COMBUSTION A N D F L A M E 57:33-39 (1984) 33

Experimental Evaluation of Fire-Induced Stratification

JEFFREY S. NEWMAN

Factory Mutual Research Corporation, Norwood, MA 02062

Experimental data for stratification in a ventilated duct fire are examined and related to a Froude number.
Chemical compound stratification is shown to follow the local temperature rise. Regions of varying degrees of
stratification are also identified in terms of specific Froude number values.

INTRODUCTION where A Tis a temperature difference associ-

ated with the stratification; Trcf is the refer-
Stratification, or layering of combustion prod-
ence temperature; I)ref is the reference flow
ucts, in ventilated duct fires is important for (1)
velocity; g is the acceleration due to gravity;
flame spread analysis; (2) fire detection, control,
and H is the characteristic dimension, e.g.,
or extinguishing requirements; and (3) fire haz-
the height of the duct.
ard evaluation, such as for human escape poten-
. For negligible heat loss, the radial mass
tials. Stratification, however, has received only
distribution of a chemical compound i (Ci) in
limited attention. For example, it is virtually
a nonreacting fire-generated plume follows
ignored in duct flame spread models [1-6], al-
the local gas temperature rise (AT), i.e.,
though the use of an average gas temperature rise
appears to predict adequately certain conditions,
Ci AT
such as those governing ignition and the extent of
(2)
flame propagation [7]. Limited studies have been
Ci,re f ATref
performed on stratification effects in mine fire
detection [81 and smoke movement [9-12]. In order to verify the relations given in Eqs.
In this paper the results of an experimental (1) and (2), experiments were performed in the
evaluation of stratification in duct fires are pre- large-scale duct ( - 2 . 4 m x 2.4 m × 61 m long)
sented and generalized relationships are devel- shown in Fig. 1. The duct is a T-shaped structure
oped. designed to simulate conditions in underground
mine passages. The ventilation air entered the
EXPERIMENTAL duct through an inlet orifice contained in a flow
Relationships given in Refs. [ 13, 14] suggest the straightening section and was exhausted through
following: an exhaust duct connected to an air pollution
control system. For all the experiments, the fire
1. Temperature stratification can be described source was placed at the location indicated by N-
by a Froude number (Fr), i.e., 3 in Fig. 1, and instrumentation stations were
located at N-5, N-I1, N-19, E-1 (corresponding
-- =f(Fr) = f , (1) to distances downstream of the fire source, l/H,
ATref [(A T/ T,~f)gHI I/2 from 2 to 22), and in the exhaust duct near the

Copyright © 1984 by The Combustion Institute

Published by Elsevier Science Publishing Co., Inc.
52 Vanderbilt Avenue, New York, NY 10017 0010-21801841503.00
 34 JEFFREY S. NEWMAN

N•I9
Section~[ • H IE~,, s,c*i0,,'-j-//~7-1-~; ~t.~
• 11 tL. ~ ~L*~ ~.
Om/ (-I t-3/

Closed

4Urn

Cl I~thaullt Duct

."~" ~ Orifice Plote

Ventilation \ L g ,.- TO Air PollutiOn Control System

Air Flow
Fig. 1. Schematic of the large-scale duct.

orifice plate. Further details are given in Ref. [7] ties: (1) the difference between the gas tempera-
and [15]. ture near the ceiling (at height h = 0.88/-/) and
The following variables were investigated in the gas temperature near the floor (at height H -
the experiments: (1) fire source and (2) ventila- h = 0.12H), ATcf; and (2) the gas temperature
tion air velocity, v0. The heat release rates from rise above ambient at height h, ATh. The temper-
the various fire sources examined ranged be- ature stratification results were later generalized
tween 10 kW and 20 MW, with associated gas through a Froude number correlation of the form
temperature rises in the range of 1-600K. The of Eq. (1). With the temperature fields well-
ventilation air velocity was varied between 0.5 defined, the chemical compound distribution
and 4.0 m/s. The heat release rates and velocities was examined to determine whether it could be
were selected to give a wide range in Froude related to the local temperature stratification, as
number used as defined by Eq. (1). The fire suggested by Eq. (2).
sources in the experiments were heptane (as a
calibration fuel), coal with kerosene, and neo-
GAS T E M P E R A T U R E S T A T I F I C A T I O N
prene (conveyor belting) with coal and with
methanol. Further details are given in Ref. [15]. Examples of temperature profiles for various
values of v0, a, and I/H for heptane are given in
ANALYSIS Figs. 2a and 2b, where o~ is defined as the fire
intensity parameter in kW/s I/3 obtained from the
The following steps were used in the analysis of actual heat release rate versus time [7]. Figure 3a
the experimental data for assessing stratifica- presents a crossplot among the quantities ATavg ,
tion. A Tel, and A Th for heptane. The data in the figure
First, the global effects of a fire in a duct on the reveal two distinct regions; (1) region I, given by
average gas temperature rise relative to ambient values of AT~f/AT=,,g > 1.7, and (2) region II,
(AT~vg) were examined as a function of time, given by ATce/ATa,,g < 1.7. Region II can be
distance downstream of the fire source, average defined through the following equation:
duct velocity, and duct dimensions. Details are
reported in Ref. [7]. ATcf [ ATof1 o.77
The postignition fire environment was next = 0.67 (3)
analyzed for two additional temperature quanti- ATh L ~--~avgJ
FIRE-INDUCED STRATIFICATION 35

I0
mo
o
o 05 462
o I0 436 aOc3 o
20 593
08 0 40 389

06

.c
O4
© o(25

02

~o

J 0'5 ' ' L A J

0 I0 15 20
Th/L~ Tovg

F i g . 2 a . V e r t i c a l t e m p e r a t u r e p r o f i l e s f o r h e p t a n e p a n fires at I / H = 16.

I0~
x~12 o
.~IH xQl~ o
o 2
x ~aO
o 8
,~ 12
08 0 16
x 22

xt~
06

O4
~x

02

i ~ I I i 115 I
05 I0 20
ATh/~Tavg

F i g . 2b. V e r t i c a l a n d h o r i z o n t a l t e m p e r a t u r e p r o f i l e s f o r h e p t a n e w i t h o0 = 1 . 0 m / s a n d ~t = 4 3 . 6 k W / s ,/3
 36 JEFFREY S. NEWMAN

2 I I I I l 1 1 l I l I I l ' l I
cient r 2 = 0.90:
oHeptone Pon Fire I
I
I '/,~ c 1.5 u~ -I
= , (4)
ATavg L [(A Tcf/Tavg)gH ] I/2
Q5
where ..~g = Tavg/To/vo.
Three regions, two analogous to those previ-
ously observed in Fig. 3 and one new region, can
0.2 X o n Tr Region I
be defined in Fig. 4. An apparent "critical"
Froude number, (Fr)cr = 0.9, can be specified at
0"101. 021 I I 0.5lI I t Ili 2i I i 5i I I
the boundary between the first two regions (re-
~Tcf/ZTovg gions I and II), where ATcf/ATsvg = 1.7. Below
Fig. 3a. Gas temperature stratification heptane data. this value, ATcf = AT,, as previously shown in
Fig. 3. Therefore in region I only two tempera-
ture parameters, A T, and A Tavg, define stratifi-
where the multiple correlation, r 2, equals 0.99
cation. Region II is bounded by 0.9 < Fr <
(for the regression of ATcf on AT, and A Ta~8). In
- 10. Here, ATcf, AT,, and ATavg are all neces-
this region vertical stratification is a function of
sary to determine the degree of stratification. For
both the ceiling temperature rise and the average
Fr _> 10 (region III in Fig. 4), stratification is
gas temperature rise. As AT~f/AT~g approaches
insignificant, i.e., AT, = AT, vs.
values of the order of 0.1 or less, the flow
ATh is useful for various applications, such as
becomes well mixed and stratification can be
in assessing flame spread and detector response.
considered negligible. For region I, AT~f/AT, =
It can now be expressed in terms of T~vgand V~vg
1.0. This implies that the gas near the floor is
using closed form solutions. For region I, from
essentially at ambient conditions.
Eq. (3) and taking AT, = AT~f,
In Fig. 3b additional data are plotted for the
kerosene/coal, coal, coal/belting, and methanol gH [ A Tavg] 2
2.25
belting fire tests and oil shale rubble from Ref. ATh= Tavg L v - ~ J (5)
[8]. The solid line, from the best fit to the
heptane data, shows the additional data are also For region II, from Eqs. (3) and (4),
well fit to within + 10%.
From the data in Fig. 3, it appears that region I
AT, = 1.8 ATavg].23. (6)
is associated with relatively low Froude num- Tavg Oavg2
bers, i.e., buoyancy dominating temperature
stratification. Region II appears to be governed
by large Froude numbers; i.e., there is a signifi-
I I I I I i i ii i I I I I I I
cant interaction of the ventilation velocity with • Kerosene/Cool I
the fire-induced buoyancy. To examine these • Cool I
I eCoollBelting(Neoprene) ~ :: :~ ;:
effects further, a Froude number (Fr) was used • Methonol/Belting
based upon the average " h o t " velocity across
the duct (va~s), as the reference velocity, ATcf,
and the average gas temperature (T,,g) as the
reference temperature. This is shown in Fig. 4,
where ATcf/AT.,g is plotted as a function of Fr 0.2 / Region "11" Region z
for the data in Fig. 3. (The specific conditions / -
0.1
and data used for each of the plotted points are I I I I I Illl , I I I I I

0.I 0.20.5 I 2 5 I0
given in Ref. [15].) The data in Fig. 4 suggest the /'~TCf//x TOVg
following relationship with a correlation coeffi- Fig. 3b. Gas temperature stratification data for other fires.
FIRE-INDUCED STRATIFICATION 37

stratification is sufficient to specify the stratifi-

I0 ~ 0 HeF
ptoine
r rPo
~nme~ r ~
..... T i I • Kerosene/Cool cation of i. The average concentration of i for a
5 Rey,v,, x ! l,cool material in a ventilated duct is given by
% I J• Cool/Belting(Neoprene)
o~,3~,ol I • Methonol/Belting
Oi
Ci,avg = , (7)
,2
PoVoA

where Oi is the mass generation rate of i, Po is the

- o5 i4 ambient gas density, and A is the cross-sectional
area of the duct. Oi can be expressed by the
following equation:

Oi = Yi ( Q A / H A ) , (8)
OI 0.5 I Vovg I/Z5 I0 20
where Yi is the fractional yield of i; QA is the
Fr= F~g.1
actual heat release rate; and HA is the actual heat
L "ovg J of combustion of the material.
Fig. 4. (3as temperature stratification versus Froude num- Figure 5 shows a plot of C/Ccavg as a function
ber. of AT/ATavg for CO and CO2. All the CO data
were obtained from direct measurements in the
CHEMICAL COMPOUND STRATIFICA- duct for h/H = 0.94. For CO2, data from Refs.
TION [16] and [17] for premixed propane burners were
used for h/H between 0.32 and 0.68. A very
According to Eq. (2) the mass concentration of good correspondence (r = 0.99) between the
any chemical compound i (Ci) follows the local chemical compound and local gas temperature
gas temperature rise; thus the gas temperature stratification can be noted. Figure 5 suggests

30 ! ! ! ! i /

25 0 Heptone
[] Pre-Mixed Propone D6,17l
/
20 %7 Cool/Neoprene Belting
Methonol/Beltinq

<
1.0

0.5

I I i I I
0.5 1.0 1.5 2.0 25 30
AT IZ~ Tovg
Fig. 5. Chemical compound stratification versus local gas temperature stratitication.
38 JEFFREY S. NEWMAN

that, while the heat losses downstream of the fire Fr Froude number
source are not negligible [7], (1) the local gas g acceleration of gravity (m/s 2)
temperature and chemical compound stratifica- generation rate of chemical compound
tion are governed by the same transport process, (kg/s)
i.e., Froude number dependence, and (2) Eq. (2) h height for position of interest (m)
is applicable to the fire environment downstream H ceiling height of duct (m)
of a fire source in a ventilated duct. Since a HA actual heat of combustion (kJ/kg)
nonreacting system has been assumed, it is ex- 1 distance downstream of fire source (m)
pected that these relationships will also hold for Q^ actual heat release rate (kW)
smoke and other compounds. T temperature (K)
Since the ceiling concentration (Ci,h) is impor- AT temperature difference or temperature
tant for many applications such as fire detection, rise (K)
it is useful to relate Ci,h to Ci,av~ through the v gas velocity (m/s)
following expression: Y mass yield of chemical compound (kg/kg)

Ci,h ATh
- - = . (9) Greek
Ci,avg ATavg
ot fire intensity parameter (kW/s 1/3)
Combining Eqs. (7)-(9) and rearranging yield p gas density (kg/m 3)
Gi (ATn
Subscripts

avg average
It should be noted from Eq. (10) that Ci,h ap-
cf ceiling to floor
proaches Ci,avg as heat is lost from the gas and
cr critical
temperature stratification is reduced (ATh/ATavg
h height
~1). i individual chemical compound
o ambient air
SUMMARY ref reference
1. Stratification in a ventilated duct fire can be
The author wishes to thank Drs. A. Tewar-
quantified by a Froude number in combina-
son and G. Heskestad of Factory Mutual and
tion with the average gas temperature rise
Dr. A. F. Cohen o f the U.S. Bureau o f Mines,
and two temperatures associated with the
Pittsburgh, for many valuable discussions
stratification.
during the course o f this work. The financial
2. Based on experimental results it is found
support o f the United States Bureau o f Mines,
that, when Fr < 0.9, the combustion prod-
Pittsburgh, PA, under Contract No.
ucts in the duct are highly stratified, while
HO113017 is deeply appreciated.
for Fr > 10, the combustion products are
well mixed.
3. The mass concentration of any chemical REFERENCES
compound is shown to follow the local gas
1.
Roberts, A. F., and Clough, G., Combust. Flame
temperature rise. 11:365 (1967).
2. de Ris, J., Combust. Sci. Technol. 2:239 (1970),
NOMENCLATURE 3. Roberts, A. F., Research Report No. 263, Safety in
Mines Research Establishment, Buxton, England,
A cross-sectional area of duct (m 2) 1970.
C concentration of chemical compound (kg/ 4. Hunter, L. W., and Favin, S., Combust. Flame 42:7
kg) (1981).
FIRE-INDUCED STRATIFICATION 39

5. Edwards, J. C., Perlee, H. E., and Chaiken, R. F., 13. Heskestad, G., Eighteenth Symposium (Interna-
ASME Paper No. 76-HT-61, 1976. tional) on Combustion, The Combustion Institute,
6. Habib, I. S., J. o f Heat Transfer 102:110(1980). Pittsburgh, PA, 1981, p. 951.
7. Newman, J. S., and Tewarson, A., Combust. Flame 14. Heskestad, G., Factory Mutual Research Corporation,
51:347 (1983). Norwood, MA, Serial J.I. 0FOEI.RU, August 1980.
8. Litton, C. S., Hertzberg, M., and Furno, A. L., 15. Newman, J. S., Factory Mutual Research Corpora-
Eighteenth Symposium (InternationaO on Combus- tion, Norwood, MA, Serial J.I. OG2N4.RC, July 1982.
tion, The Combustion Institute, Pittsburgh, PA, 1981, 16. Newman, J. S., and Croce, P. A., Supplementary
p. 633. Report, Third Annual Report, Factory Mutual Re-
9. Hwang, C. C., Chaiken, R. F., Singer, J. M., and Chi, search Corporation, Norwood, MA, Serial No. 22501,
D. N. H., Sixteenth Symposium (InternationaO on July 1979.
Combustion, The Combustion Institute, Pittsburgh, 17. Stavrianidis, P., Supplementary Volume, Final and
PA, 1977, p. 1385. Fifth Report, Factory Mutual Research CorpoTation,
10. Lee, C. K., Chaiken, R. F., and Singer, J. M., Norwood, MA, Serial J.I. 0FON3.RA, June 1982.
Combust. Sci. Technol. 2:59 (1979).
11. Thomas, P. H., Fire Research Note 723/1968, Fire
Research Station, Borehamwood, Hefts, England,
1968.
12. Pucher, G. K., Osterreichischelngenieur-Zeitschrift
22:92 (1979). Received 9 November 1983; revised 13 February 1984