You are on page 1of 23

Fire Safety Journal 37 (2002) 83105

Post-ashover res for structural design


Roger Feaseya, Andrew Buchananb,*
a

Opus International Consultants, Auckland, New Zealand


University of Canterbury, Private Bag 4800, Christchurch, New Zealand

Received 2 October 2000; received in revised form 16 February 2001; accepted 16 February 2001

Abstract
One of the major diculties in designing structures to resist res is the selection of postashover design res. This paper proposes modications to the parametric equation for postashover res given in the Eurocode. The proposal is based on design res obtained from the
COMPF2 computer program, after calibration to results from realistic test res using
judicious characterisation of the fuel in the input data. The proposed new equation gives a
much better t to the real re data than the existing Eurocode equation, which predicts
temperatures lower than those measured in tests. r 2002 Elsevier Science Ltd. All rights
reserved.

1. Introduction
1.1. Overview
This paper has three major sections, which are:
1. Description of the COMPF2 post-ashover re model, and its calibration to full
scale test res.
2. Use of the calibrated model to produce a series of design res, for various fuel
loads, ventilation openings and construction materials.
3. Modication of the Eurocode parametric re equation to give a good t to the
calculated design res.

*Corresponding author. Tel.: +64-3-364-2250; fax: +64-3-364-2758.


E-mail address: andy@civil.canterbury.ac.nz (A. Buchanan).
0379-7112/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 7 9 - 7 1 1 2 ( 0 1 ) 0 0 0 2 6 - 1

84

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

1.2. Background
This paper has grown from concerns that there are serious deciencies in the
parametric equations for post-ashover res given in the Eurocode [1] for structural
design. For short duration res, the Eurocode equations give temperatures that are
lower than those measured in many tests, and the decay rates are inconsistent for
very large and small ventilation openings. The new equations presented in this paper
were derived by making simple modications to the Eurocode formula, in order to
give a good empirical t to output from the COMPF2 program [2], which had been
calibrated to real re test results as described in a comprehensive report by Feasey
[3]. The COMPF2 output was used rather than the test results because this allowed a
wider range of re conditions to be compared with the proposed parametric curves.
There are still many unanswered questions about post-ashover res, and further
research is needed.
1.3. Flashover
When a re in a small room is allowed to grow without intervention, assuming
sucient fuel in the burning item, temperatures in the hot upper layer will increase,
with increasing radiant heat ux to all objects in the room. If a critical level of heat
ux is reached, all exposed combustible items in the room will begin to burn, leading
to a rapid increase in both heat release rate and temperatures. This transition is
ashover. The re is then referred to as a post-ashover re, a fully developed
re or a re which has reached full room involvement. The re behaviour
described in this paper represents typical observations of res in small rooms of
regular geometry; the behaviour will be less predictable in rooms which have large
oor areas, high ceilings or irregular arrangement of fuel or openings.
1.4. Burning rate
In a typical room re before ashover, the burning rate depends on the type of fuel
and the geometry of the burning item. The temperatures in the room can be
characterised with a two-zone model. The behaviour of the re changes dramatically
after ashover. The ows of air and combustion gases become very turbulent. The
high temperatures and radiant heat uxes throughout the room cause all exposed
combustible surfaces to pyrolyse, producing large quantities of combustible gases,
which burn where there is sucient oxygen. The temperatures are dependent on the
heat release rate, which in turn depends on the rate of pyrolysis or evaporation of the
fuel and the available air supply to provide oxygen for combustion of the gaseous
fuel. The rate of burning and consequent heat release rate may be ventilationcontrolled or fuel-controlled.
1.4.1. Ventilation-controlled burning
In rooms with small or medium sized windows, post-ashover res are ventilationcontrolled, so the rate of combustion depends on the size and shape of ventilation

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

85

openings. It is usually assumed that all window glass (other than wired glass or re
resistant glass) will break and fall out at the time of ashover, as a result of the rapid
rise in temperature. If the glass does not fall out, the re will still be ventilationcontrolled, but because of the smaller openings it will burn for a longer time at a
lower rate of heat release.
In a ventilation-controlled re, the rate of combustion is limited by the volume of
cool air that can enter and the volume of hot gases that can leave the room. There is
insucient air ow for all the combustible gases to burn inside the room, so the
ames extend out the windows and additional combustion takes place where the hot
unburned gaseous fuels mix with outside air. For a room with a single opening,
Kawagoe [4] used many experiments to show that the mass loss rate of burning wood
fuel m can be approximated by
p
m 0:092Av Hv kg=s;
1
where Av is the area of the ventilation opening (m2), and Hv the height of the
ventilation opening (m).
p
The dependence of burning rate on Av Hv has been observed in many studies.
This is often expressed in terms of an opening factor Fv given by
p
Fv Av Hv =At m0:5 ;
2
where At is the total area of the bounding surfaces of the room (m2).
Eq. (1) is widely used, but not always accurate. Even if the burning rate is known
precisely, the calculation of heat release rate is not accurate because an unknown
proportion of the pyrolysis products burn as ames outside the window rather than
inside the compartment. Other sources of uncertainty arise because some proportion
of the fuel may not be available for combustion, and the re may change to fuel
control after some time.
Drysdale [5] shows how Eq. (1) can be derived by considering the ows of air and
combustion products through an opening as shown in Fig. 1, for fuel of wood cribs.
In a ventilation-controlled re there are complex interactions between the radiant
heat ux on the fuel, the rate of pyrolysis (or evaporation) of the fuel, the rate of
combustion of the gaseous products, the inow of air to support the combustion,
and the outow of combustion gases and unburned fuel gases through openings. The

Fig. 1. Window ows for ventilation-controlled re.

86

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

interactions depend on the shape of the fuel (cribs or lining materials), the fuel itself
(wood or plastic or liquid), the shape of the room and the ventilation openings.
Some tests have shown departures from Eq. (1). Following a large number of
small scale compartment res with wood cribs reported by Thomas and Heseldon
[6], Law [7] proposed a slightly more rened equation
p for burning rate, nding that
the burning rate is not directly proportional to Av Hv but also depends on the oor
shape of the compartment. Thomas and Bennetts [8] have shown that the burning
rate also depends heavily on the shape of the room and the width of the window in
proportion to the wall in which it is located. If the width of the window is less than
the full width of the wall, the burning rate is seen to be much higher than predicted
by Eq. (1) because of increased turbulent ow at the edges of the window.
Eq. (1) applies to a single ventilation opening in one wall of the compartment. If
there is more than one opening, the same equation is often used, with Av being the
total area of all the openings and Hv being the average height of all the window and
door openings, weighted by the area of the openings [9]. If the openings are on
several walls, the use of Eq. (1) implies an assumption that the air ow is similar in
all openings and there is no cross ow through the room.
Although Eq. (1) is not universally applicable, it is a useful approximation to
typical behaviour, and it has been the basis of most computational studies into postashover res, including most of those described in this paper.
1.4.2. Fuel-controlled burning
Not all post-ashover res are ventilation-controlled. The rate of burning may
sometimes be controlled by the surface area of the fuel, especially in large, well
ventilated rooms containing fuel items which have a limited area of combustible
surfaces. In this case, the rate of burning will be similar to that which would occur
for the fuel item burning in the open air, with enhancement from radiant feedback
from the hot upper layer of gases or hot wall and ceiling surfaces. Depending on the
fuel location, most res become fuel-controlled in the decay period when the exposed
surface area of the fuel decreases and the thicker items of fuel continue to burn [10].
In fuel-controlled burning, all of the heat is released inside the room, with no ames
projecting out of the windows.
For both ventilation- and fuel-controlled burning, not all of the combustible
material in the room may be available for immediate combustion. For this reason,
many researchers introduce a fuel fraction or combustion eciency factor by which
the heat of combustion or available fuel is reduced. Babrauskas [11] suggests a value
in the range of 0.50.9.

2. Temperatures
Temperatures in post-ashover res are essential input to structural design for re
safety. Unfortunately, these temperatures cannot be calculated precisely. This paper
describes measured and predicted temperatures from various studies, and proposes a
modied method of estimating temperatures for design purposes. Maximum

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

87

temperatures in post-ashover res are often at least 10001C. The temperature at any
time depends on the balance between the heat released within the room and all the
heat losses through openings by radiation and convection, and by conduction into
the walls, oor and ceiling.
2.1. Measured temperatures
Several experimental studies have measured temperatures in post-ashover res.
There is considerable scatter among the results of dierent studies. An empirical
equation for the maximum temperature during the steady burning period for a large
number of wood crib res in small scale compartments has been developed by Law
[7], based on experiments reported by Thomas and Heseldon [6], and summarised by
Walton and Thomas [12].
Arnault et al. [13,14] have carried out a wide ranging series of tests in full size
compartments. This test data (subsequently referred to as the NFSC data) was used
in this study to calibrate the post-ashover computer re model COMPF2.
These test res had fuel either of wood cribs or a mixed re load of timber,
furniture and paper. Compartment sizes were generally modest with the majority of
res carried out in a compartment of 3.38  3.68  3.13 m high. Compartment
construction was typically of mixed masonry, concrete and lightweight concrete,
with and without internal ceramic bre linings. A single ventilation aperture in one
of the walls provided a range of opening factor Fv from 0.157 to 0.014.
Most of the NFSC re tests were carried out with opening factors of 0.055 and
0.091. Fire loads were most often in the range of 1560 kg of wood per m2 of oor
area, although a few tests had higher re loads of up to 90 kg of wood per m2 of oor
area. Converting these into a net re load at a rate of 15.1 MJ/kg (the default value in
COMPF2 for a typical representation of wood fuel), this corresponds to re loads of
2251360 MJ/m2 of oor area. For the test re compartment of the geometry
described above, this range of test re loads corresponds to 50310 MJ/m2 of total
surface area (using the Swedish characterisation of post-ashover res, as
discussed below).
2.2. Computer models
A number of computer models have been developed for calculating temperatures
in post-ashover room res. Most of these are single-zone models which consider the
room to be a well-mixed reactor. It is possible to use two-zone models for postashover res, but these are not generally considered appropriate because many of
the pre-ashover assumptions are no longer valid [15]. Field models are not easily
applied to post-ashover res because of excessive turbulence.
All computer models for post-ashover res are based on heat balance. Fig. 2
shows the main components of heat ow in a simple compartment re. The heat
produced by combustion of the fuel qc is balanced by the heat losses; the main
components being heat conducted into the surrounding structure qw ; heat radiated
through the opening qR ; and heat carried out of the opening by convection of hot

88

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Fig. 2. Heat balance for a post-ashover room re.

gases and smoke qL : The computer models quantify this heat balance and solve the
conservation equations to predict the temperature of the gases within the
compartment in successive time increments. Most single-zone models assume that
all combustion takes place within the compartment, temperatures are uniform within
the compartment, and heat ow into the surrounding structure is identical on all
walls and the ceiling. A description of many dierent models is given by Feasey [3].
A diculty for all models is the calculation of the burning rate. Most models
simply assume ventilation-controlled burning as given by Eq. (1), but some also
include fuel-controlled burning. None of the commonly available models include the
eects of horizontal openings in the ceiling.
2.2.1. Swedish curves
The most widely referenced timetemperature curves for real re exposure are
those of Magnusson and Thelandersson [16], shown in Fig. 3. These are often
referred to as the Swedish re curves. They are derived from heat balance
calculations, using Kawagoes equation Eq. (1) for the burning rate of ventilationcontrolled res. Each group of curves is for a dierent opening factor Fv with fuel
load as indicated. Note that the units of fuel load are MJ/m2 of total surface area At
(not MJ/m2 of oor area Af which is more often used in design calculations). These
curves have been derived for p
a
compartment withpthermal
properties representative
of lightweight concrete, with thermal inertia krcp 1160 W s0:5 =m2 K; where
k is the thermal conductivity (W/m K), r is the density (kg/m3) and cp is the specic
heat (J/kg K). The rising branch of the curve for an opening factor of 0.04 is very
similar to the ISO 834 standard timetemperature curve.
When calibrating their model, Magnusson and Thelandersson [16] manipulated
the heat release rate to produce temperatures similar to those observed in short
duration test res. The resulting shape of the heat release rate curve used in those
calculations is shown in Fig. 4. The peak rate of heat release is the theoretical rate for
ventilation-controlled burning given by Eq. (1), assuming a caloric value of

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

89

Fig. 3. Timetemperature curves for dierent opening factors and fuel loads (MJ/m2 of total surface area).
[16]

Fig. 4. Burning rate and resulting temperatures used in the Swedish curves [16].

10.78 MJ/kg which is lower than the gure used by most other authors. Magnusson
and Thelandersson extrapolated their results to much higher fuel loads and a wider
range of opening factors than the available test data on which the computer model
was calibrated.

90

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

3. COMPF2
3.1. Summary
COMPF2 [2] is a widely used computer program for calculating temperatures in
post-ashover room res. COMPF2 is a single-zone model which solves the heat
balance equations to calculate gas temperatures during the re as described by
Feasey [3]. The program represents the ventilation of the compartment through a
single vertical ventilation aperture, with the ventilation
p limited combustion being
represented as a xed coecient multiplied by Av Hv as in Eq. (1), but with a
slightly higher coecient as described later. Compartment thermal properties must
be common for all surfaces, but the thermal properties such as conductivity and
specic heat can be varied as function of temperature. The program continually
compares the ventilation-controlled burning rate with several fuel-controlled
burning rates, and uses the lower of these to calculate the heat release rate at each
time step, and hence the gas temperatures in the compartment.
The COMPF2 program has three dierent means of characterising the fuel source:
liquid pool res, stick res, and wood crib res with dierent stick geometry and
with varying crib porosity. In theory, any material can be burnt in the program,
by dening the appropriate carbon: hydrogen: oxygen: nitrogen ratios for the fuel.
In the calibration of COMPF2 against real res, Feasey considered only cellulosic
material (primarily wood) as fuel, using both the crib re and stick re modes of
operation. Babrauskas [11] has developed a closed form approximation to the results
of COMPF2. The procedure, which is a little cumbersome for design purposes,
involves modifying a reference temperature with ve non-dimensional multiplicative
factors to allow for the combustion chemistry, transient and steady state wall losses,
window opening height, and combustion eciency accounting for both fuel rich and
fuel lean res. A summary is provided by Walton and Thomas [12].
3.2. The eect of fuel geometry
When running COMPF2 with wood fuel, it is immediately evident that
characterisation of the fuel geometry is critical to obtaining a reliable simulation.
For most opening factors, if the fuel has a small surface area to volume ratio, the re
is fuel surface controlled with a long slow temperature increase and decay, and low
maximum temperatures in the compartment. Alternatively, if the surface area to
volume ratio is very high, all the fuel is rapidly consumed, and COMPF2 will
calculate high gas temperatures which drop very rapidly after the fuel has burned
away.
For example, Fig. 5 shows several attempts to model the results of one particular
re test using COMPF2. In the test, the re load was 60 kg of wood cribs/m2 of oor
area, with an opening factor Fv of 0.055, and room construction of mixed masonry
materials. The low, mean and maximum measured temperatures in the room are
shown by the solid lines. The marked lines show the range of COMPF2 results which
can be obtained by varying only the stick diameter, and the crib spacing to height

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

91

Fig. 5. Variability in COMPF2 output caused by changing the input description of the fuel.

ratio. The change in possible fuel geometry leads to representations of the


experimental data which range from good to abysmal. Feasey [3] used such curves
to investigate the best way of characterising the fuel to get a good prediction of the
test results.
3.2.1. Burning of wood in COMPF2
The available mechanisms for wood burning in COMPF2 are:
*
*

crib res (where both the stick size and crib spacing to height ratio can be varied),
stick burning with dierent shape factors.

Within COMPF2, stick burning mode is automatically selected if the user species a
surface regression rate for the fuel. This value is typically 1.0  106 m/s, but there is
no restriction on the numerical value which can be used. The shape factor denes
whether the stick fuel has a small diameter compared with its length (F 2), or the
diameter and length are equal (F 3). These values correspond to four sided
burning or six sided burning respectively. The input surface regression rate is used by
COMPF2 along with the input stick diameter and shape factor to calculate the fuel
surface controlled burning rate.
Crib res are automatically selected by the program if the user species the surface
regression rate to be zero. In this case, the stick diameter and the crib spacing to
height ratio variable are user dened.
Babrauskas [2] presents dierent correlations for fuel surface controlled or
porosity controlled crib res which are embedded within COMPF2. For the former,
the surface regression rate for the wood in the crib is calculated within COMPF2
from the stick dimensions rather than specied by the user. For the latter, the fuel
surface regression rate is not explicitly used in the calculations. For both crib re
mechanisms, the pyrolysis rate is independent of the shape factor F used to describe
stick burning res.

92

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Whichever wood burning mechanism is selected, once the initial fuel pyrolysis rate
is calculated, the course of the re is fundamentally set, with the program calculating
the change from ventilation limited to fuel-controlled burning as required, until all
the fuel is consumed.
As coded within COMPF2, the stick burning fuel pyrolysis rate m is given by
1=F

m 2Vp F=Dm11=F M0

kg=s;

where Vp is the initial surface regression rate of the stick fuel (m/s), F the shape
factor (F 1; 2 or 3) for plane, stick-like or cubic fuel elements, D the stick diameter
(m), m the mass of fuel at any time (kg), and M0 the initial mass of fuel (kg).
As coded within COMPF2, the ventilation limited pyrolysis rate m is given by
p
4
m 0:12Av Hv kg=s:
3.3. Diculties with COMPF2
There are some diculties in using COMPF2. The COMPF2 computer code (in
Fortran) is not always totally consistent with the various papers describing its
background and operation [17,18,2]. For example, the code representing crib res
incorporates the stick burning shape factor F into the equation to calculate the fuel
pyrolysis rate whereas the theory contains no such parameter.
Another diculty is that COMPF2 starts with the compartment in a postashover condition, with initial temperatures up to 10001C and occasionally higher.
Given that post-ashover conditions are usually assumed to exist once compartment
temperatures reach about 6001C, a part of the post-ashover temperature prole can
be missing, and the fuel consumption required to reach this status is ignored. Heat
transfer to the surroundings only begins at the time of this initial calculated
temperature, possibly leading to a misrepresentation of the total energy absorbed
into the compartment surroundings compared with a real re which grows from
ambient temperature.
It is very evident that COMPF2 is a complex tool for commercial engineering
design purposes. Because it is over twenty years old, its interface is archaic, and both
input preparation and output processing are very time consuming. It is more suited
as a research tool than a cost-eective design tool for structural engineers. As a
research tool however, its predictions (when calibrated) can be used to investigate
alternative design res.
3.4. Calibration of COMPF2
A set of the NFSC test res from Arnault et al. [13,14] was selected, covering the
widest possible range of re loads, ventilation parameters and types of construction,
as described in Section 2.1. These tests were carried out with arrays of thermocouples
0.7 m below the ceiling and 1.05 m above the oor. For most tests, three time
temperature curves were published as shown in Fig. 5. The maximum curve
represented the highest temperature at any thermocouple for each time interval

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

93

throughout the re. The minimum curve represented the lowest temperature at
any thermocouple, and the middle curve represented the average of all
thermocouple readings for each time interval. The maximum curve was used as
calibration data for COMPF2, providing a slightly conservative representation of
the timetemperature history at any particular location in the re compartment. The
timetemperature curves for the maximum temperature in all of these res are shown
in Fig. 6, compared with the ISO 834 standard re curve.
The test res were simulated using all possible wood burning mechanisms encoded
into the COMPF2 program to determine whether any particular fuel characterisation gave superior results. Of the possible wood burning mechanisms, the stick
burning simulations with shape factor F 3 consistently gave the most reliable
simulations of the re test data regardless of whether the test fuel was wood cribs or
actual wood, paper and furniture mixtures. The other wood fuel representations
within the program provided good re simulations on many occasions, but not as
consistently as the stick burning model with nominally cubic elements (shape
factor F 3), especially in the temperature decay phase. For these simulations, a
relatively high combustion eciency factor of 0.98 provided good results for res
which were adequately ventilated. Within COMPF2, the heat released during the
fuel surface controlled combustion phase is the product of the mass generation rate
of pyrolysates, the combustion enthalpy of the pyrolysates and the combustion
eciency factor. Within reason, the higher value of combustion eciency can be
compensated by lower values for the mass ow rate of pyrolysates (which are a
function of fuel element geometry). For poorly ventilated res, much lower values
for the combustion eciency factor were required, to get good simulation of the test
re.
In an attempt to rationalise this data, the initial fuel pyrolysis rate was evaluated
for the best simulations of the test data using stick res with a shape factor F 3; as
a function of re load (kg wood per m2 of oor area) and opening factor Fv : For this
evaluation, an initial fuel pyrolysis rate ratio was calculated by dividing the initial

Fig. 6. NFSC res used for calibrating COMPF2.

94

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

fuel pyrolysis rate as calculated by the COMPF2 Eq. (3) by the initial ventilation
limited pyrolysis rate as calculated by COMPF2 Eq. (4).
All the parameters in Eq. (3) are eectively user-specied in the input data except
for m, the mass at any particular time. Naturally at the start of the re (time t 0),
m and M0 are equal. The program then calculates both the stick burning and
ventilation limited pyrolysis rates, and uses the lower of the two to calculate the heat
release rate within the compartment and then the temperature by energy balance.
The remaining mass of fuel m is progressively reduced based on the previous
pyrolysis rate m and the specied time increment, and the pyrolysis rate is
recalculated.
A systematic relationship between the initial pyrolysis rate ratio for time t= 0 of
the best simulations, and the independent variables of re load and opening factor
was evident. The best pyrolysis rate ratio was a weak function of opening factor at
high re loads, becoming a very strong function of both opening factor and re load
at low re loads. The systematic nature of this relationship indicated that relative to
the test data, reliable simulations could be achieved for other re loads and opening
factors, by selecting a fuel element size and surface regression rate to provide the
required initial pyrolysis rate ratio.
For design purposes, the initial pyrolysis rate ratios for the desired range of
opening factor and re load were determined by interpolation from the values
obtained from simulations of the NFSC test data. The initial assessment of the
preferred initial pyrolysis rate ratios presented in Feasey [3] has been re-visited. A
seven term polynomial has been tted to the calibration data, allowing a more
accurate prediction of the initial pyrolysis rate ratio for the range of design
parameters desired. Table 1 shows the preferred initial pyrolysis rate ratio for a range
of design res with re loads from 200 to 1200 MJ/m2 of oor area, and opening
parameter Fv from 0.02 to 0.12.
As originally noted [3], the test data for low opening factors were particularly
limited. In addition, COMPF2 did not reproduce the experimental temperature
curves as reliably for the limited data available for low opening factors. Hence the
initial pyrolysis rate ratios listed in Table 1 for calculating design res at opening
factor Fv 0:02 must be considered less reliable than those at higher opening
factors.
It can be seen that the initial pyrolysis rate ratio to be used in COMPF2 is greater
than 1.0 for res with high fuel loads and small opening factor. This means that the
Table 1
Initial pyrolysis rate ratio as a function of re load and opening factor
p
Opening factor Fv Av Hv =At
Fire load (MJ/m2 of oor area)

0.02
0.04
0.08
0.12

200

400

800

1200

0.99
0.62
0.23
0.18

1.37
1.00
0.61
0.45

1.53
1.16
0.77
0.62

1.57
1.20
0.81
0.66

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

95

initial pyrolysis rate must be signicantly more than the ventilation-controlled


pyrolysis rate for these res if good calibration is to be achieved. Conversely the ratio
becomes very small for small re loads and larger ventilation openings.
It is important to note that the above characterisation of the fuel as wood sticks of
a particular geometry is not a description of the actual fuel used in the re tests. The
fuel in some re tests was wood cribs, and real furniture in others. The
characterisation used in this study followed an investigation into which range of
fuel input variables in the COMPF2 model would give the most realistic time
temperature curves across the range of various fuel types. This characterisation was
then used to generate design res which are believed to be applicable to a wide range
of cellulosic fuel geometries.
3.5. Recommended use of COMPF2
Following the calibration exercise described above, the following recommendations are given for use of COMPF2 for the prediction of post-ashover res.
*
*

*
*

*
*

dene all the compartment geometric and thermal properties,


enter the ventilation aperture properties (aperture area Av and aperture height
Hv ),
enter the fuel load in units of kg of wood per m2 of oor area (this parameter
multiplied by the oor area within the program gives the initial fuel load M0 ),
set the fuel shape factor to F 3;
select stick burning mode by specifying the fuel surface regression rate Vp to be
greater than zeroFa typical realistic value is 1  106 m/s,
using the known re load energy density and opening factor, determine the
required initial pyrolysis rate ratio from interpolation of Table 1,
calculate the ventilation-controlled pyrolysis rate using Eq. (4),
calculate the initial fuel pyrolysis rate m by multiplying the initial pyrolysis rate
ratio from Table 1 with the ventilation limited pyrolysis rate from Eq. (4),
using Eq. (3) select a stick diameter D which coupled with the specied surface
regression rate Vp provides the required initial pyrolysis rate m ;
set the ventilation aperture discharge coecient Cd equal to 0.68 for all res
except those where the aperture is the full width or near to the full width of the
wall, in which case reduce the coecient Cd to 0.340.40,
set the combustion eciency factor to 0.98 unless the opening factor Fv is less
than 0.03, in which case set the value to 0.80.
run the COMPF2 program.

3.6. Design res


Based on runs of COMPF2 with the initial pyrolysis rate ratios from Table 1,
design res were developed for compartments with two dierent construction
materials, three fuel loads and four opening factors. The re loads are 400, 800 and
1200 MJ/m2 of oor area. The calculations were made for a room 5 m  5 m plan

96

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

area and 3 m high, with a single opening in one wall. The resulting design res are
shown in Fig. 7.
The upper half of Fig. 7 shows design res for compartments of lightweight
construction. The thermal properties are representative of gypsum board, with wall
density r 720 kg/m3, conductivity k 0:2 W/m K and specic heat cp 1130 J/
kg K, providing
a value
of b 400 W s0.5/m2 K, where b is the square root of thermal

p
inertia (b krcp ). The lower half of Fig. 7 shows design res for compartments of
heavy construction. The thermal properties are representative of concrete [20] with
density r 2300 kg/m3, conductivity k 1:6 W/m K and specic heat cp 980 J/
kg K, providing a value of b 1900 W s0.5/m2 K. A typical commercial oce
building with a mixture of these materials on the walls and ceiling would give curves
between these two, similar to a building made from lightweight concrete.

4. Eurocode parametric re curves


The Eurocode [1] gives an equation for parametric res, allowing a time
temperature relationship to be produced for any combination of fuel load,
ventilation openings and wall lining materials. The Eurocode parametric re curves
were derived to give a good approximation to the burning period of the Swedish
curves shown in Fig. 3.
4.1. Equation for burning period
The Eurocode equation for temperature T (1C) is
T 13251  0:324e0:2t *  0:204e1:7t *  0:472e19t * ;

where t is a ctitious time (h) given by


t * Gt;

where t is the time (h) and the modication factor gamma G can be expressed as
G

Fv =Fref 2
;
b=bref 2

where Fv is the opening factor given by


p
Fv Av Hv =At ;

p p
opening
factor,
b
the
thermal inertia krcp and
Fref is the reference value ofpthe

bref the reference value of krcp (W s0.5/m2 K).


Eq. (5) is a good approximation to the ISO 834 standard re curve for
temperatures up to about 13001C. Hence the Eurocode parametric re curve is
close to the ISO 834 curve for the special case where Fv Fref and b bref : Larger
ventilation openings or more highly insulated compartments will result in higher
room temperatures. Smaller openings and poorly insulated compartments will result
in lower temperatures. The Eurocode species a lower limit of b 1000 W s0.5/m2 K

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Fig. 7. Design res, from COMPF2 output, calibrated to NFSC res.

97

98

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

which is far higher than the typical value for gypsum plaster board of about
400 W s0.5/m2 K.
In the Eurocode, the value of Fref is 0.04 and the value of bref is 1160 such that
GEC

Fv =0:042
:
b=11602

4.2. Duration of burning period


The equation for the duration of the burning period td (h) in the Eurocode
simplies to
td 0:00013et =Fv ;
0:00013E=Av

9
p
Hv ;

10

where et is the fuel load (MJ/m2 total surface area), E is the total energy content of
the fuel (MJ).
It is interesting to compare this duration with the theoretical duration for
ventilation-controlled burning obtained from Eq. (1). Assuming a value for heat of
combustion of DHc 15:5 MJ/kg, the duration of the burning period given by
Eq. (10) is only 67% of the theoretical duration. The reason for this is not stated in
the Eurocode, but it is probably intended to allow for some burning to take place
during the decay period of the re. If the heat release rate is constant during the
burning period with a linear decay rate, the implied curve of heat release rate vs. time
is shown in Fig. 8 where the duration of the decay period is equal to the duration of
the burning period.
4.3. Decay rate
The Eurocode uses a reference decay rate dT=dtref equal to 6251C/h for res with
a burning period less than half an hour, decreasing to 2501C/h for res with a

Fig. 8. Heat release rate implied by Eurocode parametric re.

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

99

burning period greater than 2 h, as shown in Fig. 9, based on the ISO 834 testing
standard [19]. In the Eurocode, the reference rate of temperature decay is modied
by using the ctitious time from Eqs. (6) and (8), which can lead to actual decay rates
being unrealistically fast or slow for certain combinations of opening factor and
construction materials.
4.4. Comparison with COMPF2 output
Fig. 10 shows a comparison of the Eurocode equation with the design res
produced by COMPF2 for both concrete and gypsum-lined rooms. The solid heavy
lines are the COMPF2 output (the same curves as on Fig. 7). The solid thin lines are
the timetemperature curves predicted by the Eurocode equation. It is clear that the
Eurocode equations often predict temperatures which are too low. The Eurocode
equation gives extremely fast decay rates for large openings in well insulated
compartments and extremely slow decay rates for small openings in poorly insulated
compartments.
The re parameters in Fig. 10 are the same as those in Fig. 7. The top half of
Fig. 10 shows the curves for lightweight construction typical of gypsum plaster
board (b 400 W s0.5/m2 K), and the lower half shows the curves for heavyweight
construction typical of normal weight concrete (b 1900 W s0.5/m2 K).
4.5. Modication of Eurocode formula
In an attempt to produce a simple design equation, the Eurocode equation has
been manipulated to produce a better t. The Eurocode equation for the burning

Fig. 9. Reference rate of temperature decay in Eurocode parametric res.

100

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Fig. 10. Comparison of Eurocode parametric res with COMPF2 output.

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

101

period assumes that a re in a room with an opening factor of 0.04 and intermediate
building materials (b=1160 W s0.5/m2 K) will have a temperature curve close to the
standard curve. Inspection of the COMPF2 output shows that the re temperatures
in a room of dense concrete (b 1900) is closer to the standard curve.
This leads to a simple modication to the equation, using a value of bref 1900:
The recommended post-ashover design re is therefore obtained from Eqs. (5)(7),
using
GDESIGN

Fv =0:042
:
b=19002

11

The ctitious time used in the Eurocode for the burning period has never been
justied for use in the decay period. It has been shown above (Fig. 10) that the
resulting decay rates are not realistic. In an attempt to make a simple modication, it
has been found by trial and error that if the squared terms in Eq. (7) are changed to
square roots, a much better t is obtained. The resulting decay rate is given by
dT=dt dt=dtref GDECAY ;

12

where
GDECAY

p
Fv =0:04
p :
b=1900

13

This is equivalent to using a second ctitious time, similar to that in the growth
period, but using square root rather than squared terms to give a much better t to
test results and computer simulations.
Fig. 11 shows the modied parametric timetemperature curves plotted for the
same range of opening factors, fuel loads and materials as used in Fig. 10. The
temperatures in the burning period have been calculated from Eqs. (5), (6) and (11).
The rate of temperature decay is from Eq. (12). The Eurocode suggestion of limiting
the lower value of b to 1000 W s0.5/m2 K has not been followed.
The match between the modied parametric re curves and the COMPF2 output
is very good, considering the empirical nature of the equations and the large number
of parameters involved. The worst t is for two curves with a fuel load of 400 MJ/m2
in a gypsum-lined room with big windows (opening factor Fv 0:08 and 0.12) but
these are on the safe side. There are some other noticeable dierences, especially at
peak temperatures, but otherwise the modied equations match the design res very
well. Inspection of Fig. 11 shows that the slope of the parametric timetemperature
curve in the decay phase is a far better t than the Eurocode formulation in Fig. 10.
These modied equations are recommended for design purposes [21].
In this process of assessing the proposed parametric re curves, COMPF2 output
was used rather than the raw test results because this allowed a wider range of re
conditions to be considered, and permitted comparisons at several clearly dened
steps in the range of input variables.

102

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Fig. 11. Comparison of modied parametric res with COMPF2 output.

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

103

5. Future developments
Work is in progress to address some of the issues raised in this paper. Yii [22] is
modifying the COMPF2 program to incorporate new material which has become
available since the code was originally written. A better description of fuel items is
necessary in order to calculate the decay period more accurately. Other ventilation
openings and progressive burning are also being investigated.

5.1. Other ventilation openings


Ventilation-controlled res are very sensitive to the size and location of openings.
The presence of a ceiling opening allows combustion products to exit the ceiling
opening while cool air enters the window. This signicantly increases the ventilation
to the re as shown in Fig. 12.
A room with openings on two opposite walls may have cross ventilation, especially
if there is a wind blowing as shown in Fig. 13, producing increased rates of burning.
No research has been done on this type of scenario, other than some estimates for
external steel structures exposed to res from windows [23].
A recent series of tests in a long narrow room is described by Kirby et al. [24]. This
room was approximately 6 m wide and 20 m deep with uniformly distributed wood
cribs as fuel. Even when the re was ignited at the end farthest from the window,
burning moved quickly to the window end, then progressed slowly from that end
towards the back as shown in Fig. 14. Clifton [25] has proposed a model for re
spreading within a large compartment. Thomas and Bennetts [8] reported similar
behaviour in small scale experiments, nding that most res burn progressively with
the object nearest the window burning rst, and often delaying burning of other
items in the room. These ndings have signicant implications for modelling of postashover res, and are leading to renewed research eorts.

Fig. 12. Vent ows for room with ceiling opening.

104

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

Fig. 13. Vent ows for two windows, with wind blowing.

Fig. 14. Progressive burning in deep room with one window.

6. Conclusions
This paper proposes modications to the Eurocode parametric re curves to give a
better estimation of temperatures in post-ashover compartment res. The
recommended changes to the Eurocode parametric re are:
p
1. Change the reference value of
krc
to 1900 W s0.5/m2 K.
p from 1160
p
0.5
2. Remove the lower limit of krcp 1000 W s /m2 K, and use the actual value
calculated for the construction materials.
3. Calculate the rate of temperature decay using a new ctitious time which is based
on square root rather than squared terms.
These modications are based on design res calculated using the COMPF2
program, calibrated using results from a large number of realistic test res.
The use of COMPF2 for calculating temperatures in post-ashover compartment
res has been described, with emphasis on the characterisation of the fuel in the
input data, and the initial pyrolysis rate. Recommendations have been given for
using COMPF2 to achieve realistic results for a wide range of opening factors, fuel
loads and construction materials.
The paper supports the urgent need for further research into post-ashover re
behaviour.

References
[1] EC1. Eurocode 1: basis of design and design actions on structures. Part 2-2: actions on structures
exposed to re. ENV 1991-2-2, European Committee for Standardization, Brussels, 1994.
[2] Babrauskas V. COMPF2: a program for calculating post-ashover re temperatures. NBS Technical
Note 991, National Bureau of Standards, 1979.

R. Feasey, A. Buchanan / Fire Safety Journal 37 (2002) 83105

105

[3] Feasey R. Post-ashover design res. Fire Engineering Research Report 99/6, University of
Canterbury, New Zealand, 1999.
[4] Kawagoe K. Fire behaviour in rooms. Report No. 27, Building Research Institute, Tokyo, 1958.
[5] Drysdale D. An introduction to re dynamics, 2nd ed. Chichester, UK: Wiley, 1998.
[6] Thomas PH, Heselden AJM. Fully developed res in single compartments. CIB Report No. 20, Fire
Research Note 923, Fire Research Station, UK, 1972.
[7] Law M. A basis for the design of re protection of building structures. Struct Eng 1983;61A(1).
[8] Thomas IR, Bennetts ID. Fires in enclosures with single ventilation openingsFcomparison of long
and wide enclosures. Proceedings of the Sixth International Symposium on Fire Safety Science,
Poitiers, France, 1999.
[9] Shields TJ, Silcock GWH. Buildings and re. Longman Scientic and Technical, UK, 1987.
[10] Yii J. Eect of surface area and thickness on re loads. Research Report 00/13, University of
Canterbury, New Zealand, 2000.
[11] Babrauskas V. A closed form approximation for post-ashover compartment res. Fire Saf J
1981;4:6373.
[12] Walton WD, Thomas PH. Estimating temperatures in compartment res. SFPE Handbook of Fire
Protection Engineering, Society of Fire Protection Engineers, USA, 1995 (Chapter 36).
[13] Arnault P, Ehm H, Kruppa J. Rapport Exp!erimental sur Les Essais Avec des Feux Naturels Ex!ecut!es
Dans La Petite Installation de Maizi!eres-l"es-Metz. Centre Technique Industriel de la Construction
M!etallique Report No. CECM 3/73611-F, 1973.
[14] Arnault P, Ehm H, Kruppa J. Incendies Naturels avec des Meuble set du Papier Ex!ecut!es Dans La
Petite Installation de Maizi!eres-l"es-Metz. Centre Technique Industriel de la Construction M!etallique
Report No. 2.10.20-3, 1974.
[15] Buchanan AH. Modelling post-ashover res with FASTLite. J Fire Prot Eng 1997;9(3):111.
[16] Magnusson SE, Thelandersson S. Temperaturetime curves of complete process of re development;
Theoretical study of wood fuel res in enclosed spaces. Acta Polytech Scand, Civil Engineering and
Building Construction Series 65, 1970.
[17] Babrauskas V, Williamson RB. Post-ashover compartment res: basis of a theoretical model. Fire
Mater 1978;2(2):3953.
[18] Babrauskas V, Williamson RB. Post-ashover compartment res: application of a theoretical model.
FireMater 1979;3(1):17.
[19] ISO. Fire resistance testsFelements of building construction. ISO 834, International Organization
for Standardization, 1975.
[20] EC2. Eurocode 2: design of concrete structures. ENV 1992-1-2: general rulesFstructural re design.
European Committee for Standardisation, Brussels, 1993.
[21] Buchanan AH. Structural design for re safety. Chichester, UK: Wiley, 2001.
[22] Yii E. Post-ashover compartment res. Ph.D. thesis, University of Canterbury, New Zealand, 2001
(in preparation).
[23] Law M, OBrien T. Fire safety of bare external structural steel. UK: Constrado, Croydon, 1981.
[24] Kirby BR, Wainman DE, Tomlinson LN, Kay TR, Peacock BN. Natural res in large scale
compartmentsFA British Steel Technical, Fire Research Station Collaborative Project. British Steel
Technical, Swinden Laboratories, UK, 1994.
[25] Clifton GC. Fire models for large recells. HERA Report R4-83, New Zealand Heavy Engineering
Research Association, Auckland, New Zealand, 1996.

You might also like