You are on page 1of 44

Fire Safety Journal 18 (1992) 139-182

Scale Modeling of Fires with Emphasis on Room


Flashover Phenomenon

S. J o l l y & K. S a i t o *

Department of Mechanical Engineering, University of Kentucky, Lexington, Kentucky


40506, USA

(Received 31 August 1990; revised version received 23 September 1990; accepted 18


November 1990)

ABSTRACT
The difficulties associated in predicting fire behavior by numerical
techniques necessitates the use of experimentally determined values.
Lower cost and convenience in conducting small-scale tests encourage
the use of such models in predicting actual fire behavior. Such
prediction is possible only when the outcome of a prototype and its
model are related by appropriate scaling laws. The aim of this work is
to focus on experimental and theoretical studies conducted in order to
develop and test these scaling laws. Although major emphasis is given
to studies aimed at enclosure flashover phenomenon, studies on scale
modeling of open fires are also discussed to enhance understanding of
fire modeling techniques. Despite the fact that considerable amount of
work has been done in developing these scaling laws, success has been
achieved only for certain fuels, fuel locations, interior finish materials
and room geometry. More research is needed in order to extend the
range of applicability of scaling laws to predict fire behavior in an
enclosure.

NOTATION

Af Floor area
Ai A r e a of the ith finish material
Ao A r e a of the opening
* To whom all correspondence should be addressed.
139
Fire Safety Journal 0379-7112/91/$03.50 © 1991 Elsevier Science Publishers Ltd,
England. Printed in Northern Ireland
140 S. Jolly, K. Saito

Au Area of the hot upper walls


Area of the walls and ceiling in convective contact with the hot
gases
Cpa Ambient specific heat
Cpg Specific heat of the gas
D Pan diameter
E Irradiance received by a radiometer at geometrically similar
points, scaled according to the size of the fuel bed
f Frequency of flame pulsation
f, Flame covered area to burner heat output constant
g Acceleration due to gravity
h Scale factor
ha Ratio of heat loss per unit area to the temperature rise
hc Heat of combustion of the fuel
hk Effective heat transfer coefficient through the walls and ceiling
Overall heat transfer coefficient = h . . . . . . rive "4- hradiative
H R o o m height
Hf Flame height
Ho Height of the opening
I Externally imposed radiative heat flux
k Mass flow rate constant
L Ratio of the summation of conduction heat loss through the
interior surfaces of the room and radiation heat loss through
the opening to the temperature rise
tc Crib stick length
Lw Flame depth
m Gasification rate, mass per unit per second of vapors leaving
the fuel surface
•h a Mass flow rate of air
ms Pre-exponential factor
rhv Mass burning rate of the fuel
rh,¢,$ Stoichiometric burning rate
nc Number of cribs
q Heat release rate
qc Heat release rate per unit mass of crib materials
qf Heat release rate per unit mass of liquid fuel
qi Heat release rate of ith finish material
Q Net heat of combustion of burning materials
Q Fire heat release rate
0r Radiant heat feedback rate to the fuel surface
Q, Total heat feedback rate to the fuel surface
R Fire bed radius
Room flashover phenomena in scale modeling of fires 141

R0 Universal gas constant


t Time
t~ Characteristic time
T Temperature
L Ambient temperature
T, Flame temperature
T. Wall temperature
AT Temperature rise relative to ambient
u Lateral velocity component of wind
w Vertical velocity component at a specified point in the flames
V Velocity of incoming air through the opening
v, Regression rate of liquid fuels
Velocity of flame spread
Volumetric flow rate
Wo Width of the opening
Y Fraction of oxygen consumed, rhv/rhv,s

Ot Qthv,s/pCa V
Pa Ambient density
Pc Density of crib materials
Pf Density of liquid fuel
.g* Non-dimensional flashover time

Superscripts
Corresponding values for the model

1 INTRODUCTION

Fire has been known to mankind since prehistoric times and is


considered to be the single most important contributor to the develop-
ment of m o d e m civilization, yet it is not a very well understood
phenomenon.14 The study of fires did not receive much attention from
researchers until the middle of the twentieth century. In the 1950s the
Building Research Institute of Japan performed a series of unique
experimental studies on building fires, 7-~° which provided momentum
for future fire studies. There are two main reasons for this lack of
research: ~ firstly, the available technology was insufficient to explain the
complexity of problems associated with fires; and secondly, insufficient
funding prevented extensive study in this area, since benefits derived
from this research were obscure. Accidental fires take nearly 10 000
142 s. Jolly, K. Saito

lives each year in the United States and cost billions ( x 10 9) of dollars in
property damages. T M Thus, the study of fires is important not only for
safety reasons but also for economic aspects. A proper understanding
of fire p h e n o m e n a in enclosures would assist in the design of fire-safe
interior finish materials, 8 would be a guide in the selection of more
effective fire suppressants, 9 and would help to determine concentrations
of toxic gases and smoke so that their effects on human beings could be
studied. 12-21This study would also be beneficial for designing protective
clothing for fire fighters 22-24 and would define the time constraints
during which the occupants could be safely rescued.
The major objective of this paper is to present a review of some
experimental and theoretical studies carried out in an effort to develop and
test scaling laws for room flashover p h e n o m e n o n . Scaling attempts
made to model 'open' fires are also discussed to elucidate scaling of
'enclosure' fires. The term 'open' fire is defined as fire occurring under
unconstrained spatial conditions, such as wildland fires and open tank
fires. In contrast, the term 'enclosure' fire implies fire occurring under
constrained spatial conditions, such as fire in a room or in a vehicle.
Flashover, an important indicator for fire safety, is defined as the
room fire condition when the heat flux levels at various locations in the
room are so high that it is believed that such high fluxes would ignite
most of the combustible materials kept in the room, thus causing the
entire room to be enveloped in flame. Flashover condition could be
defined either in terms of the temperature of the hot upper gas layer or
in terms of a certain heat flux value at a particular location in the room.
Flashover time, defined as the time between fire initiation and the time
when this particular value of temperature or heat flux is achieved, is an
important parameter used to describe the degree of fire development
and to determine the time available for fire suppression. A typical
temperature history measured at the center of a room is shown in Fig.
1, in which flashover time is indicated as shown. 9 As shall be seen in the
later sections of this paper, different definitions of flashover have been
used by various authors, thereby indicating the complex nature of the
phenomenon.
Ideally, tests performed on full-scale models would give accurate
predictions of fire behavior, but it is expensive, time consuming and
often dangerous to conduct full-scale tests. Accidental occurrence of
wind, which is difficult to control in full-scale tests, may change the
nature of fire and further complicate the testing procedure. ,In addition,
numerous possible combinations of internal arrangements of fuels in a
room, exacerbated by a large number of parameters involved, often
make conducting of full-scale tests almost impractical. There is a need,
Room flashover phenomena in scale modeling of fires 143

~ lome

~ r T herrnocoupie
Locotion

I
Fire Growth I Developed Fire Decoy
Process I Fire I Process
I
I

( Initiation ( F. O. Time )
of Fire)
Time After Fire Initiation

Fig. 1. A typical temperature history at the center of r o o m . 7

therefore, to develop reduced-scale models and to perform these tests


under well-controlled laboratory conditions.

2 D I F F E R E N T M O D E L I N G SCHEMES

The studies relating to room fires can be divided into four groups:
stochastic models, field equation models, zone models and scale
models. 25 The stochastic modeling approach is used to predict the
behavior of fire with the help of mathematical models developed from a
series of experiments considered to be probabilistic events under the
assumption that such experiments can be expected to resemble those
under actual circumstances. In field equation modeling, the conserva-
tion equations are solved (most often numerically) by applying ap-
propriate boundary conditions. In zone modeling, the fluid mechanics,
heat transfer and combustion processes are mathematically coupled to
characterize the fire growth behavior. 26 The field equation modeling
144 s. Jolly, K. Saito

approach gives a point by point solution, whereas the zone modeling


approach gives a solution for each of the zones considered in the
analysis and interrelations among them. Therefore, the field modeling
technique gives greater detail, while the zone modeling technique is
simpler to apply to practical situations. 27 Stochastic models, field
equation models and zone models face several difficulties in predicting
the fire behavior in an enclosure. According to Kanury, z5 lack of
complete knowledge of the chemistry of combustion and c o m p o n e n t
processes pose major difficulties associated with field and zone models,
respectively. The scale modeling technique is an attempt to predict the
behavior of full-scale tests from the results obtained from reduced- or
enlarged-scale tests. Scaling laws correlate the data of different sizes of
tests conducted. When scale modeling is used, the difficulties associated
with the stochastic, field equations and zone models can be avoided, z5
The use of scale models is justified only if it is able to provide
information relevant to physical as well as chemical processes of the
full-scale tests. These processes can be defined in terms of equations
which contain various parameters, each of which may contribute to the
behavior of the process. The determination of the scaling laws requires
understanding of the importance of each of these parameters on the
desired outcome. Partial modeling is the 'art' of identifying important
parameters so that any desired outcome could be predicted with a
reasonable degree of accuracy. The concept of partial modeling was
first introduced and discussed by Spalding. 28 U n d e r certain conditions, a
strong parameter may become weak. For example, the regression rate
of a transparent fuel burning in an open pool becomes constant once
the diameter of pool exceeds 1 m . 29 Yetter et al.,3° suggested the use of
sensitivity analysis to evaluate the role of various parameters on the
desired outcome to serve as a guide in the selection of strong
parameters. The approach to designing a model for combustion
problems is to select significant groups from a core subset of dimen-
sionless groups and to keep these selected groups constant for the
prototype and its model. According to Williams, 3~ the degree of control
an experimenter has over the model depends upon the number of
groups selected--the fewer the n u m b e r of groups, the lesser is the
degree of control. The general requirement of the scale modeling
approach is:
:~i = st; (1)
where
i=1,2,3 ..... n,

:ri represents the ith dimensionless group, n is the total n u m b e r of such


groups involved in a p h e n o m e n o n and prime (') represents correspond-
Room flashover phenomena in scale modeling of fires 145

ing values for the model. There are different approaches to obtaining
the scaling correlations; the most significant are the law approach, the
equation approach and the parameter approach. 3°'32 The suitability of
any particular approach depends upon the complexity of the problem
involved.
In fire and combustion problems, scaling or simulating transient
heterogeneous surface pyrolysis reactions is difficult to achieve. A
simplified assumption for a prototype and its model is that the overall
reaction follows the Arrhenius type: 1
m = m s e -e~/R°r (2)
where m is the gasification rate, ms is a pre-exponential factor, and Es is
the overall activation energy for surface pyrolysis. The overall chemical
process then remains the same if the concentration and temperature of
species at a corresponding time and place are the same. Similarity of
chemically reacting flow systems was first introduced and discussed by
Damk6hler, 33 who introduced four important dimensionless numbers,
known as Damk6hler's similarity group. These numbers are defined as
ratios of process rates: 33 DI = (chemical reaction)/(bulk flow), D . =
(chemical reaction)/(molecular diffusion), Din = (heat liberated)/(heat
transported by bulk flow), a n d / ) i v = (heat liberated)/(heat transported
by conduction). In practical applications, usually, it is difficult to satisfy
all the above dimensionless groups simultaneously. Therefore, some
approximations have to be made in order to reduce the requirements
for similarity of chemical reactions between the prototype and its
model. For example, in large-scale fires, there is enough time for the
reactive gases to reach chemical equilibrium. For such cases, the
simplifying approximation can be made that the mixing process controls
the heat release rate. However, for small-scale laminar fires, this
approximation may not be valid, since the order of chemical reaction
time may become the same as that of mixing time, and the chemical
equilibrium may no longer be achieved. In this paper, chemical reaction
scaling is not discussed; it is assumed that the chemical reactions remain
the same for the prototype and its model if the temperature and species
concentrations are the same.

3 O P E N FIRES

3.1 Pool and crib fires

So that the basic characteristics of fires could be elucidated, experi-


ments were performed on pool and crib f i r e s . 29"34"35 Spalding 36 explained
146 s. Jolly, K. Saito

that the burning rate of buoyancy-controlled turbulent pool fires is


governed by natural convection rather than by radiation. Following
Spalding's explanation, a series of large pool fire tests were performed.
The results demonstrated that an increase in the pool diameter
increased the contribution of radiation to the total heat flux received by
the fuel surface. Figure 2 shows a plot of the contribution of radiation
to the rate of total heat released which was received by the fuel surface
versus pool diameter. 37 For example: a recent series of kerosene pool
fire experiments using 30-m, 50-m and 80-m diameter ground pools 38
showed that the radiation accounted for nearly 90% of the rate of total
heat feedback to the fuel surface for the 50-m diameter pool, whereas
experiments using the 30-m and 80-m diameter pools yielded unreliable
results, due to instrumentation error and wind effects. Thus, radiation
is a primary mode of heat transfer which influences the burning rate of
large pool fires; this result is supported by Hottel, 39 Thomas et al., 4°
Yumoto, 41'42 Modak & Croce, 43 Alger et al., 44 de Ris, 45 Ndubizu et al., 27
Mudan, 46 Shinotake et al. ,47 and Brosmer & Tien. 48
Emori & Saito 49,5° developed scaling laws based on physical hypoth-
eses for non-propagating pool and crib fires. They introduced a
two-length scaling concept for pool fires and a one-length scaling
concept for crib fires, and found fourteen dimensionless groups for the
pool fires and seventeen for crib fires. Only groups of significance were
retained to establish criteria for partial similarity between the full-scale
model and its counterparts. For pool fires, the form of the functional
relationship was found to be:

tp 'D' v uz'pf~Vf =0 (3)

The scaling laws predicted the pulsation frequency, f, to be inversely

i
.A ETHYL-BENZENE [ 3 7 ]
O BENZENE ] 3 5 ]
• J P - 5 [44]
0 NETHANOL [44]
• KEROSENE [38]

0.5

I I I I
0•!• o.1 o.'5 , lo
o (m)
Fig. 2. A plot of the contribution of radiation (Qr) to the rate of total heat feedback
(Qt) to the fuel surface for different fuels versus pool diameter (D). 37
Room flashover phenomena in scale modeling of fires 147

proportional to the square root of the pan diameter, D; the velocity


(horizontal component u, and vertical component, v) to be directly
proportional to the square root of the pan diameter, D; and irradiance,
E, measured at geometrically similar locations (outside flame), to be
independent of pool diameter. For crib fires, they obtained a different
functional relationship of the form:

(,, I-1, Log


L~ t¢, u2
eto
=0 (4)
u ' Lc' u ' p~LcqJ

which includes the additional requirement that the time of burning and
irradiance is proportional to the square root of the length of the crib, Lc
(see Fig. 3). The scaling laws proposed are valid only for open pool and
crib fires and for non-charring materials. The validity of the derived
scaling laws for pool fires was confirmed by experimental results for
several different single-component liquid fuels for tank diameters
ranging from 1 m to 10 m, and for crib fires with crib lengths ranging
from 0.2 m to 1 m.
Further increases in pool diameters, however, caused significant
generation of smoke from the flames, as shown in Fig. 4. The generated
black smoke enclosed the luminous flame zone and blocked radiation
emitted by the luminous flame? 1 As a result, the radiative heat flux,
measured outside the flame core at geometrically similar locations,
decreased significantly (see Fig. 5). These results served to invalidate
the proposed scaling laws. For such cases, the smoke generation rate
depends on fuel type, 52,53fuel geometry54 and mixing conditions caused
by the surrounding oxidizer flow?5 The mechanism of radiation

~3

CRIB LENGTH
K 2O
--'-- 60
(8~L = 14) .... 80

I I I
°b I 2 3 4
l x ~ L~/L (min)

Fig. 3. Scaled irradiance as a function of scaled time for crib fires (Lc is the crib
length, L3 = 90 cm, and tSc is the horizontal distance from the center of crib to the
location of radiometer). 49
4~
O~

(a) (b) (c)

(d) (e) (f)


Fig. 4. Photographs of laminar and turbu|ent liquid hydrocarbon diffusion flames. (a) D = 0-016 m (benzene); (b) D = 0.05 m
(benzene); (c) D = 0.6 m (crude oil); (d) D = 6 m (crude oil); (e) D = 10 m (crude oil); (f) D = 30 m (crude oil).
Room flashover phenomena in scale modeling of fires 149

0.5
[] [] []
[] [] Z~ []

0.1 Q
°0" A
0.05
O HEPTANE
• A
• CRUDE OIL
A KEROSENE

0.01 i i i i I t , , , I , , , 1
0.1 0.5 1 5 10 50
(m) D
Fig. 5. A plot of the contribution of radiation (Qr) to the total heat release rate (Q)
for different fuels versus pool diameter (D). 51 The total radiative output was calculated
from 0R = 4~L 2E using the measured irradiance, E, at the distance, L, from the center
of the tank. The total heat release rate was calculated from QR = st(D/2)ZVfpHc using
the measured fuel regression rate, Vt, at the steady-state burning condition. Here, p is
the density of fuel and Hc is heat release rate per unit mass of liquid fuel.

screening, caused by smoke for large turbulent diffusion flames, is not


well understood.27'5 !'56'57
Efforts have b e e n m a d e to correlate species profiles b e t w e e n laminar
and turbulent diffusion flames by employing the mixture fraction, Z.
This mixture fraction is purely a fluid dynamical mixing p a r a m e t e r 58-61
and is given as
Y, -
z - - - (5)
Y. -

where Y, is the mass fraction of e l e m e n t i and the subscripts 1 and 2


refer to fuel and oxidizer respectively. F r o m eqn (5) the mixture
fraction for different e l e m e n t species can be calculated. In normal
h y d r o c a r b o n - a i r combustion systems, the element species are C, N, O
and H. For different types of laminar and turbulent diffusion flames, a
high correlation was found b e t w e e n concentrations of final steady
products and mixture fraction based on carbon (Z¢) (see Fig. 6 ) . 5 8 ' 5 9 ' 6 2 - 6 6
However, correlation b e t w e e n intermediate and high molecular weight
products, which are related to the formation of soot, is significantly
weak (see Fig. 7), as the mixture fraction did not account for chemical
reactions. 6. E v e n for a candle-like laminar h y d r o c a r b o n - a i r diffusion
flame, the chemical and t e m p e r a t u r e structures are complicated and not
fully understood. A n example of such a flame is given in Fig. 8 for (a)
visual observation and (b) species and t e m p e r a t u r e profiles.
150 S. Jolly, K. Saito

1500 I 9

P
n. -

~ ~ooo
rr ~g
uJ
n

i,i
i
500
f
.~__SCALE
I CHANGE

TICK'S LOCATION
SYMBOLS INVtESTI- IN (h/H)
GATORS (BOTTOM)(SIDE)(TOP)
TSUJI AND
HALF-SOLID YAMC~KA

0.2 SOLID MrFCHELL~Ol 021 041 086(H,SSmm)


O¢:~N PRESENT 014 02.50~ (H=~Smc'n):
V,O RK

FOR CO AND COt 14

>_~ o.i

0 0.2 1.0

i, YCH

::,_~0.1 o~

=\

~~0.2 0.4i o'.6 o18 " 1.0


MIXTURE FRACTION Zc
Fig. 6. Profiles of temperature (T) and of mass fraction (Y) of CH4, CO2 and 02
versus mixture fraction based on C (Z¢) for a 1.6 cm diameter CH4-air coflow laminar
diffusion flame with 3-8 cm flame height, for the flame of Mitchell et al. 65 and for the
flame of Tsuji & Yamaoka 6~ (from Ref. 64).
Room flashover phenomena in scale modeling of fires 151

0.15
FLAM[
SYMaOLS HEIC,HT BCV[STK~ATOR$
• ] (nv,r,) ~ AgO "ItMA~A

0.10
o ~ (19861
X
>_e
Q05

4
,~ 0.20
X

6 0.10
#1
>_,-,

0
0.15
b
\

,'2
o× 0.10
o
6
6 0.05 \
0
?
o 0.4
X
oi

>_,-, 0.2

? 0.2
0
X
'-' 0.1
>._,-,
0
1.0
o
x 0.5

MIXTURE FRACTION Zc
Fig. 7. Profiles of mass fraction (Y) of allene, methyl acetylene, benzene and
acetylene versus mixture fraction based on C (Zc) for a 1.6 cm diameter CI-L-air coflow
laminar diffusion flame with 3.8 cm flame height, for the flame of Mitchell et al. 65 and
for the flame of Tsuji & Yamaoka 66 (from Ref. 64).
152 s. Jol~, K. Saito

Air entrainment into the flame interior was studied for a laminar
coflow m e t h a n e - a i r diffusion flame using a 1.6-cm diameter burner, 67
for turbulent (gaseous and liquid) diffusion flames using a 30-cm gas
burner, 68 and for open tanks of diameters 30 c m , 51'69 60 cm, 1 m, 2 m
and 6 m. 51 These studies found that m a j o r air entrainment occurred
near the flame base. H o w e v e r , it is not clear w h e t h e r the air
entrainment occurs by penetration through the flame sheet by turbulent
mixing or by diffusion through the quenching region (a dark annular
area formed b e t w e e n the burner rim and flame edge, i.e. in the
quenching region). These two different mechanisms are schematically
illustrated in Fig. 9. In order to study these two mechanisms,
experiments were p e r f o r m e d on a kerosene pool of 50 m diameter. 38
10 m o l % oxygen was found 1 m from the fuel surface and along the
flame centerline at 300 s from ignition. This finding was not predicted
by the smaller pool fire experiments. For example, 1 m o l % oxygen was
found at locations similar to that of kerosene pool m e n t i o n e d above
when experiments were p e r f o r m e d on a 1.6-cm diameter m e t h a n e - a i r
coflow diffusion flame. 67 The only possible mechanism to explain this
result for the 50-m diameter pool fire is that a large a m o u n t of air was

~rn)
2
3
:4

THIN

THIN

THICK BLt

t
AIR
I
CH 4

Fig. 8. (a) A schematic diagram of a laminar methane-air diffusion flame. Burner


diameter is 1-6 cm, fuel flow rate is 2.2 cm/s, and air flow rate is 9.8 m/s. For different
hydrocarbon fuels, a two-color zone structure is normally observed. In the yellow zone,
a polymerization reaction takes place, which causes the soot particle temperature to be
higher than the gas temperature, whereas in the orange zone, the soot particle
temperature and the gas temperature remain the same.
H2 MOLE PERCENT
0
I i , v CzC-C, C=C=C,C:C-C, @
0 ~ X
I l I I
0 2 , C2H z , C2 H4, CzHs
0 O 0 P O P P P O
I I I I I I t I I i
MOLE PERCENT x
CH4, CO,CO2
5 o o ~ ~
8 r T T

o ? ~--..i', .---o
8-

,,, ~=
..( 0 t~
m ~'
0

g g

o :/'/°
-- :'U --

oO
~o
~..ao go
TEMPERATURE ( K )
154 S. Jolly, K. Saito

/
/f'~~jlj FIorne Sheet

J
. ,,, ,,oioo
(MECH. B) i ~ . ~ L _ . I __ r #
/ Fuel 1
r'~ Open Tonk
Fig. 9. Schematicsof two different mechanisms of air entrainment into a turbulent
diffusion flame: (1) a mechanism associated with turbulent mixing which can penetrate
the flame sheet and carry air into the flame interior (mechanism A) and (2) a
mechanism associated with diffusion of air through the quenching zone (mechanism B).

entrained through the large quenching zone extending several centi-


meters in height. Since the flame shape was cylindrical, with a flame
height to pool diameter ratio of three, the turbulent mixing effects of
flame sheet were not likely to reach the center of the flame.
Flame height measurement for such a large pool fire is difficult to
achieve, 7° since the flame becomes filled with black smoke and pulsates
periodically. Koseki & Y u m o t o 51 measured the center line thermo-
couple temperature in their unique indoor test facility for 6-m diameter
heptane pool fires. They found that the location of the m a x i m u m
temperature was approximately one-half of the visible flame height
centerline. Interestingly, the location of the m a x i m u m temperature was
also found to be nearly one-half of the visible flame height for a 0-7-cm
diameter laminar coflow hexane-air diffusion flame. 71 Williams 1 ob-
tained a high correlation between non-dimensional ratio of flame height
to horizontal scale of the fire and Froude (Fr) n u m b e r (extending over
15 orders of magnitude). This correlation can be used to estimate the
amount of heat liberated from the fires. Therefore, in general, both
experimental and analytical studies are required for further under-
standing the structure of large-scale, buoyancy-controlled turbulent
diffusion flames.
Alpert 72 applied the technique of pressure modeling3L73 to the
transient process of fire growth and decay in pine wood cribs. In the
Room flashover phenomena in scale modeling of fires 155

fuel surface-controlled regime, a fairly high correlation was found


between the sizes of fires tested and the peak burning rates as well as
the decay in burning rates. However, the attainment of peak burning
rate was delayed in the model tests due to non-scaling of fire spread
rate. The accuracy of the results predicted decreased when the ambient
pressure was raised from 12 to 30 atmospheres. For the ventilation-
controlled regime, the entire burning rate time history could be
accurately modeled, since, in this region, the effect of flame spread
produced less severe burning rates.

3.2 Mass fires

A thorough discussion of mass fire phenomenology is given by


Williams. 1 In one study, Williams obtained 29 dimensionless para-
meters for mass fires by non-dimensionalizing the governing equations
and boundary conditions. 31 According to Williams, mass fires are those
fires which involve more than one sizeable structure. He showed that
the dimensionless groups representative of radiative and convective
heat transfers could not be satisfied simultaneously unless adjustments
were made in the composition and temperature of ambient atmosphere
and fuel properties. This simultaneous scaling of radiation and convec-
tion was much more complex than in the case where either radiation or
convection was to be satisfied. Even in this case, when simultaneous
scaling of radiation and convection was applied, the radiation scaling
was only approximate, since the dimensionless group representing
radiation absorption length was not satisfied. The high adjustments in
pressure requirement impose a limit on the size range over which the
scaling laws can be applied. These pressure requirements can be offset
by allowing the acceleration due to gravity, g, to vary. However, this
would require a centrifuge capable of accelerating the chamber at
10-100 g. At the same time, the chamber must be capable of handling
internal pressures higher than 10 atmospheres, and there must be a
means of varying the temperature and composition of gases within the
chamber.
The main advantage of using pressure modeling is that both the
dimensionless groups for the fluid flow, Froude (Fr) number, (inertia
force)/(buoyancy force), and Reynolds (Re) number, (inertia
force)/(viscous force), can be satisfied simultaneously, de R i s e t al. 73
applied the pressure modeling technique to model mass fires at
atmospheric pressure. Here, convection was assumed to be the main
mode of heat transfer, and radiation was assumed to be of secondary
importance. The radiation was assumed to be proportional to the
156 S. Jolly, K. Saito

burning rate of fuel. The Grashof (Gr) number, [(buoyancy


force)/(inertia force)][(inertia force)/(viscous force)] 2, was the govern-
ing dimensionless group for the fluid flow. Similarity between the
full-scale tests and small-scale laboratory tests was achieved when the
pressure was adjusted to be proportional to (length) -3/2. Two different
time scales were defined: one for the gas phase and the other for the
solid phase. The purpose of defining the two time scales was to
separately model the steady state and the transient fire p h e n o m e n a . A
high correlation was achieved between burning rates and Gr number
for both steady and transient flows. The modeling of fire spread and
solid-phase heat and mass transfer was also found to be successful.
Alpert TM was partially successful in his effort to apply pressure
modeling technique to fires dominated by radiative heat transfer. The
buming rates and radiation for large-scale polymethylmethacrylate
(PMMA) wall and pool fires were compared to the model results. For
wall fires, the burning rate was found to model when the operating
pressure in the design chamber system was below 30 atmospheres.
However, at very high pressures (30 to 35 atmospheres) the burning
rates were lower than expected for P M M A walls. He also showed that
the net radiative feedback (difference between the flame radiation and
surface reradiation) to the walls was approximately modeled only for a
Gr number (based on the distance from the bottom of wall) of
5 x 101°. At Gr numbers significantly different from 5 × 10 ~°, radiation
did not change according to predictions of the modeling theory. At
higher Gr numbers, flame radiation did not increase sufficiently, while
at lower Gr numbers, surface reradiation was not accounted for by
the model. For pool fires, the modeling of burning rate was found to be
successful for pool sizes less than 40 cm in width.
Apart from pressure modeling, attempts have been made to partially
model mass fires in order to simulate only a few parameters of interest.
Lee 75 applied the modeling concept suggested by Williams 3~ to predict
temperature fields and air flow velocities for large fires. The Froude
number was taken to be the main dimensionless group for the fluid
flow, since the viscous forces as compared to the inertia and buoyant
forces were considered to be unimportant. It is valid to neglect the
viscous forces when considering many fire problems. However, for fires
where the boundary layer plays an important role (such as laminar
flame spread phenomena), the viscous forces become significant. With
this assumption, the velocity of wind would scale as the square root of
the characteristic length, and gas temperature would remain the same.
Two different types of models were employed: a gas model using
methane as fuel, and an electrical model using an electric heater as a
Room flashover phenomena in scale modeling of fires 157

heat source to simulate fire. Attempts to use electrical models were


very successful in simulating velocity and temperature profiles for
large-area fires, but considerable variations were found when the gas
model was used in place of the electric model. The main reasons for the
deviations in velocity profiles for the gas model were thought to be due
to scatter in the model data at early times and incomplete adjustment of
wind during the early phase of prototype fires. Non-scaling of radiation
was thought to be the main reason for failure to obtain similar
temperature profiles for the gas model.
Due to high costs and difficulty of controlling boundary conditions,
experimental data on large-area fires is limited (e.g. Refs 76 and 77).
An experiment on a 50-acre mass fire was performed to determine
whether the induced air velocity is proportional to the cube root of fire
heat release rate. At the peak of the fire, the velocity of the wind was
4.2 m/s, the convective heat release rate was 2 × 107 kW, and radiative
heat release rate was 4 × 106 kW. The radiative heat flux was 20% of
the convective heat release rate. If the induced wind velocity is
proportional to the cube root of fire heat release rate, then the
predicted wind velocities of 2.8 m/s or 3 - 0 m / s (depending u p o n the
relationship used) were much lower than 4-2 m/s measured experimen-
tally. These results indicate that the induced air velocity must be a
stronger function than the cube root of energy release rate.

3.3 Fire plumes and fire whirls

Classical and extensive studies on the behavior of fire plumes generated


by large turbulent diffusion flames were done by Yokoi, ~° Rankine, TM
and Taylor. 79 Fire whirls generated in urban and wildland fires have
important aspects associated with environmental safety problems.
Soma 8°-a2 and coworker 83 conducted a series of extensive experimental
studies on large- and small-scale fire whirls. They found that all the data
correlated with Fr n u m b e r and Rossby (Ro) number. Here, F r =
uEax/dg and Ro = 2~Umaxd/F, where Umax is the maximum tangential
velocity of whirl, d is the whirl core diameter, and F is a measure of
vortex circulation. Soma & Saito 83 collected historical records of fire
whirls and categorized t h e m into three groups: Hifukusho-Ato fire
whirl ( H A F W ) , H a m b u r g fire whirl (HFW) and Dessen's fire whirl
(DFW) (see Fig. 10). Examples of fire whirls generated in large mass
fires have been reported elsewhere. 84-87 Since the full-scale fire whirls
are too huge to be tested, reduced-scale model tests have been
performed to understand the mechanism of the prototype fire
whirls. 75'8L88-9° However, very little has been done to develop scaling
158 S. Jolly, K. Saito

....,............
:.:.:.:.:.:.:+:
...,............
......,..........
:~:~:~:!:!:!:1%
...,......,..
.................
...............


:.:.:.:.:.:.:.:,:
.................
..........,....

.................

U S H0 - A T 0 ~~ ~ "'''''
,.............

':':':':':':':':"
.........,,

,-.,.....-.,.
....,.,...,

.....,......
,................
,,............

................ ~ .:.:.:.:.:.:.:.
,..............
...,............

::::::::::::::::: .~
ili!i!iiiii
iiNN!i!i!i!iiiiiili!i!i!iiiiiiilili!iii!iil
o

tYPE-

tllllUlll,,

TYPE- DESSENS

Fig. 10. Schematics of three different types of fire whirls.84 The generated whirls are
indicated by arrow lines.

laws for fire whirls. Emmons & Ying 88 performed a small-scale fire
whirl test. They suggested that Fr number and Ro number are
important dimensionless groups to be considered in the modeling of fire
whirls, a conclusion in agreement with Soma's results. Emori and
coworkers 9°'91 discussed scaling laws and scaling techniques concerning
fluid dynamical and thermal aspects of fire plumes. They suggested Fr
number scaling, which is also in agreement with Soma's result. Grashof
number scaling has also been applied successfully to model buoyancy-
controlled combustion of spherical gas clouds. 92 This scaling, however,
can be reduced to Fr number scaling under the fully developed
turbulent conditions, since the viscous term becomes negligible. There-
fore, Fr number and Ro number are useful tools for fire plume and fire
whirl simulations.
3.4 Flame spread
Excellent reviews of flame spread phenomena are available in many
references (e.g. Refs 93-95). However, very few attempts have been
Room flashover phenomena in scale modeling of fires 159

made to apply the scale modeling approach to flame spread phenom-


ena. In this paper, attention has been limited to scale modeling related
studies only.
Emori et al. 91 attempted to develop scaling laws for turbulent flame
spread by selecting out the governing physical quantities. The purpose
of the study was to provide better understanding of the turbulent flame
spread over horizontal and up-slope surfaces so that more effective
methods to predict urban and wildland fires could be developed. Paper
strips coated with candle wax and embedded in Marinite board were
used to simulate horizontal and up-slope flame spread for stem fires,
while excelsior of different sizes were employed to simulate grassland
fires. Experiments were performed to measure flame spread over these
two models. It was found in the study that the pyrolysis line was nearly
vertical for the paper strip model and about 45 ° for the excelsior model.
Larger entrainment rate to the upper part of the pyrolysis plane than
that to the lower part, which was due to porosity of fuel used, was
thought to be the main reason for the inclined pyrolysis line in the case
of the excelsior model.
For the paper strip model, it was found that the spread rate and
irradiance were independent of the size of the fuel bed when the fuel
bed tilt angle, re, was equal to 0°. The independence of irradiance from
fuel bed size suggested the applicability of the scaling laws developed
for pool fires to these kinds of fires. For 300 - c r - 45 °, the obtained
results suggested the applicability of scaling laws for crib fires to these
kinds of fires. However, for 0°< a~ < 30°, no clear classification was
available. The spread rate and irradiance measurements for excelsior
models suggested that the scaling laws developed for the crib fires could
be used for this type of flame spread. Experiments conducted on
prototype grassland (20m x 17.5 m) and corresponding scale models
(1/10 and 1/25) confirmed that Fr number scaling, Vfs/Vf's = L~w/L'w,
was sustained between flame spread rate, Vfs, and flame width, Lw.
Here prime (') indicates corresponding values for the model. It was
found that convective heat transfer was more dominant than radiative
heat transfer for the grassland fires. Thus, the result was in agreement
with the modeling theory requirement.

4 E N C L O S U R E FIRES

The behavior of fires in enclosures differs from that of open fires. The
complexity of the problem increases because of changes in fluid
dynamic aspects and heat transfer characteristics due to the presence of
openings, combustible walls and ceilings.
160 S. Jolly, K. Saito

4.1 Mechanism

Consider a combustible material kept on a room floor. When this fuel is


ignited, exothermic reactions take place which cause a rise in the
ambient temperature and thermal radiation. This energetic conversion
produces a plume of hot gases and smoke which rises to the top of the
room due to the buoyancy effects and impinges on the ceiling to form a
ceiling jet. The ceiling jet moves along the ceiling until it reaches one of
the walls. The temperature and velocity of the fluid in the propagating
ceiling jet decreases because of heat loss and wall shear, respectively.
After striking the wall, the ceiling jet is deflected, and it moves in a
downward direction along the wall. The vertical distance traveled by
the fluid depends upon its flow characteristics. If the fluid momentum is
greater than its buoyancy, then the fluid would flow down the wall
before its momentum is overcome by buoyant forces. The greater the
downward momentum of the fluid, the greater is the distance traveled
by the fluid and the higher is the amount of mixing. If the buoyant
forces in the fluid are dominant, then the fluid would tend to move back
towards the plume under the ceiling jet. This would cause less mixing
and a high degree of stratification in the medium. 96
The layer of smoke and hot gases trapped in the upper part of the
room contains soot particles and toxic products. Radiation is emitted
from the smoke and hot gas layer which heats the unignited fuel. This
heating of unignited fuel causes a rapid increase in the amount of fuel
taking part in the combustion process. The radiation emitted by the
flame is a function of wavelength, time interval from initiation of fire,
size of fire, type of burning material and local conditions. 97 Radiation
emitted by the ceiling and hot upper walls is partially attenuated by the
hot gases, smoke and soot particles in the upper part of the room.
Simultaneously, the combustible walls are heated by radiation and
convection from the flames, hot gases and burning neighboring walls.
The effect of this heating is greater on the upper walls than on the
lower portion of the room because of the proximity of the upper walls
to the hot gases. When the walls are heated sufficiently, rapid pyrolysis
occurs and the flashover phenomenon is enhanced. 25 Heat loss occurs
from the walls to the surroundings by conduction. The hot gases escape
through the top of the opening, and cool air from the surroundings
flows in through the lower part of the opening. The cool air tends to
cool the floor and lower portions of the walls. The amount of smoke
and hot gases accumulated under the ceiling depends upon the soffit
height, which is the distance between the top of the opening and the
ceiling, and on the rate of burning of fuel and combustible walls.
Room flashover phenomena in scale modeling of fires 161

Convection is dominant in the early stages of the fire but radiation


becomes dominant as the fire progresses. When the temperature of the
hot layer reaches about 920K or above, it is assumed that there is
sufficient radiation to the lower part of the room to ignite practically all
combustibles present. 98

4.2 Relationship between rate of burning and ventilation factor

Kawagoe 99 found that the burning rate of wooden piles in an enclosure


is proportional to AoH~ r2. Thomas 1°° has shown that the rate of weight
loss of fuel in an enclosure is not a simple function of the ventilation
factor, AoH~ ~2, but involves the shape and surface area of the enclosure.
Gross & Robertson 1°1 conducted experiments to measure burning rate,
temperature profiles and gas compositions in different sizes of en-
closures used. They observed that the burning rate was influenced by
the ventilation factor until a limiting value of AoH~/2 was achieved.
Beyond this limiting value, the burning rate had a constant value which
was equal to the rate of burning of cribs in an open environment. When
cribs with different surface areas were used, it was observed that the
effect of fuel surface area on burning rate was small until the critical
fuel surface area was achieved. Above this critical value, the effect was
quite large. Similar results were also reported by Quintiere et al. 102 for
PMMA pool fires burning in an enclosure. They found that for small
doorway widths, the rate of mass loss increased with increase in
ventilation factor, since more oxygen was available for combustion to
take place. At higher values of the ventilation factor, the mass loss rate
became a stronger function of fuel area and radiative heat transfer,
since excess oxygen was present to support higher burning rates.
Babrauskas 1°3 came up with a simple relationship between the heat
release rate required to achieve flashover and the ventilation factor.
The minimum heat release rate required to achieve flashover was given
as

0 = 600AoH~/2(kw) (6)

where the enclosure dimensions are expressed in meters. On the other


h a n d , L e e 96 found that eqn (6) predicted a minimum heat release rate
of 1170 kW to cause flashover for one of his experiments, whereas he
found experimentally that 300 kW was sufficient to cause flashover.
McCaffrey et al. lo4 developed a relationship between the heat release
rate of fire and room ventilation, room geometry and thermal pro-
162 S. Jolly, K. Saito
perties of the room. The expression was given as
1/2
0 = Vgcpgpa T2(AT~3~ {hkAoV~o} '/2 (7)
a\480,/ J
Lee 98 applied this correlation to find the minimum heat release rate
required to achieve AT of 873 K. The predicted value of 450 kW was
consistent with the experimental value of 475 kW required to obtain A T
of about 873 K in a room with the burner kept at the center of the
room. In developing this correlation, McCaffrey et al. did not include
data from tests involving combustible walls and ceilings, and the
correlation may therefore not hold well for corner or wall fires.

4.3 Experimental studies

Fang 1°5 conducted a study in order to determine the effects of various


interior finish materials on the fire buildup process in an enclosure. To
simulate enclosure fires from a single furnishing, wooden cribs were
used as fuel. The rate and extent of fire involvement in the room was
evaluated by using different types of interior finish materials and crib
sizes. Ignition of newspaper and plywood specimens kept at various
locations in the room were used as criteria to determine the degree of
fire involvement. Among the limited range of interior finish materials
considered in his analysis, a maximum upper room temperature of
1076 K was achieved when melamine-finished hardboard was used as
the covering for the walls and ceiling. In these experiments, the flux
measurements at the center of room floor at the time of ignition were
found to vary from 1-7W/cm 2 to 2-5W/cm 2 for the newspaper
specimen and from 2.1 W/cm 2 to 3.3 W/cm 2 for the plywood specimen.
The corresponding values for the upper gas temperatures were 731 K to
911K and 804K to 956K, respectively. The mean horizontal flame
spread rate along the ceiling was found to be higher than the mean
vertical flame spread rate up the walls. An attempt was made to
empirically correlate the following: (1) maximum upper temperature in
a room and flame spread measurements on small-scale laboratory test
specimens; (2) maximum upper temperature in a room and heat release
measurements on small-scale test specimens; (3) ignition time in a room
and in a test specimen made up of same interior finish material; and (4)
smoke generation in a room and in a smoke chamber. Data obtained
for the ignition time, flame spread and heat release rate correlated to a
fairly high degree. However, no clear correlation was found for the
smoke-generation capabilities. The main drawback of developing such
empirical correlations is that they may not be applicable to other
Room flashover phenomena in scale modeling of fires 163

interior finish materials and experimental settings. Convective heat loss


was found to be the major m o d e of energy loss from the gases leaving
the room through the opening, and these gases became a threat to
neighboring rooms and buildings. Studies on the spread of fire from an
enclosure to the neighboring rooms, corridor and buildings, were
reviewed by Quintiere 1°6 and are not subjects of this discussion.
Quintiere & McCaffrey 1°7 conducted experiments on wood and
plastic cribs in an enclosure. The objective of their study was to analyze
the behavior of cellular plastics, to determine the effects of fuel type,
fuel size and ventilation conditions on fire behavior, and to establish a
correlation among them by developing mathematical models. The
burning of plastic cribs was found generally to correlate empirically as
i,hv ~5/4aoI a Lrl/2~,
= ,,o j . ono ) (8)

and the mass flow rate of air was found to correlate empirically as
rh a ~ .,L." l l 4 / A /J1/2\1/2
r,,v b,-to.'ao ) (9)
The peak rise in temperature of the upper hot gas layer was found
empirically to be

A T = 24"2~7Q,,,~oo ! (10)

It can be seen from eqn (10) that the temperature increase is a much
stronger function of burning rate, Q, than the ventilation factor,
WoH3o/2. Z o n e modeling approach was used to develop a mathematical
model in order to compare theoretically predicted fire growth behavior
with the experimental values. These theoretical results were found to
be in fair agreement with the experimental results, but the concept of a
two-layer model and the assumption of steady-state conduction through
the walls restrict the use of this model. Quintiere & McCaffrey also
reported that for cribs and room geometry considered in their analysis,
the effect of radiation feedback and oxygen deficiency on the pyrolysis
of cribs was not more than 25%. Also, the likelihood of flashover, i.e.
incident radiation flux of 2 W / c m 2 at any point on the floor, for burning
plastic cribs was twice that of w o o d e n cribs. Since the walls did not
burn, the same conclusions may not be valid for enclosure fires with
burning walls.
Experiments were conducted by Tu & Babrauskas TM to measure
parameters of significance in enclosure fires and to compare these
parameters with theoretically predicted results. The dimensions of the
room were 3.52 m x 3.40 m x 2.44 m (length × width x height) with
doorway dimensions of 0.91 m x 2.13 m (width x height). A diffusion
164 S. Jolly, K. Saito

flame burner was used to simulate preflashover conditions with peak


temperature of 573 K outside the hot plume. Readings were taken after
steady-state conditions were achieved. The walls and ceiling were made
up of type X gypsum board and the f o o r was made up of a concrete
slab. The floor, walls and ceiling were then covered with a 13 m m
cement-asbestos board lining. In general, the highest temperature was
recorded at about 0 . 1 m below the ceiling; the temperature at the
ceiling was found to be about 1 0 - 4 0 K below the maximum. The
measured temperatures were fairly uniform over the lower half of the
room, and the floor temperatures were greater than the surrounding
air. The heat loss from the walls and ceilings was found to be 3 - 6 % of
the fuel heat release rate. This low value of heat loss to the wall may be
due to the good insulating properties of the cement-asbestos board
used as the interior finish material. Heat and mass balances in the
enclosure could not be achieved because of uncertainty involved in the
measurement of the doorway velocity profile. Even in this case, the
walls did not take part in burning, so these conclusions may not be
applicable to enclosure fires with burning walls.
Ostman & Nussbaum 1~ found an empirical correlation between the
time to flashover for a full-scale test and the heat release rate measured
on small-scale test specimens for eleven different materials. The
dimensions of the small-scale test specimens were 0 . 1 m × 0 . 1 m ,
whereas the dimensions of the room were 3 . 6 m × 2 - 4 m × 2 - 4 m
(length x width × height) with doorway dimensions of 0.8 m x 2-0 m
(width x height). The relationship is given as
a/ignitionWr'p
/flash .... -- "~- b (11)
qss
where tflash is the flashover time for full-scale test (s); tignition is the
....

ignition time for small-scale tests at 25 kW/m2(s); a is a constant,


2.76 x 106 J(gm)-°5; b is a constant, - 4 6 s; qss is the heat release rate
when the small-scale test specimen is subjected to a heat flux of
50kW/mZ(j/m2), and p is the density (kg/m3). However, the ap-
plicability of this correlation for other materials and room configura-
tions was not investigated by Ostman & Nussbaum.

4.4 Application of modeling techniques

Quintiere 11° derived 14 dimensionless numbers from appropriate gov-


erning equations in a procedure similar to that used by Williams. 3~ He
stressed partial modeling in which only a few significant parameters of
interest are retained and proposed three practically useful modeling
Room flashover phenomena in scale modeling of fires 165

schemes: Froude modeling, analog modeling and pressure modeling.


Pressure modeling, which is certainly a useful technique but complex
and expensive, has already been discussed. Analog modeling is used to
simulate fire induced flow by using different density fluids. Froude
modeling is the technique most commonly used for correlating en-
closure fires of different sizes. In Froude modeling the inertia and
buoyant forces are dominant and the viscous forces are negligible.
Quintiere discussed some typical fire problems, for example: fire
plume, ceiling jet, burning rate and flame spread. In all of these, the
three modeling schemes mentioned have produced useful results.
Steckler et al. 11~ studied movement of smoke and hot gases in
multicompartment enclosures by using salt water modeling. The scaling
laws simulating fire plume behavior, by injecting salt water into fresh
water, were developed by non-dimensionalizing conservation of mass,
momentum and energy equations, subject to appropriate boundary
conditions. These scaling laws were developed on the assumption that
viscous and heat transfer effects were small. It was found that the salt
water flows in reduced-scale tests simulated gas flows in full-scale tests
quite well for four different geometric arrangements by comparing time
and layer interface height. The interface height is defined as the height
where the temperature is 10% of ceiling temperature rise above
ambient. In order to simulate actual fire behavior, it was required that
the salt water plume in the model is turbulent, and is thus governed by
buoyancy and inertia forces. This was achieved over 80% of the plume
height. However, due to initial momentum flux of salt water being
greater than its buoyant flux, the requirements for buoyancy and
turbulence could not be met near the salt water source.
Waterman 112 suggested that even when scaling criteria were met, the
relative volume occupied by instrumentation may become an important
factor in deciding the scale-factor to be used: for small absolute model
sizes, the volume occupied by instrumentation may become significant.
He used a one-eighth scale model to predict the effects of room size,
window height, fire location and fire size on flashover condition and
time to the flashover. 113 His criterion for the flashover was a total heat
flux of about 1.25 W/cm 2 at the wall opposite to the fire. Table 1 lists
the room and window sizes used in the experiments. A propane burner
was used to simulate real furniture fires. A ventilation factor of WoH3o/2
was used for the doorway to scale the volumetric air flow rate. The size
of the fire was found to be the most important variable to determine
flashover. If the size of the fire was increased beyond the minimum
required to achieve flashover, considerable reduction in flashover time
was observed. For flashover to occur, the burner located near the
166 S. Jolly, K. Saito

TABLE 1
Dimensions of Various Room and Window Sizes in the Experim-
ental Study lt3

Model Full-scale
(cm) (rn)

Room depth

t, 5 137.16 10.97
3 4 114.30 9.14
3 91.44 7.32
2 68.58 5.49
1
1 45.72 3-66
m

Top View Window height


L 34.29 1.37
M 22.86 0-91
S 11.43 0.46

Front View Room height 60.96 2-44


Room width 45.72 3-66
Window width 15.24 1.22
Sill height 15-24 0-61

middle of the long wall required more heat input than that required by
the burner located near the corners. This is because of restriction of
entrainment of convective heat current arising from fire burning in a
corner which heats up the upper walls and ceiling more efficiently than
in the case where the fire is located near the middle of a room. Even
though the convective heat transfer near the stagnation point above the
corner burner is high, as noted previously, the ceiling jet temperature is
high at all radii for a corner burner. The higher upper wall and ceiling
temperatures cause an increase in radiative heat flux. In addition there
is an increase in radiative heat flux, due to multiple reflections, when
the burner is located near a corner. For small and m e d i u m size
windows, the effect of room depth on flashover was significant only up
to twice the initial room depth. However, the effects were found to be
minimal when the room depth was increased three times.
Heskestad 114 proposed an approximate modeling technique for crib
fires in an enclosure. The scaling laws he suggested were only partly
non-dimensional. Quintiere 11° has shown that the scaling laws proposed
by Heskestad can be obtained by applying the Froude modeling
approach to the governing equations and boundary conditions pre-
sented by him. Heskestad demonstrated that the two conditions
required to obtain similar temperature profiles for the wall of the
Room flashover phenomena in scale modeling of fires 167

prototype and its model could not be satisfied simultaneously. One


condition required an increase in wall thickness for the model, whereas
the other condition required a decrease in wall thickness for the model.
The burning rates, percentage of gas species and the fire behavior were
reproduced within reasonable correlation limits. However, the n u m b e r
of experiments conducted was small.
Additional tests were performed by Croce 115 to check the validity of
the scaling laws proposed by Heskestad. H4 Croce's experiments sup-
ported Heskestad's results. The crib burning rate, gas and wall
temperatures and gas species concentrations correlated within 20%,
although the largest deviations were observed in simulating gas species
concentrations. The reproducibility of fire behavior was poor for low
ventilation values (Ho/H <-0-2). In most cases, for Ho/H <-0.05, the
combustion of cribs could not be sustained because of the non-
availability of sufficient oxygen. In all of the above cases, the walls did
not burn. In cases where the walls take part in burning, the contribu-
tion of the wall lining to the total heat flux is a function of fuel amount,
fuel location and ventilation opening. ~°° W h e n a reduced-scale model is
used, the surface area of the model cannot be scaled accordingly, so the
contribution of heat flux from the burning walls will be greater for the
reduced scale model, since 0 ~ h5/2.98 Therefore, this scaling technique
is inapplicable to model enclosure fires with burning walls.
Kanury 25 applied mass, m o m e n t u m and energy equations in order to
identify important dimensionless groups, which were then used to
correlate experimental data. These groups were: initial fire characteris-
tic, q~ = thv/kWoH3o/2; window geometry, ~p = Wo/Ho; soffit depth, ), = A
/WoHo; ceiling heat transfer, f2 = h/kcpgH~o~2; and wall burning, fl =
fl~(Tf- Tw)q/L In addition, he defined the dimensionless flashover time
l'* as h:t*/kpc in which the properties were averaged over the wall
linings and ceiling. Three of these groups, g2, ), and ~0, were kept fixed,
while the product q~(1 + f l ) was plotted against the dimensionless
flashover time r*. It was found that when flashover occurred, the
flashover time l'* decreased with an increase in q~(1 + fl). This result
was expected, since large initial fires or highly combustible walls
aggravate fire and lead to an early flashover. In cases where flashover
did not occur, large discrepancies attributed to incomplete scaling
correlations, lack of interpretation of results and variation in properties
were found.
Parker & Lee H6 conducted a series of experiments in order to predict
the behavior of fires in full-scale tests from results obtained from
one-quarter-scale tests. The temperature of the upper gas layer was
assumed to be the criterion for determining the magnitude of fire
severity in an enclosure. After applying the heat balance in the
168 S. Jolly, K. Saito

enclosure, assuming negligible radiative losses through the opening,


and neglecting radiation from the upper hot layer of gases to the room
surfaces, Parker & Lee developed an expression for the upper gas layer
temperature:
T - Ta = YO' cpa/Cpg
1 + (Cpa/Cpg)(ha/pcpa)(Au/Af)/(Vg/Af) (12)
where Y, the fraction of oxygen consumed from the incoming air, is
given by
rhv E qi(Ai/A,)
Y - -- - (13)
thv,s pCpaO'(Vg/Af)
According to Parker & Lee, eqns (12) and (13) between the prototype
and its model were preserved if the following conditions were satisfied.
For ventilation controlled fires, the viscous forces are insignificant, and
the Fr number is the governing dimensionless group for the fluid flow.
Therefore,
Vg = A o v o¢A o V ~ o = W o H 3'2 (14)
In order to preserve the term Vg/Ar, the ventilation factor, Vg, must
scale as the floor area, Af. This is achieved by allowing 141oto vary as the
square root of the scale factor and Ho to vary as the scale factor itself.
Since the burning rate is proportional to the ventilation factor, WoH3o/2,
the burning rate should also scale as the floor area. In addition, the fuel
surface area should also scale as the floor area in order to maintain the
same heat release rate per unit area. A i / A f could be preserved by
geometric scaling. The thickness of the wall material was kept the same
in order to maintain the same heat transfer from the walls of the
prototype and its model to the surroundings. Parker & Lee suggested
that if these scaling criteria were met, then the same temperature for
the upper gas layers would be achieved between the prototype and its
model. Lauan walls and gypsum board ceiling were used for the
full-scale and the one-quarter-scale tests. For the one-quarter-scale
tests, the velocities at different locations along the doorway were
indeed found to be one-half of those measured at similar locations for
the full-scale tests, thus preserving the Froude number. High similarity
was found for the temperature profiles for the full-scale and one-
quarter-scale tests. The relationship between the temperature of the hot
upper layer of gases and the oxygen depletion rate expressed by eqn
(12) was verified by experiments. However, experimental data were
limited and the range of materials considered was very small.
In another study, additional tests were performed by Lee 98 in order
to test the applicability of the above-mentioned scaling laws. The
Room flashover phenomena in scale modeling of fires 169

criterion used for flashover was based on thermal radiation mainly from
the upper part of the hot walls and gases to the lower part of the r o o m
( 2 . 2 W / c m 2 for the full-scale and 2.1 W / c m 2 for the one-quarter-scale
model). Equation (12) was slightly modified to obtain a relationship
which is more convenient to use and easy to understand. In a m a n n e r
similar to the previous study, 116 Lee assumed the r o o m to be divided
into two temperature zones, i.e. T for the upper portion of the room
and Ta for the lower portion of the room. After applying heat balance
in the room, the following expression for the temperature rise was
obtained:
(~/Af + ~ q~A~/Af
T- Ta = (15)
pgcpgVg/Af + L/Af
where Q/A, was preserved by controlling the heat input to the burner,
and Ai/Af was preserved by geometric scaling. Again, Fr n u m b e r was
taken to be the governing dimensionless group for the fluid flow, and
the term Vg/Af was preserved as discussed before. The geometric
scaling of L/At was only approximate, since the convective and
radiative heat fluxes were not modeled. The scaling techniques men-
tioned had some problems associated with them: flame height and
lateral flame spread were too high in the model; radiation and
convection were not scaled; a large portion of the total heat was lost to
the surroundings; and too little heat was released from one of the walls
due to an increase in the width of the opening required to preserve the
volumetric flow rate.
Three different kinds of materials were selected by Lee: 98 fibrous
glass, nitrile foam rubber and plywood paneling. A schematic of the

® HEIGHT

L. wIDTH .J
Fig. 11. S c h e m a t i c of the r o o m . ~2~
170 S. Jolly, K. Saito

TABLE 2
Full-scale and One-quarter-scale Dimensions TM

Full-scale Model Model Model


I II III

Length and width 3000-00 750-00 750-00 750-00


of room (mm)
Height of room (mm) 2400.00 600.00 600.00 600-00
Doorway height (mm) 2014.00 503.00 468.00 433-00
Doorway width (mm) 732.00 366-00 406.00 460-00
Scaled doorway Full-size 1.00 0-93 0.86
height (mm)

test room is shown in Fig. 11, and its dimensions are given in Table 2.
He found that most of the one-quarter-scale experiments conducted did
not reach flashover conditions while the full-scale tests did. Even in
those situations where flashover was achieved for one-quarter-scale
tests, the flashover time was longer than that in the full-scale tests.
When the height of the opening was lowered to allow the thickness of
the hot gases trapped in the upper part of the room to increase, the
number of one-quarter-scale tests leading to flashover increased. This is
because of increase in thermal radiation due to the presence of a
thicker layer of hot gases and soot particles in the upper part of the
room. Similar results were also reported by Lee and coworkers nT'ns in
similar kinds of experiments which he performed.
Dingyi n9 conducted several one-quarter-scale tests for ten full-scale
tests using the same scaling techniques applied by Lee. 9a The results he
obtained agreed with those obtained by Lee: the one-quarter-scale tests
had longer flashover time, lower heat release rate, lower temperature
of gases trapped in the upper part of the room, and lower combustion
efficiency (CO2/CO) as compared to the full-scale tests. However, for
most of the cases, the ordering of interior finish materials was
consistent with those of full-scale tests. Similar results were also
reported for reduced-scale models of mobile home fires; 12° i.e. flashover
time was longer for the reduced-scale tests.
Jolly et al. 121 gave two main reasons for the discrepancies in the
results of full-scale and one-quarter-scale tests conducted by Lee: 9s
radiation effects from soot particles and hot gases and the heat loss
from the interior surfaces to the surroundings were neglected in the
scaling attempts. Their criterion for flashover was a radiative heat flux
of 2 W / c m 2 at the floor. The major aim of the study was to focus on the
effects of various parameters on radiative heat transfer, since radiation
Room flashover phenomena in scale modeling of fires 171

was presumed to be the dominant criterion for flashover to occur in a


room. In order to identify important radiation parameters, the radiative
transfer equation was simplified by using the Prapproximation. The
resulting elliptic partial differential equation was solved using the
ELLPACK c o d e . 122A23 M e d i u m temperature profiles for full-scale and
one-quarter-scale tests were taken from the literature, 98 so that the
predicted radiative fluxes could be compared to the experimental
values. Some of the required data not found in the literature was
assumed. Figures 12 and 13 show average radiative flux to the floor
using nitrile foam rubber and fibrous glass as the interior finish
material, respectively. It can be seen from Fig. 13 that the average
radiative flux to the floor is greater for the one-quarter-scale test when
fibrous glass is used as the interior finish material. This is because
fibrous glass coated walls did not burn, and the heat lost to the
surroundings because of increased surface to volume ratio for the
one-quarter-scale test was less due to good insulating properties of
fibrous glass. Figures 14 and 15 show the effect of soot volume fraction
on the radiative heat flux to the floor. It can be concluded from these
figures that the volume fraction of soot particles is a very important
factor influencing the net radiative heat flux to the floor. Small

E -~ ¢ * F u l l - S c a l e

--- --, .-. O n e - Q u a r t e r - S c a l e


o
o

E-

o
<
X O o 0 [] O 0 O 0 0
J

>

<
E
<

I I I I
20 40 60 80 100

PERCENTAGE OF ROOM LENGTH


Fig. 12. Average radiative flux distribution at the floor along the width of the room
versus percentage of room length in the x-direction (using nitrile foam rubber as
interior finish material).lz~
172 S. J o l l y , K . S a i t o

1 .00

* * , Full-Scale
,,/ = = = One-Quarter-Scale
0
0 0.75
rT
L~

L3
Z
© 0.50
.3
<
M [] [] [] O D Q O D
.3
r,
0.25
>
7
<

<

0.00 I I I I
20 40 60 80 100

PERCENTAGE OF ROOM LENGTH


Fig. 13. A v e r a g e radiative flux distribution at the floor along the w i d t h of the r o o m
versus p e r c e n t a g e of r o o m length in the x - d i r e c t i o n (using fibrous glass as i n t e r i o r finish
material). TM

* * * Full-Scale
E

4 One-Quarter-Scale

rr.
3

b"
<
X

0 I I I I I
0.0 0.5 ~.0 ~.5 2 0 2.5 3.

VOLUME FRACTION (10 "6)


Fig. 14. A v e r a g e radiative flux at t h e floor as a function of v o l u m e fraction of soot
particles in the u p p e r h o t layer (using nitrile f o a m r u b b e r as interior finish material). TM
Room flashover phenomena in scale modeling of fires 173

1 .00

E *-*'-* Full-Scale
- --- --- One-Quarter-Scale

O
© 0 • 75

E-
l.-
< 0.50
X

i 0.25
.<

0.00 I I I I I
0.0 0 5 1.0 1.5 2.0 2.5 3.

VOLUME FRACTION (10 ~)


Fig. 15. A v e r a g e radiative flux at t h e floor as a f u n c t i o n of v o l u m e f r a c t i o n of soot
particles in t h e u p p e r h o t layer (using fibrous glass as i n t e r i o r finish m a t e r i a l ) . TM

inaccuracies in the simulation of surface temperatures, surface emis-


sivities and single scattering albedo (ratio of scattering coefficient to the
sum of scattering and absorption coefficients) resulted in radiative heat
flux variations, but not to the extent due to variations in soot volume
fraction measurements. Equation (15), developed by Lee, was not
modeled to reproduce similar temperature profiles for different sizes of
tests conducted, because his assumption that the radiation effects due to
the presence of the medium (especially the soot particles) is insig-
nificant, is not valid.

5 SUMMARY

Flashover is a critical phenomenon related to a developing fire in a


compartment. Several attempts have been made in the past to predict
flashover conditions and flashover time. The number of parameters and
the many combinations of fuel locations possible in an enclosure make
the prediction of fire behavior a very complex task. The difficulties
associated with stochastic models, field equation models and zone
models encourage the use of experimentally determined results. Under
such conditions, conducting full-scale tests would be an ideal approach.
However, full-scale tests are time consuming and expensive; these
174 s. Jolly, K. Saito

limitations make the use of reduced-scale tests desirable. Predicting


behavior of full-scale tests from the reduced-scale tests is possible only
when the different sizes of models are related to each other by
applicable scaling laws. In combustion problems, it is often impossible
to satisfy all the conditions for complete similarity. Out of these
dimensionless parameters, some have strong, others weak, influence on
the desired results. Neglecting the weak parameters may cause minor
variations in desired results, but this technique of partial modeling
greatly simplifies the problem.
Even when the partial modeling technique is applied to study
flashover phenomenon, the dimensionless groups representing sig-
nificant parameters may still be large enough to introduce complexities
which would discourage their use. However, the number of dimension-
less groups could be further reduced by focusing on the specific
situation involved, e.g. if the walls are not burning and the fire source is
a burner, then the dimensionless groups representing flame spread
phenomenon could be simplified or eliminated. The scale modeling
technique has been successful in predicting flashover phenomenon only
for certain fuels, fuel locations and interior finish materials. This range
is further reduced for cases where the enclosure walls take part in the
burning mechanism. More research is needed in order to simulate
flashover conditions over a wider range of fuel types, fuel locations,
enclosure geometry and interior finish materials. The results of success-
ful cases could be implemented into the prediction models. At the same
time, other modeling approaches could be used to enhance understand-
ing and development of scaling laws: e.g., mathematical modeling
approaches could be applied to identify important parameters in
flashover phenomenon. It is important to note that when modeling
enclosure fires, especially when walls take part in combustion,
radiation--a difficult mechanism to simulate--plays a dominant role
near the flashover conditions, and it therefore can not be neglected.
Over all, in the engineering approach, 'observation comes first, then
modeling'. This philosophy is surely valid for studies of compartment fires.
In a room fire, one broken window could change the whole history of
the fire growth process, and thus makes any theoretical modeling work
inpractical. Knowledge obtained through experiments is the essence for
modeling. However, mere experiments without the aim of modeling
can not feed the desire to predict fire behavior in compartments. It is
obvious, in a sense, that experiments and mathematical models should
be designed in a way that their results can supplement each other. As a
result, a general prediction model for the compartment fires will be
accomplished.
Room flashover phenomena in scale modeling of]ires 175

ACKNOWLED GEMENTS

The authors wish to acknowledge technical discussions with Henry


Mitler concerning c o m p a r t m e n t fire modeling and partial support from
the National Institute of Standards and Technology under Grant No.
60NANB7D0739 and the Center for Robotics and Manufacturing
Systems, University of Kentucky, under Grant No. 2-04002.

REFERENCES

1. Williams, F. A., Urban and wildland fire phenomenology. In Progress in


Energy and Combustion Sciences, Vol. 8. Pergamon Press, Oxford, 1982,
pp. 317-54.
2. Albini, F. A., Wildland fires. American Scientist, (1984) 590-7;
Alibini, F. A., A model for the wind-blown flame from a line fire.
Combustion and Flame, 43 (1981) 155-74.
3. Emmons, H. W., The needed fire science. In First Symposium
(International) on Fire Safety Science, Hemisphere Publishing Company,
Washington DC, 1985, pp. 33-53.
4. (a) de Ris, J., A scientific approach to flame radiation and meterial
flammability. In Second Symposium (International) on Fire Safety
Science; Tokyo, Japan, 1988, pp. 29-46.
(b) de Ris, J., Spread of a laminar diffusion flame. In Twelfth Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1969, pp. 241-52.
5. Thomas, P. H., P. H. Thomas Fire Research Station 1951-1986, Selected
Papers. Building Research Establishment, Garston, Watford, 1986.
6. Quintiere, J. G., State of fire research and safety. In Second Symposium
(International) on Fire Safety Science, Tokyo, Japan, 1988, pp. 15-28.
7. Yokoi, S., A study on dimensions of smoke vent in fire-resistive
construction. Report of the Building Research Institute, Japan Ministry
of Construction, Shinjuku-ku, Tokyo, March 1959, No. 29, pp. 1-10.
8. Sekine, T., Room temperature in fire of a fire-resistive room. Report of
the Building Research Institute, Japan Ministry of Construction,
Shinjuku-ku, Tokyo, March 1959, No. 29, pp. 13-24.
9. Kawagoe, K., Damage of structures in full-size fires. Report of the
Building Research Institute, Japan Ministry of Construction, Shinjuku-
ku, Tokyo, March 1959, No. 29, pp. 27-41.
10. Yokoi, S., Study on the prevention of fire-spread caused by hot upward
current. Report of the Building Research Institute, Japan Ministry of
Construction, Shinjuku-ku, Tokyo, November 1960, No. 34, pp. 1-42.
11. Conflagrations in America since 1900, National Fire Protection Associa-
tion, Boston, MA, 1961.
12. Gross, D., Loftus, J. J., Lee, T. G. & Gray, V. E., Smoke and gases
produced by burning aircraft interior materials. National Technical
Information Service, Springfield, VA, February 1969, Document #P-B-
193736.
176 S. Jolly, K. Saito
13. Seader, J. D. & Einhorn, I. N., Some physical, chemical, toxicological
and physiological aspects of smoke fires. In Sixteenth Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1977, pp. 1423-45.
14. Kawagoe, K., Japanese research program on toxicity of fire effluents with
the cooperation of the United States and Canada. In Proceedings of the
8th Joint Panel Meeting, UJNR Panel on Fire Research and Safety,
Tsukuba, Japan, 13-21 May 1985.
15. Bukowski, R., Toxic hazard evaluation of plenum cables. In Proceedings
of the 8th Joint Panel Meeting, UJNR Panel on Fire Research and Safety,
Tsukuba, Japan, 13-21 May 1985.
16. Rutkowski, J. E. & Levin, B., Acrylonitrile-butadiene-styrene copoly-
mers (ABS): Pyrolysis and combustion products and their toxicity--a
review of the literature. Fire and Materials, 10 (1986) 93-105.
17. Braun, E., Levin, B. C., Paabo, M., Gurman, J. L., Clark, H. M. &
Yoklavich, M. F., Large-Scale Compartment Fire Toxicity Study: Com-
parison with Small-Scale Toxicity Test Results, National Bureau of
Standards, Gaithersburg, MD, 1988, NBSIR 88-3764.
18. Braun, E., Levin, B. C., Paabo, M., Gurman, J., Holt, T. & Steel, J. S.,
Fire toxicity scaling. National Bureau of Standards, Gaithersburg, MD,
1987, NBSIR 87-3510.
19. Braun, E. & Levin, B. C., Nylons: A review of the literature on products
of combustion and toxicity. Fire and Materials, 11 (1987) 71-88.
20. Levin, B., EXIT-I?--A simulation model of occupant decisions and
actions in residential fires: Users guide and program description. National
Bureau of Standards, Gaithersburg, MD, 1987, NBSIR 87-3591.
21. Gurman, J. L., Baier, L. & Levin, B. C., Polystyrenes: A review of the
literature on the products of thermal decomposition and toxicity. Fire and
Materials, 11 (1987) 109-30.
22. Behnke, W. P., Predicting flash fire protection of clothing from labora-
tory tests using second degree burn to rate performance. Fire and
Materials, 8 (1984) 57-63.
23. Shalev, I. & Barker, R. L., Protective fabrics: A comparison of
laboratory methods for evaluating thermal protective performance in
convective/radiant exposures. Textile Research Journal, 54 (1984) 648-
54.
24. Kransy, J., Rockett, J. & Dingyi, H. Protecting fire fighters exposed in
room fires: Comparison of results of bench scale test for thermal
protection and conditions during room flashover. Fire Technology, 24
(1988) 5-19.
25. Kanury, A. M., Scaling correlations of flashover experiments. National
Bureau of Standards, Gaithersburg, MD, October 1983, NBS-GCR-83-
448.
26. Quintiere. J., A perspective on compartment fire growth. Combustion
Science and Technology, 39 (1984) 11-54.
27. Ndubizu, C. C., Ramaker, D. E., Tatem, P. A. & Williams, F. W., A
model of freely burning pool fires. Combustion Science and Technology,
31 (1983) 233-47.
28. Spalding, D. B., The art of partial modeling. In Ninth Symposium
Room flashover phenomena in scale modeling of fires 177
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1963, pp. 833-43.
29. Blinov, V. I. & Khudyakov, G. N., Diffusion Burning of Liquids.
Academy of Science, Moscow, 1961, pp. 93-100.
30. Yetter, R. A., Saito, K. & Emori, R. I., On the role of sensitivity
analysis in scale model development and evaluation. In First Symposium
(International) on Scale Modeling, Tokyo, Japan, July 1988, pp. 380-7.
31. Williams, F. A., Scaling mass fires. Fire Research Abstract and Reviews,
11 (1969) 1-23.
32. Emori, R. I. & Schuring, D. J., Scale Models in Engineering, Pergamon
Press, Oxford, 1977.
33. Damk6hler, G., Einfluss von Diffusion, Stromung und Waimetransport
anf die Ausbeute bei chemisch-technischen Reaktionen. Z.
Elektrochem., 42 (1936) 359-485.
34. Emmons, H. W., Some observations on pool burning. In The Use of
Models in Fire Research, ed. W. G. Berl. National Academy of Science,
National Research Council, Washington DC, NAS-NRC Publication No.
786, 1961, pp. 50-67.
35. Burgess, D. S., Strasser, A. & Grummer, J., Diffusion burning of liquid
fuels in open trays. Fire Research Abstract and Reviews, 3 (1961) 177-92.
36. Spalding, D. B., The burning rate of liquid fuels from open trays by
natural convection. Fire Research Abstract and Reviews, 3 (1962) 234-6.
37. Arai, M., Saito, K. & Altenkirch, R. A., A study of boilover in liquid
pool fires supported on water, Part I: Effects of water sublayer on pool
fires. Combustion Science and Technology, 71 (1990) 25-40.
38. Akita, K. Report on Large Open Tank Fire Experiments. ed. Committee
of the Large Oil Tank Fire Experiments, Japan Society of Safety
Engineers, Yokohama, Japan, 1981 (in Japanese).
39. Hottel, H. C., A Review of Blinov, V. I., and Khudyakov, G. N. (1957).
Fire Research Abstract and Reviews, 1 (1959) 41-4.
40. Thomas, P. H., Baldwin, R. & Heselden, A. J. M., Buoyant diffusion
flames: Some measurements of air entrainment, heat transfer and flame
merging. In Tenth Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, PA, 1965, pp. 983-96.
41. Yumoto, T., Heat radiated from oil tank fires. Japan Society of Safety
Engineering, 10 (1971) 143-52.
42. Yumoto, T., Heat transfer from flame to fuel surface in large pool fires.
Combustion and Flame, 17 (1971) 108-9.
43. Modak, A. T. & Croce, P. A., Plastic pool fires. Factory Mutual
Research, Norwood, MA, 1976, Serial No. 22361-3.
44. Alger, A. S., Corlett, R. C., Gordon, A. S. & Williams, F. A., Some
aspects of structures of turbulent pool fires. Fire Technology, 15 (1979)
143-56.
45. de Ris, J., Fire radiation--A review. In Seventeenth Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1979, pp. 1003-16.
46. Mudan, K. S., Thermal radiation hazards from hydrocarbon pool fires.
Progress in Energy and Combustion Science, 10 (1984) 59-80.
47. Shinotake, A., Koda, S. & Akita, K., An experimental study of radiative
178 S. Jolly, K. Saito
properties of pool fires of intermediate scale. Combustion Science and
Technology, 43 (1985) 85-97.
48. Brosmer, M. A. & Tien, C. L., Radiative energy blockage in large pool
fires. Combustion Science and Technology, 51 (1987) 21-37.
49. Emori, R. I. & Saito, K., A study of scaling laws in pool and crib fires.
Combustion Science and Technology, 31 (1983) 217-31.
50. Emori, R. I. & Saito, K., A unified view of fire scaling. Bulletin of
JSME, 29 (1986) 494-500.
51. Koseki, H. & Yumoto, T., Air entrainment and thermal radiation from
heptane pool fires. Fire Technology, 24 (1988) 33-50.
52. Schug, K. P., Manheimer-Timnat, Y., Yaccarino, P. & Glassman, I.,
Sooting behavior of gaseous hydrocarbon diffusion flames and the
influence of additives. Combustion Science and Technology, 22 (1980)
235-50.
53. Glassman, I. & Yaccarino, P., The temperature effect in sooting diffusion
flames. In Eighteenth Symposium (International) on Combustion, The
Combustion Institute, Pittsburg, PA, 1981, pp. 1175-83.
54. Sidebotham, G., An inverse co-flow approach to sooting laminar
diffusion flames. PhD Thesis, Department of Mechanical and Aerospace
Engineering, Princeton University, Princeton, NJ, 1988.
55. Kitano, M. & Otsuka, Y., Suppression effects of stretching flow on soot
emission from laminar diffusion flames. Combustion Science and
Technology, 42 (1985) 165-83.
56. Akita, K., Report on Large Open Tank Fire Experiment. ed. Committee
of the Large Oil Tank Fire Experiments, Japan Society of Safety
Engineers, Yokohama, Japan, 1979 (in Japanese).
57. Koseki, H., Soot and thermal radiation from hydrocarbon pool fires. In
Extended Abstracts, Joint Conference of Western States and Japanese
Section, The Combustion Institute Meeting, Honolulu, Hawaii, 22-25
November 1987, 250-2.
58. Bilger, R. W., Turbulent jet diffusion flames. Progress in Energy and
Combustion Science, 1 (1976) 87-109.
59. Bilger, R. W., Reaction rates in diffusion flames. Combustion and Flame,
311 (1977) 227-48.
60. Bilger, R. W., Reaction rates in diffusion flames. Combustion and Flame,
311, (1977) 277-84.
61. Williams, F. A., Combustion Theory, 2nd Edition. Benjamin and
Cummins, Menlo Park, CA, 1985.
62. Orloff, L., de Ris, J. & Delichatsios, M. A., General correlations of
chemical species in turbulent fires. In Twenty-First Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1986, pp. 101-9.
63. Gore, J. P. & Faeth, G. M., Structure and radiation properties of
luminous turbulent acetylene/air diffusion flames. Transactions of
ASME: Journal of Heat Transfer, 110 (1988) 173-81.
64. Saito, K., Williams, F. A. & Gordon, A. S., Structure of laminar coflow
methane-air diffusion flames. Transactions of ASME: Journal of Heat
Transfer, 1118 (1986) 640-8.
65. Mitchell, R. E., Sarofim, A. F. & Clomburg, L. A., Experimental and
Room riashover phenomena in scale modeling of fires 179
numerical investigation of confined laminar diffusion flames. Combustion
and Flame, 37 (1980) 227-34.
66. Tsuji, H. & Yamaoka, I., Structure analysis of counterflow diffusion
flames in the forward stagnation region of a porous cylinder. In
Thirteenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1980, pp. 723-31.
67. Saito, K., Gordon, A. S. & Williams, F. A., Effects of oxygen on soot
formation in methane diffusion flames. Combustion Science and
Technology, 47 (1986) 117-38.
68. Delichatsios, M. A., Air entrainment into buoyant jet flames and pool
fires. Combustion and Flame, 70 (1987) 33-46.
69. Weckman, E. J., The structure of the flow field near the base of a
medium-scale pool fire. PhD Thesis, Department of Mechanical Engi-
neering, University of Waterloo, Ontario, Canada, 1987.
70. McCaffrey, B. J., Momentum implications for buoyant diffusion flames.
Combustion and Flame, 52 (1983) 149-67.
71. Spengler, V. G. & Kern, J., Untersuchungen an Diffusionsflammen
Konzentrationsverteilung in einer Hexan-Diffusionsflamme. Brennstoff-
Chemie, Zeitschrift Fur Chemie und Technik Von Kohle, Ol und Gas, 50
(11) (1969) 321-52.
72. Alpert, R. L., Pressure modeling of transient crib fires. Combustion
Science and Technology, IS (1977) 11-20.
73. de Ris, J., Kanury, A. M. & Yuan, M. C., Pressure modeling of fires. In
Fourteenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1973, pp. 1033-41.
74. Alpert, R. L., Pressure modeling of fires controlled by radiation. In
Sixteenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1977, pp. 1489-500.
75. Lee, B. T., Laboratory scaling of the fluid mechanical aspects of large
fires. Combustion Science and Technology, 4 (1972) 233-9.
76. Countryman, C. M., Palmer, T. Y., Storey, T. G., Bush, A. F.,
Leonard, J. J. & Yundt, W. H., Project Flambeau--An investigation of
mass fire (1964-1967). US Department of Agriculture, Final Report,
1969.
77. Adams, J. S., Williams, D. W. & Treyellas-Williams, J., Air velocity,
temperature and radiant heat measurements within and around a large
free-burning fire. In Fourteenth Symposium (International) on
Combustion, The Combustion Institute, Pittsburgh, PA, 1973, pp.
1045-52.
78. Rankine, A. O., Experimental studies in thermal convection. The
Proceedings of the Physical Society, 63 (364B) (1950) 225-51.
79. Taylor, G. I., Fire under influence of natural convection. In The Use of
Models in Fire Research, ed. W. G. Berl. 1961, National Academy of
Science, National Research Council, Washington DC pp. 10-31. NAS-
NRC Publication No. 786.
80. Soma, S., A study of fire whirl generated in Hifukusho-Ato. Journal of
Geology in Japan, 84 (1975) pp. 204-17 (in Japanese).
81. Soma, S., Fire whirls generated in a city. Technology and Human Beings,
71 (1977) 71-8 (in Japanese).
180 S. Jolly, K. Saito
82. Soma, S., unpublished.
83. (a) Soma, S. & Saito, K., A study of fire whirl on mass fires using scale
models. In First Symposium (International) on Scale Modeling, Tokyo,
Japan, 1988, pp. 353-60.
(b) Soma, S. & Saito, K., Reconstruction of fire whirls using scale-
models. Combustion and Flame (to appear).
84. Dessens, J., Man made tornadoes. Nature, 193 (1962) 13-14.
85. Ebert, I4_. V., Hamburg's fire storm weather. Quart. National Fire
Protection Association, (1963) 253-60.
86. King, A. R., Characteristics of a fire induced tornado. Australian Met.
Magazine, (44) (March 1964) 1-9.
87. Luke, R. H. & McArthur, A. G., Bushfires in Australia. Australian
Government Publication Service, Canberra, 1978.
88. Emmons, H. W. & Ying, S. J., The fire whirl. In Eleventh Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1967, pp. 475-88.
89. (a) Lee, S. L. & Otto, F. W., Gross vortex activities in a simple
simulated urban fire. In Fifteenth Symposium (International) on
Combustion, The Combustion Institute, Pittsburgh, PA, 1974, pp.
157-62; (b) Lee, S. L. & Garris, C. A., Formation of multiple fire whirls.
In Twelfth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1969, pp. 265-73; (c) Lee, S. L., Axisymmetri-
cal turbulent swirling natural convection plume, Part 1 and 2.
Transactions of ASME: Journal of Applied Mechanics, 33 (September
1966) 647-61.
90. Emori, R. I. & Saito, K., Model experiment of hazardous forest fire
whirl. Fire Technology, 18 (1982) 319-27.
91. Emori, R. I., Iguchi, Y., Saito, K. & Wichman, I. S., Simplified scale
modeling of turbulent flame spread with implications to wildland fires. In
Second Symposium (International) on Fire Safety Science, Tokyo, Japan,
1988, pp. 263-73.
92. Roper, F. G., Jaggers, H. C., Franklin, D. P., Slaven, N. & Campbell,
A., Factors controlling scaling laws for buoyancy controlled combustion
of spherical gas clouds. British Gas Corporation, London Research
Station, London, 1986.
93. Markstein, G. H. & de Ris, J., Upward fire spread over textiles. In
Fourteenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1973, pp. 1085-97.
94. Williams, F. A., Mechanism of flame spread. In Sixteenth Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1976, pp. 1281-94.
95. Fernandez-Pello, A. C. & Hirano, T., Controlling mechanisms of flame
spread. Combustion Science and Technology, 32 (1983) 1-31.
96. Zukoski, E. E., Fluid dynamic aspects of room fires. In Fire Safety
Science, Proceedings of the First International Symposium, Hemishere
Publishing Co., Washington DC, 1985, pp. 1-30.
97. Portscht, R., Studies on characteristic fluctuations of flame radiation
emitted by fires. Combustion Science and Technology, 10 (1975) 73-84.
Room flashover phenomena in scale modeling of fires 181
98. Lee, B. T., Quarter scale modeling of room fire tests of interior finish
materials. National Bureau of Standards, Gaithersburg, MD, 1982,
NBSIR 81-2453.
99. Kawagoe, K. L., Fire behavior in rooms. Japan Ministry of Construction,
Building Research Institute, September 1958, Report No. 27.
100. Thomas, P. H., Behavior of fires in enclosures--some recent progress. In
Fourteenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1973, 1007-20.
101. Gross, D. & Robertson, A. F., Experimental fires in enclosures. In Tenth
Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, PA, 1965, pp. 931-42.
102. Quintiere, J. G., McCaffrey, B. J. & Den Braven, K., Experimental and
theoretical analysis of quasi-steady, smaU-scale enclosure fires. In
Seventeenth Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, PA, 1980, pp. 1125-37.
103. Babrauskas, V., Estimating room flashover potential. Fire Technology,
16 (1980) 94-104.
104. McCaffrey, B. J., Quintiere, J. G. & Harkleroad, M. F., Estimating
room temperatures and the likelihood of flashover using fire test data
correlations. Fire Technology, 17 (1981) 98-119.
105. Fang, J. B., Fire buildup in a room and the role of interior finish
materials. US Department of Commerce/National Bureau of Standards,
Gaithersburg, MD, 1975, NBS Technical Note 879.
106. Quintiere, J., The spread of fire from a compartment--a review. In
Design of Buildings for Fire Safety, ed. E. E. Smith & T. Z. Harmathy.
American Society for Testing and Materials, Washington DC, 1979,
ASTMSTP 685, pp. 139-68.
107. Quintiere, J. & McCaffrey, B., The burning of wood and plastic cribs in
an enclosure, Vol. I. National Bureau of Standards, Washington DC,
1980, NBSIR 80-2054.
108. Tu, K. M. & Babrauskas, V., The calibration of a burn room for fire tests
on furnishings. National Bureau of Standards, Washington DC, 1978,
NBS Technical Note 981.
109. Ostman, B. A. L. & Nussbaum, R. M., Correlation between small-scale
rate of heat release and full-scale room flashover for surface linings. In
Fire Safety Science, Proceedings of the Second International Symposium,
Hemishere Publishing Co., Washington DC, 1989, pp. 823-32.
110. Quintiere, J. G., Scaling applications in fire research. In First Symposium
(International) on Scale Modeling, Tokyo, Japan, July 1988, pp. 361-72.
111. Steckler, K. D., Baum, H. R. & Quintiere, J. G., Salt water modeling of
fire induced flows in multicompartment enclosures. In Twenty-First
Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, PA, 1987, pp. 143-9.
112. Waterman, T. E., Room flashover--scaling of fire conditions. Fire
Technology, 5 (1969) 52-8.
113. Waterman, T. E., Scaled room flashover. Final Technical Report,
Contract D A H C 20-70-C-0308, OCD Work Unit 2534G, Prepared for
the Office of Civil Defence, Washington DC, April 1971.
114. Heskestad, G., Modeling of enclosure fires. In Fourteenth Symposium
182 S. Jolly, K. Saito
(International) on Combustion, The Combustion Institute, Pittsburgh,
PA, 1973, pp. 1021-30.
115. Croce, P.A., Modeling of vented enclosure fires, Part I. Quasi-steady
wood-crib fire source. Factory Mutual Research, Norwood, MA, 1978,
FMRCJ.I.7AOR5.GO.
116. Parker, W. J. & Lee, B. T., A small-scale enclosure for characterizing
the fire buildup potential of a room. National Bureau of Standards,
Washington DC, 1975, NBSIR 75-710.
117. Lee, B. T. & Breese, J. N., Submarine compartment fire study-fire
performance evaluation of hull insulation. National Bureau of Standards,
Washington DC, 1979, NBSIR 78-1584.
118. Lee, B.T. & Parker, W.J., Naval shipboard fire risk criteria--berthing
compartment fire study and fire performance guidelines. National Bureau
of Standards, Washington 1976, NBSIR 76-1052.
119. Dingyi, H., Evaluation of quarter-scale compartment fire modeling for
constant and stepped heat inputs. Fire and Materials, U (1987) 179-90.
120. Klein, D. P., Reduced-scale modeling of mobile home fires: A progress
report. National Bureau of Standards, Washington DC, 1981, NBSIR
81-2333.
121. Jolly, S., Menguc, M. P., Saito, K. & Altenkirch, R. A., Scaling
flashover phenomena in compartment fires. In First Symposium
(International) on Scale Modeling, Tokyo, Japan, July 1988, pp. 373-9.
122. Menguc, M. P. & Viskanta, R., Radiative transfer in three-dimensional
rectangular enclosures containing inhomogenous, anisotropically, scatter-
ing media. Journal of Quantitative Spectroscopy and Radiative Transfer,
33 (1985) 533-549.
123. Rice, J. R. & Boisvert, R. F., Solving Elliptic Problems Using
ELLPACK. Springer Verlag, New York, 1984.

You might also like