Download The Development and Validation of a CFD-based

Natural Fire Safety Concept –
The Development and Validation of a CFD-based Engineering
Methodology for Evaluating Thermal Action on Steel and
Composite Structures
EC Proposal N° : P4169/ Contract N° : 7210-PR-184
Project Co-ordinator :
Dr Suresh Kumar, BRE - Watford (GB)
Project Partners :
Building Research Establishment Ltd (GB)
ProfilARBED S.A. (LU)
VTT Building Technology (FIN)
LABEIN (ES)
Arbeitsgemeinschaft Brandsicherheit AGB (DE)
1. ABSTRACT
The design of steel/composite structures is defined in the structural Eurocodes, which set out the
standard calculation procedures. Further, in recent years significant research has gone into the
development of the 'natural fire safety concept' which aims to provide a more realistic representation
of thermal response under 'natural fire' conditions (cf NFSC1 and NFSC2 programmes). The current
project aims to build on this approach by exploiting the techniques of computational fluid dynamics
(CFD) to develop, verify and apply a comprehensive engineering methodology for the prediction of
thermal actions on steel and composite structures.
Model development is currently underway and is focussing on the implementation of a multi-block
mesh scheme in the SOFIE CFD code. This will enable blocks of fine mesh to be applied in key
locations, e.g. around any steel components of interest or in the fire source. Also, initial investigations
have been made on the issues associated with the computation of heat transfer to the structural
envelope, for example, ray number effects and the representation of the radiative properties of the
participating gas-phase medium, including particulate soot and combustion products.
A progressive model validation exercise is being undertaken. This involves simulation of the thermal
response of steel/composite members in fire tests for which experimental data is available - a localised
beam fire test, standard fire-resistance furnace tests and full-scale tests involving natural fires. The
initial stage of the model verification exercise (WP 2) has focused on the localised beam fire test case
(WP 2.1) and data preparation for the BRE full-scale fire test case (WP 2.3).
The localised beam fire study (work package 2.1) has established baseline results for the performance
of the key physical and numerical models which are relevant in this application. This concerns mainly
the optional parameters in the discrete transfer radiation model, such as ray number effects and the
representation of the radiative properties of the participating gas-phase medium. By this means, the
requirements for the radiation model description for this case have been clarified.
Detailed studies and comparisons of flowfield and heat transfer parameters has further provided quite
an insight into the nature of the physical phenomena of relevance in this problem. No major
weaknesses were found in the model representations and overall, the predictions are found to be fair
and physically plausible. Nevertheless, some discrepancies between experiment and prediction
remain.
Initial work on work package 2.3 has focussed mainly on data analysis and preparation. Data obtained
from the additional instrumentation installed on the BRE fire test series on full-scale compartments,
under the NFSC2 programme, has been processed and will eventually be used to help define the
problem setup for the simulations and for the verification of the CFD model for this application.
Ultimately, model predictions will be used to determine equivalent values for the convective heat
transfer coefficient, resultant emissivities and other empirical factors used in heat transfer calculations.
Detailed assessment of the results will then be made to identify the critical design parameters affecting
the thermal action on the steel/composite structures.
The procedures set out in the Eurocodes will be reviewed in the light of the CFD predictions.
Recommendations will be made on improvements to the parameters used in the standards procedure
and/or extension of the methodology as necessary.
2. INTRODUCTION & OBJECTIVES
2.1. Introduction
The standard procedures for prediction of thermal actions on steel/composite structures are defined in
the structural Eurocodes. In addition, recent and current ECSC research programmes (NFSC1, NFSC2)
address issues related to the performance of steel structures in natural fire conditions, with a view to
the development and extension of the Eurocodes methodologies.
The modelling technique of computational fluid dynamics (CFD) is being extensively used in a range
of engineering disciplines for simulation of fluid flow and heat transfer processes. In the past twenty
years, CFD has found increasing application in fire modelling, and this has contributed to the
establishment of the discipline of fire safety engineering. However, there has so far been no
coordinated attempt to exploit the potential of the technique in prediction of thermal behaviour of
boundary materials due to the effects of fire in buildings. This project is seeking to apply CFD
techniques to develop, validate and apply a comprehensive engineering methodology for prediction of
thermal actions on steel and composite structures, and to contribute to the development of the firerelated Eurocodes.
2.2. Objectives
The objective of the project is to develop an engineering methodology, exploiting the advanced
capabilities of computational fluid dynamics (CFD), for determining the thermal behaviour of
structural elements in steel/composite-framed buildings. Specific objectives of the project are as
follows:

To develop a verified CFD-based engineering methodology for simulating the thermal action on
steel/composite structures,

To apply the methodology for evaluating the effect of fire loading, ventilation and compartment
construction on the thermal action on steel/composite structures,

To identify the essential elements of the methodology developed and provide guidance on its
'correct' use, i.e. defining the range of applicability and the sensitivity to various input parameters,

To apply the model for the assessment of the calibration and sensitivity of empirical design
parameters, such as the convective heat transfer coefficient and safety factors used in the design
guides (cf. Eurocodes EC1 and EC3).

To contribute to the development of the design guides.
3. PROGRESS & ACHIEVEMENTS
3.1. Model development
SOFIE is a single-block structured code,
which means that the entire modelled
geometry is represented inside a single threedimensional curvilinear mesh.
Fig. 1
illustrates this in simplified way for an
imaginary compartment fire scenario (shown
on two-dimensions only here).
Here the compartment, the fire source, an IFig. 1 - Compartment with single-block mesh
beam and a section of the outside environment
are all discretised with a single mesh block. Although the mesh is shown as Cartesian, it could be a
general curvilinear mesh, i.e. with the mesh lines curved to fit the geometry. Heat transfer into the Ibeam is necessarily modelled using the same mesh.
The main weakness of this approach is that the necessary compromise between resolution and number
of mesh elements may limit the usefulness of the simulation for practical problems. In respect of
typical building structure applications, one of the main limitations of a single-block mesh is that the
essential structural elements cannot be sufficiently well-resolved because a fine mesh cannot be
defined locally at, say, the I-beam. A fine mesh is necessary here in order to accurately model heat
transfer into the solid. A further weakness of a single-block mesh is that regions of the geometry that
are unimportant, e.g. the solid region above the compartment, are included unnecessarily in the
computational domain.
A method of bypassing these problems is to use
a multi-block structured mesh, so that each
sub-block is discretised using a separate mesh
of its own. The overall CFD solution is
obtained by the appropriate coupling between
meshes. Fig. 2 shows how the geometry of
Fig. 1 could be divided into 11 separate blocks
of mesh, with three blocks at the I-beam alone.
Note that no mesh elements have been wasted
in the solid region above the compartment.
Fig. 2 - Compartment with multi-block mesh
The coding for the multi-block model is under development, the first stage of which was to convert the
source code from FORTRAN77 to FORTRAN95.
3.2. Model validation - localised beam fire
3.2.1 Experimental setup
The experiments were performed at the Building
Research Institute (BRI) in Japan and are described
for instance in Pchelintsev et al. (1997) [1]. The steel
beam is suspended below a mineral-fibre ceiling slab
and the whole assembly is exposed to the effects of a
localised fire issuing from a burner located directly
below (see Fig. 3). The overall length is 3.6m and the
I-beam dimensions are 75mm wide by 150mm deep.
Comprehensive measurements of temperature and
total heat flux were obtained.
Beam and ceiling assembly
Gas Burner
Propane
Fig. 3 - Experimental layout [1]
The experiments of interest covered 6 different combinations of fire size and beam height, ranging
from 95 to 200 kW and with the beam located either 0.6m or 1.0m above the fire.
3.2.2 Results
A comprehensive set of predictions has been obtained which clearly reveal the particular sensitivities
to various variable model parameters for the scenario in question. For the 100kW burner/1m beam
height combination, the effects studied include:
radiation model discretization scheme
number of discrete transfer radiation model rays
radiation transfer equation (RTE) / absorption coefficient models
'weighted sum of grey gases' model coefficient set
geometric effects - i.e. raised burner and static pressure boundary location
In addition, routine grid sensitivity studies
were undertaken. The conclusions from this
phase of the work were that the predicted flux
distributions are in general fairly insensitive
to most of the parameters varied. For
example, Fig. 4 shows the results obtained
with the different radiation transfer equation
/absorption coefficient models.
Total heat flux v distance
RTE solution/absorption coefficient model
25
Total heat flux to gauge (kW/m 2)





lumped transmissivity/CAC
20
lumped transmissivity/tpWSGG
lumped absorptivity/CAC
15
lumped absorptivity/tpWSGG
WSGG/tpWSGG
10
WSGG banded
The baseline results for the 100kW/1m case
5
were in fair agreement with the experimental
0
values in terms of the overall magnitude of
0
0.2
0.4
0.6
0.8
1
the heat flux, but the profile was rather
Distance (m)
different. The flux just outside the plume
region tended to be underpredicted with that Fig. 4 - Effect of RTE solution/absorption
around the stagnation point and at more coefficient model on predicted flux distribution
remote locations being much nearer the
experimental value. In order to investigate this, the influence of various parameters affecting the
plume spreading rate was studied. In general, the impact of the changes was found to be small, and
the most noticeable improvement was achieved using the model parameters recommended by Nam
and Bill (1993) [2]. However, caution must be exercised when modifying fundamental and otherwise
well-established parameters and this parameter set was not employed in other simulations.
Computations were performed for each of the experimental fire size and beam height combinations
and the agreement between prediction and experiment was found to vary somewhat from case to case,
see Fig. 5. The largest discrepancies were noted for the 0.6m beam height cases, where the plume
region fluxes were generally underpredicted.
Investigations were made into various aspects of the flowfield predictions, so as to better understand
the overall performance of the models and the reasons for any discrepancies. Using empiricallyderived correlations for flame extension, it was found that the implied experimental value was quite
short (c. 0.4m), agreeing well with the computed value. Further, a theoretical estimate of the ceiling
jet thickness (due to Alpert (1975) [3]) suggested that this would be considerably less than the beam
height, so that the lower flange of the steel beam might not be expected to be exposed to hot flow
away from the impingement point, again in agreement with the predictions.
Detailed analyses of the breakdown of the heat fluxes to the underside of the beam-ceiling assembly
showed that the influence of convection is relatively unimportant in this application. Also, a
simulation with a radiative heat source at the burner location and no fluid flow revealed a more-or-less
symmetrical flux distribution, thus verifying the macroscopic performance of the radiation model.
Diagnostics performed on the gas-phase radiative transfer predictions showed a peak flux generation
in the centreline cell, some absorption occurring in the adjacent cell and thereafter, the flux
propagating outwards with no further modification. This gas-phase transfer is substantially symmetric
and physically plausible.
Fig. 5 - comparison of prediction with experiment for each fire size/beam height combination
3.3. Model validation - full scale compartment fire tests
The test data obtained from the BRE series of full-scale compartment fires under NFSC2, including
velocities, gas/surface temperatures, heat fluxes and the mass loss defining the heat release rate, has
been processed in preparation for the final phase of the model validation exercise (WP 2.3).
4. CONCLUSIONS
Model development is underway and is focussing on the implementation of a multi-block mesh
scheme in the SOFIE CFD code. Initial work on model verification has focused mainly on the
localised beam fire test case. Baseline results for the performance of the relevant key physical and
numerical models have been established. These concern mainly the key parameters in the discrete
transfer radiation model, such as ray distribution specifications and the representation of the radiative
properties of the participating gas-phase medium.
A good understanding has been obtained of the physical phenomena exhibited in this application,
including the characterisation of the flowfield and the predominant heat transfer regimes. Careful and
systematic comparison against model predictions for a wide range of physical and numerical
modelling choices reveal no major weaknesses in the models adopted and suggest that by-and-large
the models are providing a good and physically-plausible representation of the experiment.
Nevertheless, some discrepancies remain between the predictions and the experiment, particularly for
the case of the reduced beam height. Thus, further work will be needed to develop a better
understanding of the reasons for these discrepancies.
REFERENCES
1. Pchelintsev, A., Hasemi, Y., Wakamatsu, T. and Yokobayashi, Y. (1997) Proc. of the Fifth Int.
Symp. on Fire Safety Science, 1153-1164
2. Nam, S. and Bill, R.G. Jr. (1993) Fire Safety Journal, 21, 231-256
3. Alpert, R.L. (1975) Combust. Sci. and Tech., 11, 197-213