnitude of applied loads leads to further challenges in determining how the structural system
will respond in an actual fire scenario.

Figure 1. Propagation of Uncertainty in the Structural Fire Simulation

Characterize Random sampling and Propagation of samples Failure

uncertain pairing (based on LHS) through the thermo- Analysis
parameters structural model
fX(Xi) x11 x12 x1 N pf =
x 21 x 22 x2 N

x n1 xn 2 x nN

Figure 2. Stochastic Model for Fire-Structure Interaction

To capture this propagation of uncertainty, a stochastic model of the system can be employed. Figure 2
is shows the extension of the Monte Carlo simulation technique [18] to a structural fire application. As
illustrated in Fig. 2, the model is defined in terms of n uncertain parameters X = [X1, X2, ..., Xi, ..., Xn],
each of which is characterized by its statistical properties, such as the mean, variance, and probability
distribution. For MCS, a sample size of N is prescribed, which requires that N values for each uncertain
parameter Xi be defined. In classical MCS, all possible combinations of random parameters are
considered, thus requiring Nn total simulations. In the present study, Latin Hypercube sampling (LHS)

Volume 4 · Number 1 · 2013

40 Stochastic Analysis of Structures in Fire by Monte Carlo Simulation

is used to reduce the total number of simulations [19]. The Latin Hypercube sample is given as

  x 
 x x 
 
11 12 1N

 x x  x 
x= 
21 22 2N

   x  
 ij
 (2)
 
 x x  x 
n1 n2
 nN

where xij = the random value of parameter i in sample j. The parameter i ranges from 1 to the total
number of parameters n, while the sample j ranges from 1 to the total number of samples N. Thus, each
column of the matrix x represents one sample of parameters that is propagated through the thermo-
structural model. In the following study, three stochastic analyses were carried out to evaluate the fire
behavior, the thermal response of the structure, and the mechanical response of the structure. Random
values were generated for each uncertain parameter based on probabilistic characteristics that were
either assumed or based on existing data in the literature. Details about the analysis are provided in the
following section.


To illustrate the stochastic simulation of structural performance in fire, Monte Carlo simulations were
conducted for a protected steel beam exposed to natural fire. As illustrated in Fig. 3a, the beam was
simply supported and carried a uniformly distributed load w. In addition to the applied loads, the beam
also supported a concrete slab, which was assumed to act non-compositely with the beam. The steel had
a nominal yield strength of 345 MPa. A cross-section of W28x8 was required to resist the assumed
design load based on the U.S. steel design specification [20] and to meet the ANSI/UL 263
requirements for prescriptive fire resistant design in the U.S. The beam’s cross-section is shown in Fig.
3b. The beam was protected by a spray-applied fire resistant material such that the beam provided a 1h
fire resistance.

Concrete Slab

Fire Resistant
W8x28 Material (SFRM)
tw = 7.2 mm d = 204.7 mm
L = 4.88 m

(a) tf = 11.8 mm

bf = 166.1 mm

Figure 3. Protected Steel Beam Exposed to Fire: (a) Loading, and (b) Cross-sectional

The purpose of the analysis was to evaluate the response of the protected steel beam exposed to natural
fire given uncertainties in the fire, material, and loading parameters. A schematic of the general

Journal of Structural Fire Engineering

Kaihang Shi Qianru Guo and Ann E. Jeffers 41

procedure is shown in Fig. 4. Natural fire exposure was modelled using the Eurocode parametric fire
curve as modified by Buchanan [21]. To evaluate the thermo-structural response, two sequentially
coupled analyses were conducted using Abaqus [16]. Heat transfer over the cross-section was modelled
using two-dimensional continuum elements. The mechanical response was subsequently modelled
using two-dimensional beam elements. Temperatures in the flanges and web were obtained from the
heat transfer analysis and transferred directly into the structural model by specifying the flange and web
temperatures as predefined fields in Abaqus.
One thousand Monte Carlo simulations were carried out using Latin Hypercube sampling to reduce
the total number of simulations required for the analyses. As illustrated in Fig. 4, two parametric studies
(i.e., one for the heat transfer analysis, one for the structural analysis) were run in Abaqus, each of
which utilized a Python script file that generated an Abaqus model for each combination of random
parameters. To perform the parametric study, the input file (.inp) for the finite element analysis was
written in terms of the uncertain parameters associated with the particular heat transfer or structural
analysis using the *PARAMETER command. Random values for each parameter were generated in
Matlab [22] using the appropriate, mean, covariance, and probability distribution. Values for the
random parameters were then entered in the Python script file along with commands to define the
combination of parameters for each case and options for executing the analysis. The analysis was
executed by running the Python script from the Abaqus command prompt. Once the analysis was
completed, a separate Python script was executed to compile the results from each of the simulations.

Samples of random Parametric fire Heat transfer analysis Structural analysis

parameters analysis
x 11 x 12 x1 N
x 21 x 22 x2N

x n1 xn2 x nN

Heat transfer input: Structural input:

1 Python script file (.psf) 1 Python script file (.psf)
Matlab Matlab 1 input file (.inp) 1 input file (.inp)

Heat transfer output: Structural output:

N output database files (.odb) N output database files (.odb)

Abaqus Abaqus

Figure 4. Schematic of the Stochastic Structural Fire Simulation

3.1 Fire Analysis

Natural fire exposure was modelled using the temperature-time curve recommended by Buchanan [21].
In particular, the fire temperature Tf (oC) is given as

T = 1325(1 − 0.324e
−0.2 t*
− 0.204e −1.7 t*
− 0.472e −19 t*
), (3)

where t* is a fictitious time given by

t = Γt .

Here, t is the time (hours) and Γ is given as

Volume 4 · Number 1 · 2013

42 Stochastic Analysis of Structures in Fire by Monte Carlo Simulation

( F / 0.04) ,
Γ= v
( b / 1900)

where Fv is the ventilation factor and b is the thermal inertia (i.e., k ρC ) of the surroundings. The
duration of burning td can be determined based on the fire load per total area et and the ventilation factor
Fv, i.e.,

t = 0.00013e / F
d t v

After time td the fire is assumed to decay linearly according to the rate defined in [21].
For stochastic simulation of the fire, the fire parameters were assumed to have realistic mean values
and probability distributions based on the works of Culver [23] and Iqbal and Harichandran [11].
Specifically, the ventilation factor Fv was assumed to have a mean value of 0.04, a coefficient of
variation (COV) of 0.05, and normal distribution. It was also assumed that the walls and ceiling were
made of gypsum board, which has a mean value of b = 423.5 Ws1/2/m2K, a COV of 0.09, and normal
distribution. Based on an assumed compartment geometry, the fire load per total area et was determined
to have a mean value of 132.54 MJ/m2 with a COV of 0.62 and Gumbel (Extreme Type 1) distribution.
Random values for each of the fire parameters were generated in Matlab and subsequently inserted
in the fire model to obtain a series of natural fire curves that represent the range of potential compartment
fires expected in the current case study. As illustrated in Fig. 5, fires varied in duration and intensity, with
the expected (mean) fire reaching a maximum temperature of approximately 1100 °C and burning
steadily for 15 min before decay. Due to the high variability in the input parameters, the maximum and
minimum time-temperature curves deviate considerably from the mean fire temperature.

Figure 5. Predicted compartment temperatures

3.2 Heat Transfer Analysis

The fire temperatures computed from the previous step were subsequently used in a two-dimensional
heat transfer analysis of the beam’s cross-section, which is shown in Fig. 3b. Heat is transferred from
the fire to the structure by convection and radiation. Mean values for the convection heat transfer
coefficient and effective emissivity e were taken from Eurocode 1 [24], while both constants were

Journal of Structural Fire Engineering

Kaihang Shi Qianru Guo and Ann E. Jeffers 43

assumed to be normally distributed with coefficients of variation of 0.10 due to lack of existing data.
To achieve the 1h fire resistance rating, the SFRM was required to have a thickness of 11.1 mm. The
mean value for the SFRM thickness was taken as the nominal thickness plus 1.6 mm, resulting in a
mean thickness of 12.7mm [11]. The COV was assumed to be 0.20 and variability that follows a
lognormal distribution. The density, thermal conductivity, and specific heat for the SFRM were
assumed to be independent of temperature using mean values reported in the Eurocode and probability
distributions given by Iqbal and Harichandran [11].
Conduction at the steel-concrete interface was modelled by treating the concrete as a semi-infinite
medium with constant temperature on the top surface of 20 °C [25]. Thermal properties for the concrete
were assumed to be constant and independent of temperature, while thermal properties for the steel
were assumed to follow the temperature-dependent Eurocode models. Variability in the thermal
properties for steel and concrete was ignored in the present analysis for simplicity.
There were a range of steel temperatures obtained due to the variability in the fire temperatures and
thermal properties of the SFRM. Average temperatures from the heat transfer analysis are shown in Fig.
6. Note that the maximum temperatures in the steel averaged around 500 °C for the expected fire load.
However, more severe fire loads combined with low fire protection thicknesses resulted in the
possibility of steel temperatures in excess of 1000 °C.

Figure 6. Average temperatures in the steel flanges and web in comparison to the
mean fire temperature

3.3 Structural Analysis

The steel temperatures from the heat transfer analysis were subsequently specified in the mechanical
model of the beam that is shown in Fig. 3a. The stochastic simulation considered uncertainties in the
yield strength and magnitude of the applied loads. While the beam was designed for a nominal yield
strength of 345 MPa, a statistical analysis of data presented by Wainman and Kirby [26] showed that
this grade of steel has a mean strength of 380 MPa and a COV of 0.08 (normal distribution). The
uniformly distributed dead and live loads had design values 5.15 kN/m and 3.65 kN/m, respectively,
based on typical office loading in U.S. construction. For stochastic simulation, arbitrary-point-in-time
dead and live loads were used based on the calculations of Ellingwood [5]. Thus, the dead load had a
mean value of 5.41 kN/m, COV of 0.10, and normal distribution, and the live load had a mean value of
0.88 kN/m, COV of 0.60, and followed a gamma distribution.
The mid-span displacement for each simulation was measured at each time step in the analysis.

Volume 4 · Number 1 · 2013

44 Stochastic Analysis of Structures in Fire by Monte Carlo Simulation

Results are illustrated in Fig. 7. Note that the expected (mean) mid-span displacement reached a
maximum value of 58 mm before cooling. However, due to the range of potential fire loads, material
parameters, and magnitudes of applied loads, there were many cases in which the deflections became
excessively large.

Figure 7. Predicted mid-span displacement in the beam

The data generated by the 1,000 Monte Carlo simulations can subsequently be analyzed to evaluate the
probabilistic characteristics of the system and assess the probability of failure. To evaluate the
probability of failure, limit state criteria must be defined based on strength, stability, and/or
serviceability requirements. For the beam considered here, it is fairly obvious that the structure will fail
by the formation of a plastic hinge at mid-span. While this is a criterion related to the strength of the
system, it is difficult to derive a closed-form statement for the plastic moment capacity the structure
due to the temperature dependence of the material properties. Therefore, for the purposes of assessing
the reliability of the structure, failure was defined as the time at which the mid-span displacement
exceeded a limiting value of L/30 in the present study.
Based on this limit state criterion, the probability of failure can be computed by observing the
number of simulations in which the maximum deformation exceeded the limiting value of L/30 = 162
mm and substituting this value for Nf into Eq. 1. Thus, the failure probability for the beam was found
to be 1.3 percent. Despite the potential for catastrophic failure based on a potential “worst-case”
scenario, the beam (as designed according to the current prescriptive codes) appears to be sufficiently
designed to resist the natural fire, which is expected. Note that the analysis only considered 1,000
Monte Carlo simulations, and so the margin of error in calculating the probability of failure is high.
Therefore, the findings of this study are inconclusive at this point in time.
The failure probability can subsequently be implemented in a risk analysis to assess the adequacy of
the design based on anticipated costs associated with the failure of the structural member. It is important
to note that design alternatives can readily be explored and compared on the rational basis of structural
reliability. For example, one could consider alternative fire protection measures (e.g., reducing the fire
protection thickness while installing sprinkler systems to mitigate the fire hazard) to increase the economy
of the design while achieving a target level of safety. Furthermore, the methodology allows the designer
to define alternative performance criteria for the structure depending on the desired performance
objectives (e.g., imposing a more stringent limitation on deformation for a special-use facility). Thus, the
flexibility offered by reliability-based design is a necessary component to performance-based fire resistant
design, and the methodology therefore offers much promise to the future of structural fire engineering.

Journal of Structural Fire Engineering

Kaihang Shi Qianru Guo and Ann E. Jeffers 45

This paper presents a preliminary study into the stochastic simulation of structures in fire. Specifically,
the Monte Carlo simulation method was used in conjunction with a sequentially coupled finite element
analysis in Abaqus to evaluate the response of a protected steel beam given uncertainties in fire load
and structural resistance. While the application shows much promise for future investigations into the
probabilistic mechanics of structures at extreme temperatures, the computational demands required to
perform three sequentially coupled Monte Carlo simulations with embedded finite element simulations
calls for a more computationally efficient approach. Furthermore, the study presented here
demonstrates a significant need for empirical data on the uncertainty associated with various model
parameters, including (but not limited to) the thermal boundary conditions imposed at the surface of the
structure and the temperature-dependent material properties of construction materials. The extension of
the methodology to evaluate system level performance requires clear definitions of failure in various
structural components, including beams, columns, composite floor systems, and beam-to-column
connections, as well as improvements in the computational efficiency.

This research was supported by the U.S. National Science Foundation under Grant No. CMMI-
1032493. Any opinions, findings, conclusions or recommendations are those of the authors and do not
necessarily reflect the views of the sponsoring agency.

[1] Lundin, J. (2005). Development of a Framework for Quality Assurance of Performance-based
Fire Safety Designs, Journal of Fire Protection Engineering, 15, 19-42.
[2] Watts, J.M., and Hall, J.R. (2008). Introduction to Fire Risk Analysis, The SFPE Handbook of
Fire Protection Engineering, 4th Ed., National Fire Protection Association, Quincy, MA.
[3] Fellinger, J.H.H., and Both, C.K. (2000). Fire Resistance: Reliability vs. Time Analyses,
Proceedings of Composite Construction in Steel and Concrete IV, Hajjar, J.F., Ed., American
Society of Civil Engineers, USA, 816-827.
[4] Lange, D., Usmani, A., and Torero, J. (2008). The Reliability of Structures in Fire, Proceedings
of the 5th International Conference on Structures in Fire, Tan, K.H., Kodur, V.K.R., and Tan, T.H.,
Eds., Research Publishing Services, Singapore, 760-770.
[5] Ellingwood, B.R. (2005). Load Combination Requirements for Fire-Resistant Structural Design,
Journal of Fire Protection Engineering, 15, 43-61.
[6] Hietaniemi, J. (2007). Probabilistic Simulation of Fire Endurance of a Wooden Beam, Structural
Safety, 29, 322-336.
[7] Sakji, S., Soize, C., and Heck, J.V. (2008). Probabilistic Uncertainty Modeling for
Thermomechanical Analysis of Plasterboard Submitted to Fire Load, Journal of Structural
Engineering, 134, 1611-1618.
[8] Jensen, E., Van Horn, J., and Eamon, C.D. (2010). Variability of Fire and Concrete Temperatures
and the Associated Uncertainty in Structural Behavior, Proceedings of 6th International
Conference on Structures in Fire, Kodur, V., and Franssen, J.M., Eds., DESTech Publications,
USA, 959-966.
[9] Van Coile, R., Annerel, E., Caspeele, R., and Taerwe, L. (2011). Probabilistic Analysis of
Concrete Beams during Fire, Proceedings of the International Conference on Application in
Structural Fire Engineering, Wald, F., Horova, K., and Jirku, J., Eds., Prazska Technika, Prague,
[10] Magnusson, S.E., and Pettersson, O. (1980/81). Rational Design Methodology for Fire Exposed
Load Bearing Structures, Fire Safety Journal, 3, 227-241.
[11] Iqbal, S., and Harichandran, R.S. (2010). Capacity Reduction and Fire Load Factors for Design
of Steel Members Exposed to Fire, Journal of Structural Engineering, 136, 1554-1562.

Volume 4 · Number 1 · 2013

46 Stochastic Analysis of Structures in Fire by Monte Carlo Simulation

[12] Iqbal, S., and Harichandran, R.S. (2011). Capacity Reduction and Fire Load Factors for LRFD of
Steel Columns Exposed to Fire, Fire Safety Journal, 46, 234-242.
[13] Vaidogas, E.R., and Juocevicius, V. (2008). Reliability of a Timber Structure Exposed to Fire:
Estimation using Fragility Function, Mechanika, 73, 35-42.
[14] Huang, P.T., and Delichatsios, M.A. (2010). Quantitative Risk Analysis for the Response of Steel
Beams in Fires, Journal of Structural Fire Engineering, 1, 231-241.
[15] Haldar, A., and Mahadevan, S. (2000). Reliability Assessment Using Stochastic Finite Element
Analysis, John Wiley and Sons, New York.
[16] Abaqus, v.6-10 (2010). Dassault Systemes, Providence, RI.
[17] Au, S.K., Wang, Z.-H., and Lo, S.-M. (2007). Compartment Fire Risk Analysis by Advanced
Monte Carlo Simulation, Engineering Structures, 29, 2381-2390.
[18] Kong, D., Lu, S., Feng, L., and Xie, Q. (2011). Uncertainty and Sensitivity Analyses of Heat Fire
Detector Model based on Monte Carlo Simulation, Journal of Fire Sciences, 29, 317-337.
[19] Bergmeister, K., Novak, D., Pukl, R., and Cervenka, V. (2009). Structural Assessment and
Reliability Analyses for Existing Engineering Structures, Theoretical Background, Structure and
Infrastructure Engineering, 5, 267-275.
[20] AISC (2005). Steel Construction Manual, 13th Ed., American Institute of Steel Construction,
[21] Buchanan, A.H. (2002). Structural Design for Fire Safety, John Wiley and Sons Ltd, England.
[22] MATLAB, v.7.11 (2010). MathWorks, Inc., Natick, MA.
[23] Culver, C.G. (1976). Survey Results for Fire Loads and Live Loads in Office Buildings, National
Bureau of Standards, Washington, D.C.
[24] EC1 (2009). Eurocode 1 – Basis of Design and Actions on Structures, Part 1.2: Actions on
Structures – Actions on Structures Exposed to Fire, BS EN 1991-1-2, British Standards
Institution, London.
[25] Incropera, F. P., and DeWitt, D. P. (2002). Fundamentals of Heat and Mass Transfer, 5th Ed., John
Wiley and Sons, Inc., New York.
[26] Wainman, D. E., and Kirby, B. R. (1988). Compendium of UK Standard Fire Test Data:
Unprotected Structural Steel - 1, Report No. RS/RSC/S10328/1/87/B, British Steel Corporation,

Journal of Structural Fire Engineering