Escolar Documentos
Profissional Documentos
Cultura Documentos
BACKDRAUGHT
IN FIRE ENGINEERING
BY
P.M. COURTNELL.
ACKNOWLEDGEMENTS Vill
SUMMARY ix
1. INTRODUCTION 1
2.2(i) Flashover. 10
2.2(ii) Backdraught. 12
1
3. A CONCEPTUAL APPROACH TO THE
FLASHOVERANDBACKDRAUGHT
PHENOMENA. 22
3.1 Semenov's Theory. 22
Fires. 27
Flashover. 34
Backdraught. 37
Current 40
DIFFUSION. 44
ii
4.2 What's Burning? 52
5 EXPERIMENTAL EQUIPMENT. 54
5.1 Introduction 54
5.2 Equipment 55
~ EXPERIMENTAL RESULTS 62
6.1(i) Introduction 62
Box. 64
with Oxygen(Air). 65
(U.E.L-Experiment 2) 67
iii
Propane prior to Ignition 69
7 CONCLUSIONS 90
APPENDICES 96
Appendix A Flammability Limits 96
REFERENCES. 100
BIBLIOGRAPHY 109
IV
ILLUSTRATIONS AND TABLES
ILLUSTRATIONITABLE TITLE
PAGE
Compartment Fire
Compartment
Exothermic Reaction
Compartment Fires
v
Figure 11 Quasi-Steady State Analysis of 37
Backdraught
Speed
Propane
Ignition
Table 1 Results 63
VI
PHOTOGRAPHS AND CHARTS
Box l02a-102g
Thermal Images
Vll
ACKNOWLEDGEMENTS
Thanks must be given)o the following people for their invaluable assistance during the
course of this dissertation. Members of Greater Manchester Fire Service, namely L.Fm
Chris Killeen and Sub.Off Alan Cleasby of the Breathing Apparatus School. Also Fm
Phil Sweeney of the video production unit at Thompson St Training Centre.From the
University of Central Lancashire, the research student Tsz Man Yau who provided the
technical expertise in use of the data logger and associated equipment. Thanks must also
go to Mrs F.E. Courtnell for assistance in typing of the dissertation and Mrs C.M.
viii
SUMMARY
The purpose of this dissertation is to review the current knowledge of the phenomenon of
backdraught. The review begins by examining early references to such an occurrence and
continues to present day where modern building practices have made the likelihood of an
event of this nature more probable. As a result the terms of backdraught and flashover
have been brought to the fore in the training of fire service personnel.
The definitions of both backdraught and flashover are then examined, particularly in
consequence of the concepts introduced by Giselsson and Rosander of the Swedish Fire
and Rescue Service. Training techniques introduced by these authors to the British Fire
Service has brought about the use of their conflicting terminology with that used by fire
scientists.
Thomas et al for quasi-steady state analysis are examined and the mechanistic theories of
C.M.Fleischmann. This introduces the experimental part of the dissertation where the
where it is proposed that the use of a gravity current in the mechanistic theory of a
IX
1. INTRODUCTION.
One of the earliest references to the backdraught phenomenon dates from as early as
1914. The National Fire Protection Association (N.F.P.A.) quarterly ran an article
smoke.
In Britain one of the first references was made by the former Chief Fire Officer of
London Fire Brigade, Sidney Gompertz Gamble as reported by Paul Grimwood[ 1].
In his book of 1931 , "Outbreaks of frre, Their Causes and Means of Prevention". He
"Backdraught is the sudden ignition of inflammable dust in the air caused by organic
substances that have become heated by the frre. Owing to a lack of oxygen,
combustion is delayed until a window is broken or a door opened. When the inrush
of cold air containing its oxygen causes the sudden ignition of the heated air and an
outburst of flame with such force as to give the effect of an explosion ... . a dense
mass of black smoke is usually seen issuing from the building a few moments before
phenomenon occurring with some compartment frres some 70 years ago. It implies
recognised by the frequent presence of dense smoke. However, it does not make
clear whether ignition of the flammable decomposition products occurs through pilot
ignition or auto-ignition and the affects that temperature of the decomposition
The term backdraught was brought into the public eye after the Universal Studios
motion picture release of that name, in 1991. However, other events in the real
world had already resulted in the inclusion of backdraught awareness in the training
of personnel in many fire brigades at some level. Significantly, the 1973 mattress
store fIre at Chatham Dockyard, in which, tragically two fuemen lost their lives.
Following this tragedy a literature survey was completed by W.M. Croft with regard
to fues involving explosions. This review covered both the United Kingdom and the
1. More explosions were associated with developing fues than with smoldering
with smoldering fues due to personnel entering the premise believing the fire to be
3. The most likely purpose group for a premise where smoke explosions may be
4. More explosions resulted from cellulosic material than any other type of material.
However, it should be borne in mind that this could also be due to the pattern of
usage of such materials, at the time of study, rather than a particular characteristic.
2
More recently, and perhaps more poignantly for fire fighters, was the terraced house
fire of 14 Zephaniah Way, Blaina, Gwent in which two fIre fIghters were killed on
the 1st February 1996. This fire was important for many reasons . Investigation into
the incident by Dr. Martin Thomas [3] and the Fire Experimental Unit concluded that
the phenomenon that killed the fire fighters was indeed a delayed backdraught . This
reported in the Fire Safety Journal of 1980 [2] . Indeed the fIre of Zephaniah Way
inadvertently create conditions that are ideally suited for the formation of
Modern buildings are continually being better insulated and the use of plastic
materials and man made organic based fibres are used ever increasingly. Coupled
with modern double glazed window units and close fItting draught free doors, many
domestic premises are now unable to 'breathe' by natural ventilation in the event of
a fire.
consume most of the oxygen within the room suppressing combustion and possibly
extinguishing it, with perhaps a still glowing ember. The hot gases of combustion
will also descend from ceiling height eventually filling the compartment helping to
extinguish the fire. The lower oxygen content prevents complete combustion of the
3
ever diminishing fue and further pyrolysis products are produced by the fuel,
continually undergoing pyrolysis due to the heat energy within it from the earlier free
burning fire. If a new source of air (oxygen) is then introduced to the compartment,
the fue may be rekindled, or the hot embers act as an ignition source, to the
products may be rapid and cause a deflagration within the compartment which may
From the above we may see that Fire Scientists and Fire Brigades have over recent
years become more aware of the possible hazards of a backdraught occurring when
tackling compartment fires within a building, or indeed the whole building if smoke
logging is severe. Since the Zephaniah Way incident of 1996 and the subsequent
Health and Safety Executive improvement notice issued to the Gwent Fire Authority
(July 1996), the terms 'backdraught' and 'flashover' are used frequently in the
and 'flashover' . The two terms do not describe the same phenomenon [4] . Indeed
training within the Swedish Fire and Rescue Service based upon work of Giselsson
and Rosander used defmitions which are not in total agreement with those used by
4
The aim of this study is a review of the current knowledge of both flashover and
Theoretical models are also considered with a view to predicting the time for a
experimental part of the dissertation is then concerned with the practical application
of Fleischmann's work and comparisons are made with models for describing the
ingress of fresh air into a compartment by infiltration and natural dilution ventilation.
relationship exists between flame speed and the incoming gravity current.
5
2. FLASHOVER AND BACKDRAUGHT.
room volumes of up to 100 m 3 , any larger and the geometry of the room will greatly
affect the fire development [4] due to the subsequent development of any hot gas
The first stage of any fire is the ignition of a combustible material. This fire will
then grow in size as it develops, chiefly as a result of flame spread across the surface
of the item first ignited. During this period the fIre behaves as though it is burning in
the open, and if ventilation within the compartment is adequate the fIre will continue
to grow.
Eventually the fIre will reach a size when the confmes of the compartment will have
an effect upon the fire's development. Hot gases from combustion rise to collect
below the ceiling, their buoyancy forming a ceiling jet that forces them radially
outward from the fire plume. Upon reaching the confines of the compartment (walls)
this ceiling jet flow is forced downwards against its natural buoyancy eventually
causing the thickening of a hot gas layer at ceiling level. At this initial stage of fire
growth the rate of burning is fuel controlled. However, as the fire continues to
6
develop, and if there is sufficient fuel and mechanisms exist for fue spread, it slowly
As the fire continues to grow, and if ventilation is adequate, the ever increasing hot
gas layer at ceiling level starts to affect the combustion processes. This hot gas layer
begins to radiate heat to the compartment below. This in combination with radiation
from the walls and ceiling enhances fire growth. Accelerated fire growth due to
I) Incident radiation upon the item first ignited pre heats the surface of the fuel
where they eventually begin to pyrolyse. This pyrolysis may be sufficient for pilot
ignition to occur, or if temperatures are high enough ( "" 500°C) spontaneous ignition
Eventually all combustible items within the compartment become involved in fue
and the fire is said to be fully developed. At this stage the fue is ventilation
controlled and maximum temperatures within the compartment are attained. This
transition from growth of the initial fire to total room involvement, or a fully
developed fire, is often referred to as a flashover. However, this definition is not the
sole definition and we will discuss this in more detail later on.
7
The 'flashover' transition should not be thought of as an isolated single event such as
ignition but as a dynamic and rapid transition from the growth period of a fire to a
Fully DeVelope,
Fire i Decay
over
Time
compartment fire
The above assumes that ventilation within the compartment is adequate to support
and sustain the growing fire. The heat from the fue perhaps causing a window to
fail, thus maintaining the ventilation for the fue's increasing demand for oxygen.
However, if the compartment is such that ventilation is poor, the rate of burning may
be suppressed sufficiently to extinguish the fire or reduce the rate of burning to give
8
slow combustion or even a smoldering fire. In this instance, if sufficient heat energy
is maintained in the material first ignited, pyrolysis may continue filling the
mixed with air in the correct proportions form a flammable mixture which may be
readily ignited. Ignition of this mixture may occur by entry of fire service personnel
into the compartment causing an inrush of cold air from outside. This inrush of cold
air and the subsequent movement of the hot gases within the compartment to the
outside, due to their natural buoyancy, forms a gravity current within the
compartment [5]. The subsequent mixing, and perhaps the rekindled fue from the
vitiated air, may lead to potentially catastrophic consequences. The fue may now
form an ignition source, the vitiated air creating a flammable mixture, with the
vast amounts of heat energy causing their expansion. As a result of this expansion
other flammable gases may be forced out of the compartment, undergoing turbulent
mixing with air outside. This has the subsequent effect of producing a fue ball or jet
of flame at the exit of the compartment, the duration of which may vary between
backdraught. I will again discuss this definition in more detail below to ensure that
terminology remains consistent. I will also discuss other situations that can lead to a
9
2.2 Definitions of Flashover and Backdraught.
So far we have discussed both flashover and backdraught phenomena together when
considering compartment fires . This has not been accidental, but rather by design as
the two terms though generally definitive in scientific circles have become entwined
in the concepts of fire service training of personnel. The work of Giselsson and
Rosander of the Swedish Fire and Rescue Service describe the phenomenon of
noted even the spelling is different between the United Kingdom and American
Various definitions for flashover are given in the U.K. by various respected sources,
Dougal Drysdale in his publication 'an introduction to fire dynamics ' [4] proffers
10
1) The transition from a localised fire to a situation of general conflagration
3) The sudden propagation of flame through the unburned gases and vapours
All of the above definitions, except the very last, describe the transition from the
growth period to the fully developed period of fire behaviour. R. Chitty in his survey
should highlight that this transition must be sustained for flashover to have occurred.
propagation of flame through the unburned gases and vapours collected under the
ceiling. This definition itself does not state whether there is total compartment
involvement after such an event, and indeed according to the theories of Giselsson
and Rosander such an event can occur for a short duration and be self extinguishing.
Other definitions commonly encountered have been put forward by various research
at floor level was required for flashover to occur. In these series of experiments he
used paper targets positioned on the floor at different locations relative to the flre of
origin. Incident radiation of this intensity was found to be sufficient to ignite the
11
Work by Hagglund et al, 1974; and Fang, 1975 also suggested that an upper ceiling
temperature will also be a function of height of the ceiling above the fIre. In the
average size compartment this temperature may be assumed to be the case, i.e. 2.5 m
high ceiling.
For the rest of this dissertation I will consider the term flashover to mean the total
involvement of all combustible surfaces within the compartment, i.e. the transition
from an isolated fIre to a fully developed fIre involving all combustible items within
implication of this is that for flashover to occur there must be other combustible
items within the compartment. The exception to this defInition, however, is the
discussion of terminology used by the Swiss Fire Service where the word flashover
define the term backdraught. In Britain a definition is given by the LF.E. (Institute
of Fire Engineers)[l2].
12
"An explosion of greater or lesser degree, caused by the inrush of fresh air from any
source or cause, into a burning building, where combustion has been taking place in
a shortage of air".
"The explosive or rapid burning of heated gases that occurs when oxygen is
introduced into a building that has not been properly ventilated and has a depleted
Both definitions describe the cause of a backdraught quite succinctly but they are not
complete. Both statements infer that there has to be limited ventilation to the fire
causing a build up of hot pyrolysis products within the compartment. However, the
possibility that these gases may be forced out of the compartment into another part of
the building should also be considered. Indeed such an occurrence may have more
products form a stoichiometric mixture with air within another part of the building.
An ignition source, as in the Blaina incident [3], may then present itself causing a
The supplement to Book 1 and 11 of the Manual of Firemanship [14] describes three
burning at a reduced level. Entry of fire fighting personnel (or indeed failure of a
13
window etc.) introduces fresh air in the form of a gravity current. Air moves into the
compartment as hot gases leave the compartment via the upper reaches of the
opening. If the gases are hot enough they may auto-ignite, the burning restricted in
some sense by the rate of mixing of the air with the flammable gases.
If the gases are relatively cool the gravity current of fresh air may rekindle the fIre
providing an ignition source for the flammable gases. The flame produced will
depend upon the degree of mixing between the "tongue" of cold air that has entered
the room and the hot gases above. Upon ignition turbulent mixing may occur as the
heat from combustion causes the gases to expand forcing them out of the
compartment. This may produce a ball of flame outside the compartment as further
In some instances the fue in a compartment may have almost died out and the fue
gases cooled down. When the compartment is ventilated air will diffuse into a
ignition source may then involve an explosion within the compartment. This is
The final scenario occurs when the flammable gases of partial combustion are forced
into other parts of the building and an ignition source causes a backdraught. This
may occur upon failure of the compartment which has been containing the fire
14
igniting gases within other parts of the building. Obviously this endangers personnel
The above definitions demonstrate current thinking within the fire scientist
community with regard to flashover and backdraught. I will include the following
section containing definitions and concepts put forward by Giselsson and Rosander
of the Swedish Fire and Rescue Service. R. Chitty [11] in his review, is rather
critical of some of the explanations and definitions given by these two authors.
in their training. I will therefore, give a precise' of their terminology to try and
clarify the definitions for fire fighting personnel with a view to avoiding confusion
fue and the use of their terminology should be avoided. Some alternatives may be
15
2.3 The Concepts of Giselsson and Rosander with regard to
Compartment Fire.
pressure differential from floor to ceiling. At the ceiling there is an over pressure,
and towards the floor an under pressure. Dividing the two zones exist a neutral zone.
There is no reference to the natural buoyancy of the hot combustion gases causing
+ve overpressure
neutral plane(zone)
16
Figure 2 Under and Overpressure of a Fire Compartment.
It is quite conceivable to observe this effect when the compartment is vented, but in a
The position of the neutral zone is very significant, when the neutral zone * remains
high at ceiling level a 'lean flashover' is possible. The combustion gases are near
their lower flammability limits. As this neutral zone descends near to the floor it is
upper flammability limits are eventually reached when the neutral zone reaches 0.5
Terminology used here is in conflict with many academics of fire science, but it is
it here. Giselsson and Rosander's ideas come under criticism for the use of pseudo-
scientific arguments and the lack of references or evidence with which to support
their theories. However, some of the phenomena they have observed have not been
explained with any degree of satisfaction by any other author, so from this viewpoint
their ideas are interesting and invite specific scientific research into this area.
• The neutral zone as described by Giselsson and Rosander is synonymous with the neutral plane used
in conjunction with a two zone model of a compartment fIre.
17
Lean flashover is the result of combustion gases and pyrolysis products (from
secondary heating) collecting under the ceiling. As the thickness of this layer
increases, the heat from the fire increases the overall room temperature and thus the
subsequent level of flammable gases reaching the ceiling layer. The flammable
gases eventually reach their lower flammability limit (appendix A) when it is argued
the initial fire ignites these gases causing a lean flashover. This is usually of short
duration due to its oxygen consumption and then produces an over carburated
mixture (above its upper flammability limit). It is also presumed this flashover
extracts many combustion gases from the wall and ceiling material.
An observation could be made at this point with regard to the last statement. Many
dwelling houses in Britain have paper on gypsum plaster for wall and ceiling lining
materials. It could not be envisaged that these materials will produce vast quantities
of pyrolysis products. However, it emphasizes the point that the upper gas layers in a
compartment may behave very differently with various lining materials. The lining
of the ceiling with polystyrene tiles could significantly contribute to the gaseous
mixture at ceiling level. The possibility of auto-ignition of these gases may also be
18
2.3 (iii) Rich Flashover.
Post 'lean flashover' the atmosphere within the compartment becomes oxygen
deficient and the now reduced fire continues to produce pyrolysis products until a
rich mixture (over carburated) is produced. If lean flashover is not experienced then
it is argued that a rich over carburated mixture of flammable gases will still be
produced. This rich potentially flammable mixture may then produce various 'rich
will fall within its flammable range. If an ignition source is present a rich flashover
will occur. This is synonymous with our definition of backdraught given in section
2.2 (ii).
If the temperature of the combustion gases within a rich mixture is sufficiently high,
then ventilation of the compartment may lead to auto-ignition. As the gases ignite a
fire ball will develop from the ventilation opening and propagate into the
given earlier.
19
In this scenario ignition does not take place until an igniting flame flares up from the
initial fire. This is related to a concealed source of ignition for training purposes. In
such a situation the delay allows the flammable mixture to mix thoroughly with air
In this instance Giselsson and Rosander describe energy rich substances that do not
ignite immediately air is introduced to a fuel rich atmosphere. This is due to the
greater requirement of air compared with "normal" combustion gases which often
'A small initial fire begins in a closed room, the temperature increases, water in the
the room falls and replacement oxygen can enter. The fue then continues to grow,
the hot combustible gases collecting at the ceiling, and combustible ceiling materials
begin to pyrolyse. A lean flashover may now occur forming an oxygen deficient
20
atmosphere. The temperature within the room is now elevated to an extent, more
fuel pyrolyses developing a fuel rich mixture. As a consequence the fire is reduced
due to over carburation, and the temperature falls within the compartment, as the
gases cool the pressure reduces and entraining fresh air from outside. A 'rich
flashover' then occurs, consuming the oxygen and elevating the temperature and
pressure within the compartment, This last cycle is then repeated forming a
pulsation process with combustion gases being forced out of the compartment, and
The positive pressure within the compartment in this instance is attributed to the
21
3. A CONCEPTUAL APPROACH TO THE FLASHOVER AND
BACKDRAUGHT PHENOMENON.
A fonn of quasi-steady state analysis has been put forward by Thomas et al with
regard to thennal instabilities within a compartment containing a fIre. The basis for
this approach are the theories postulated by Semenov with regard to thennal
at a temperature, T o ' the initial gas temperature within that vessel will also be T o .
Thus the rate of heat release from the initial chemical reaction will be[15, 16]:-
q, = Hc·W
H c = Heat. oj. combustion. J. MoZ-'
W = Rate. oj. Chemical. Reaction. Mol.s -'
The rate of heat transfer to the vessel walls will be given by:-
V=Volume
22
T=Gas Temperature within the Vessel.
The rate of temperature increase of the reactant mixture within the vessel is thus
given by:-
Where:-
The effect of combustion products being ignored at this point, we will consider
For the reaction to proceed at a constant rate the temperature inside the vessel will be
23
dq/dt
Temp
24
Initially when the reaction proceeds there will be no heat loss to the vessel walls as
the reactant mixture is at the same temperature. This causes the temperature of the
mixture to rise due to heat from the reaction. Losses of heat energy to the vessel
walls also begin to increase as (T-T )becomes significant in the relationship for q 2 .
0
At a certain point the gas temperature within the vessel reaches a temperature T F •
At this temperature heat losses are equal to heat gains and the temperature of the
mixture remains constant. This results in a 'constant reaction rate'. If we also study
the graph it may be seen that a small increase in temperature results in heat losses
exceeding heat gains, hence tending to restore the reaction mixture to temperature
T F. The point F is a stable point, and a gas will react for a long time at this steady
However, if we consider point G, we may see that this temperature the point shows
instability. If the temperature is lowered slightly, heat losses exceed heat gains and
the temperature will fall to point T F (Steady State). For a slight temperature
increase heat gains exceed heat losses accentuating the rise in temperature. This in
tum causes a rise in the reaction rate and may result in a thermal explosion.
From the above we can see that the reaction rate strongly depends upon the nature of
the vessel it occurs in. Changes to the heat transfer conditions of the vessel wall will
alter the slope of the heat loss line, thus possibly preventing the ability for a stable
25
dq/dt
Temp
Figure 4 The Effect of Wall Composition and Geometry Upon Reaction Rate.
A thermal explosion occurs when the heat loss line lies below the heat release curve.
26
A similar effect is noted by changing the initial conditions of wall temperature. See
figure 5 below.
dq/dt
To -00 T o +00
Figure 5 The Effects of Different Initial Vessel Temperatures upon Reaction Rate
These ideas have been used by Thomas et al in the study of compartment fues[17].
27
3.2 Quasi-steady State Theory for Compartment Fires.
When considering a compartment fire, initially the fire will bum 'freely' consuming
oxygen within the room, the rate of heat release being fuel controlled. As the fire
grows in size the rate of oxygen consumption increases until natural ventilation
within the compartment can no longer supply sufficient fresh air. At this point the
fire becomes ventilation controlled, the mathematical model for such a regime is as
small enough to restrict the flow of air into the compartment sufficiently to govern
H = Height of opening[m]
N.B . This model was obtained by the burning of wood cribs in a compartment with
materials.
This transition from a fuel controlled regime to ventilation control signifies the
However, not all compartment fires will flashover, leading to conditions that may
induce a backdraught.
28
Using Semenov's ideas the temperature within a compartment (average gas
temperature) is governed by the heat input from the fIre and heat losses from the
compartment itself. These losses may be due to conduction through the walls,
Heat gains to the compartment may be attributed to the heat from combustion and the
29
QLC
.... QLC Qc
~QR
...
QLCV = Heat Content of Outflowing Gases QLC= Heat Loss Through Walls
The above assumes that any changes within the compartment occur slowly, a quasi-
steady state is achieved for a short period of time, the gas temperature remaining
30
If we now consider the heat release rate for the fIre, we must observe both the fuel
For the fuel controlled regime, a simple model used by Thomas shows that the heat
4
Qa(T - T/)
T = compartment. temperature
Tf = fuel . surface. temperature
This assumes that heat energy incident upon the fuel surface causes vapourisation of
the fuel, which may be true for liquid pool fires, but is an approximation for solid
fuel fires. Further work by Hasemi has shown that the reaction rate, and hence heat
31
R =Ideal gas constant[J mol - I K -I ]
In either case, both show that for fuel controlled fires, heat release is strongly
dependent upon temperature. For the ventilation controlled fire the heat release rate
may be related to the mass flow of air through the opening, the combustion
efficiency, the heat of reaction and the stoichiometric ratio of fuel to air.
Q= X·ma He
r
where:
X = combustion. efficiency
ma = mass. flo w. of. air.int o. the. fire. [kg / s]
He = heat.of·combustion.[J / kg]
r. = Stoichiometric. ratio. fuel. to. air
The mass flow rate in this expression may be described using Bernoulli' s
relationship, the flow rate through a ventilation opening within a compartment given
by:-
where
C = opening. disch arg e. coefficient
W = width. of .opening.
H = height. of . opening
Hn = height. of . opening. through. which. the. air. enters == H / 2
To = ambient. temperature
T = compartment.temperature(average )
32
Therefore it may be deduced from this expression that the mass release rate increases
quickly with temperature T, then decreases at large T, but remains fairly independent
of T between 400 and l000K. Thus, for a typical compartment fire the heat release
33
dq/dt
TEMP
To complete our thermal balance for the compartment it has been shown by Thomas
Q losses a (T - r:)
This takes into account losses by radiation, convection and conduction; the constant
temperature. Hence, higher wall temperatures will result in lower rates of heat loss
From the above we may now consider that for quasi-steady state conditions to exist
heat loss rates must equal heat gains, for both ventilation control and fuel control
regtmes:-
34
Overall Rate of heat gain ( .t = (a.Q)
a.
a Q)
T
at • T(Juel )
'(a.aQt )
• T(air)
+ Go
Or more simply;
G(T)=L(T)
Where L(T) is equal to the total heat losses from the compartment. We may represent
35
a. Q )
( J.t T
TEMP
a Compartment Fire.
36
3.2(i) Quasi Steady State Analysis of Flashover.(See Fig 9)
If we now consider point Al ,for a fuel controlled fire, as being a quasi-steady state;
a small increase in temperature allows heat losses to exceed heat gains. This tends to
allow the rate of heat gains to exceed the rate of heat loss, again tending to maintain
point B I . An increase in temperature would permit heat gains to exceed heat losses
increase until the quasi-steady state at point C, for a ventilation controlled fire, is
achieved.
However, our compartment fire is dynamic, and the wall temperatures will increase,
this in tum bring about a decrease in the slope (rate of rate of change) of heat loss
progresses.
37
G(T)FUEL
d.Q1
( d. t )T
.......~ ...
..6-_ . . . . . . . . . . . ..
-~.-/-1\
:: / · /L ....................:ltfMP .................j
~ ~
~ i
TEMP
This is again emphasized by the above diagram. As the slope decreases for the line
representing heat losses i.e. with increasing compartment wall temperature; the
points A 2 and B 2 coincide. This situation is thermally unstable and results in a jump
represents a flashover.
Flashover may also be considered when a ventilation controlled fire is further vented
38
L(T).-OPEN
---------------------------------------~-~-
(a.T)
..................
a.tT
----~::::;,<:::-------------
.............. G(T) OPEN.,
.,..,/.,........./ / ---'"
./ /......... t-Transition --
--------------~~~:-------------- :
......... -------~--------~------------
TEMP
39
3.2(ii) Quasi-Steady State Analysis of Backdraught
If we consider that the fire although ventilation controlled still produces pyrolysis
products at a rate equal to a fuel controlled fire at the same temperature, we may
consider the chemical potential energy stored in such a gaseous mixture. This may
------------------------
(d.Qld.t)T
L(T)cLOsED
TEMP
This energy difference could possibly be used to calculate the quantity of unburned
pyrolysis products within a compartment and hence any possible energy release
obviously worthy of more research. This quasi-steady state analysis also leads to
predictions in the development of a fIre if the fue loading and heat of combustion are
known as well as the rate of heat loss mechanisms and their magnitude.
40
3.3 Modeling of a Backdraught.
Little research work has been carried out into the backdraught phenomenon with
regard to estimating the magnitude of the deflagration, size of the fire ball outside of
the compartment, delay in the time to ignition and potential pressure wave caused by
such an event.
Most recent work and the most specific to date is that carried out by C.M.
used was based upon half scale to that of a typical domestic premise, i.e. 1.2 m x 1.2
m x 2.4 m. The series of experiments involved the ignition of a gas burner within
the compartment positioned at one end. The flame was allowed to burn until
extinguishment took place due to depletion of oxygen within the atmosphere. The
level of unburned fuel was then allowed to increase within the compartment for
variable periods of time. After a time delay the fuel supply was cut off (to prevent
disturbance to the atmosphere within) and a vent opened at the opposite end of the
to the original fuel burner then ignited the gases when the fuel/air mixture within the
41
The purpose of the above experiments was to demonstrate that time to ignition of the
fuel within the compartment depended upon the velocity of air entering the
that the resulting flame front occurred at the fuel/air interface where eddy currents
propagated through the compartment the expanding combustion products drive the
remaining fuel and air out of the compartment forming a fire ball outside.
Methane was chosen as the fuel in many of the experiments to represent hot
combustion products of a compartment fire, its density is less than that of air (0.714
kg/m3 & 1.29 kglm 3 @ S.T.P. respectively). This methodology is akin to the
backdraught scenario described in section 2.1 (ii) where dense ambient air enters a
fIre compartment along the floor underneath the hot combustion gases, forcing them
out of the upper reaches of the ventilation opening. Propane is also used in some of
the experiments, its density being greater than that of air (1.96 kglm 3 @ S.T.P.).
This may be akin to a delayed backdraught (section 2.1 (ii) ) where air diffuses into a
compartment (or indeed the denser gas spills out) forming a flammable mixture. The
completed some salt water modeling tests to demonstrate a gravity current scenario
42
within a compartment where there is a density difference of fluids inside and outside
1I8th scale of the backdraught experiments (i.e. 1I16th full scale) and filled with
water. Denser salt water was used as the external (ambient) fluid and the geometry
of the opening into the compartment was the same as for the backdraught
experiments.
A difference in pH of the compartmental water and the external salt solution was
the two fluids when the compartment opening was uncovered. The gravity current
formed (defined as the flow of one fluid into another caused by a difference in
backdraught trials. The less dense methane filling the compartment and the denser
ambient air moving in along underneath. This salt water modeling was necessary as
the flow field of a gravity current is usually sufficiently complex to prove very
Fleischmann' s basis for his description of a gravity current relies on previous work
by other research fellows as far back as 1955, this includes the work by B.R. Morton,
43
Fleischmann' s model is based upon an semi-infinite horizontal box within which a
fluid is contained. At a specific moment in time one end is removed and a denser
ambient fluid allowed to enter the box. Obviously, assuming incompressible laminar
flow , as the denser fluid enters the box the internal fluid leaves via the upper reaches
of the same opening. The driving force behind this flow is the density difference of
the two fluids . The buoyancy force is represented by the normalised density
difference 13.
Using the laws for the conservation of mass, momentum and energy for an inviscid
V\h\ =V2 ~ ................. .... ... .... .. ......... ..... ... ....... .. ... ... Mass
V\2h\ + f3 .g.h\2 = 2V22h2 + f3 .g. hi ..... .... ....... .... .... ... Momentum
V22 = 2. f3 .g(h\ - ~ ) ..... .. ..... ...... .. .................. .. .... .. . Energy
44
That is the height of the gravity current (h o ) is equal to the height of the exiting
current for a perfect liquid with no energy losses. The gravity current(h 0) refers to
From the above is developed a non-dimensional velocity or Froude number for the
Hence for scaling purposes a characteristic dimension and velocity may be used
45
Therefore, a characteristic time may be given for the geometry of the box (or
opening), hence:-
This characteristic time may be described as the time taken for the fluid in the
46
4. CHEMICAL KINETICS VERSUS DIFFUSION
In the preceding text we have discussed the characteristics of a compartment fire that
with a higher or lower intensity due to the nature of the flammable gases within the
compartment and their subsequent mixing with the external atmosphere. The four
(i) Backdraught a) - Hot combustion products auto ignite due to their temperature
when the gases within the compartment are vitiated with the atmosphere outside; due
their auto ignition temperature. When an opening is created to the compartment air
enters via a gravity current causing layering and mixing of the gases at the interface
of the cold air and hot compartment gases. A rekindled fire or other ignition source
(ii) Delayed Backdraught a) - The gases within the compartment may have cooled
down and the fire almost extinguished. Air entering the room may then diffuse into
the compartment and thoroughly mix with the flammable gases contained within.
An ignition source may present itself causing combustion of the gases to propagate
47
b) - Flammable gas may be forced out of the compartment containing the
fire to other parts of the building. These gases may be relatively cool and mix freely
with air contained within these other areas of the building. A flammable mixture
may subsequently be formed, which, with a suitable ignition source may be ignited.
The different physical effects of the above may be explained by considering the
chemical kinetics and physical transfer processes (heat and mass) with regard to the
combustion process.
following categories:-
(ii) Combustion that is controlled by diffusion, flow and other mixing processes.
and the mixing processes of fuel and oxidant, the effects of both being of similar
importance.
48
If we consider a mixture of fuel, oxidant and combustion products fonning a
throughout, the rate of reaction will entirely depend upon the chemical kinetics of
that reaction. The combustion process is said to be kinetically controlled, and is due
to the fact that the chemical kinetics are slow compared to the rates of heat and
species diffusion. In this instance there are no physical transfer processes as the
concentration and temperature gradients within the compartment, but the rate of the
transfer processes still exceed the chemical kinetics, there will be adequate time for
species and temperature to smooth out any spatial variations. This again will be
kinetically controlled combustion and will occur more or less uniformly in the
reaction space.
At the other end of the scale chemical kinetics for combustion are usually very fast
and any spatial non-unifonnity's of species concentration and temperature are unable
to equalise out. Distinct gradients of species and temperature are formed, diffusion
causing the transfer of heat and species to areas of lower temperatures and species
concentrations respectively. Reactants will diffuse into the combustion zone and
products will diffuse from it. This form of combustion is said to be diffusion
space.
49
From the above we may observe that the backdraught scenarios described lie
a diffusion controlled combustion, i.e. gravity current. This accounts for the
When we are considering a backdraught scenario we are concerned with the rate of
propagation of the flame front through the compartment, or the deflagration. The
rate of propagation, or the burning velocity, is dependent upon fluid flow, heat
Deflagrations are considerably slower flames than detonations typically in the range
of 10° to 10 2 cm/Sec. As the flame front progresses, gases in front of the flame are
heated by the combustion processes, expanding and accelerating (as a result of this
expansion), to accommodate the mass flux through the flame. When we are
considering a laminar flame in a long horizontal tube it may be shown [16] that the
50
Pf =Pressure of Products.
ps = Density of Reactants.
Pf = Density of Products.
Us = Velocity of Reactants.
Where typically for hydrocarbons this pressure drop is small and a deflagration is a
Flame speed must also be related to any movement of the unreacted fuel loxidant
flame front in a direction normal to itself and with respect to the reactant mixture.
fundamental flame speed and the velocity of the fresh gas mixture u R by:-
The flame speed is strongly influenced by the reaction kinetics, presence of dilutents,
Flame speed shows a strong dependency upon the composition of the reactant
mixture. Near the stoichiometric mixture results in maximum flame speeds and may
be correlated with the mixture that corresponds to the maximum flame temperature
51
for a adiabatic mixture. If the mixture contains too little fuel or too little oxidant the
FLAME SPEED
cmls
o 20 40 60 80 100 % FUEL
Work by Gerstein, Levine and Wong has shown that for various families of
flammability [4, 16]. By implication, this results in a narrowing of the curve for
flame velocities with respect to fuel oxidant mixtures, However, for the lower
alkanes (methane, propane, butane etc.) the dependence of the number of carbon
atoms is not shown, the flame speeds remaining roughly constant at 70 cm/Sec.
52
2~ Figure 141nfluence of the Type of Fuel.
1~
a
1 2 3 4 5 6 No of Carbon
Atoms
Lewis studied the effects of pressure on flame speeds for various hydrocarbon
mixtures was found to be dependent upon the relative flame speed itself. For flame
speeds <50 cm/s, the flame speed increases with decreasing pressure. In the range 50
- 100 cm/s, it is independent of pressure and > 100 cm/s. decreases with decreasing
pressure. This has been attributed to the overall reaction order of the combustion
process, being less than two for flame speeds less than 50 cm/s., equal to two for
flames in the range 50 - 100 cm/s. , and greater than two for the latter.
53
4.1 (iv) Influence of Initial Mixture Temperature.
Burning velocities are usually quoted for fuel oxidant mixtures at an initial
initial reactants.
TR Increasing
Uo / em 2-1
s
%Fuel
54
A general relationship for methane, ethane and propane was found by Dugger [16] as
shown below.
40
I I I I I I I
o 100 200 300 400 500 600 700
m
uOa .TR
Where.m.is. in. the. region.1Sto2.0
Zabelakis (1965) also showed the following expression for methane, propane,
55
At high temperatures of 800 K, mixtures of these gaseous fuel and air undergo
Pre-mixed turbulent flames will propagate much faster than laminar ones. However,
once a flame becomes turbulent, what is measured is not the fundamental flame
the reacting mixture, but the flame speed (burning speed). This burning speed is
usually dependent upon the dimensions of the vessel where combustion takes place.
Transport of heat and mass in turbulent flames does not solely rely upon molecular
diffusion phenomenon c.f. laminar flames, but also upon eddy mixing which is a
function of geometry.
From a practical fire fighting point of view the signs for a possible backdraught
scenario are well documented in fire service training manuals [14] . These include
such things as oily deposits on windows, the pulsation of smoke in and out of
ventilation openings, and heavy smoke logging of the building. However, little is
56
In some instances the presence of blue ghosting flames has been reported. This has
been attributed to the combustion of carbon monoxide, produced at high levels due
to poor ventilation. However, can these high levels of carbon monoxide be solely
responsible for the highly flammable nature of such gases or are other partial
An initial investigation was undertaken into the pyrolysis products of chipboard with
the aim of identifying which flammable pyrolysis products may be contained in the
smoke from the above. The analysis involved the pyrolysis of chipboard under
The results were inconclusive and it was determined that the use of such techniques
would be worthy of a dissertation in its own right. However, the preliminary results
57
5. EXPERIMENTAL EQUIPMENT.
5.1 Introduction.
In the light of what has been said previously, with regard to the Modeling of a
C.M. Fleischmann in his work, showed that a backdraught often depended heavily
the gases methane and propane were used to create a fuel rich atmosphere within a
scaled compartment, which was then vented allowing a gravity current of ambient
The two items of equipment were the Giselsson box and Firebox. Video camera
footage was used to analyse the use of the above equipment along with thermal
54
5.2 Equipment.
The Giselsson box was developed by Giselsson and Rosander of the Swedish Fire
and Rescue Service for training of fire service personnel. The original intention for
The tank comprises of two glass sides and flaps across the top to release any
pressure build-up within the tank itself. There is also a fan located at floor level, to
mix the gases within the tank, a gas inlet and an ignition source in the form of an
electric arc. The tank may be filled with a variety of flammable gases, the supply of
which is at a constant flow with an electrically operated valve. The ignition source
is manually operated by the means of a remote switch. The tank may be vented by
the removal of an inspection cover located on one end of the tank. There is also a
crude form of timer on the front of the tank to show the approximate time the gas
supply remains open, hence an estimation of the quantity of gas supplied to the tank
maybe made.
Details of the tank dimensions are given below. The actual tank may be seen on the
55
Side Elevation En-d Elevation
695mm 303mm
500mm 500 25
245mm
Plan View
695mm
o
gas source
303mm
The results of different experiments were analysed using video camera footage at a
speed of 25 frames per second. This could be slowed on play back to approximately
one quarter of the filming speed. Bum speeds for the flame front were obtained on a
56
frame by frame basis, one frame being equivalent to a time step of 1I24th of a
second.
conditions to fire fighting personnel within the fire service, The box comprises of a
metal box lined with duraboard to act as insulation. Over the duraboard is then
fixed chipboard panels 15 mm thick on all internal surfaces within the box. The
57
Fire Box
550mm
o
o
o
o
o =Position of Thermocouples
Figure 18 The Fire Box
58
Section Through Wall Side Elevation Showing Location of
Thermocouples
REAR
OF
BOX
A small kindling fire (comprising of chipboard) is then ignited using a blow torch at
the rear of the box. The fIre within the box slowly develops, flashover occurs and a
fully developed fue then exists. The development of a neutral plane,i.e. the
boundary between hot buoyant gases and cooler entrained gases, can clearly be seen.
After the fire has been burning for some time the box is "closed down" by manually
holding a door over the aperture at the front of the box. This causes the fire to self
extinguish, but the excess heat energy generated within the box maintains rapid
pyrolysis of the chipboard, The generation of the pyrolysis products eventually fills
the box creating a positive pressure in relation to the ambient pressure, forcing gases
59
out from behind the door seal, where the fit is less than perfect. If the gases are hot
When the door cover is removed air is immediately entrained into the box as the hot
gases exit due to their buoyancy. The fire is rekindled and the temperatures within
the box in any case, cause a rapid deflagration forming a substantial fire ball outside
the box.
Both video footage of this sequence of events and temperature variations were
were logged every 2 seconds using an analogue recorder, this data was then
The thermal image camera was also used in an attempt to gain thermal images of gas
fluctuations . The results of these thermal images were disappointing due to reasons
Thermal image pictures were obtained for both the Giselsson box and the fIre box.
This camera uses 256 x 256 platinum silicide focal plane array detection. Images
produced may be highlighted in colour to show isotherms within the field of view or
60
provide spot temperatures at a precise location within the image. The image itself
comprises of 256 x 256 pixels (picture elements), the output of which may be
directed to a television or V.c.R. Digital signals may also be used for data analysis
using a P.e.
The camera was used with the intention of showing thermal irregularities, and hence
gravity currents, within both the Giselsson box and firebox experiments. A flame
filter was used over the lens (3.9Jl m) which eliminates wavelengths common to
luminous flames entering the camera. This allows the camera to measure the
However, problems were encountered with the use of the camera. Measurements of
temperature for the Giselsson box proved ineffective due to the reflection of light
from the glass sides of the box. Also software for use with the camera only permits
Problems were also encountered with the fIre box due to the fact that the linings of
the box were at a higher temperature than the gases within it. The camera thus gave
61
5.2 (v) Thermocouples and Data Logger.
In the fIre box test temperatures were recorded using thermocouples in tandem with
a data logger. The location of the thermocouples was as shown in the diagram for
the fire box, the ends of each thermocouple extending to an imaginary central
Mineral insulated.
't (63% of step change) typical 10m secs. Temperature range 0 - 1,OOO°C.
62
6. EXPERIMENTAL RESULTS.
The purpose of the group of experiments completed with the Giselsson Box was to
demonstrate the ventilation of a compartment in which resided an over rich fuel air
mixture. The delay to ignition of the flammable gas was measured from the moment
of ventilation and the subsequent flame speeds and modes of propagation recorded.
The sequence of events was recorded using the digital video camera as described in
the experimental equipment and attempts were also made to analyse the subsequent
gas temperatures, internally and externally of the box, using the thermal imaging
camera.
An initial experiment was completed to determine the gas flow rate into the
compartment of the box (Experiment 1, Table 1). In this experiment propane was
allowed to enter the box with continuous stirring and the ignition source (electric
is assumed, the point at which ignition prevails may be taken as the instant at which
the lower explosive limit is reached. Hence, the flow rate of propane into the box
could be calculated.
62
The results of further experiments for the Giselsson box are shown in table I below,
Experiment 2 was completed to demonstrate the upper explosive limit. The gas
supply was maintained without the ignition source. Mixing of the compartment gas
was achieved using the stirring mechanism, and ignition delayed to a point
experience of previous tests carried out with this apparatus. A similar methodology
the table.
Experiments 4,8,9,10,11 were used to demonstrate the backdraught scenario, the gas
compartment was then vented to allow fresh air to enter. The time to ignition in
each of these experiments being the delay between venting of the compartment and
subsequent flame propagation throughout the mixture. The ignition source was
suppression of the lower limit with an elevated temperature of the box. Experiment
63
6.1 (ii). Analysis of the Results from the Giselsson Box.
open
ignition source.
open. of box
open.
open.
open.
TABLE 1
64
6.1 (iii). Overall Combustion Reaction of Propane with Oxygen (Air).
I
26 x 100 = 3.8%.v / v .... . Oj .Propane.in.Air
However, the lower flammability limit of the alkanes is approximately twice the
stoichiometric quantity of air in relation to the alkane gas ; hence, the lower
flammability limit:-
65
L.E.L (Lower Explosive Limit)=lISI x 100
[18, 19] stated that the hydro-carbon concentration (mass fraction) needed to be
greater than 10% for a backdraught to occur. This, however, was measured in the
=6.8% v/v
It may be deduced from this that a backdraught may still be experienced when the
conditions. Combustion prohibited due to the depletion of oxygen which affects the
66
6.1 (iv) Calculation of Volume Flow Rate into the Tank. (Experiment 1).
= 105dm 3
=2.21 x 10-3 m 3
The experiment conducted estimated a duration of the gas supply required for the
box to reach the upper explosive limit (9.1 % v/v) from the experience of previous
tests. However, a more scientific approach may be realized using the previously
67
3
Vol = 0.092 x 101 =9.292 dm
=8.8 %
This value of 8.8% v/v is close to the literature value of 9.5% v/v and shows the
In this experiment the gas supply was maintained with the stirring mechanism to
ensure a homogeneous mixture. After a time interval of 50 seconds the gas supply
4.6xlO-3
% v/v Propane in Air = 0.105 x 100 = 4.4%
The actual theoretical stoichiometric mixture for propane from the previous
flame temperatures and hence flame speeds (pre-mixed) are obtained when the gas is
68
slightly above its theoretical stoichiometric mixture. Hence, for practical purposes
flammable gas within a compartment above its explosive limits. When the
compartment is vented air is introduced diluting the mixture within the compartment
Ideally Methane would have been the gas selected for the experiments as it is less
dense than air and would have been analogous to hot compartment fire gases.
cold smoke explosion, may occur which is probably more analogous to Propane, the
Using the gas flow rate estimated from the lower explosive limit the following may
be calculated:-
Experiment 4.
69
Gas supply duration = 150 sec.
:. %v/v Propane=13.1 %
This calculation was repeated for the other experiments where delayed ignition was
TABLE 2
When the Giselsson box was being filled the mixture was continually stirred in
order to obtain a homogenous mixture throughout the box. This situation may be
products have cooled and species diffusion within the compartment has occurred.
Once sufficient gas had entered the compartment the stirrer mechanism was
70
switched off before ventilation commenced. This was done in order to eliminate
Density Calculation.
86.9 )
44+ ( -x28.8
.. 13.1
Average Molecular WeIght MIxture 7.634 = 30.8. g / mol
PV=nRT
Where:
R = 8.314J.K- 1.mor1
P = 1.atm = 1.013 x 105 N / m 2
nR.T 3
Vol@12Ceicius= - - = 0.02339m = 23.39dm3
P
. 30.8 X 10-3 -3
:. Denslty = 3 = 1.317 kg.m
23.39 x 10-
71
If the calculation is repeated using 13.5% v/v we obtain a density of 1.319 kgm - 3 ,
so for accuracy that may be expected from such a calculation I will assume 1.32
The theories of Fleischmann have been discussed in the previous text, however, a
useful relationship for the gravity current velocity is given in his paper Exploratory
It has been assumed that p is equal to the mean of the two densities. However, if
this is incorrect the error will be small and consistent by a small factor. This
For our full door opening (Experiment 8) this equates to a volume flow rate of:-
72
Assuming half of the opening is available to the entering flow;
Q =A.u ..... ................. ... ....... ... Where A and u are the Area and Velocity
offlow respectively.
The results of similar calculations for the other experiments are shown in the table
of Results. Table 3.
The CIBSE Guide describes a calculation for the infiltration of ambient air into a
room with only one opening[22]. The driving force to this infiltration being the
difference in air temperature between the inside and outside of the building. The
model that is applicable to the situation within the Giselsson Box is that of a single
room with one opening to the outside, i.e. such as a classroom in summer adjoined
to other rooms by a corridor, but with all internal doors closed. An example
Hence:-
73
O.5
A tl.t.ho.g
Q=Cd
[---....:::....-:::....
3 t+273 J
A =Area of Opening[m 2 ]
-
t = Mean of the External Ambient and Internal Air Temperatures
Where we may use the following relationship to substitute density for temperature
[21].
PI = Reference Density
1
Where f3.: - And ~ Is the Reference Temperature.
~
openings)
74
A =Area of Opening [m 2 ]
Q ==5.21 x 10-3 .m 3 / s
= 5.2.dm 3 / s
The above relationship assumes a neutral plane forming between the incoming and
outgoing flow of fluids from the compartment and is based upon principles
This gives us an initial flow rate into the compartment of 5dm 3 /s. Thus if no mixing
occurs between the incoming and outgoing fluids and the buoyancy potential could
seconds. However, the fluids are not perfect and mixing will occur at the interface
between the fluids . This will eventually affect the density difference (the driving
force to buoyancy) between the fluids and will decrease the rate of ventilation.
However, study of the table of results for delayed ignition shows that for a full door
75
If we assume an initial inflow of 5dm 3 1 s and an initial gas concentration of 13.1 %
vlv, a very broad approximation to attainment of the upper explosive limit and hence
U.E.L=9.5 %
3 3
== 0.095 dm in 1 dm
== 10.0 dm 3 in 105 dm 3
== lOdm 3 • C3 H g.in.l05dm 3
Loss of propane required from total volume of the box =3.8 Litres
Loss rate=0.475dm 3 S - 1
internal and ambient gases and the buoyancy driven flow is maintained at a constant
rate. However, the minimum time to ignition could be expected to be in the region
Flow rates for the other experiments calculated by this method are again shown in
Table 3.
76
If the assumption is made that when the aperture to the box is first opened,
ventilation takes place but the gravity current is far from ideal and rapid mixing and
diffusion occurs to form a diluted mixture; then the extreme of this situation may be
represented by:-
Co=Initial Concentration
v =Ventilation Rate
t= Time [sec]
V =Volume of Box
77
6.1.(xii)Ventilation Flow Rates and Time to Ignition
The previous calculations were repeated for each of the backdraught experiments,
the results of which are shown below. Ventilation rates are derived for each of the
mathematical models described in the above text. These values are then compared
to the delay in ignition from the instant of opening the vent to the compartment, and
the corresponding vent area or ventilation factor. The latter is calculated from the
area of the vent and its vertical height using the expression given in Table 3.
78
VENTll-ATION RATESldm 3
Vent Ventilation
~onIS
AHl /2
100
iN~; ~ I";
':.L -""-
TABLE 3
79
Figure 19
3.5
3
. . '.
.
...0
-
I.
2.5
( .)
.! 2
c::
0
;;
~ 1.5
;;
c::
41)
,
> '.,
0.5 • .,CC
,.
'"
• .
o
o 10 20 30 40 50 60
Time/s l.serieS1 1
The rates of Propagation of the flame front were measured for each of the
experiments with the Giselsson box. Measurements were obtained by using the
necessary as the centre of Propagation is not always clearly defined, and the distance
of flame front travel is obtained by using the enclosing dimensions of the box.
80
EXPERIMEN COMMENT
2 U.E.L
3 Stoichiometric
9 Ghosting Flame
10
TABLE 4
81
6.2 Discussion of Results.
I will firstly consider the results produced by the Giselsson Box. The objective of
this series of experiments was to appraise whether a time to ignition for a potentially
and thus accounts for the mechanism by which such an event may occur.
The only current research into this area has been completed by C.M. Fleischmann et
al [18, 19] where theories for a gravity current have been postulated. In his
experiments both Methane and Propane were used, although results are only given
for Methane. As previously stated Methane was chosen due to its relative density to
that of ambient air. In his paper theories are based upon work previously undertaken
on the fluid theory of Convective flows. However, he often refers to the velocity of
the gravity current depending upon the height of the compartment (based upon the
model for a semi-infinite horizontal box) but concludes that the characteristic
velocity is entirely dependent upon the opening geometry. It was therefore assumed
that 'h' in the expression used, referred to the opening geometry, as would be
anticipated, and was not related to the compartment height. Bearing this in mind it
completeness and further research into his work is essential for modeling to be based
on the above. Indeed it may be concluded that for any significance to be founded
upon results from Fleischmann's theories, compartment and opening geometry must
82
be similar. Scaling is accounted for by using characteristic dimensions of velocity
etc.
Therefore, unfortunately our model with the Giselsson Box does not really apply.
The compartment geometry being very different from Fleischmann' s gravity current
compartment, the opening being sited well above the bottom of the tank. Also due
to the unavailability of methane at the time of the tests, propane produced a reverse
gravity current to that of Fleischmann's, the incoming fluid being less dense than the
However, in his work Fleischmann does not appear to concern himself with any
contributions from the 'step' of the opening when the current enters the
compartment and any subsequent effects this may have upon its velocity, i.e. slot
and window opening geometry' s. Therefore, taking this into consideration it was
felt that an initial velocity for a gravity current may be calculated. Results using this
methodology were very similar to those obtained using the emSE calculation for
the rate of ventilation into a room with a single ventilation opening. Upon
the buoyant stack effect and assumes a neutral plane. This assumption of a neutral
The application of these two models was of limited use as any gravity current can
possibly it may be anticipated to be reflected off the far wall and return towards the
83
opening. Eventually, diffusion and turbulence at the interface of the fluids, and the
break up of the head of the gravity current due to the formation of lobes and bellows
Fleischmann's model relies upon the movement of the gravity current across the
floor to the ignition source. In the Giselsson Box, however, the ignition source and
the position of the ventilation opening prevent the time to combustion being wholly
reliant upon a gravity current. The time to combustion may be considered as being
Consideration of this relationship will show that the time to induction is negligible
in comparison for the time for air to enter the compartment and subsequently mix.
From the calculations it was estimated that an initial ventilation rate could be
anticipated. However, the degree of mixing could not be calculated. A very 'broad
brush' approach was then taken, assuming the amount of propane lost per second
was at the initial ventilation rate. The time to reduce a homogeneous mixture down
to its U .E.L. at this loss rate was then used to calculate a time to ignition. The
84
Both of the theoretical models based upon buoyancy driven flow depend upon the
fluids not mixing to any appreciable extent. However, mixing is required for
Consideration was given to the idea that mixing may occur at a much earlier instant
in time when the compartment is ventilated. Indeed if it is assumed that the ambient
gas enters and forms a homogeneous mixture with the gas within the compartment,
occurs using fresh uncontaminated air. This methodology was used to derive a
theoretical ventilation rate for the mixture of propane and air to be reduced to its
D.E.L. This was achieved by regarding the point of combustion as attainment of the
upper explosive limit. Results achieved by such calculations were remarkably close
to figures obtained for buoyancy driven flows. However, some experiments showed
Examination of the times obtained to ignition shows that there is reasonably large
85
Ghosting Flames as reported by Sugawa, Kawagoe and Oka [23] and also
Fleischmann [18] may be partly responsible for some of the variations of the times
to combustion. A ghosting flame is a flame that has become detached from the fuel
surface (in the case of liquids and solids) and exist suspended in the combustion
compartment. In our case, however, there is no fuel surface and a ghosting flame
compartment is low and the subsequent oxygen concentration is also very low.
combustion may occur where under normal ambient conditions it could not be
heating by the electric arc allows combustion to commence probably by a shift in the
flammability limits. The strength of the ignition source (total energy available) will
also have an effect as described by Dougal Drysdale in his book 'Introduction to Fire
Dynamics' [4].
The result of this is a vertical flame extending from the ignition source to the top of
Eventually due to further entrainment of ambient air and the subsequent thermal
effects of such a ghosting flame a general combustion takes place. This phenomenon
varies greatly with the initial gas temperature and it was realised that the heating of
86
the tank from previous experiments occurred to an extent to effect the flammability
limits of the propane. Changes in temperature would also alter the density of the
fue1Jair mixture within the box to some extent resulting in a change of the
ventilation rates.
Although the results obtained upon initial inspection do not show consistency in
relation to the time of ignition and the size of ventilation opening, it may be
It was found that significant heating of the box occurred between experiments
effectively altering the flammable limits and density of the propane/air mixture
contained within. It may be observed from the graphical representation of the time
87
occurring sequentially to each other(See Fig 19). The large discrepancy in times for
each pair of results may possibly be attributed to the time delay between
Consideration would have to be given to future work repeating a more detailed and
the ventilation factor AHI I2 and the time to ignition. It would be necessary,
the fuel loxidant ratio. Such analysis would be in keeping with the theories of
Fleischmann and the relationship given in the CIBSE Guide. Such an approach
factor for the location of the ventilation opening in relation to the compartment.
assumes a degree of pre-mixing (diffusion flames do not have flame speeds). The
rate of flame Propagation through a mixture is dependent upon the velocity of the
combustible medium itself and may be related to the fundamental flame velocity by
The intention of measuring the rate of flame Propagation through the Giselsson Box
was to observe whether any dependency of the flame velocity may be attributed to
88
the size of the ventilation opening and hence the incoming' gravity current' or flow
of gases. Flame velocities were obtained by use of the digital video camera (Table
some of the tests, often the centre of combustion was removed from the ignition
source.
Examination of the table of results shows that in many instances the rate of flame
Propagation is close to the literature value of O.4m1s for propane at its stoichiometric
ratio with air for a laminar pre-mixed flame[4]. Deviations from this figure are due
Fire Box experiments were performed in order to see if a thermal image may be
obtained of air entering the box, and hot gases leaving the box prior to and during a
the chipboard lining. A fully developed fire ensues within the open compartment
displayed by the temperature time curves obtained from thermo-couple data. It was
then found if the box was 'closed down' by the manual application of a door across
89
the aperture, backdraught conditions were realised. Pyrolysis of the chipboard
flammable gases collecting within the box. When the box was opened, often auto-
ignition of the hot gases occurred, or the inrush of ambient air re-ignited the
chipboard providing an ignition source. Video evidence also showed that if the
gases were sufficiently hot, auto-ignition could occur when the hot gases were
forced out of the compartment past the imperfect door seal due to the pressure
within.
Although these fire box tests demonstrate clearly the mechanism for a backdraught
to occur, analysis of the time temperature curves proves difficult. 't (response time
63% of step change) for the thermo-couples used is typically 10m secs. but
higher heat output) of the linings of the box when the compartment is vented.
were a 'snapshot' in time and although clearly demonstrating the high temperatures
obtainable in a compartment fire and the subsequent extent of a fire ball produced
by a backdraught, they could not show the origin and movement of the fire ball
produced. A moving thermal image would be required, the software for such
90
Thermal images of the fire box did not in all cases show the temperature of the gases
within. When the door is removed, hot gases escape via the upper reaches of the
opening whilst cool ambient gases enter along the floor of the compartment. This
results in rapid cooling of the gases often below the temperature of the radiating
surface linings of the box, the thermal image subsequently given is that of the
lining.
temperature over the height of the box even with such small dimensions.
Differences could also be noted front to back and inspection of the graphs also
regions of the box. This possibly may be caused by the combustion of the floor
Any further analysis of this type would need greater control. The frequency of door
openings to the compartment would need to be reduced and perhaps a smaller fire
concluded in tandem with flammable gas experiments, i.e. Giselsson type box to
identify whether any correlation could be found. Prohibitive to this type of work are
the health and safety implications and the availability of a safe area to carry out such
91
research. These latter limitations being prohibitive to the repetition of the above
experiment.
7. CONCLUSIONS.
The initial aim of this dissertation was to review the current knowledge and thinking
with regard to backdraught. It has been shown that backdraught has recently come
to the fore in fire service training in the light of the recent fatal incident at Blania,
Gwent [3]. In his investigation, Dr. Thomas of the F.R.D.G. stated that the cause of
As a result of this incident, and the subsequent H.S.E. improvement notice issued
upon Gwent Fire Authority, backdraught and flashover training have become the
concern of all Fire Brigades throughout the country. However, as this dissertation
has shown the two terms often appear synonymous with each other, particularly
It was found that terminology used by these authors was conflicting with concepts
used by Fire Scientists in Britain and America, and the suggestion made that their
92
At this juncture the concept of a backdraught has been fIrmly established. This
Theoretical concepts were also considered. Dr. Martin Thomas et al [17] developed
the idea of quasi-steady state analysis with regard to explaining the growth and
development of a compartment fIre. It was shown that this analysis is based upon
the ideas of Semenov with regard to a thermal explosion; and in tum lead to the
Indeed it was shown that the potential energy of a backdraught may possibly be
An appraisal of work by C.M. Fleischmann [18, 19,20] was also carried out. Ideas
entering the said compartment. It was found that further investigation would be
some of the concepts he has used in this paper. However, it was elucidated that his
model depended heavily upon the compartment and opening geometry, relationships
93
given breaking down if such geometry's were departed from to any extent. It was
also found that relationships given by C.M. Fleischmann were very similar to semi-
empirical relationships given in the CIBSE Guide for the ventilation of ambient air
relationships are governed by the stack effect and describe a buoyancy driven flow
with a neutral plane. Indeed, calculations using such a relationship were compared
Experimental work carried out for the dissertation also showed that these
relationships could not describe ventilation of the compartment beyond the time that
the gravity current ceased to exist, i.e. when mixing becomes significant. It was
suggested that the time to combustion depended upon the time for the gravity
current to enter and the subsequent time for mixing. The relationship below may be
considered: -
The predominance of the two variables T AIR.TO.ENTER and T AIR .TO.MIX. ' upon the time to
combustion will depend upon the compartment geometry and relative position of the
temperature and composition of the flammable gas. This was demonstrated to some
94
Consideration was also given to the relationship governing dilution ventilation and
the calculation of a ventilation rate necessary to achieve the U.E.L. at the time of
combustion. It was recognised that this model was incorrect but lay at the other
Further experimental work was also performed using a Fire Box. It was hoped that
work carried out by analysis of this item of equipment may clearly show the
development of a gravity current and ignition of the gases at the interface boundary.
However, due to reasons explained in the previous text the work was inconclusive
with respect to this methodology. It was felt that the use of such equipment may be
analysis.
The construction of a scaled compartment where heat losses and heat outputs may
validate the ideas of Dr. Thomas et al. The use of variable ventilation openings and
the formation of gravity currents and rates of ventilation, when compared to the time
to combustion. The use of realistic fuel sources with known heats of combustion
95
In conjunction with the use of realistic fuel sources, analytical chemical techniques
smoke. The use of such techniques would involve either direct line sampling or
Appendix C). However, the limitations of such analysis should be bourne in mind
The further use of instrumentation, such as that used in the course of this
dissertation, for instance the thermal image camera and real time video footage,
should also be extended. Both of these items of equipment arte powerful tools for
the analysis of flame and fIre ball propagation. In the case of the thermal image
Further investigative work could also be carried out with similar experiments to
those of the Giselsson Box, and those of C.M. Fleischmann using flammable gas/air
mixtures and perhaps comparable salt water models. The use of variable
compartment geometry and ventilation openings coupled with the ability to relocate
the ignition source could be used to investigate the time to combustion when
compared with ventilation factor AH II2. The possibility of this approach would
allow a geometric factor to be introduced for the size and the geometry of the
compartment.
96
From the above it can be seen that there is still a great deal of investigation possible
to the deflagration and the extent of such a flame front. Indeed with modem
building practices it is suggested that this type of scenario will become more
prevalent endangering the lives of Fire Service personnel. From this viewpoint
alone, without regard to potential property loss and damage the phenomenon is
97
APPENDIXA .
FLAMMABILITY LIMITS
Any flammable gas when mixed with air will show upper and lower flammability
limits (U.E.L's and L.E.L's respectively). The most extensive review was by
Zabetakis (1965) [16] using Bureau of Mines apparatus, the results of which are still
Flammability limits are affected by both pressure and temperature, the main effect
being that of temperature. For a flame to propagate through a medium there must be
the flammable gas is reduced in air, the amount of heat energy released from the
propagation of the flame is maintained, i.e. too much heat energy is lost to the
ambient air for reactant molecules to maintain their activation energy. We can see
from this that if the ambient temperature increases less energy is lost by the system
(reactant molecule) to the surroundings, hence flame propagation may occur at much
lower concentrations. The same argument may be used for the uppers limits,
however, in this case the reactant flammable gas itself becomes the heat sink. It may
98
APPENDIXB
GLOSSARY OF TERMS
Pyrolysis:- The word Pyro -lysis means "death due to heat", or more idiomatically,
within the chemical structure, to release volatile substances which are able to burn.
The chemical and physical state of the original substance is permanently changed.
When the heat source is removed the liquid usually remains chemically unchanged.
Flame:- A spatial domain where a rapid exothermic chemical reaction takes place,
subsequent heating of the gas by the pressure wave initiates the reaction resulting in
or compression and hence pressure differentials are negligible across the flame front.
99
APPENDIXC
CHEMICAL ANALYSIS
A preliminary investigation was made into the use of gas chromatography and mass
backdraught scenario.
passing over a stationary adsorbent phase. Components of the mobile phase are
adsorbed onto the surface of the stationary phase at different rates depending upon
the nature and size of the molecular species. Release from the adsorbent surface also
occurs at different rates. This allows separation of the components of the fluid phase
to be made, depending upon their molecular size and weight, by passage of the fluid
through a long column packed with a suitable stationary phase. The subsequent
peaks as each component is eluded from the column may then be used as a
However, if the output of the column is connected in line with a mass spectrometer,
analysis of each peak may be carried out when each component is eluded from the
electron gun. The molecules of each component are either ionised remaining intact,
ratio(m/e). These ions are then separated in the mass spectrometer by acceleration
between two charged plates and passage through a magnetic field. The difference in
100
the mass/charge ratio of each different type of ion permits separation by the
magnetic field into individual ion beams which are focused onto a detector. The
detector then measures the relative intensity of each beam and displays the results as
a mass spectrum.
A computer data base of library results of mass spectra of different compounds may
then be used to identify the components of the fluid separated by the gas
chromatograph. This is possible due to the fact that many compounds give distinct
mass spectra, the peaks of mass/charge ratio occurring at specific relative intensities.
In the preliminary tests both chipboard and polyurethane foam were analysed, both
would probably have affected the nature of the pyrolysis products, however, for a
substantially reduced.
The inert atmosphere(helium)also acted as the carrier gas for the column of the gas
small sample on wires formed from Curie Point alloys. These alloys are of nickel
alloy, and when subjected to a high frequency induction coil, are heated to a specific
101
This is known as the curie point and is the temperature when the alloy becomes
paramagnetic, and its energy intake drops, thus holding the temperature constant.
Column details:-
Temperatures used:- 358, 480,510,610, and 770°C. Each sample held at this
Results from the chipboard were negative. This was because of the difficulty
small contact area of sample being against the curie point wire. This indicated that
small furnace.
Results obtained for the polyurethane foam were more encouraging, spectra being
obtained over several Curie Point temperatures. However, identification of the peaks
102
and subsequent parent molecules proved difficult even using the extensive data base
It was felt at this point that for such analysis to be continued any further would be
beyond the scope of this dissertation. The amount of work involved would be
worthy of a dissertation its own right. However, these analytical techniques offer a
powerful tool and there are many texts available on them[24]. It is felt that results
obtained from such analysis would require careful consideration when extrapolating
to fires within a compartment. Regard would need to given as to the nature of the
sample and pyrolysis techniques used. Investigation into the use of these techniques
may include the effects of the size and shape of the sample, pyrolysis temperature
103
REFERENCES.
K, Rosander M.
106
11. Survey of Backdraught. - Pub. by F.RD.G. Publication 5/94.
Author R Chitty.
13. Fire Terms. - A Guide to their Meaning and Use. By the N.F.P.A.
Pub. by H.M.S.O.
G.M. Makhhviladze.
H., Bullen M.L. , Quintere J.G. , McCaffrey BJ. Pub. Combustion and
348.
107
20. Salt Water Modeling of Compartment Gravity Currents. - Author
F.R.S. , and l.S. Turner. Pub. Royal Society Proceedings 1956 pages 1-23.
fire.
Author. Osami Sugawa, Kunio Kawagoe, Yasushi Oka and Iichi Ogahara.
24. Pyrolysis -Gas Chromatography. Author. R.W. May, E.F. Pearson, &
108
BIBLIOGRAPHY
Beck. Pub. Center for Environmental Safety and Risk Engineering, Victoria,
University, Australia.
7. Preparing for the Worst. Pub Fire April 1997 pages 27-28
Firefighter.
Atle William Heskstead & Per Jostein Hovde. Pub Fire Safety Science-
109
10. A Model for Predicting Concentrations of Carbon Monoxide in
Building Fires. Author. Shigeru Yamada & Takeyoshi Tanaka. Pub. Fire
Pages 539-550.
11. 'Flashover Training'. Author. John Taylor. Pub. Fire Engineers Journal
Author. F Jia, E.R. Galea & M.K. Patel. Pub. Fire Engineers Journal Vol
110
19. Radio Spares Catalogue. Sensorrrransducers Page 1-869
20. Essex Fire and Rescue Service-Realistic Fire & Flashover Training
11 1
APPENDIXD.
The following appendix contains the time/temperature curves for the fire box
experiments. To enable easier analysis of the graph on page lOla, it has been
expanded along the 'x' axis, time intervals of 300sec considered on each of the
104
TEMPERATUREICELCIUS
~ ~
N ~ O'l ex> 0 N
0 0 0 0 0 0
0 0 0 0 0 0 0
0
42
84
126
168
210
252
294
336
378
420
462
504
546
588
630
672
71 4
756 ~
798 i
-f
m "'Tl
840 "tJ
~ m :;0
I!! 882
(J) ;;0 m
CD
-"
0
CD
(') 924 5 0
UI C
ru 966 0
><
I -i
1008
~
co m
(j)
0 -i
1050 0 (j)
en
1092 CD
11 34 I
(J)
IIII
(J) (J) (J) (J)
I
(J) (J) (J)
I I
(J)
(")
til
1176 ~ ~ ~ ~ ~ ~ ~ ~ ~
iii' iii' iii' iii' iii' iii' iii' iii' iii'
(Jl (Jl (Jl (Jl (Jl (Jl (Jl (Jl (Jl
1218 <0 ex> -"j O'l (J1 ~ c...> N ~
1260
1302
1344
1386
1428
1470
1512
1554
1596
1638
1680
1722
1764
FIRE BOX TESTS
1000
- - Series1
900 - - Series2
- - Series3
- - Series4
800 - - Series5
- - Series6
- - Series7
700 - - Series8
- - Series9
(/)
::>
U 600
...J
W
(.)
iii
~
::> 500
I-
C2
W
Il.
:E 400
w
I-
/
300
200
100
-
o ~==~~~~~~~~~~
~~~~~~~~~~~~~~~~~~~#~~~$~~~~~~~~~~~$~~
TIME/Secs
101b
FIRE BOX TESTS
900
800
700
I./)
::> 600
(3
...J
W
U
W
a:: 500
::>
I-
~
W
0..
~ 400 - - Series1
w
I- - - Series2
Series3
- - Series4
300
- - Series5
- - Series6
- - Series?
200
- - Series8
- - Series9
100
o
##~4#~~#~~~#~~~~~~~#~~~~###~¢#~~$~~###
TIME/Secs
101c
FIRE BOX TESTS
1000
900
800
700
en
::l
U 600
-I
W
U
W
a:: 500
::l
t-
~ - - Series1
W
Q.
~ 400 - - Series2
w
t- Series3
- - Series4
300 - - Series5
- - Series6
- - Series7
200 - - Series8
- - Series9
100
##~~##~,#~#"#~~~~~~~~~~~#,~~###,#~,,,
TIME/Sees
101d
FIR E BOX TE STS
1000
900
800
?OO
(/)
:2 600
0
...J
W
0
iLl
~ 500
~
~
w
Q.
~ 400 - - Series1
~
- - Series2
Series3
300 - - Series4
- - Series5
- - Series6
200 I - - Series?
- - Series8
- - Series9
100
o ,~--------------------------------------------------- __________-J
##~~#~~$#~~~~~
~ ~ ~ ~ ~ ~ ~
~ ~###~$$####~~~$~~~$$~~~~
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
TlMElSecs
101e
FIRE BOX TESTS
900 r-----------------------------------------------------------------------------------------------------~
800
700
600
en
=>
(3
..J
W
u 500
ill
0:::
=>
I-
~ 400 - - Series1
w
Q.
:E - - Series2
w
I- Series3
- - Series4
300
- - Series5
- - Series6
- - Series7
200
- - Series8
- - Series9
100
o
,,');~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*
,,'); ,,'l.- ,,'l.- ,,'l.- ,,'V ,,'V ,,~ ,,'); ,,'l.- ,,'V ,,'V ,,'); ,,"3 ,,~ ,,~ ,,~ ,,~ ,,":r ,,"S ,,"3 ,,"3 ,,~ ,,OJ ,,"S "b< "b< ~ "b< "b< "bI "bI "b< "b< "b< ~ "bI "b<
TIME/Sees
101f
TEMPERA TURE/CELCIUS
->.
N ~ en ()) 0 N
0 0 0 0 0 0
0 0 0 0 0 0 0
1500
1506
1512
1518
1524
1530
1536
1542
1548
1554
1560
1566
1572
1578
1584
1590
1596
1602
1608 ~
3:
1614 m
"'C 11
1620 m
::0 ::0
-t 1626
0 m
0
~ 1632 0
.....
CD
0
co W1638 c.n ><
0 -i
f/I
1644 0
.....
I m
(f)
1650 co -i
0 (f)
0
1656 CJ)
CD
1662 n
1/1
1668
1674
1680
1686
1692
1698
1704
1710
1716 IIIIII
C/) C/) C/) C/) C/) C/) C/)
II
C/) C/)
1722 ~ ~ ~ ~ ~ ~ ~ ~ ~
0)' 0)' 0)' 0)' 0)' 0)' 0)' 0)' 0)'
C/I C/I C/I C/I C/I VJ VJ VJ VJ
1728 <D ()) -..J en c..n ~ VJ N ~
1734
1740
1746
1752
1758
1764 I
J
APPENDIXE
105
APPENDIXF
106
APPENDIXG
107
(