Professional Documents
Culture Documents
in Compartment Fires 30
William D. Walton, Philip H. Thomas,
and Yoshifumi Ohmiya
Temperature rise
Fully developed fire
Ignition Decay
Growth
Time
generation or burning rate can be found in pyrolized than can be burned with the oxygen
Chap. 26. If sufficient fuel and oxygen are avail- available in the compartment. In this case, the
able, the fire will continue to grow, causing the fire is said to be ventilation controlled. If there
temperature in the compartment to rise. Fires are openings in the compartment, the unburned
with sufficient oxygen for combustion are said fuel will leave the compartment in the gas flow
to be fuel controlled or well-ventilated. and may burn outside of the compartment. It will
cause fire spread to upper floors and neighboring
Flashover Flashover is generally defined as the buildings. During the fully developed stage, the
transition from a growing fire to a fully devel- environment within the compartment has a sig-
oped fire in which all combustible items in the nificant effect on the pyrolysis rate of the burning
compartment are involved in fire. During this objects.
transition there are rapid changes in the compart-
ment environment. Flashover is not a precise Decay Stage Decay occurs as the fuel becomes
term, and several variations in definition can be consumed, and the heat release rate and temperature
found in the literature. However the onset of within a compartment decline. The fire may change
flashover should be estimated for considering from ventilation to fuel controlled during this
fire safety. Most have criteria based on the period.
temperature at which the radiation from the
hot gases in the compartment will ignite all of
the combustible contents. Gas temperatures of Compartment Fire Phenomena
300–650 C have been associated with the
onset of flashover, although temperatures of Compartment Fire Model
500–600 C are more widely used [2]. The igni-
tion of unburnt fuel in the hot fire gases, the In order to calculate or predict the temperatures
appearance of flames from openings in a and other properties generated in a compartment
compartment, or the ignition of all of the fire, a description or model of the fire phenomena
combustible contents may actually be different must be created. This model will be described in
phenomena all related to flashover. terms of physical equations that can be solved to
predict the temperature in the compartment. Such
Fully Developed Fire During this stage, the a model is, therefore, an idealization of the com-
heat release rate of the fire is the greatest. The partment fire phenomena. Consider a fire that
fire behavior in this stage is influenced by starts at some point below the ceiling. It releases
conditions of enclosure such as the size and con- energy and products of combustion at a rate that
struction materials, size and form of openings, may change with time. The hot products of
type amount and distribution of fuel in the enclo- combustion form a plume that, due to buoyancy,
sure. Frequently during this stage more fuel is rises toward the ceiling above heat source. As the
998 W.D. Walton et al.
not well understood, but for two fire types some Conservation of Mass
simplifying assumptions can lead to useful
methods for approximating the energy released The mass flow into the compartment and the flow
by the fire. out are related by
Fuel-controlled fires are defined as those in
which the pyrolysis rate and the energy release m_ g ¼ m_ a þ m_ f ð30:3Þ
rate are affected only by the burning of the fuel where ṁf is the mass burning rate of the fuel
itself and not by the room environment, analo- (kg/s).
gous to a fire burning outdoors on a calm day. The mass flow rate of hot gas out of a window
Babrauskas has provided data on free-burning or door is given by Rockett as [7]:
fires in Chap. 26. This data is most useful for
estimating burning rates of primarily horizontal 2 T1 T 1 1=2
m_ g ¼ Cd W o ρ1 2g 1 ðH o XN Þ3=2
fuels in preflashover fires, where the primary 3 Tg Tg
heating of the fuel is from the flames of the ð30:4Þ
burning item itself. Vertical fuels, such as wall
linings and fuels located in the upper hot gas where
layer, will likely be influenced by the ṁg ¼ Mass flow rate of hot gas out an opening
preflashover room environment. (kg/s)
Ventilation-controlled fires are defined as Cd ¼ Orifice constriction coefficient
those in which the energy release rate in the (typically 0.7)
room is limited by the amount of available Wo ¼ Width of opening (m)
oxygen. The mass flow rate of air or oxygen Ho ¼ Height of opening (m)
into the room through a door or window can be ρ1 ¼ Ambient air density (kg/m3)
calculated from the expressions described g ¼ Acceleration due to gravity, 9.8 m/s2
below and in Chap. 23. For most fuels [5], the XN ¼ Height of neutral plane (m)
heat released per mass of air consumed is a Tg ¼ Temperature of the hot upper gas layer (K)
constant approximately equal to 3000 KJ/kg. T1 ¼ Ambient temperature (K)
Therefore, the rate of energy release of the The mass flow rate of air into a door or
fire can be approximated from the air window is given by
inflow rate.
The amount of energy released by the fire 2 T 1 1=2
m_ g ¼ Cd W o ρ1 2g 1 ðXN Xd Þ1=2
that enters the hot upper layer is a function of 3 Tg
the fire, layer conditions, and geometry. For ðXN þ Xd =2Þ
most fires, approximately 35 % of the energy ð30:5Þ
released by the fire leaves the fire plume as
radiation [6]. (A discussion of flame radiation where
can be found in Chap. 23.) In a compartment Xd ¼ Height of the interface (m).
fire, a fraction of the radiated energy reaches the The expressions for mass flow in and mass
upper layer. The majority of the remaining flow out cannot be solved directly for Tg since
energy released by the fire is convected into the height to the neutral plane and interface are
the upper layer by the plume. As the plume unknown. The complete solution of these
rises, it entrains air from the lower layer, thus equations requires expressions for plume entrain-
reducing its temperature and increasing the ment and additional energy equations and is nor-
mass flow rate. For a first approximation, it can mally carried out only in computer fire models. If
be assumed that all of the energy generated the mass burning rate of the fuel is small com-
by the fire is transported to the upper layer. For pared with the mass flow rate of air into the
a complete discussion of fire plumes see compartment, the mass flow out of the opening
Chap. 13. may be approximated as equal to the mass inflow
30 Estimating Temperatures in Compartment Fires 1001
rate. Flows out of vents in the ceiling are The Conseil International du Bâtiment (CIB)
discussed in Chap. 66. experiments upon which Law [11] has based her
For preflashover fires in compartments with method shows a dependence on AT. It seems
typical doors or windows, the neutral plane and possible that the wide use of Equation 30.7 is a
interface can be approximated at the midlevel of result of a concentration of experimental fires in
the opening. This approximation can only be rooms of a limited range of
made after the initial smoke filling of the com-
partment is complete, and flow in and out of the AT
pffiffiffiffiffiffi
opening is established. Ao H o
For fires nearing flashover and postflashover
where
fires, the interface between the upper and lower
AT ¼ Total area of the compartment enclosing
layers is located near the floor, and the flow
surfaces (m2)
reaches a maximum for a given upper gas temper-
Traditionally, energy balances were often
ature. Rockett [7] has shown the temperature
stated in terms of the energy produced by the
dependence on the flow becomes small above
burning fuel and, thereby, led to an effective
150 C and the flow into the compartment can be
pffiffiffiffiffiffi heat of combustion of the fuel. However, this
approximated as a constant multiplied by Ao H o . practice in principle leads to the same result—
Rockett calculated values for this constant of the energy produced is related to the air flow
0.40–0.61 kg/s · m5/2, depending on the dis- for ventilation-controlled fires. Kawagoe [9]
charge coefficient of the opening. Thomas and and Magnusson and Thelandersson [12] used
Heselden estimate the value of this constant at 10.75 MJ/kg for the effective heat of combustion
0.5 kg/s · m5/2, which is the value most com- of wood in the flaming phase for fully developed
monly found in the literature [8]. The resulting compartment fires. With 16.4 MJ/kg for the heat
approximation is then of combustion of wood volatiles, this setup
pffiffiffiffiffiffi corresponds to a combustion efficiency of
m_ a ¼ 0:5Ao Ho ð30:6Þ 10.75/16.4, which is virtually identical to the
0.65 used in several computer models.
where
By far most data are based on experiments in
Ao ¼ Area of opening (m2)
which the fuel was cellulosic, and much of the
Ho ¼ Height of opening (m)
pffiffiffiffiffiffi experimental data are based on wood in the form
The term Ao H o is commonly known as the of cribs. For the post-flashover burning of a dif-
ventilation factor. The first use of this type of ferent fuel with a different chemistry, the burning
opening flow analysis for evaluating rate expressions may still be used, as long as the
postflashover fire test data is attributed to fuel is a hydrocarbon producing approximately
Kawagoe [9]. From early work analyzing such 3000 kJ for each kg of air consumed in the
data, the empirical observation was made that combustion process. Because different fuels
wood fires in rooms with small windows appeared react differently to the thermal environment and
to burn at an approximate stoichiometric rate. will pyrolyze at different rates according to the
Although flames emerging from the windows energy requirements to produce volatiles, one
implied that some fuel was burning outside, can only estimate temperatures by evaluating
calculations often suggested that enough air was the differences or obtain maximum temperatures
entering the fire for stoichiometric burning. by using stoichiometry. Fuels more volatile than
Empirical observations on wood fires [9] led to wood will probably produce lower temperatures
pffiffiffiffiffiffi inside a compartment, even if the excess fuel
m_ f ¼ 0:09Ao H o ð30:7Þ
produces a greater hazard outside the compart-
There is now a body of data [10] that modifies ment. The assumptions that the energy is related
this simple proportionality between ṁf and to the air flow and that the fuel is in stoichiomet-
pffiffiffiffiffiffi ric proportion will give an upper estimate of
Ao H o .
1002 W.D. Walton et al.
Table 30.1 Heat balance measured in experimental fires in a compartment of 29 m2 floor area with a fire load
of wood cribs
2.4 m3/s of air (5000 cfm). Perform the calculation Quintiere, and Harkleroad for naturally
for t>tp. The fire size is given as 1000 kW; ambi- ventilated compartments. Beyler offers an
ent air conditions at 300 K. Using Equation 30.17, improved correlation for compartments where
!0:72
the forced-ventilation flow rate is known
ΔT g Q_ hk AT 0:36 [17, 18]. This method begins by applying the
¼ 0:63
T1 m_ g c p T 1 m_ g c p conservation of energy in the upper layer of a
compartment. Combining Equations 30.8 and
where 30.9 yields
Q_ ¼ 1000 kW
T1 ¼ 300 K Q_ ¼ m_ g c p T g T 1 þ hk AT T g T 1
cp ¼ 1.0 kJ/kgK ð30:18Þ
AT ¼ 4 (5 4) + 2 (5 5) ¼ 105 m2
ṁg ¼ (2.4 m3/s) (1.18 kg/m3) ¼ 2.8 kg/s where:
Calculate hk for t > tp. For 0.025-m-thick Q_ ¼ Energy (heat) release rate of the fire (kW)
concrete, ṁg ¼ Gas flow rate out the opening (kg/s)
δ ¼ 0.025 m cp ¼ Specific heat of gas (kJ/kgK)
ρ ¼ 2000 kg/m3 Tg ¼ Temperature of the upper gas layer (K)
k ¼ 1.4 103 kW/mK T1 ¼ Ambient temperature (K)
cp ¼ 0.88 kJ/kgK hk ¼ Effective heat transfer coefficient
(kW/m2K)
ρc
δ2
AT ¼ Total area of the compartment enclosing
tp ¼
k 2 surfaces (m2)
Rearranging Equation 30.18 yields
ð2; 000Þ ð0:88Þ 0:025 2
¼
1:4 103 2
Q_
¼ 196s for t > t p ΔT g ¼ ð30:19aÞ
m_ g c p þ hk AT
k
hk ¼ or
δ
1:4 103 ΔT g m_ g c p 1
¼ ¼ ð30:19bÞ
0:025 Q_ 1 þ ðhk AT Þ=m_ g c p
¼ 0:056 kW=m2 K
where ΔT g ¼ T g T 1 .
0:72
ΔT g 1, 000 A nondimensional temperature rise is
¼ ð0:63Þ
T1 ð2:8Þ ð1Þ ð300Þ defined as
ð0:056Þ ð105Þ 0:36 ΔT g m_ g c p
ΔT* ð30:20Þ
ð2:8Þ ð1Þ
Q_
rffiffiffiffiffiffiffi !
kρc k δ ¼ 0.25 m
hk ¼ 0:4max ; ð30:22Þ ρ ¼ 2000 kg/m3
t δ
k ¼ 1.4 103 kW/mK
where c ¼ 0.88 kJ/kgK
k ¼ Thermal conductivity of the compartment
k 1:4 103
surface (kW/mK) hk ¼ 0:4 ¼ 0:4
ρ ¼ Density of the compartment surface (kg/m3) δ 0:25
c ¼ Specific heat of the compartment surface ¼ 0:0224 kW=m2 K
material (kJ/kgK)
δ ¼ Thickness of the compartment surface (m) (b) Calculate the compartment temperature
t ¼ Exposure time (s) using Equation 30.19a.
The expression switches from transient to 1000
steady state at a thermal penetration time of T g 300 ¼
ð2:8Þ ð1:0Þ þ ð0:224Þ ð105Þ
tp ¼ (ρc/k)δ2 rather than tp ¼ (ρc/k)(δ/2)2 used
by McCaffrey et al. and Foote et al. For the data T g ¼ 494 K
set Beyler and Deal evaluated, the standard error
for their method was 29 K as compared to 51 K Method of Peatross and Beyler
for the method of Foote et al., even though the
equation uses only one fitting constant. The correlations used in the McCaffrey,
Beyler and Deal demonstrated that this Quintiere, and Harkleroad method and the Beyler
method works for ventilation to the lower part and Deal method are based on the assumption of
of the compartment (with or without a plenum) normal insulating wall materials. For highly con-
as well as for ventilation to the upper part of the ductive walls such as steel, Peatross and Beyler
compartment. The Beyler and Deal method was suggest the use of an alternative heat transfer
based on data up to 2000 s into fire tests. At coefficient [19]. Using a lumped mass analysis
longer times, the heat loss model breaks down. for heat transfer through the wall that is appro-
priate for a highly conductive wall yields
Example of Beyler and Deal Method Estimate
the temperature in a 5 5 m floor area and 00 dT w
4-m-high compartment with 0.025-m (1-in.) mw c ¼ hg T g T w h1 T w ð30:23Þ
dt
thick concrete walls. The forced-ventilation rate
is 2.4 m3/s of air (5000 cfm). Perform the calcu- where
00
lation for t > tp. The fire size is given as ṁw ¼ Mass per unit area of the wall (kg/m2)
1000 kW; ambient air conditions at 300 K. c ¼ Specific heat of the wall (kJ/kgK)
Using Equation 30.19a, Tw ¼ Wall temperature (K)
t ¼ Time (s)
Q_ hg ¼ Heat transfer coefficient on the hot side of
Tg T1 ¼
m_ g c p þ hk AT the wall (kW/m2K)
Tg ¼ Upper layer temperature (K)
where
H1 ¼ Heat transfer coefficient on the ambient
Q_ ¼ 1000 kW side of the wall (kW/m2K)
ṁg ¼ (2.4 m3/s) (1.18 kg/m3) ¼ 2.8 kg/s Solving for the wall temperature with the ini-
cp ¼ 1.0 kJ/kgK tial condition of the wall at ambient temperature
T1 ¼ 300 K yields
AT ¼ 4(5 4) + 2(5 5) ¼ 105 m2
hg T g hg þ h1
Tw ¼ 1 exp 00 t ð30:24Þ
(a) Calculate hk for t > tp. For 0.25-m-thick hg þ h1 mw c
concrete,
1008 W.D. Walton et al.
00
The heat transfer through the wall, q_ , may be The hk calculated with this method can be used
expressed in terms of the heat transfer to the hot directly in the Beyler and Deal method. It must
side of the wall or in terms of an overall effective be multiplied by 2.5 for use in the McCaffrey,
heat transfer coefficient, hk. Quintiere, and Harkleroad method to account for
the 0.4 fitting constant in the hk in the Beyler and
00 Deal method.
q_ ¼ hg T g T w ¼ hk T g T 1 ð30:25Þ
50
hk ¼ 30 18 1 exp t
ρδc
50
¼ 30 18 1 exp 200
ð7833Þ ð0:00635Þ ð0:465Þ
¼ 23:7 W=m2 K
Beyler applied a nonsteady energy balance to the which include the fitting coefficient. Beyler
closed compartment expressed by the differential used data with a maximum temperature rise of
equation [17] 150 C to develop this correlation.
dT
mc p ¼ Q_ hk AT ΔT g ð30:28Þ Example of Beyler Method Estimate the
dt temperature in a 5 5 m floor area and 4-m-high
where “closed” compartment having 0.025-m (1-in.)
Q_ ¼ Energy (heat) release rate of the fire (kW) thick concrete walls. Perform the calculation for t
m ¼ Mass of the gas in the compartment (kg) ¼ 120 s. The fire size is given as 100 kW; ambient
cp ¼ Specific heat of gas (kJ/kgK) air conditions at 300 K. Using Equation 30.29,
ΔTg ¼ Tg – T1 2K 2 pffi pffi
cient leaks to prevent pressure buildup, but the 2 0:4 1:4 103 ð2; 000Þ ð0:88Þ ð130Þ
leakage is ignored. The mass of the fuel is ¼
ð118Þ ð1:0Þ
ignored, and the initial temperature is assumed
¼ 1:3834
to be ambient temperature. For constant heat
release rate, the solution to Equation 30.28 is where
2K 2 pffi pffi
m ¼ (100 m3) (1.18 kg/m3) ¼ 118 kg
ΔT g ¼ 2 K 1 t 1 þ eK1 t ð30:29Þ cp ¼ 1.0 kJ/kgK
K1
ρ ¼ 2000 kg/m3
where k ¼ 1.4 103 kW/mK
pffiffiffiffiffiffiffi c ¼ 0.88 kJ/kgK
2ð0:4 kρcÞAT AT ¼ 130 m2
K1 ¼ ð30:30Þ
mc p (b) Calculate K2 using Equation 30.31.
Q_ Q_ 100
K2 ¼ ð30:31Þ K2 ¼ ¼ ¼ 0:84746
mc p mc p ð118Þ ð1:0Þ
where where
k ¼ Thermal conductivity of the compartment m ¼ (100 m3) (1.18 kg/m3) ¼ 118 kg
surface (kW/mK) cp ¼ 1.0 kJ/kgK
c ¼ Specific heat of the compartment surface (c) Calculate the compartment temperature
material (kJ/kgK) using Equation 30.29.
0.4
Fuel rich
(1 – θ1) = 0.05 (lnφ) 1.67
0.2
0
–1.6 –1.2 –0.8 –0.4 0 0.4 0.8 1.2 1.6
lnφ
30 Estimating Temperatures in Compartment Fires 1011
(1 – θ1)
0.1
0.01
0.1 1.0 10
(–lnη)
where ð30:45Þ
pffiffiffiffiffiffi
Ao H o 0:5Δh p and is shown in Fig. 30.6.
η¼
ð30:44Þ
Af rσ T 4 T 4
g b
Wall Transient Losses, θ3 For the transient
where case, Equation 30.45 predicts the asymptotic
Δhp ¼ Heat of vaporization of liquid (kJ/kg) temperature value. An additional time-dependent
Af ¼ Pool area (m2) factor, however, is needed (Fig. 30.7).
1012 W.D. Walton et al.
0.
0.
00
0.5 01 L
(m2·0 C/W)
5
0.
0 K
02
0. .05
0.4 1
(1 – θ2)
0.
0.
2
1. 5
0.3 2. 0
0
0.2
0.1
0
0.001 0.01 0.1
A h (m1/2)
Aw
02
0.
05
0.3
0.2
0.1
0
0.001 0.01 0.1
A h (m1/2)
Aw
(1 – θ4) = 0.205Ho−0.3
0.1
0.1 1.0 10
Window height, Ho (m)
Fig. 30.9 Effect of bp, the 0.5
maximum combustion
efficiency
(1 – θ5)
0.10
0.05
0.1 1.0
(1 – lnbp )
compartments
600
0
30 60
At – Ao
(m–1/2)
A o Ho
0.05
Ao Ho
mf
0.00
0 30 60
A t – Ao
(m–1/2)
Ao H o
30 Estimating Temperatures in Compartment Fires 1015
This equation represents an upper limit of fire value of ṁf depends on L and the type of fuel.
temperature rise for a given Ω. However, if the For example, domestic furniture has a free-
fire load is low, this value may not be obtained. burning fire duration of about 20 min, giving
The importance of the effect of fire load also τ ¼ 1200 s and m_ f ¼ L=1200.
depends on Ao and AT, and can be expressed as The temperatures discussed above are
averages measured during the fully developed
T g ¼ T gðmaxÞ 1 e0:05Ψ ð CÞ ð30:51Þ period of the fire. It is assumed that all fires are
ventilation controlled, with the simple relation-
where
ship for rate of burning given by Equation 30.53,
Tg ¼ Average temperature in the compartment
which is near stoichiometric burning, and it is
( C)
assumed that combustion of 1 kg of wood
where
releases 18.8 MJ in total.
L
Ψ¼
½Ao ðAT Ao Þ 0:5 Method of Delichatsios et al.
Delichatsios’ proposed a method assuming
where L is the fire load (wood) in kg. that [23];
The effect of the fire on the structure depends 1. determination of uniform maximum gas tem-
not only on the value of Tg but also on the perature in the compartment for adiabatic
duration of heating. The effective fire duration, conditions
τ, in seconds, is given by 2. determination of an average heat flux to the
compartment boundary during the develop-
L
τ¼ ð30:52Þ ment of the fire
m_ f 3. a transient thermal model for the response
where ṁf is the rate of burning measured in kg/s. of the compartment boundary to account
Equation 30.7 implies that the smaller the for heat losses to the boundary of the
pffiffiffiffiffiffi compartment.
value of Ao H o the lower the rate of burning
First, gas temperature in the compartment is
and the longer the duration. Assuming a com-
defined from the following equation at quasi-
plete burnout, therefore, the effect on the struc-
steady conditions:
ture tends to be more severe for large values of Ω
pffiffiffiffiffiffi
The mass flow rate of hot gas out of the opening instances, the energy release must be less than
can be given by the equation below. stoichiometric. The method does not take into
pffiffiffiffiffiffi account that the actual mass loss rate may be
m_ g ¼ 0:5Ao Ho ð30:56Þ greater than stoichiometric, with the excess fuel
where burning outside the compartment. A computer
Ho ¼ Height of opening (m) program, SFIRE (versions 1 through 3), is avail-
Substituting Equations 30.55 and 30.56 into able to perform this method. The results from
Equation 30.54, the following energy balance the computer program have been compared with
equation is obtained for the case of ventilation a large number of full-scale fire experiments,
controlled fires. both in the fuel- and ventilation-controlled
pffiffiffiffiffiffi pffiffiffiffiffiffi
regimes, with good agreement between theory
1, 500Ao Ho ¼ 0:5Ao H o C p T *g T 1 and experiment. It should be added, however,
that most of the experiments involved wood crib
þ σAo T *4
g T 4
1 fires, which inherently burn slower and
produce less excess fuel load than furnishings
ð30:57Þ
and other combustibles found in practical
Equation 30.57 can be solved by using simple fire loads. In the Swedish method, the fire
numerical inversion to obtain the adiabatic tem- load is expressed in relation to AT as Q_ ¼
perature. As for Equation 30.57, the adiabatic 18:8 L=AT MJ=m2 .
pffiffiffiffiffiffi
temperature depends only on Ao H o . The design curves approved by the Swedish
From the dimensional analysis, authorities were computed on the basis of
0 1 systemized ventilation-controlled heat-release
pffi
Tg T1 t Q _c
¼ function@pffiffiffiffiffiffiffiffiffiffiffiffiffi
A: curves taken from Magnusson and Thelandersson
T *g T 1 ðkρcÞw AT T * T 1 [12]. Figure 30.12 shows some typical curves.
g
The curves are calculated for wall, floor, and
ð30:58Þ ceiling materials with “normal” thermal
where properties from an energy balance which assumes
AT ¼ Total area of the compartment enclosing a uniform temperature in the compartment.
surfaces (m2)
Compared to the experimental data, the model Japanese Method
for gas temperature in the enclosure is proposed
below for the growing period of fire before burn- The McCaffery’s method (Equation 30.12)
out occurs, was originally derived for fuel-controlled fires
0 11=2 [24]. However, the formula was extended to
pffi
Tg T1 t Q_c calculate the temperature of ventilation controlled
¼ 0:5@pffiffiffiffiffiffiffiffiffiffiffiffiffi
A fires as well [25]. The fire temperature in the com-
T *g T 1 ðkρcÞw AT T * T 1
g partment and fire duration are calculated by
ð30:59Þ !2=3
Q
pffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T f ¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi t1=6 þ T 1
AT kρc Ao H o
Swedish Method ð30:60Þ
1200
Ao Ho /At = 0.02 m–1/2 Ao Ho /At = 0.04 m–1/2
1000
600
50 100 150 200 100 200 300 400
400 37.5 75
25 50
200 12.5 25
Tg (°C)
1 2 3 4 5 6 1 2 3 4 5 6
Time (hr)
1200
Ao Ho /At = 0.08 m–1/2 Ao Ho /At = 0.12 m–1/2
1000
Qt = 1000 MJ/m2 Qt = 1500 MJ/m2
800
600
200 400 600 800 300 600 900 1200
400
225
150 150
200 100
50 75
1 2 3 4 5 6 1 2 3 4 5 6
Time (hr)
pffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 30.12 Examples of gas temperature-time curves of area, At, and the opening factor Ao H o =At . Fire compart-
postflashover compartment fires for different values of the ment, type A—from authorized Swedish standard
fire load density Q_ t MJ per unit of total internal surface specifications [12]
Q ¼ Heat release rate by combustion (MW) The heat release rate is calculated by the
AT ¼ Internal surface area of compartment burning type index (fraction of ventilation factor
enclosure (m2) to surface area of fuel),
pffiffiffiffiffiffiffi
kρc ¼ Thermal inertia of compartment enclo- pffiffiffiffiffiffi
Ao H o
sure (kW.s1/2/m2.K) χ¼ : ð30:62Þ
Afuel
Ao ¼ Area of window opening (m2)
Ho ¼ Height of window opening (m) The fuel surface area was assumed to follow the
T1 ¼ Initial and ambient temperatures ( C) following formula [26].
FL ¼ Fire load density (MJ/m2) 1=3
Ar ¼ Floor area of the room (m2) Afuel ¼ 0:26FL Aroom ð30:63Þ
tD ¼ Fire duration (min.) Using burning type index, the heat release rate is
calculated by
8
>
> 1:6χ ðχ 0:081Þ
<
Q ¼ Afuel 0:13 ð0:081 < χ 0:1Þ ð30:64Þ
>
>
:
2:5χexpð11χ Þ þ 0:048 ð0:1 < χ Þ
1018 W.D. Walton et al.
1=3
Afuel ¼ 0:26 FL Aroom ¼ 0:26 7201=3 ð3 3Þ ¼ 21:0 m2
pffiffiffiffiffiffi pffiffiffiffiffiffiffi
Ao Ho ¼ 0:6 1:8 1:8 ¼ 1:45 m5=2
The burning type index is For example, the fire temperature at fire
pffiffiffiffiffiffi duration is
Ao H o 1:45
χ¼ ¼ ¼ 0:069
Afuel 21:0 T f ¼ 522 46:61=6 þ 20 ¼ 1, 009 C
As χ < 0.081, fire is ventilation-controlled. The
heat release rate and fire duration are
Predicting Flashover
Q ¼ Afuel 1:6χ ¼ 21:0 1:6 0:069 ¼ 2:32MW
1 FL Ar 1 720 3 3 One use of predicted compartment fire
tD ¼ ¼ ¼ 46:6 min:
60 Q 60 2:32 temperatures is estimating the likelihood of
flashover. The methods used are similar to those
Combining the results, the fire temperature in the
used in the prediction of temperature. In one case,
compartment can be calculated by the following
that of McCaffrey et al., the method is simply an
equation as shown in Fig. 30.13.
extension of the temperature calculation.
!2=3
Q
pffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T f ¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi t1=6 þ T 0
AT kρc Ao H o Method of Babrauskas
2=3
2:32
¼ 1, 280 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffi t1=6 þ 20 Babrauskas uses the energy balance for the upper
31:32 1:75 1:45
layer given in Equation 30.8, where the gas flow
¼ 522t1=6 þ 20 rate out of the opening is approximated by [27]
800
600
400
200
0
0 10 20 30 40 50
Time(min)
30 Estimating Temperatures in Compartment Fires 1019
pffiffiffiffiffiffi
m_ g 0:5Ao Ho ð30:65Þ Q_ ¼ 0:5 Q_ stoich
The primary energy loss is assumed to be which, substituting into Equation 30.68 yields
radiation to 40 % of the wall area, which is at pffiffiffiffiffiffi
Q_ ¼ 750Ao Ho ð30:69Þ
approximately ambient temperature:
The 33 test fires used had energy release rates
qloss ¼ εσ T 4g T 41 ð0:40AT Þ ð30:66Þ from 11 to 3840 kW, with fuels primarily of wood
pffiffiffiffiffiffi
and polyurethane. Ventilation factors Ao H o
where ranged from 0.03 to 7.51 m5/2, and surface area
ε ¼ Emissivity of the hot gas to ventilation factor ratios
σ ¼ Stefan-Boltzmann constant 5.67 1011
kW/m2K4 AT
pffiffiffiffiffiffi
Combining Equations 30.8, 30.65 and 30.66, Ao H o
using a gas temperature for flashover of 873 K, a ranged from 9 to 65 m1/2.
specific heat of air of 1.0 kJ/kgK, an emissivity
of 0.5, and assuming the correlation between Example of Babrauskas’s Method Calculate
compartment wall and opening area of the heat release rate necessary to cause flashover,
AT using the method of Babrauskas. Assume the
pffiffiffiffiffiffi 50 same room as in the McCaffrey et al. method
Ao H o
example for predicting compartment fire
yields a minimum Q_ required for flashover, temperatures. From Equation 30.69
pffiffiffiffiffiffi
pffiffiffiffiffiffi Q_ ¼ 750Ao Ho
Q_ ¼ 600Ao Ho ð30:67Þ
where
The airflow into the compartment has been Ao ¼ 1.08 m2
approximated as Ho ¼ 1.8 m
pffiffiffiffiffiffi Q_ ¼ (750)(1.08)(1.8)1/2 ¼ 1087 kW
0:5Ao Ho
_
Q¼ g c p ρ1 T 1
2
h k AT Ao H o
480
From this derivation, it is shown that the min-
ð30:70Þ
imum Q_ required for flashover equals 0:4 Q_ stoich .
Comparing these results with fire tests, Selecting an upper gas temperature of 522 C
Babrauskas found that the data fall within a and ambient temperature of 295 K or ΔTg ¼
range of Q_ ¼ 0:3 Q_ stoich to Q_ ¼ 0:7 Q_ stoich . 500 C for flashover, and substituting values for
A best fit of the data suggests the gravitational constant (g ¼ 9.8 m/s2), the
1020 W.D. Walton et al.
specific heat of air (cp ¼ 1.0 kJ/kgK), and the Method of Thomas
density of air (ρ1 ¼ 1.18 kg/m3), and rounding
607.8–610 yields Thomas uses the energy balance for the upper
pffiffiffiffiffiffi
1=2 layer shown in Equation 30.8, where the gas flow
Q_ ¼ 610 hk AT Ao H o ð30:71Þ rate out of the opening is approximated by [2]
pffiffiffiffiffiffi
where m_ g 0:5Ao Ho ð30:73Þ
hk ¼ Effective heat transfer coefficient
Thomas develops an expression for q_ loss which
(kW/m2K)
assumes the area for the source of radiation for
AT ¼ Total area of the compartment surfaces (m2)
roughly cubical compartments is AT/6:
Ao ¼ Area of opening (m2)
Ho ¼ Height of opening (m) AT
A
T
Using Equation 30.13 yields a slightly differ- q_ loss hc T g T w þ εσ 2T 4g T 4floor
2 6
ent value, 623.6 rounded to 620, of the leading ð30:74Þ
coefficient because of the difference in the value
used for the specific heat of air: where
pffiffiffiffiffiffi
1=2 AT ¼ Total area of the compartment-enclosing
Q_ ¼ 620 hk AT Ao H o ð30:72Þ surfaces (m2)
hc ¼ Convective heat transfer coefficient
The use of either 610 or 620 is acceptable (kW/m2K)
within the accuracy of the expression. Tw ¼ Temperature of the upper walls (K)
Tfloor ¼ Temperature of the floor (K)
Example of McCaffrey et al.’s Method Estimate From experimental data, Thomas developed
the energy release rate required for flashover of a an average for q_ loss of 7.8 AT. Using an upper
compartment. Assume the same room as in the layer temperature of 577 C or a ΔTg of 600 C
McCaffrey et al. method example for predicting for flashover criterion and cp ¼ 1.26 kJ/kgK
compartment fire temperatures. Assuming yields an expression for the minimum rate of
ΔTg ¼ 500 C as a condition for flashover, and energy release for flashover:
air properties at 295 K, use Equation 30.71 and pffiffiffiffiffiffi
assume the compartment has heated for a period Q_ ¼ 7:8AT þ 378Ao Ho ð30:75Þ
of time that exceeds the thermal penetration time.
pffiffiffiffiffiffi
1=2
Q_ ¼ 610 hk AT Ao Ho Comparison of Methods for Predicting
Flashover
where
Babrauskas has compared the effect of room wall
k 0:48 103 area on the energy release required for flashover,
hk ¼ ¼ ¼ 0:03 kW=m2 K
δ 0:016 using the above methods [28]. The results of his
comparisons, along with some experimental data
for rooms with gypsum board walls, are shown in
AT ¼ 45.72 m2
Fig. 30.14. The graph shows the energy required
Ao ¼ 1.08 m2
for flashover as a function of compartment wall
Ho ¼ 1.8 m
area, both normalized by the ventilation factor
Therefore, pffiffiffiffiffiffi
Ao Ho . The method of Babrauskas used in this
pffiffiffiffiffiffiffi1=2 graph is based on Equation 30.32 with T1 ¼ 25
Q_ ¼ 610 ð0:03Þ ð45:72Þ ð1:08Þ 1:8
C and Tg ¼ 600 C. Babrauskas observes that
¼ 860 kW over the range of compartment sizes of most
30 Estimating Temperatures in Compartment Fires 1021
Q /Ao Ho (kW·m–5/2)
1200 0.8
1000
0.6
800
600 0.4
400
0.2
200
0
0 10 20 30 40 50 60 70
Aw
(m–1/2)
Ao H o
interest, all of the methods produce similar k thermal conductivity of the wall
results. The method of McCaffrey et al. diverts (kW/mK)
from the others for small room sizes. Babrauskas L fire load, wood (kg)
notes that all of the methods are a conservative m mass of the gas in the compartment
representation of the data. (kg/s)
ṁa mass flow rate of air into an opening
(kg/s)
Nomenclature ṁg gas flow rate out the opening (kg/s)
ṁf mass burning rate of fuel (kg/s)
Aceiling area of compartment ceiling (m2) ṁf,st stoichiometric mass burning rate of
Af pool fire area (m2) fuel (kg/s)
Afloor area of compartment floor (m2) m_ 00w mass per unit area of the wall (kg/m2)
Ao area of openings (m2) q_ loss net radiative and convective
AT total area of the compartment heat transfer from the upper gas layer
enclosing surfaces (m2) (kW)
Awalls area of compartment walls (m2) _
Q energy (heat) release rate of the fire
bp maximum combustion efficiency (kW)
c specific heat of the wall (kJ/kgK) _
Q stoich
stoichiometric heat release rate (kW)
Cd orifice constriction coefficient
t time (s)
cp specific heat of gas (kJ/kgK)
tp thermal penetration time (s)
D compartment depth (m)
Tb liquid boiling point (K)
g acceleration due to gravity, 9.8 m/s2
Tfloor temperature of the floor (K)
hc convective heat transfer coefficient
Tg temperature of the upper gas layer
Δhc effective heat of combustion of the
(K)
fuel (kJ/kg)
Tp thermal penetration time (s)
hg heat transfer coefficient on the hot side
Tw wall temperature (K)
of the wall (kW/m2K)
T1 ambient temperature (K)
hk effective heat transfer coefficient
W compartment width (m)
(kW/m2K)
Wo width of opening (m)
h1 heat transfer coefficient on the ambient
Xd height of the interface (m)
side of the wall (kW/m2K)
XN height of neutral plane (m)
Ho height of opening (m)
1022 W.D. Walton et al.
22. V. Babrauskas, “COMPF2—A Program for Calculat- 28. V. Babrauskas, “Upholstered Furniture Room
ing Post-Flashover Fire Temperatures,” NBS TN 991, Fires—Measurements, Comparison with Furniture
National Bureau of Standards, Washington, DC (1979). Calorimeter Data, and Flashover Predictions,” Jour-
23. M. Delichatsios, Y. P. Lee, P. Tofilo (2009) A new nal of Fire Science, 2, pp. 5–19 (1984).
correlation for gas temperature inside a burning enclo-
sure. Fire Safety Journal 44(8):1003–1009
24. Ministry of construction of Japan, Notification 1430, William D. Walton is retired from the Building and Fire
2000 Research Laboratory, National Institute of Standards and
25. K. Matsuyama, T. Fujita, H. Kaneko, Y. Ohmiya, Technology.
T. Tanaka, T. Wakamatsu, “A Simple Predictive
Method for Room Fire Behavior”, Fire Science and Philip H. Thomas was with the Fire Research Station,
Technology, Vol. 18 (1998) No. 1 Borehamwood, England.
26. K. Aburano, H. Yamanaka, Y. Ohmiya, K. Suzuki,
T. Tanaka, T. Wakamatsu, “Survey and Analysis on Yoshifumi Ohmiya is a professor in the department of
Surface Area of Fire Load”, Fire Science and Tech- architecture, graduate school of science and technology,
nology, Vol. 19 (1999) No. 1 Tokyo University of Science.
27. V. Babrauskas, “Estimating Room Flashover
Potential,” Fire Technology, 16, 2, pp. 94–104 (1980).