Experiments and Large Eddy Simulation of Underventilated Pool Fires

Experiments and Large Eddy Simulation of Underventilated Pool Fires
S. E. Yakush *1 , G. M. Makhviladze 2 , A.V. Chamchine 2 , P. Oleszczak 2
1 A.Yu. Ishlinskii Institute for Problems in Mechanics, Moscow, Russia.
2 University of Central Lancashire, Preston, UK
Abstract
Results of experimental and computational studies on the transient flame behavior in a compartment fire in the conditions
of limited ventilation are presented. Hydrocarbon pool fires in a small-scale fire box with a single rectangular
ventilation opening are considered. In the conditions where the oxygen supply into the compartment is insufficient
to maintain steady combustion (underventilated fires), the flame becomes essentially transient and can be thrown out
of the ventilation opening into the atmosphere. In the experiments, temperature-time curves as well as video recordings
demonstrating the flame exhaust are obtained for four opening geometries. Large Eddy Simulations (LES) of
underventilated fires are carried out and compared with the experiments.
Introduction
Flame behaviour in compartment fires has been a
focus of numerous experimental and theoretical works.
It is recognized that in the underventilated conditions,
which are encountered when the oxygen supply is insufficient
to maintain the required fuel burning rate, the
flame becomes essentially transient. In some cases,
flame extinction occurs, while a more dangerous scenario
is when the flame reaches the ventilation opening
and is thrown out into the atmosphere [1-3]. This leads
to external flaming and can promote rapid fire spread
over a multi-compartment building, or pose the danger
of igniting nearby structures. To assess the fire spread
hazards, it is important to know the relationship between
the time from ignition to the flame exhaust, and
the fire source parameters, geometry of the compartment
and openings, position of fire source etc.
Flame exhaust in underventilated fires has mostly
been studied experimentally and theoretically using the
burner fire approach, when a gaseous fuel is supplied at
a prescribed mass flowrate [4-6]. In real fires, however,
the burning rate is not constant, but is controlled by the
positive energy feedback from the flame to the fire bed.
Thus, an important problem is to study the flame behavior
in underventilated conditions for realistic fire
sources.
In this paper, recent experiments on transient pool fire
development carried out in a small-scale fire box with a
single rectangular opening are presented. Metal pans of
different size filled with liquid hydrocarbon fuels (petroleum
ether) served as a fire source. Temperature
measurements were carried out by several thermocouples
to provide data on hot layer development and
flame emergence out of the ventilation opening. The
main purpose of the work was to obtain the critical
conditions for flame exhaust and to evaluate the induction
time in different conditions. Also, numerical simulations
of pool fire development in the conditions of
insufficient ventilations are carried out and compared
with the experimental observations. *
* Corresponding author: yakush@ipmnet.ru
Proceedings of the European Combustion Meeting 2009
Experimental Setup
In Fig. 1, the general view of the experimental facility
available at the University of Central Lancashire
(UCLAN) which was used for the studies of flame exhaust
in the current work is shown.
Fig. 1. Experimental facility at UCLAN
The main element of the facility is the fire box having
the internal dimensions of 0.72×0.48×0.48 m, which
corresponds to 1/5 of the standard ISO room. The
25 mm thick walls of the box are made from Monolux
500 insulating material. The box has a rectangular opening
in its front wall, centred horizontally and abutting
the box floor; the sizes of the opening can be changed
by using appropriate inserts. In the experiments, four
openings were used with the sizes (height H, width W,
and area A) listed in Table 1.
Table 1. Geometry of ventilation openings
No. H, m W, m A, m 2
1 0.15 0.125 1.88·10 -2
2 0.14 0.16 2.24·10 -2
3 0.18 0.08 1.44·10 -2
4 0.19 0.16 3.04·10 -2

The box is equipped with a system of thermocouples
located on side walls (14 thermocouples), ceiling (4
thermocouples) and in the opening (4 thermocouples).
Positions of the thermocouples are shown in Fig. 2. The
thermocouple signals are collected in Thempscan-1100
data logger and forwarded to an acquisition system
installed on PC computer.
Fig. 2. Positions of thermocouples in the firebox
A square metal container was used as a fire source.
Five sizes of the container were used in the tests, with
the length of the container side equal to 60, 100, 120,
150 and 200 mm. The height of the container walls was
equal to 25 and 50 mm.
Different kinds of liquid fuels were used in the studies:
methanol, acetone, petrol, and petroleum ether.
After some initial tests, petroleum ether was chosen for
further research, since it gave more repeatable results.
Some of the experiments with pool fires were repeated
in the fire box with a glass side wall in order to visualize
the fire development and flame exhaust process.
Experimental Results
In all the experiments on pool fires, the pan was
filled with an initial volume of fuel equal to 100 ml. The
fuel container was located at the centre of the fire box
floor. Temperature records from all thermocouples were
recorded, and the induction time (i.e., the time between
the ignition and appearance of visible flame in the opening)
was obtained from the temperature curves as well
as from visual observations. Typically, it took between
1.5 and 5 minutes for the flame exhaust to occur.
In Fig. 3a the time histories of the ceiling temperatures
obtained for opening 1 (see Table 1) are plotted for
different pool sizes. In Fig. 3b, the temperature histories
recorded by the top of the four thermocouples located in
the opening are presented. The time to flame exhaust
(induction period) is indicated by the peak on the opening
temperature graph (the corresponding times for
different pool sizes are denoted in Fig. 3 by vertical
dashed lines). The dependence of the flame exhaust
time on the pool size is shown in Fig. 4 for the two pan
height, 25 and 50 mm.
2
T, [C]
T, [C]
500
400
300
200
100
0
0
500
50 100 150 200 250 300 350
120 mm
(b)
400
300
200
100
120 mm
150 mm
150 mm
100 mm
(a)
100 mm
0
0 50 100 150 200 250 300 350
Time, [s]
Fig. 3. Temperature histories for opening 1: (a) ceiling,
(b) opening
Flame exhaust time, [s]
300
250
200
150
100
Pan height
25 mm
50 mm
100 120 140 160 180 200
Pool Size, [mm]
Fig. 4. Dependence of flame exhaust time on pool size
for opening 1

For each experiment, the average mass loss rate was
estimated; also, the specific mass loss rate (per unit area
of the pool) was also determined. In Fig. 5a,b, these are
plotted against the pool size. It can be seen that, with the
increase in the pool size, the absolute mass loss rate
grows, but not proportionately to the pool area: the
specific mass loss rate decreases with the pool size,
probably due to the decrease in the radiative feedback.
Different depths of the pool could also result in different
times necessary to heat the pan contents to boiling temperature,
which could also contribute to the differences
in the average specific evaporation rates.
The experiments have shown that the smallest size
of fuel pan required to obtain flame exhaust is equal to
100 mm. It was observed that for a pan with side length
of 200 mm and height of 50 mm there were strong oscillations
of the flame and its extinction. However, for a
container with height of 25 mm the flame exhaust was
observed.
The data in Figs. 4 and 5 demonstrate the typical degree
of data scatter observed in the experiments performed
for the same initial conditions. This is due to the
intrinsic randomness of the flame exhaust phenomenon.
Similar results were obtained for other openings (see
Table 1). In Figs. 6-8, the temperature curves under the
ceiling and in the opening, as well as the flame exhaust
times and mass loss rates (absolute and specific) obtained
in the experiment with the largest opening 4 are
presented.
Mass loss rate, [g/s]
Specific mass loss rate, [g/m 2 s]
0.40
0.35
0.30
0.25
0.20
0.15
20
18
16
14
12
10
8
6
(a)
100 120 140 160 180 200
(b)
Pan size
25 mm
50 mm
100 120 140 160 180 200
Pool size, [mm]
Fig. 5. Dependence of absolute (a) and specific (b) mass
loss rate for opening 1
An interesting feature of the mass loss rate data is
that for the smaller opening (Fig. 5), as well as for openings
2 and 3, the absolute mass loss rate for the largest
pan size (200 mm) turns out to be lower than for pan
size of 150 mm. This was not observed in the experiments
with opening 4 (Fig. 8), where the mass loss rate
increased monotonically with the pan size. Also, for
opening 4, no flame exhaust occurred for the fuel pan of
25 mm height, so only results for 50 mm high pan are
plotted.
3
T, [C]
T, [C]
500
400
300
200
100
0
0
600
50 100 150 200 250 300 350
500
150 mm 120 mm (b)
400
300
200
100
150 mm
200 mm
200 mm
120 mm
(a)
100 mm
100 mm
0
0 50 100 150 200 250 300 350
Time, [s]
Fig. 6. Temperature histories for opening 1: (a) ceiling,
(b) opening
Flame exhaust time, [s]
250
200
150
100
Pan height
50 mm
100 120 140 160 180 200
Pool size, [mm]
Fig. 7. Dependence of flame exhaust time on pool size
for opening 4

Mass loss rate, [g/s]
Specific mass loss rate, [g/m 2 s]
0.50
0.45
0.40
0.35
0.30
0.25
0.20
30
25
20
15
10
(a)
100 120 140 160 180 200
(b)
Pan size
50 mm
100 120 140 160 180 200
Pool size, [mm]
Fig. 8. Dependence of absolute (a) and specific (b) mass
loss rate for opening 4
Thus, it can be seen that the flame exhaust time depends
on ventilation opening geometry. The difference
between the pool fires and burner fires is that in the
latter case the fuel supply rate is a controlled parameter,
whereas in the former case it depends on the feedback
between the flame and fuel pool. It is interesting to see
how the time to flame exhaust behaves as a function of
mass loss rate. Such a dependence, summarizing the
results of all experiments performed, is presented in
Fig. 9.
Flame exhaust time, [s]
350
300
250
200
150
100
50
Opening Pan height
1 50 mm
2 50 mm
3 50 mm
4 50 mm
1 25 mm
2 25 mm
3 25 mm
0.2 0.3 0.4 0.5
Mass loss rate, [g/s]
Fig. 9. Summary of flame exhaust time dependence on
mass loss rate in experiments with different openings
and fuel pans
Visualization of Flame Exhaust
To reveal the features of transient flame development
in undeventilated conditions, several experiments
were carried out with one side wall of the fire box replaced
by a glass wall (the remaining walls were made
from Monolux 500, as in the previously described experiments).
Also, the internal dimensions of the fire box
and other experimental conditions (opening geometry,
fuel source etc) remained the same.
Due to its transparency, the glass wall increased the
heat losses from the fire box, which affected the time to
flame exhaust. For some combinations of the opening
and pool sizes, it was even impossible to observe flaming
in the opening. Nevertheless, there were found configurations
where the results obtained in the glass box
were very similar to those obtained in the Monolux box
in terms of the flame exhaust times. For these configurations,
video recording was carried out.
In Fig. 10, six frames of such a video recording are
presented. The configuration corresponds to the fuel pan
size of 150 mm and ventilation opening 3 (see Table 1).
Note that this opening is rather narrow and tall (width is
8 cm and height is 18 cm) and has the smallest vent are
of all openings studied. At the same time, the fuel pan is
rather large. Therefore, oxygen starvation inside the box
is reached early, and flame exhaust was reliably obtained
in all experiments carried out for this configuration.
Stronger oscillations of the flame were observed
with fast ejection of the flame out of the box to distances
up to 0.5 m from the box.
Fig. 10. Snapshots of flame exhaust video recording for
fuel pan size 150 mm and opening 3
4

Large Eddy Simulations
Numerical modeling of underventilated pool fires
was performed using Fire Dynamics Simulator (FDS),
Version 5, developed at NIST, with the accompanying
Smokeview viewer for visualization of results [7-9].
Earlier, FDS was successfully applied to modelling of
underventilated burner fires [6]. In this work, the fire
source was provided by evaporating liquid fuel, and the
fuel supply rate is determined by the heat flux on the
pool surface due to conduction and radiation. Thus, the
evaporation rate cannot be set ad hoc, rather, it is determined
in the course of the solution.
The combustion model used in all LES simulations
is that available in FDS by default and based on the
mixture fraction approach in which a transport equation
for the mixture fraction (defined to be 1 in the fuel and 0
in the oxidizer) is solved [7, 8]. The state relations for
the fuel (heptane was used to simulate the petroleum
ether which is a mixture of several hydrocarbons) are
applied to restore the species concentrations and assess
the volumetric heat release rate. The reaction is assumed
to proceed at an infinite rate, so that the fuel and oxidized
cannot co-exist on both side of the surface where
the mixture fraction corresponds to the stoichiometric
conditions. It should be noted that, according to this
model, the fuel ignites immediately upon entering the
atmosphere, rather that after reaching the ignition
source.
The calculations have shown that simulations by
FDS give adequate qualitative picture of the process, as
was the case when burner fires were modeled previously
[6]. However, quantitative characteristics of the process,
most importantly, time to flame exhaust, are very sensitive
to such parameters as absorption coefficient and
thermal properties of pool liquid, which are somewhat
uncertain because petroleum ether is a mixture of several
hydrocarbons. Also, convection in the burning pool is
known to affect the burning rate, but is not taken into
account in the FDS model. More detailed studies are
required to assess the sensitivity of flame exhaust time
to each parameter. In this paper, we present the results
of preliminary calculations which demonstrate the predictive
capabilities of LES model.
In Figs. 11 and 12, the results obtained for the geometry
of fire box used in the experiments described in
the previous sections are presented. The opening was
chosen to be 0.18 m high and 0.1 m wide, the fuel
source was a square pan with the side of 150 mm.
In Fig. 11, the distributions of volumetric heat release
rate are presented (visualized by Smokeview
viewer [9]) at four time instants, representing the main
stages of the process: initial fuel-controlled combustion
at t = 7 s, oxygen starvation and slumping of flame to
the floor where oxygen is still available at t = 70 s,
flame propagation towards the ventilation opening and
its emergence outside the fire box at t = 150 s, and,
finally, developed external flaming at t = 240 s. The
temperature fields in the plane of symmetry are shown
at the same instants in Fig. 12. The time to flame exhaust
obtained in the calculations (about 140 s) is of the
same order as observed experimentally with the same
pool size for openings 2 and 3, but more detailed comparison
and analysis have yet to be done.
Conclusions
Experiments with underventilated pool fires have
shown that the time to flame exhaust strongly depends
on the size of the fuel source, compartment geometry
and ventilation factor. The temperatures measured inside
the box during the experiments were relatively low
(in the range of 220÷310°C). The height of the fuel pan
walls also affects the time to flame exhaust.
In the pool fire experiments it was found that flame
exhaust process is much more repeatable than in the
burner fire experiments. It seems that pool fires are
“self-organizing”: the fuel supply rate is controlled by
the radiative feed from the flame. For the same fire box
configuration, it results in more stable and repeatable
flame development, at least in terms of flame exhaust.
CFD modeling is capable of capturing the main features
of underventilated pool fires, although detailed
parametric study of coupled physical phenomena has
yet to be done.
Acknowledgments
The research was supported by EPSRG (Grant No.
GR/S69122/01) and Russian Science Support Foundation.
References
1. Bullen, M. L., and Thomas, P. H., Proc. Comb.
Inst., 17: 1139-1148 (1978).
2. Drysdale, D., An Introduction to Fire Dynamics
(2 nd edition). Wiley, Chichester, UK, 1999.
3. Chamberlain, G. A., Trans. IchemE, B, 72(B2):211-
219 (1994).
4. Sniegirev, A. Yu., Makhviladze, G. M., Talalov,
V. A., and Chamchine, A. V., Combustion, Explosion
and Shock Waves, 3(1):1-10 (2003).
5. Sniegirev, A. Yu., Makhviladze, G. M., Talalov,
V. A., Isaev, S. A., and Chamchine, A. V., Proc. 3 rd
Conf. On Heat Transfer RNKT-3, Moscow, Russia,
3: 227-230 (2002).
6. Makhviladze, G. M., Shamshin, A. V., Yakush,
S. E., and Zykov, A. P., Combustion, Explosion,
and Shock Waves, 42(6):723-730 (2006).
7. McGrattan, K., Hostikka, S., Floyd, J., Baum, H.,
Rehm, R., Mell, W., McDermott, R., Fire Dynamics
Simulator (Version 5): Technical Reference
Guide. NIST Special Publication 1018-5 (2008 Edition);
92 p. July 2008.
8. McGrattan, K., Klein, B., Hostikka, S., and
Floyd, J., Fire Dynamics Simulator (Version 5):
Users Guide. NIST Special Publication 1019-5
(2008 Edition); 188 p. July 2008.
9. Forney, G., User's Guide for Smokeview Version 5:
A Tool for Visualizing Fire Dynamics Simulation
Data. NIST Special Publication 1017-5 (2008 Edition),
142 p. July 2008.
5

Fig. 11. Large eddy simulation of flame exhaust in underventilated
pool fire: distributions of volumetric heat
release rate are shown at t = 7,
70, 150 and 240 s after
ignition (top to bottom)
Fig. 12. Large eddy simulation of flame exhaust in underventilated
pool fire: temperature distributions in the
central plane at t = 7,
70, 150 and 240 s after ignition
(top to bottom)
6