.
- . ‘4; h
fifimtf {‘“r‘,

$3.: ~.‘ ‘Q
i ‘ “
3“”;
“1.4" 3 “ "' a”

 

vi ‘ o-
‘.. 37".. “c"

‘7 JP... q, .-
‘W W. WM,
ma; we,

.‘u 1.4

 

"3“,;
- .xzs.‘ ~
w ‘M, 1 g.

x "if

.fi‘ 4' —, 3
’5: 133*»,-
{3‘7

‘ w
urn , m
”.mp4": ‘

lv-I‘

fiJLEi‘TrQiE
-. a i
‘3'}.‘3g-NJ:

 

IICJCJ

m mums.
', '2'21'tgrll ~~
rt, w w

9m '<~ Qua-.1.

31". 5';
.. m...

- .
. 1c, :3
r ‘1‘"!‘1; . .

'r. :13??? 'I .

, .‘ >H' ,'...‘ ._ ‘ " "
‘ . ..‘.;,‘.".',v‘5 w ' ‘ . x -- -. ' v , “ * 5.23.5,“ “ ‘-

u , K , .. , . w . '3‘ . ~ . '0 T , mung-gr“; .

..’ ' n "'2.“

 

12"“ .13”
xi“ 1' .
7‘ “.5 J

v
A

 

“htow

MICHIGAN TATE UNIVERSITY L

imuumzni mumummunm’ril‘lliil

I!
J
3 93 009117106

 

 

This is to certify that the

dissertation entitled

WIND—AIDED FLAME SPREAD OVER
CHARRING AND VAPORIZING SOLIDS

presented by

KAMEL EL MEKKI

has been accepted towards fulfillment
Of the requirements for

DOCTOR OF PHILOSOPHY degreein MECHANICAL ENGINEERING

 

 

 

flzmfl 417% E

Date 6/xx/q/

MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771

 

 

 

, \
LIBRARY
Michigan State
University
K r

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

   
  

DATE DUE DATE DUE DATE DUE

 

 

"4 L‘ 3 a A.’ ’
in g 2 £13};
,_ I

:l l-
fil_|l——l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l J

 

 

 

MSU Io An Affirmative Action/Equal Opportunity Institution
' omens-pt

 

 

WIND-AIDED FLAME SPREAD
OVER CHARRING AND VAPORIZING
SOLIDS

By

Kamel El Mekki

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Mechanical Engineering

1991

ABSTRACT

WIND-AIDED FLAME SPREAD
OVER CHARRING AND VAPORIZING
SOLIDS

By
Kamel El Mekki

This work presents the results of a detailed experimental investigation and
analysis of laminar forced flow wind-aided flame spread over wood and PMMA in
the ceiling configuration. This investigation includes measurements of flame spread

rate and heat and mass transfer.

In the flame spread work, the speed of propagation of the pyrolysis front and
the flame front and the production rates of the major chemical species were measured
as a function of time. The effect Of the free stream velocity and the oxygen mass
fraction on the flame spread rate and the production and depletion rates Of major
chemical species were investigated for both wood and PMMA. The effect of
external radiation was investigated only for wood, because use of external radiation
for PMMA results in excessive melting and dripping. It was found that all these
environmental parameters, and especially the oxygen mass fraction, control the flame

spread rate and the chemical species production rates.

In the heat transfer part of this work, the effect of wind speed and ambient
oxygen mass fraction on heat transfer to the surface underneath and ahead of the
flame tip during wind-aided flame spread were investigated experimentally. The data
were correlated according to a simple theoretical model. The experiments were also
performed in the ceiling configuration. High-temperature ceramic solids containing
surface and in-depth thermocouples were used downstream of the bumin g PMMA

sample to measure the net heat transfer to the surface. Simultaneous measurements

of the PMMA surface temperature (to determine the pyrolysis front), ceramic solid-
phase temperatures (to determine the conduction heat transfer), gas-phase
temperatures (to determine the convective heat transfer), and the flame tip location
were made. Since the excess pyrolyzate (which produced a flame front that was only
slightly ahead of the pyrolysis front) was low for all experiments, the heat transfer
measurements underneath the flame were conducted with a natural gas diffusion
flame in the boundary layer over the ceramic solids mounted in the ceiling
configuration. For these laminar flames, it was found that convection by the hot
gases is the dominant mode of heat transfer to the surface ahead of and underneath
the flame. The radiation component becomes more significant as the free stream
oxygen mass fraction increases. Surface temperatures and flame spread rate
equations were derived from these heat transfer correlations. It was found that the
flame spread rate is linearly with free stream velocity, as expected, and varies to the

1.5-power of the free stream oxygen mass fraction.

To my father; Mohammed, my mother; Zohra,
my wife; Dalila, and my children; Wafa and Karim
for their confidence, devotion, and understanding

in helping me attain this important milestone of my life.

iv

Acknowledgement

The author is greatly thankful to his advisor Professor Arvind Atreya. His
constant guidance, encouragement and direct involvement in every step of the work
was of invaluable help in the completion Of this work.I deeply appreciate his
patience and his friendship. I would also like to thank the members of my
Committee, Dr. I. S. Wichman, Dr. J. Beck, Dr. K. Mukherjee and Dr. P. Hunt for
their input and guidance.

I would like to thank my colleagues and friends; especially Sanjay Agrawal
for his help in conducting the flame spread experiment on PMMA.

This work was supported by USDA under contract number 86-FSTY-9-90192.

Table Of Contents

LIST OF TABLES ..............................................................................................
LIST OF FIGURES ...........................................................................................
NOMENCLATURE ............................................................................................

Chapter One: Introduction and Literature Review ...................................

1.1 Background ...................................................................................................
1.1.1 Pyrolysis ............................................................................................
1.1.1.1 Pyrolysis Of non-charting solids (PMMA) ...................................

1.1.1.2 Pyrolysis of charting solids (WOOD) ..........................................

1.1.2 Flame spread ......................................................................................
1.1.3 Heat transfer to the preheat zone .........................................................

1.2 Prospect Of this work ......................................................................................

Chapter Two: Experimental Apparatus and Procedure ..........................

2.1 Small-scale combustion wind tunnel ................................................................
2.1.1 Inlet Section .......................................................................................
2.1.2 Turbulence manipulation section .........................................................
2.1.3 Test section ........................................................................................
2.1.4 Extemal radiation source ....................................................................
2.1.5 Exhaust section ..................................................................................

2.2 Gas analysis equipment ..................................................................................
2.2.1 Total Hydrocarbons [THC] ................................................................
2.2.2 H20 meter .........................................................................................
2.2.3 C0—C02 meter .................................................................................
2.2.4 02 meter ............................................................................................

2.3 Data acquisition .............................................................................................

Q’i.

O‘MUJNN

12
14
16
16
18
19
23
23
25
28
28
28
28
30
30

2.4 WOOD and PMMA samples preparation ........................................................ 30

2.5 Ceramic samples preparation .......................................................................... 31

2.6 Flame spread experimental procedure ............................................................. 31

2.7 Heat transfer experimental procedure .............................................................. 33

2.8 Error analysis ................................................................................................ 36

2.8.1 Mass balance ...................................................................................... 36

2.8.2 Pyrolysis-front spread rate .................................................................. 40

2.8.3 Heat flux ............................................................................................ 40

Chapter Three: Flame Spread over Vaporizing Solids .......................... 42

3.1 Results .......................................................................................................... 43

3.1.1 Temperature measurements ................................................................. 43

3.1.2 Flame spread rates .............................................................................. 45

3.1.3 Species production rates ..................................................................... 52

3.1.3.1 Burning zone behavior during flame spread ................................ 55

3.1.3.2 Incompleteness Of combustion .................................................... 57

3.2 Discussion ..................................................................................................... 59

Chapter Four: Flame Spread over Charring Solids ................................. 60

4.1 Results .......................................................................................................... 61

4.1.1 Temperature measurements ................................................................. 61

4.1.2 Pyrolysis and flame fronts .................................................................. 63

4.1.3 Spread rates ....................................................................................... 66

4.1.4 Species production rates ..................................................................... 68
Chapter Five: Convective and Radiative Heat Transfer to the

Solid ........................................................................................................................... 79

5.1 Formulation ................................................................................................... 80

5.2 Results and Discussion .................................................................................. 82

5.2.1 Transient Heat Flux Measurements ..................................................... 82

5.2.2 Steady-State Heat Flux Measurements ................................................. 90

vii

5.2.3 Surface Temperature ........................................................................... 94
5.2.4 Flame Spread Rate ............................................................................. 97
Chapter Six: Conclusions .................................................................................. 100

Appendix A: Study of Number and Location of Thermocouples in

a Ceramic Solid for Heat Flux Computations ......... s ..................................... 103
Appendix B: Data of Flame Spread Experiments on PMMA ................ 115
Appendix C: Data of Flame Spread Experiments on WOOD ............... 120
Appendix D: Data of Heat Transfer Experiments ..................................... 134
List Of References ................................................................................................ 139

viii

Table 2.1
Table 2.2
Table 4.1

LIST OF TABLES

Mass balance for CH4 diffusion flame.

 

Mass balance for PMMA diffusion flame.

 

Empirical formula Of Poplar wood [Atreya 1984].

 

Figure 1.1
Figure 1.2
Figure 1.3

Figure 1.4
Figure 1.5
Figure 1.6
Figure 2.1
Figure 2.2

Figure 2.3
Figure 2.4
Figure 2.5

Figure 2.6
Figure 2.7

Figure 2.8

Figure 2.9

Figure 2.10
Figure 2.11
Figure 2.12

Figure 2.13
Figure 2.14

Figure 2.15

LIST OF FIGURES

PMMA surface temperature and mass flux histories during
pyrolysis experiment (Y o..=0.233. q"=3W/cm 2) [Vovelle 1987].

 

Mass flux behavior of PMMA sample under different rates of external
radiation [Vovelle 1984].

 

Typical mass flux history Of a wood sample during pyrolysis
[Nurbakhsh 1989].

 

Schematics Of different modes of flame spread relative to gravity. ...............

Schematic of upward flame spread on wood sample.

 

PMMA surface temperature history during flame spread experiment. ...........

 

Schematic diagram Of the experimental apparatus.

Velocity profile at the leading edge Of the sample (x=0) before
and after installing the frame metal screen.

 

Frame metal screen used in the turbulence manipulation section. .................

Velocity profiles at the leading edge Of the sample.

 

Optimum combination Of glass beads, honeycombs, and metal screens
used in the turbulence manipulation section.

 

Cross section of the test section and the heater assembly.

 

Screens configuration for radiation scattering, resulting into almost
uniform incident radiation along the sample surface.

 

Measured heat flux profiles along the sample (x=0 corresponds
to the leading edge Of the sample).

 

Schematic diagram Of the mixing chamber.

 

Gas analysis equipment.

 

Schematic Of wind-aided flame spread on wood.

 

Schematic of wind-aided flame spread on PMMA with ceramic solids
mounted downstream for transient heat transfer measurements.

 

Schematic Of natural gas diffusion flame over ceramic solids for
steady-state heat transfer measurements.

 

Diagram Of the porous-metal burner used in heat transfer experiments.

’I‘henaturalgascomesonlyfromthesidesofthembes.

 

Mass balance test experiment on CH1. diffusion flame.

 

20
20
21

22

27
29
32

35
37

Figure 2.16

Figure 2.17
Figure 2.18

Figure 3.1

Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5

Figure 3.6

Figure 3.7

Figure 3.8

Figure 3.9

Figure 3.10

Figure 3.11

Figure 4.1

Figure 4.2

Figure 4.3
Figure 4.4

Mass production rates histories during wind-aided flame spread on PMMA
(U..=O.9m/s. and Yo..=0.233).

 

Pyrolysis-front history on PMMA sample (U..=O.9m/s and Yo..=0.233). ...........

Measured incident and convected heat fluxes repeatability
(U..=O.9m/s, Yo..=0.233, x=321n.).

 

Measured surface temperatures during flame spread on PMMA
(U..=O.9m/s, Y...=O.233).

 

Flame-front history comparison between video records (dashed lines) and
surface temperature measurements (solid lines).

 

Pyrolysis-front and flame-front histories during flame spread

 

experiments on PMMA.

Dependence Of the pyrolysis and flame-from speeds on the free stream
velocity for PMMA (Yo..=0.233).

 

Dependence Of the pyrolysis and flame-front speeds on the free stream
oxygen mass fraction for PMMA (U..=O.9m/s).

 

Dependence Of the measured flame radiation at the center Of the test
section bottom for different oxygen mass raction for PMMA (xF70cm).
Picture of flame spreading on PMMA (a) U..=O.9m/s, Yo..=0.233

(b) U..=O.9m/s, Yo..=0.6l.

 

Dependence of the species mass production rates on the pyrolysis-front
location for different wind speed (Y o..=0.23: l-U..=O.6m/s;

 

2-U..=0.9m/s; 3-U..=l.5m/s).

Dependence of the species mass production rates on the pyrolysis-front
location for different oxygen mass fraction (U..=O.9m/s; l-Yo..=0.233;
2-Yo..=0.43; 3-Yo..=1.0).

 

Normalized meg/mm: with the corresponding mass based
stoichiomeu‘ic fraction (i.e. the value of 1 corresponds to complete
combustion). '

 

Normalized mega/mo with the corresponding mass based
stoichiometric fraction (i.e. the value Of 1 corresponds to complete
combustion).

 

Measured surface temperatures and their rate Of change with time for wood

 

during flame spread (U..=O.6m/s, Y,.=O.233, £1"=0.5W/cm2).

Measured surface temperatures for wood during flame spread
(U..=O.9m/s, Y,_=O.233, q"=1.3W/cm2).

 

Pyrolysis and flame front histories (Y 0.50.233, £1"=0.5W/cm 2).

 

Pyrolysis and flame front histories «050233, q"=O.9wIcm 2).

 

39
41

41

46

47

49

50

53

S4

58

58

62

62

Figure 4.5 Pyrolysis and flame front histories (Yo..=0.233, O"=l.3W/cm 2). ........................ 65

 

 

 

 

 

Figure 4.6 Pyrolysis and flame front histories (U..=O.9m/s, q"=0.5W/cm 2). ........................ 65
Figure 4.7 Dependence Of the pyrolysis and flame front speeds on the free stream

oxygen mass fraction (U..=O.9m/s, (1"=0.5W/cm2). 67
Figure 4.8 Dependence of the pyrolysis and flame front speeds on the external

radiation (Yo..=0.233). 69
Figure 4.9 Dependence Of the pyrolysis and flame front speeds on the free stream

velocity (Y o..=0.233). 69
Figure 4.10 Species production rates history during the 3 phases of wind-aided flame

spread on wood (U..=O.6m/s, Y,..=O.233, q"=l.3W/cm2). 70
Figure 4.11 Dependence of the species mass production rates on the pyrolysis-front

location for different wind speeds (Y,.=0.233. d"=o.SW/cm 2). ........................ 71
Figure 4.12 Dependence Of the species mass production rates on the pyrolysis-front

location for different speeds (Yo..=0.233, Q"=O.9W/cm2). - 71
Figure 4.13 Dependence Of the species mass production rates on the pyrolysis-front

location for different wind speeds (Y,.=O.233, q"=l.3W/cm 2). ........................ 72

Figure 4.14 Dependence Of the species mass production rates on the pyrolysis-front
for different oxygen mass fraction (U..=O.9m/s, q"=0.5W/cm 2). ....................... 72

Figure 4.15 Normalized Elect/limo: with the corresponding mass based
stoichiometric fraction (i.e., the value Of 1 corresponds to complete

 

combustion). 75
Figure 4.16 Picture Of a spreading flame on wood sample (U..=O.9 m/s, Yo..=0.233.
c'f'=0.5W/cm2). 75

 

Figure 4.17 Instantaneous pictures taken of a spreading flame on a wood sample
(U..=O.9m/s. Yo.=0.61,t'1"=0.5W/cm2). 77

Figure 4.18 Normalized mcoytnmo with the corresponding mass based

stoichiometric fraction (i.e. the value Of 1 correspond to complete
combustion). - 78

Figure 5.1 Schematic diagram of the model. 80

Figure 5.2 Comparison of the measured and the curve fitted gas-phase
temperature during flame spread on PMMA (U..=O.9m/s, Yo..=l.0 ,

 

 

x=33cm, film). 84

Figure 5.3 Heat flux measurements during flame spread on PMMA
(U..=O.9mls,Y.,..=0.233). 85

Figure 5.4 Heat flux measurements during flame spread on PMMA
(U..=O.9m/s, Y...=1.0). 85

 

xii

Figure 5.5

Figure 5.6

Figure 5.7
Figure 5.8

Figure 5.9

Figure 5.10

Figure 5.11

Figure 5.12
Figure 5.13

Figure 5.14

Appendix A

Figure A.l
Figure A.2
Figure A.3
Figure A.4
Figure A.5

Figure A.6

Figure A.7

Figure A.8.1

Figure 16.8.1

Nondimensional measured and predicted convective heat transfer during
flame spread on PMMA.

 

Schematic diagram of an expanding plane underneath the sample
surface.

 

Measlned and predicted flame radiation for different Yo...

 

Net, total and surface radiation heat fluxes histories
(U.=l.2m/s. xFZS.4cm, x=17.7cm).

 

Measured gas-phase temperature profiles Of a methane diffusion flame
(U..=O.9mls, Yo..=0.233. xF2S.4cm)

 

Measured convective and radiative heat fluxes underneath and
downstream of steady porous-metal burner diffusion flames
(Y 0..=0.211).

 

Nondimensional measured convective heat flux underneath and

downstream Of porous-metal burner diffusion flames (Y...=O.211). ........

Nondimensional measured and predicted PMMA surface temperatures.

Measured and predicted pyrolysis-front spread rate dependence on

...........

 

Measured and predicted pyrolysis-flout spread rate dependence on
Y... (U..=O.9mIs).

 

Experimental setup.

 

Power step input history.

 

Temperature histories at the surface and indepth

 

Comparison of measured and smoothed temperature data.

 

Computed heat flux history using one in-depth thermocouple
measurements.

 

Temperature residuals when using one in-depth thermocouple
measurements for heat flux calculations.

 

Computed heat flux history using 2 in-depth thermocouple
mearnrrements.

 

Temperature residuals when using 2 in-depth thermocouple
measurements for heat flux calculations.

 

Temperature residual of the 2lul thermocouple when using

xiii

88

88
89

91

91

93

93
96

98

98

105
105
106
106

107

107

109

109

Figure A.9

Figure A.10

Figure A.ll

Figure A.12

Figure A.13

Figure A.14

Appendix B

Figure B.Txx

Figure B.prx

Appendix C

Figure C.Txx

Figure C.prx

Appendix D

Figure Door

2 in-depth thermocouple measurements for heat flux calculations ................

Computed heat flux history when using the thermocouple at x=0 and
another in-depth.

 

Computed heat flux history when using the thermocouple at x=0 and
another 2 in-depth.

 

Effect Of the number Of thermocouples used on the computed
heat flux.

 

Effect Of 2mm location error Of the in-depth thermocouple on

 

the heat flux results (using 2 thermocouples).

Effect Of 2mm location error Of the in-depth thermocouple on
the heat transfer results (using one thermocouple).

 

Effect of properties error on the computed heat flux (using all the
thermocouple measurements).

 

Species mass production rate histories for PMMA.
Species mass production rate plotted against the pyrolysis-front
for PMMA.

Species mass production rate histories for wood.
Species mass production rate plotted against the pyrolysis-front
for wood.

Heat flux measurements during flame spread on PMMA.

xiv

110

110

111

111

113

113

114

NOMENCLATURE

C = Specific Heat

f = Mass-based stoichiometric coefficient

h = Convective heat transfer coefficient

AH = Heat Of combustion per unit mass of PMMA
k

m"

= Thermal conductivity
= Mass flux

Nu = Nusselt number

Pr = Ptandtl number

c1" = Heat flux

Re = Reynolds number

i = Time
T = Temperamre
U,V _= Velocity
x = Streamwise coordinate
y = Transverse coordinate
Y

= Mass fraction
Greek
= Thermal diffusivity
p = Density
9 = T—T.
Subscripts
f = Flame
g = Gas
0 = Oxygen
p = Pyrolysis
s = Surface
T = Total
x = at x
co = free stream

CHAPTER ONE

Introduction
and
Literature Review

Wood is, and has been, for centuries a common construction material. But unfor-
tunately, wood is combustible and is involved in hazardous building fires. In residen-
tial rooms or buildings, once wood is ignited, fire spreads in different directions
depending on the location of the wood with respect to gravity (floor, ceiling, wall).
The spread of fire in a building goes through complex interactions between physical,
heat and mass transfer, and chemical reaction processes. Thermal radiation from a
ceiling layer Of hot gases and hot walls is the primary mode of energy transfer in a fire
to the as-yet-unburned materials. The flame-spread rate su'ongly depends upon the
magnitude of this incident heat flux, which causes the surface temperature to reach the
ignition (or pyrolysis) temperature. At this temperature, the solid starts to pyrolyze
and feed combustible gases into the fire plumes. As the fire burns in a closed room,
the oxygen concentration decreases to a point where the flame can no longer exist.
Hence, the oxygen concentration is also expected to significantly afiect the rate of
flame spread Finally, as windows and doors are broken or burned, air currents

develop, which along with buoyant currents cause the flame to spread in difl'erent

directions. This results in wind-opposed (fire spread and incoming air in opposite
directions) and wind-aided (fire spread and incoming air in the same direction) modes
of flame spread. The three environmental conditions identified above, namely; incident
heat flux, ambient oxygen concenu'ation, and wind speed; are the primary external

parameters that conuol the flame-spread rate over combustible mamials like wood
[Emmons 1974].

The room-fire problem has been approached in two difl'erent ways: some investi-
gators have focused on understanding the effect Of building geometry on fire growth
[Emmons 1980, Quintiere 1980], while others have conducted experiments searching
for the flammability characterisrics and behavior of fire growth on different building
materials, such as polymethylmethacrylate (PMMA), paper, wood, etc. However, the
results of some of these experiments are different because the investigators failed to
control some of the environmental parameters [Emmons 1980]. Such independent con-
trol of wind speed and oxygen mass fraction have proved to be very helpful in identi—
fying the controlling mechanisms of wind-opposed flame spread [Fernandez-Pello et a1.
1981]. Only one similar experimental investigation of wind-aided flame-spread, on
PMMA, has been reported in the literature [Loh and Fernandez-Fella 1984]. This
investigation did not consider external radiation as one of the variables. Experimental

studies of wind-aided flame spread on (thick) charting materials like wood have not
been reported.

In this work, wind-aided flame-spread over wood is extensively studied for all the
above mentioned environmental parameters. Similar experiments are also conducted on

PMMA for comparison with the available results of Loh and Fernandez-Fella (1984).

1.1 Backgmund
1.1.1 Pyrolysis

To study flame spread over combustible solids, one has to understand the solid-
phase decomposition (pyrolysis) of the material, because the incident heat flux from

the existing flame leads to gasification of the material producing the fuel, which in turn
sustains the flame. Thus, the rate of the solid gasification is central to the rate of
flame spread. The pyrolysis Of non-charting and chaning solids represents two extreme
cases of solid-phases degradation. For PMMA, the solid is totally consumed and

essentially "vaporizes". However, for wood, a char layer is formed at the sm'face
while the pyrolysis fiont propagates into the solid. '

1.1.1.1 Pyrolysis Of vaporizing solids (PMMA)

As the surface temperature reaches the pyrolysis temperature, the solid starts to
pyrolyze and surface regression is observed as gasification continues. For a constant
incident heat flux, the mass flux was found to reach a steady state as illustrated in Fig-

ure 1.1 [V ovelle et al. 1987]. Tewarson and Plan (1976) observed that the mass loss

rate correlates as

m = —-— (1.1)

where

O = incident flux (W/cmz)
Ow = heat loss from the surface (W/cmz)

LG = heat Of gasification 0/3) + heat conducted into the solid

This linear dependency on the incident heat flux was also confirmed by Vovelle et al.
(1984). However in a real fire, the incident heat flux at the surface increases as the
flame approaches, therefore a transient mass loss rate would result. Such a study has
been carried out by Vovelle et al. (1984) when they examined the effect of known
variable incident heat flux on the mass loss rate of both PMMA and particle board in

an inert atmosphere. As expected, the mass loss rate of the PMMA sample rises with
the heat flux as shown in Figure 1.2.

400

 

Ts (°C)

 

 

10

 

4 (g/szs)

 

 

—4 E3
X
'E
*2
Olvr'fi""lr T'T'tro
O 500 1000 1500 2000
Time (see)

Figure 1.1 PMMA surface temperature and mass flux histories during

pyrolysis experiment (Y...=0.233, q"=3W/cm 2) [Vovelle 1987].

 

 

 

 

 

2E-03 .;
0 const. heat flux expt. .0
e 1.66X1O’3W/cm’.s
El 3.33X10'3W/cm’s . o
2E—03—
”a?
N .
E
o
\ 113—034
3
"E
ESE—04—
OE+OO
1 5
ll 2
q (W/ cm )
Figure 1.2 Mass flux behavior of PMMA sample under different rates Of external

radiation [V ovelle 1984].

1.1.1.2 Pyrolysis Of charring solids (WOOD)

As soon as the sample is exposed to an external heat, the so-called inert heating
stage starts. During this stage, the temperature rises without decomposition. However,
some moisture evaporation occurs. This phenomenon holds true until the pyrolysis
temperature is reached at the surface. At this moment the pyrolysis stage and the
release of volatiles are initiated. As the surface temperature continues to increase, the
volatile mass flux rapidly increases until it reaches a maximum when a thin char layer
starts to form at the surface. Then as the char layer thickness increases, the volatile
mass flux gradually decreases. This phenomenon, shown in Figure 1.3, has been
Observed by a number of investigators [Vovelle et a1. 1984, Au'eya 1983, Nurbakhsh
1989].

 

 

 

 

2E-O3
O
'1 . . 0
. 0
g 0
O O
”a? 15—034 , 'o ..
N g o '
E ‘ \s n.
< O .0 ./ ~..~
on o .o
f“ l
“E 5E-O4— :
O
O
OE+OO += , , . I
O 100 200 300 400
Time (sec) '
Figure 1.3 Typical mass flux history of a wood sample during pyrolysis

[Nurbakhsh 1989].

Due to its lower thermal conductivity than the original wood, char becomes a bar-
rier to the incoming heat flux and prevents it from passing into the inward propagating
pyrolysis zone. In addition, the re-emitted radiant flux by the char surface increases
owing to its high emissivity. This causes the conduction heat transfer to the in~depth
virgin wood to drop in time.

Thus, for these two different combustible materials, wood and PMMA, one
expects that during wind-aided flame spread: For PMMA, the flame will stay attached
to the leading edge as the flame tip propagates downstream. For wood, however, an

extinction-front would propagate behind the pyrolysis-front due to the lack of fuel
caused by the thick char-layer.

1.1.2 Flame Smead

Pyrolysis-front spread rate and heat transfer to the solid ahead and underneath the
flame have been investigated, both experimentally and theoretically, for difl'erent
materials and at different angles relative to gravity [Loh and Femandez-Pello 1984,
Atreya 1983, Saito et a1. 1987, Fernandez-Polio and Williams 1975, Femandez-Pello
and Hirano 1983, Saito et a1. 1986]. Figure 1.4 shows qualitatively the effects of grav-
ity and air flow on propagating flames. Clearly, the most hazardeous cases are wind-
aided flame spread in the ceiling configuration and upward (buoyancy driven) flame
spread. The flame in these configurations is pushed closer toward the unwa sur-

faces, thus increasing the heat transfer by the hot gases to the as-yet-unbumed sur-

faces.

Several theoretical papers have been published on the upward buoyancy driven
flame spread [Saito et al. 1986, Femandez-Pello 1978, Annamalai and Sibulkin 1979,
Markstein and deRis 1973, Orloff and deRis 1976, Sibulkin and Kim 1977,
Fernandez—Pello and Quintiere 1982, Kulkarni and Fisher 1988]. All of these models
are in qualitative agreement with the experimental results. Fernandez-Pello (197 8)
developed a model that predicred flame heights larger than the experimental data.

<— —>
s\
a) Horizontal flame spread
—> -
_> N
—> -—>
—> —>
s\ .3
—>
b W-A fl spread fl
) ame on oor d) W-A flame spread on ceiling
_>.
L” \‘
—> <—
—-> <—
s 3’
c) W—O flame spread on floor I . .
e) W-O flame spread on ceiling
is s

i

f) Upward flame spread g) Downward flame spread

Figure 1.4 Schematics of different modes of flame spread relative to gravity.

This larger predicted flame height increases the heat flux to the surface, causing higher
flame-spread rate predictions. Saito et al. (1986) conducted experiments on both
PMMA and Douglas-fir and developed a model for upward flame spread. They
predicted that the pyrolysis front spread rate goes as

 

= . (12)

where the characteristic ignition time 1: depends on the fuel properties, and the heat
flux (assumed constant) from the flame to the surface underneath, and xf and x13 are the
flame-front and the pyrolysis-front locations, respectively, as shown in Figure 1.5.
The experimental results agree very well with the predictions and previous theories for
PMMA. However for wood, the authors did not use external radiant heaters; and,
therefore, no well defined continued flame spread occured. Thus, their prediction of an
initial acceleratory spread, before reaching steady state, for charting material could not
be confirmed.

 

 

 

Figure 1.5 Schematic of upward flame spread on wood sample.

A number of models for wind-aided flame spread have also been
rier et al. 1980, Carrier et a1. 1983, DiBlasi 1987, Wichman and At
these models have focused on vaporizing solids. Recently Wichman an.
(1991) and Carrier et. a1. (1990) have also developed such models. These models
predict steady flame spread rates, and assume a steady state in the pyrolyzing zone.
Wichman and Agrawal (1991) predicted the flame spread rate as

2 2
Vp ___ p..Cp.L. Tr-Tp exp(-M2Pr) . (1.3)

 

This forrnulaticn shows the dependency on solid and gas properties, flame temperature
and wind speed. The experimental study and the model of Loh and Fernandez-Pena

(1984) for PMMA were in agreement, and the pyrolysis front spread rate was found to
be:

(1.4)

v n __w.- [my
9 °‘ p,C,3., Tp—T, '

The above equations look similar, but discrepancies exists between the results Of Loh
and Femandez-Pello (1984) and those of Wichman and Agrawal (1991) and the com-
plicated mathematical analysis and numerical solution Of integral equations by Carrier
et a1. (1980). These two latter models predict that Vp~ Yo”; however, Loh and
Femandez-Pello concluded from their analysis and experiments, that Vp~ Y3... This
certainly points out the complexity of the wind-aided flame spread process and the
need to experimentally control the external conditions carefully. Also it seems essen-

tial to evaluate the validity of the several simplifying assumptions made in the theoreti-
cal models.

Transient solutions, with finite rate gas phase kinetics, were obtained by DiBlasi
et aL (1987, 1988a) using a finite-difference scheme. They concluded that as the oxy-
gen mass fraction decreases, the flame temperature and the heat fluxes to the unburned
fuel surface drop. This causes a slower fuel production rate, and consequently, a

slower spread rate. At a very low oxygen mass fiaction, the effects of finite kinetics

10

appear mainly at the leading edge, where extinction begins.

Previous investigators have primarily focused on non-charting materiala because
of the difficulty in simultanously analyzing the solid-phase processes for charting
solids. Canier et a1. (1983) made the first and only attempt toward the development
of models for wind-aided flame spread over charring solids. They treated the char and
the virgin wood as solids with different thermal properties. The surface temperature
was considered to rise to th pyrolysis temperature at the arrival of the pyrolysis front,
then to keep rising until some temperature T, at which time char starts to chemically
erode to form gas. The transient problem was formulated, but only the steady state part
Of the solution was obtained. The numerical solution for both Blasius-type and
Oseen-type flow were also obtained. While the flame temperature was found to be
invariant under both flow-field approximations, the char-layer thickness and the flame
position under the Blasius-type flow were about twice as large as those under the

Oseen-type flow. The char layer thickness was also found to be sensitive to both the
temperature T,, and the latent heat.

DiBlasi et al. (1988b) solved numerically the problem of wind-aided flame spread.
over thin charring solids. Their results show that the pyrolysis front spread rate ini-
tially accelerates while the flame foot is still at the leading edge. As the burn-out front
starts to propagate, the pyrolysis front spread rate decelerates. Steady state is then
reached when both the pyrolysis front and the bum-out front spread rates are equal. If
the burn-out front spread rate keeps accelerating, extinction occurs. These results were
in agreement with previous experimental data. However, the dependency of the pyro-
lysis front spread rate on the flow velocity was not fully resolved. Although the
experiments indicate that the spread rate becomes constant for relatively high flow
speeds. the model predicts that the spread rate keeps increasing.

All the above models for wind-aided flame spread are for the laminar case and
can be classified as thermal models where one step infinitely fast gas-phase chemical

reaction (va+v°O——)products) is assumed. Finite rate chemical reactions were found

11

to have an effect only at the leading edge and the flame tip, since flame temperature
drops at these locations [DiBlasi et a1. 1987, 1988a, 1988b].

The flame spread rate is determined by the rate at which the surface temperature
is raised to the solid pyrolysis temperature (i.e. Vp=dxpldt where T,=Tp (or Tig) at xp).
The flame spread rate depends on how fast the solid surface temperature is raised to its
pyrolysis temperature, which depends on the convective and radiative heat flux from
the flame. The convective heat flux is proportional to the gas-phase temperature gra-
dient at the sm'face, which depends on both the flame temperature and the boundary
layer thickness. Therefore, the most important variables that affect this mode of flame
spread are the wind speed. which controls the boundary layer thickness, and the oxy-
gen mass fraction, which controls the flame temperature. The above equations show

the flame spread rate dependency on these two parameters.

Flame spread experiments over PMMA samples are conducted to compare the
results with those of Loh and Fernandez-Fella (1984) and to seek the reason(s), if any,
for the above-mentioned discrepancies.

Previous experiments on PMMA have proved that the heat flux by flame radiation
and convection by the hot gases is sufficient to bring the solid surface temperature to
its ignition temperature and allow the flame to spread. Therefore, only the wind speed
and the ambient oxygen mass fraction have been studied for PMMA. For the experi-
ments on wood, however, a minimum external radiant flux level is necessary for the
spread to occur [Saito et a1. 1987]. The magnitude of this incident heat flux and the
preheat time have a significant effect on the flame spread rate. The incompleteness of
combustion as well as flame radiation,which both affect the flame temperature, were
not previously studied and are investigated in detail in the present study. The objec-
tive is to provide a complete physical understanding of this mode of fire spread over
charring and non-charting solids.

12

The combustion products were only measured and analyzed by Atreya (1983) in
his pyrolysis, ignition, and fire spread on horizontal surfaces of wood, and by Abu-
Zaid (1988) and Nurbakhsh (1989) in their pyrolysis experiments on wood. The flame
spread results of Atreya (1983) show that the measured average mass flux is initially
proportional to the 4/3 power of the fire radius, then stays constant due to the balance
between the attenuation due to charting and increase in fire size, and finally becomes
inversely proportional to the fire radius due to char build-up. In this work, the mass

production rate of all the major chemical species are measured and analyzed.

1.1.3 Heat transfer to the solid

Figure 1.6 shows a typical surface temperature hiStory during flame spread on
PMMA. The rise of the solid surface temperature from the ambient temperature T. to
the pyrolysis temperature T1) at the pyrolysis front is due both to both the preheating
upstream of the flame tip x>xf and the preheating underneath the flame xp<x<xf (the
overfire region). Therefore, the heating of the surface to its pyrolysis temperature in
the overfire region depends on the surface temperature at the flame tip arrival x=xf.

The higher this temperatm'e,’ the less is the heating needed from the flame in the
overfire region, and hence the shorter the distance (xf—xp).

Clearly, the preheating downstream of the flame tip cannot be neglected and has
to be well understood for proper and realistic theoretical models, since in analytical

solutions, the heat flux in the preheat zone is required to develop the flame spread
model [Fernandez-Pello 1978].

Sibulkin and Kim (1977) obtained the following expression for wind-aided flame
spread

(C1192 5f
VI, = 2
kpC (Tig - s)

 

(1.5)

where kpC is the square of the solid-phase thermal responsivity, Tig is the ignition or

pyrolysis temperature, T8 is the upstream surface temperature, 8f is an effective flame

13

heat transfer distance measured from the pyrolysis front, and qf is the heat flux at the
pyrolysis front. Therefore, the flame spread rate can be found through the knowledge
of qf and 5f. Quintiere et a1. (1986) completed this cycle by performing upward flame
spread experiments where they measured the heat flux (at the wall underneath and
ahead of the flame), and the flame height for difi‘erent combustible materials. They
found that the heat flux at the wall ahead of the flame correlates as qw ~ x‘P, where
p=2.4, and x measured from the base of the sample. This is very close to the results
of Ahmad and Faeth (1979) who found that p=2.33. Ahmad and Faeth (1978) found,
both experimentally and theoretically, that the heat flux underneath the flame is
approximately constant and that the distance (xf — xp) is dependent on fuel and

ambient oxygen mass fractions.

 

 

 

 

 

 

 

 

500
Flame spread Burning zone
400— '
:8 4 . \Pyrolysis front orrivol
a) 300‘
.3
E \Flome tip orrivol
3 2004
E
a) 1
*—
100- . . r
Ignition 3
4 3);, 1: 3(8,
0 . , . , . , . 1 .
O 50 100 150 200 250

Time (sec)

Figure 1.6 PMMA surface temperature history during flame spread experiment.

14

The purpose of this work is to quantify this heat transfer for wind-aided flame
spread in the ceiling configuration. The available tools, used by Quintiere et a1. (1986),
for measuring these heat fluxes are water-cooled radiometers. However, these radiom-
eters are inappropriate because they provide total heat flux measurements to a water-
cooled surface rather than to an already-heated surface whose surface temperature is

changing with time. To overcome this problem, ceramic detectors with surface and ,

indepth thermocouples were developed.

1.2 Prospects of this work

The flame spread and the heat transfer experiments are conducted in a small scale
combustion wind tunnel. Two materials representing the extremes of a broad spectrum
of available building materials are studied These are (i) wood, which chars during
burning, and (ii) PMMA, which vaporizes during bunting. Species concentrations, and
surface temperature histories, are measured as a function of time. In addition, video
records of the flame tip are taken. The surface temperature measurements are used to
determine the pyrolysis-front arrival; and the video records show the transient flame tip
position, while the species concentration measurements provide a measure of the
evolved mass flux during the flame spread process. This mass flux and the pyrolysis-
front spread rate reveal the burning zone condition and the incompleteness of combus-
tion.

Since wood chars and PMMA melts and drips during flame spread, surface and
indepth temperature measurements are difficult and inadequate for incident heat flux
computations. Therefore, high temperature ceramic samples were cast with surface and
indepth thermocouples. The Optimum design for the number and location of the ther-
mocouples in the solid was determined experimentally. It was found that a total of
three thermocouples, one at the front surface, one at the back surface, and one indepth
near the surface exposed to the external heat flux, provide a sufficient number of tem-
perature measurements for accurate incident heat flux computations, given the thermal

properties of the ceramic (assumed linear with temperature). These ceramic detectors

15

were also painted black to make their emissivity near unity.

The heat transfer in the preheat zone ahead of the flame tip was studied by con-
ducting flame spread experiments on PMMA samples upstream of the ceramic detec-
tors. The heat transfer underneath the flame, however, was studied with the help of a

natural gas diffusion flame existing in the boundary layer over the solid surface in the
ceiling configuration. ‘

Inverse heat conduction calculations were used to obtain the net (total) heat flux
as a function of time. The convective heat transfer component was derived from the
temperature gradient at the solid-gas interface. The re-radiation heat flux component
was computed from the knowledge of the surface and surrounding temperatures and
the surface emissivity. Finally, the flame radiation component of the net heat flux was
computed by applying an energy balance at the solid- gas interface. The effect of the
free Stream velocity and the ambient oxygen concentration on the convective and radi-

ative components of the net heat transfer to the solid was studied.

The purpose of this work is to conduct a thorough experimental study of wind-
aided flame spread over charring and vaporizing materials in the ceiling configuration.
The objectives are summarized bellow:

1- To provide a physical understanding that will serve as a basis for the develop-
ment or refinement of theoretical models.

2- To provide new measurements for charring materials like wood, under
stringently-controlled test conditions.

3- To develop a model for correlating the results.

4- To provide additional measurements for vaporizing materials like PMMA, and

discover the reasons for the discrepencies, if any, between this work and available
results in the literature.

5- To provide precise measurements of the convective and radiative components of

the net heat flux to the surface ahead and underneath the difl’usion flame.

CHAP TER TWO

Experimental Apparatus
and Procedure

The flame spread and the heat transfer experiments were conducted in a small
scale combustion wind tunnel. This apparatus was designed by the principal investiga-
tor of the project. This wind tunnel is a unique facility where all the environmental
variables mentioned [11... Yo... 4"] can be easily controlled. The schematic of the
apparatus is shown in Figure 2.1. It consists of three main parts: (i) a small-scale
combustion wind tunnel, which is capable of providing a desired external radiation on
the sample surface along a flat-plate boundary-layer flow at a given free stream velo-
city and composition; (ii) a set of continuous gas analyzers for measuring the total
unburned hydrocarbons THC, the depletion of oxygen, and the production of C02, CO,
and H20; and (iii) data acquisition equipment to collect, store and process the data.

Each of these parts is described in detail in the following sections.

2.1 Small-scale combustion wind tunnel

The wind tunnel consists of three sections: (i) an inlet section where the wind
speed and composition are controlled; (ii) a test section where the flame spread over

the sample takes place; and (iii) an exhaust section where the produced species

16

17

.3333." 35.5.2.5 06 no .555 8.3.2.8 fin 95»:—

ofinoz Seemim
0:5 @33me
2923282 85385-29

 

         
 

 

 

 

 

 

 

 

 

 

 

 

   
   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. 5335‘
$82 Boi-<m _ OU
BEES» :onafiuBQBO .8» =
sassy. 803m; 3 <8u
09¢. wanna—mafia
. A
8:3 958:3: 5m :22 NO
w -
NENWneacu 552 02 i all 85238; Boo .......... ..
‘ .WZOHH.<—>mxm< DUED LE Q w ....... .
I a E 92:85 30 ....... m
x m m
awn—>250 1H. w
A— way—Em m m m
n mambwm: PZSQ<M
t m m 393.. m
2m ,4 u s 02:08
F N
(full.
W Bo Bo .E ........... .
W" n O Q I]... .... W
> \c; <m <m / . OE .
) ._. “A 7. m3

 

 

 

 

E :3 l_
V.A zo_5mm..L

FmD<Exm

 

18

mix with the core flow. Each of these sections is designed such that it can be detached
and modified separately. 2.1
2.1.1 Inlet section

The desired air flow and composition results from the mixture of air and oxygen
(or nitrogen) after flowing through sonic orifices, as illustrated in Figure 2.1. The air
supplied by the building compressor first enters a large tank in order to dampen all the
mechanical oscillations. The air then flows through a sonic orifice, of a known diame-
ter, at a certain upstream pressure. This pressure should be higher than the critical
sonic pressure, whereas the downstream pressure is close to atmospheric. However,
oxygen and nitrogen are supplied by pressurized gas cylinders to a large tank before
flowing through anorher sonic orifice. Knowing the nozzle diameter, upstream pressure

and flow density and temperature, the mass flow rate can be computed as [Holman
1984]

 

. 23c 2 7-1
=A . —L 2.
m NP“\/RTi y+l[y+1] (1)

where

m = mass flow rate (lbm/sec)

AN = area of the nozzle (in 2)

Pi = inlet static pressure (Psia)

gc = gravitational conversion factor (32.2 lbmft/lb.sec 2)
R = gas constant (ftlbf/lme)

Ti = inlet temperature (R)

y = ratio of specific heats of the gas (C p/C v)

19

The sonic nozzles were calibrated using a tracer gas technique where a known
flow rate of methane is introduced downstream of the orifice, and the mole fraction of
the methane-air mixture was measured. The discharge coefficient CD of each nozzle
was carefully determined from the flow measurements and found to be 0.97 [Abu-Zaid
1988]. The mixture flows through a 20 foot long, 3 inch inside diameter tube before

entering the settling chamber through 8 pipes, which branch out of the manifold at
equal angles.

2.1.2. Turbulence manipulation section

The flow exits the settling chamber with large and small scale eddies. Hence,
before running through the test section, the flow is first conditioned in the turbulence
manipulation section. Combinations of honeycombs, glass beads and fine screens were
tested for Blasius type of flow in the test section. The honeycombs are used to damp
the large scale eddies and the fine screens are used to dissipate the small scale eddies.
However, a two-inch space full of glass beads provides an inlet uniform flow. Unfor-
tunately, due to the high velocity near the walls of the settling chamber, the velocity
has a peak near all the four sides of the test section as shown in Figure 2.2. Different

combinations of honeycombs, glass beads and screens did lower these peaks, but didn’t

make them disappear.

Finally, after several trials, the peak in the velocity profile disappeared when a
screen frame, shown in Figure 2.3, was installed. The velocity profile measured for
different free-stream velocities is shown in Figure 2.4. These results are very close to

the Blasius-type of boundary-layer over a flat plate. The optimum combination of the
honeycomb, glass beads, and screens is shown in Figure 2.5.

20

 

V vvvvvv""
0 7

VV

 

 

 

100

80

60

4O

20

Y (mm)

Velocity profile at the leading edge of the sample (x=0) before

Figure 2.2

and after installing the frame metal screen.

 

 

ce.ere.~.e.e.e« -
e e e e e e
coo oeoeoeoe e e

 

Frame metal screen used in the turbulence manipulation section.

Figure 2.3

21

6389. 05 he on? 9532 05 no 8389 Eco—2’ in Beam..—

 

 

AEEV >
009 Ohm O_© O_.v CW 0
<an.0": a . rroo
»m\Em~.Pu3 o a -
IN.O
e
D r
o o
t t 14.0
o e a t
p p 10.0
beep» e e o > e e paeeeee o t
o lwd
o o to;
o o I
0000 OOO iN.—.
000000000 0 00 000000

 

 

(S/w) n

22

 

 

 

 

 

   

   

 
 

3...; g
E iiiilINUlllllllllllllllllllllllllllHHHHi 5:3
g niillIlHllIIIIIIIHIIIIIIIHIIIIIIHIniii} gag

   

J 1k

Optimum combination of glass beads, honeycombs, and metal screens

used in the turbulence manipulation section.

Figure 2.5

23

2.1.3. Test section

The useable portion of the tunnel test section is 0.81 m long and 0.153 m wide.
The test sample (0.76 m long, 0.076 m wide and 0.019 m thick) is placed horizontally
along the tunnel top while the bottom of the tunnel is hinged at the inlet allowing the
tunnel depth to be adjusted. The tunnel depth at the inlet is 0.1 m but can be
increased to 0.13 m at the exhaust end. This provides a maximum 30% increase in the
cross-sectional area at the exhaust end to compensate for the acceleration of the gas
core because of boundary layer grthh and gas expansion due to heat release. The
damper on the exhaust fan and the exit tunnel depth were adjusted to provide atmos-
pheric pressure in the tunnel test section to within 1x10‘4 torr. This was necessary to
prevent gas leakage in or out of the tunnel for chemical measurements and also to
maintain a nearly constant free stream velocity. A maximum 10% increase in the
free-stream velocity at the exit was observed. To further reduce the effect of variation
in the free-stream gas velocity, data for only the first 0.5 m were used. The RMS level
of turbulent fluctuations inside the tunnel was found to be less than 1% of the free-
stream velocity. Furthermore, the measurements of the velocity profile inside the tun-

nel (see Figure 2.4), show that the flow is laminar.

2.1.4. External radiation source

External radiation on the sample surface was provided from below by two types
of electrically-operated radiant heaters. The first type consists of three high-temperatm'e
(with maximum filament temperature of 1230 K) quartz electrical heaters (10 in x 10
in) placed at the bottom of the heater assembly shown in Figure 2.6. The second type
consists of six U-shaped Chromlox coil heaters (3/8 in diameter, Incoloy sheath, type
UTU, each 1.8 KW) installed above the quartz heaters. The heaters are separately con-
trolled by two 3-phase 440-Volt variable transformers. These heaters are housed in an
insulated box with the inside frame sides being highly reflective aluminum sheets. This
housing is covered by water-cooled shutters which slide out at the beginning of the
experiment.

24

Kawool
insulation W009 sample

  
 

Observation
window

Hinged
bottom

 

Water-cooled chutter

Coil heaters

Quartz electric heater

 

Ceramic fiber ' Heater temperature
insulation probe

Figure 2.6 Cross section of the test section and the heater assembly.

25

This heaters assembly is suited to simulate external radiation in building fires.

The incident radiation from the heaters passes through an optical glass window
(0.153 m wide and 0.76 m long) contained in the hinged bottom portion of the tunnel.
About 70% of the infrared radiation is transmitted to the sample surface. However,
due to the view factor. the radiation is a maximum at the center of the sample and
drops by about 30% at the two ends of the sample. To overcome this difficulty, three
screens have been installed between the heaters and the infrared optical glass window
to scatter the radiation. After a number of trials with the width and number of screens,
an optimum combination (see Figure 2.7) was configured such that the radiation meas-
ured at the sample surface was uniform to within i3% over the entire length of the

sample, as shown in Figure 2.8. The entire tunnel test section was maintained between
315 K and 335 K by cooling water.

2.1.5. Exhaust section

The main part of the exhaust section is the mixing chamber. In order to obtain a
representative gas sample for transient chemical analysis, the stratified products of
combustion have to be well mixed with the core flow. This mixing process is
acheived through the use of a combination of baffles, a series of electric tapes for
large-scale mixing, and a net of electric resistance wires for small-scale mixing. A
metal louver is added at the outlet of the mixing chamber to assure the mixing of

gases as shown in Figure 2.9. The electric tapes and the net of electric resistance are

heated to avoid condensation of heavy hydrocarbons or water.

A well mixed representative sample of gases is then extracted through the sam-
pling probe for chemical analysis. The rest of the flow then goes through a large
chamber before getting sucked by the exhaust fan. This large chamber is used to

suppress the mechanical oscillations created by the exhaust fan.

 

26

30”

 

1” 3” 1” .1

0.5”

l 1 Cooling water

 

10”

 

 

 

 

 

 

 

 

 

70

Figure 2.7 Screens configuration for radiation scattering, resulting into almom
uniform incident radiation along the sample surface.
1.0
0.9 a - t ‘
/ \
9 0.8+ // \
E
U
\ 0.7-1
E / 3 \
‘cr 0.5- M
0-4 I ' I ' I ' I I I ' T
O 10 20 3O 4O 50 60
X (cm)
Figure 2.8 Measured heat flux profiles along the sample (x=0 corresponds

to the leading edge of the sample).

27

.3220 «:38 05 he Ease 3.2.2.8 a." «Sufi

 
   
 

 

 

 

Eon been»
new on.»
men?
2.3 ”.58: 03.5
683:
V\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\_\\
\I
W
\ ‘l
f L ‘Mul
/ T
/
/ T

 

 

 

 

 

hege— 2.3
”:32 ”5.8:

28

2.2. Gas analysis equipment

For the species concentration measurement, a constant flow rate of a representa—
tive sample from the tunnel was supplied to the gas analyzers by a metal bellow
vacuum pump. To reduce errors due to condensation, the lines running to the total
hydrocarbon [THC] and H20 analyzers were heated by either electrical tape or hot
water as shown in Figure 2.10. The gas was then passed through a cold-trap at -5°C
and dried before passing through the 02 and the CO—COZ analyzers.

Prior to every experiment, the gas analyzers were first zeroed with nitrogen gas
flowing through, then they were adjusted to the proper reading for the known concen-
tration of every specie of the calibration gas running through.

2.2.1. Total Hydrocarbons analyzer [THC]

The total hydrocarbons were measured by a Flame Ionization Detector (FID) gas
chromatograph GC—3BF, which used a 40%H2—60%He mixture for fuel. The FID had

a very good response time and a time constant of 1.0 second.

2.2.2. Water analyzer

The water vapor concenu'ation history was measured by a condensation Dew-
Point hygrometer (General Eastern 1200APS). The dew-point temperature was meas-
ured by optically detecting the condensation on a temperature-controlled mirror sur-

face. The instrument had an accuracy of 10.2 °C and a time constant of 1.0 second.

2.2.3. CO-CO2 analyzer

An infrared IR702 nondispersive dual gas analyzer was used for CO and C02
concentration measurements. The flow was dried before entering the analyzer. The
meter had a good accuracy of 11% of full scale (CO : 0-3%, 0-12%; C02 : 0-6%, 0-
20%) and a time constant of 2.6 seconds.

29

Dew Point
Meter

Heated line

 

Figure 2.10 Gas analysis equiprneut.

30

2.2.4. 02 analyzer

The oxygen concentration was measured by a Beckman OM-ll polarographic

analyzer. This meter also required dry flow. The meter had an accuracy of 10.1% and
a time constant of 1.5 seconds.

2.3. Data acquisition

The analog signals from thermocouples, gas analyzers and the radiometer were
fed to a data acquisition/control unit (HP3497A). The data was then acquired by an
HP 486 personal computer using LabWindows software (National Instruments). The
handshaking commands and data transfer between the computer and the HP data
acquisition unit were assured by the use a GPIB card with a standard IEEE 488 inter-
face cable. The data was taken at 12 readings per second with a 10‘6 Volts mm.

2.4. WOOD and PMMA sample preparation

The samples used for the flame spread experiments were obtained from clear
boards of poplar and white sheets of PMMA. These were first conditioned at room
humidity and temperature (50% RH and 294 K) and then instrumented with nine ther-
mocouples on the surface 0.05 m apart. These thermocouples were made fi'om fine
chromel and alumel wires 76 pm in diameter. The method employed to install the
thermocouples on wood surface was developed by Atreya (1983). Two very fine layers
of wood were skinned off from the surface using a sharp razor blade, then a portion of
the thermocouple wire of each side on the junction was secured underneath each layer
using wood glue. A very fine drop of this glue was also put on the thermocouple junc-
tion. The assembly was then allowed to dry under a heavy weight for about 2 hours.

For the PMMA samples, however, an electric heat gun was used to heat the surface so

that the thermocouple wires stick.

31

2.5. Ceramic samples preparation

The available tools for measuring the heat flux at the surface are water—cooled
wide-angle radiometers. However, these radiometers are inappropriate because they
provide total heat flux measurements to a water-cooled surface rather than to an
already heated wall whose surface temperature is changing with time. Since wood
chars and PMMA melts and drips dming flame spread, indepth temperature measure-
ments are difficult and inadequate for incident heat flux computations. To overcome
this problem, ceramic detectors with surface and indepth thermocouples were

developed.

These high temperature ceramic samples were cast with surface and indepth ther-
mocouples (6 in length by 3 in width and l in thickness). The optimum design for the
number and location of the thermocouples was experimentally determined. It was
found that a total of 3 thermocouples, one at every boundary and one indepth near the
surface exposed to external heat flux, provide a sufficient number of temperature his-
tories for accurate incident heat flux computations. For accurate results, all the ceramic
samples have one thermocouple at each boundary and three indepth. The exact location
of the indepth thermocouples and thermal properties, at different temperatures, of each
ceramic sample, were carefully determined [Beck and Arnold 1977]. See Appendix A
for details. These ceramic detectors were also painted black to make their emissivities

near unity.

2.6 Flame spread experimental procedure

The test samples (30 in long, 3 in wide and 3/4 in thick) were placed horizontally
along the tunnel top. The wood samples were obtained from clear boards of poplar.
and the PMMA samples were obtained from white sheets of PMMA. They were first
conditioned at room temperature and humidity and then instrumented with nine ther-
mocouples two inches apart on the surface. Once all the desired conditions were set, i

the sample was ignited with a small methane porous metal burner placed at the inlet

32

with its face parallel to the sample surface as shown in Figure 2.11. The fuel flow rate
to the igniter was controlled such that the flame overhang on the sample surface was
about 0.02m. After ignition, the physical process that occurs inside the tunnel is
schematically shown in Figure 2.11. Surface temperature and species concentration
histories were collected and stored in the computer. In addition, video records of the
flame tip were taken. The surface temperature measurements provided the pyrolysis
front arrival, and the video records showed the transient flame tip position, while the
species concentration measurements provided the mass flux during the flame spread
process. This mass flux and the pyrolysis front spread rate would reveal the burning

zone condition and the incompleteness of combustion.

woon Xf ‘
SAMPLE xp¢

Igniter

 

 

 

I Thermal B.L.
WIND

‘—
DIRECTION
<—

EXTERNAL RADIATION

Figure 2.11 Schematic of wind-aided flame spread on wood.

33

2.7 Heat transfer experimental procedure

To study the heat transfer in the preheat zone ahead of the flame tip, a spreading
flame on 10 inch and 16 inch PMMA samples upstream of the ceramic detectors was
used. Time varying solid and gas-phase temperatures were measured. Video records of
the flame tip were also collected. Figure 2.12 shows a schematic of a spreading flame
on PMMA sample upstream of the ceramic detectors. The heat transfer underneath the
flame, however, will be studied for the case of a natural gas diffusion flame existing in
the boundary layer over the solid surface in the ceiling configuration, as illustrated in
Figure 2.13. The porous-metal burner used in the latter case was increased in size to 4
inches and specially constructed (see Figure 2.14) to reduce the exit velocity at the
porous metal surface for long flame experiments. A lower exit velocity allows the
flame to be more buoyantly dominated. This will certainly cause the porous metal
burner diffusion flame to have a stand-off distance close to that of the excess pyro-
lyzate in the overfire region during an actual flame spread experiment under similar

environmental conditions.

Inverse heat conduction calculations were used to obtain the net (incident) heat
flux as a function of time. The convective heat transfer component was derived from
the temperature gradient at the solid- gas interface. The re-radiation heat flux com-
ponent was computed from the knowledge of the surface and surrounding temperatures
and the surface emissivity. Finally, the flame radiation component of the total heat
flux was computed by applying an energy balance at the solid-gas interface. The
effect of the free stream velocity and the ambient oxygen concentration on the flame

convective and radiative components of the total heat transfer to the solid was studied.

34

Ceramic
solids

 

 

 

 

Water cooled plate

Schematic of wind-aided flame spread on PMMA with ceramic solids

Figure 2.12
mounted downstream for transient heat transfer measurements.

 

 

 

 

Burner
e V
Ceramic 3 g
solids ' g
x é!
.—
Flame 3 ;
i _ U
y
‘—
Water cooled plate
Figure 2.13 Schematic of natural gas diffusion flame over ceramic solids for

steady-state heat transfer measurements.

35

 

       

 

3.5”

 

 

 

Cooling-
water tube

 

 

 

 

\
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

UT .7, ”‘J

 

3”

 

Figure 2.14 Diagram of the porous-metal burner used in heat transfer experiments.
Thenaturalgascomesonlyfromthesidesofthetubes.

36

2.8. Error analysis

The experimental determination of any parameter is based upon measurements,
which usually contain errors. Two kinds of errors exist: uncertainty or random errors
and systematic errors. The uncertainty errors in the measurements can be neglected
since the data were collected by the computer through a data acquisition unit with a

10"6 Volts accuracy. Systematic errors, however, fall into one of two categories:
1. Calibration errors in the measurement devices.
2. Neglecting significant outside influences.

Since these systematic errors can exist in the measurements, a great deal of atten-
tion has been focused on calibrating the gas analyzers (errors of the 1“ kind) and keep-
ing atmospheric pressure in the test section to prevent leakage (errors of the 2“d kind).
All the experiments were repeated to confirm the results. Accuracy of the species mass
balance, the pyrolysis-front spread rate and the heat flux measurements are described

in the following sections.

2.8.1. Mass balance

The chemical measurements along with the measured mass flow rate inside the
tunnel yield the production and destruction rates of chemical species. These data were
then reduced in terms of the mass production rate of the species at the instant at which

they were produced. Investigation of the existence of systematic errors was performed

by checking the mass balance of all the species.

A known mass flow rate of methane was introduced into the burner prior to igni-
tion. To reduce errors due to incompleteness of combustion, the methane flow rate
was originally set such that a blue flame was established after ignition. The results of

such experimentation are shown in Figure 2.15, where the methane mass flow rate was
set at 0.016 g/s.

37

 

 

 

 

0.08
x THC
0 C02
0 02-depl.
A H20 6'“ ememmae
A 0 0
< 3
on g ‘3
V O
a, gymnawwmqbo
6 0.04-— 0 0°
" feMmpmman-f
(I) s O
(n o no
g 8 a
0.024 f ‘9
0 i too 200 300 400 500 600 700
Time (sec)

Figure 2.15 Mass balance test experiment on CH4 diffusion flame.

The following overall chemical reaction then describes the process (neglecting CO for-

mation; as shown in Figure 2.15, CO concentration is small)

 

CH4 + 202 —> C02 4" 2H20

The predicted and measured mass flow rate of all the species are summarized in Table

2.1. The largest error in the species measurements is about 3%, which is acceptable.

38

Table 2.1 Mass balance for CH4 diffusion flame.

 

 

 

 

 

Specie Prediction Experiment %Error
(9/8) (9/8)
CH4 0.016 0.0155
02-dep1. 0.064 0.063
C02 0 .044 0 .043
H20 0.036 0.035 2.5

 

 

 

 

The accuracy of the species mass production rate measurements during an actual
flame spread experiment can also be checked for the case of room air condition. Figure
2.16 shows the species mass production rates for the case of flame spread on PMMA
where U..=O.9m/s and Y°=0.233. For PMMA, the overall chemical reaction is
described by:

 

C5H802 + 6 02 —> 5C02 + 4 H20 .

The predicted and measured mass flow rates at t=1500 sec. (see Figure 2.16) are sum-
marized in Table 2.2. Since the total mass consumption rate of the sample is not
known, all the predictions are based on the oxygen depletion rate. The very low per-
cent errors shown in Table 2.2 show the reliability of all the gas analyzers and provide

confidence in the species measurements.

39

Table 2.2 Mass balance for PMMA diffusion flame.

 

 

 

 

 

 

Specie Prediction Experiment %Error
(9/8) (9/3)
OZ-depl. 0.05 0.05 --
C02 0.06 0.058 3.3
H20 0.0188 0.019 1.3 .

 

 

 

 

 

0.10
X THC
'1 0 C02
4' C0
0'08“ A 02-depl.
A U H20
(0
E
v 0.06“
B
O
0:
m 0.04“
(I)
O J
2

 

 

 

o ' 500 1000 1500 2000 2500

Time (sec)

Figure 2.16 Mass production rates histories during wind-aided flame spread on PMMA
(U..=O.9 m/s and Y...=O.233).

40

2.8.2. Pyrolysis-front spread rate

The accuracy of the pyrolysis-front spread rate was determined by the repeatabil-
ity of the experiments. Hence, all of the flame spread experiments were repeated to
confirm the results. Figure 2.17 shows three pyrolysis-front histories for three separate
experiments conducted for the case where U..=O.9m/s and Yo..=0.233. The deviation in

the pyrolysis-front spread rate is about 2% from its average value.

2.8.3. Heat flux

 

The accuracy of the incident heat flux measurements can be analyzed in two
ways: (i) the low level of the errors (RMS) of the output results of the inverse heat

conduction calculations, (ii) and the repeatability of the experiments.

The low RMS values show the good agreement between the measured and

predicted temperature histories, which in turn shows that there is low error in the cal-
culated heat flux.

The repeatability, however, can be investigated by looking at the heat flux results
of experiments repeated for the same condition. Figure 2.18 shows the incident heat
flux with the corresponding convected heat flux for two separate tests where
U..=O.9m/s, Yo..=0.233 and x=32in. The results of the two experiments are in very
good agreement, with an average difference of 4% for the net heat flux and 5% for the

convected heat flux.

The good repeatability of the results for species measurements, pyrolysis-front
spread rate, and heat flux measurements provide confidence that all the measured quan-

tities are within acceptable experimental error bounds.

41

 

 

 

 

60
A
E 50~
8/
x0.
+J 404
c
O
.f:
|
.2) 30”
(f7
3‘
8 Vp (cm/s)
3 20‘1
V 2.75E—2
O 2.73E-2
10 .fifi j Tar 2.72E-2

 

l I ' l ‘ T ' l l ' l H
400 600 800 1000 1200 1400 1600 1800 2000 2200
~ Time (sec)
Figure 2.17 Pyrolysis-front history on PMMA sample (U..=O.9 m/s and Y...=O.233).

0.7

 

<1“ (W/cmz)

 

 

 

 

Figure 2.18 Measured incident and convected heat fluxes repeatability
(U..-0.9m/s, Y...=0.233, x=32in.).

CHAPTER THREE

Flame Spread
over
Vaporizing Solids

This chapter presents a detailed experimental investigation on laboratory-scale
laminar wind-aided flame spread along a ceiling-mounted slab. The gas flow along the

slab is forced, and its speed and composition are controlled.

Recent models for non-charting materials [W ichman and Agrawal 1991, Carrier et
'al. 1990] have sparked a considerable controversy since they seemingly disagree with
the detailed experimental study of Loh and Femandez—Pello (1984). The models sug-
gest that the speed of the pyrolysis front varies linearly with the fiee stream oxygen
mass fraction, while the experiments suggest a quadratic dependence. This clearly
points out the need for further experimental investigation to determine whether any of
the assumptions made during the development of the models or possible experimental
errors are responsible for the discrepancy. Thus, the objectives of this work are to (i)
provide a physical understanding that will serve as the basis for the development or
refinement of theoretical models; and (ii) provide additional measurements for non-

charring materials like PMMA.

42

43

3.1. Results

 

As shown in Figure 2.11, the fuel vapors generated in the pyrolysis region which
extends from x = 0 to x = XI, are burned in the diffusion flame, which extends from
x = 0 to x = x; with xf > xp. The hot combustion products that flow downstream of
x f and the flame extension (xf - xp) help to convectively and radiatively heat the pris-
tine solid to a temperature (Tp) at which it begins to vigorously pyrolyze and contri-
bute fuel to the flame. Thus, the flame spread process consists of the spread of the
pyrolysis front. Clearly, the rate of flame spread will depend on how fast the surface
temperature of the solid is raised to its pyrolysis temperature.

3.1.1. Temperature Measurements

Figure 3.1 shows the measured surface temperatures for PMMA as the flame pro-
pagates along the ceiling-mounted sample. These temperatures are typical of all the
measurements. These measurements show a temperature plateau at about 643 K,

which is taken as the melting or vaporization temperature.

From visual observations it was found that the peak rate of change in the surface
temperature after ignition occurs is at the instant the flame tip arrives at the thermo-
couple location for both wood and PMMA. By writing a surface energy balance it can
also be demonstrated that this peak corresponds to a sharp increase in the incident heat
flux, which is caused by the flame tip arrival. Thus, the flame tip location, Xf, may be
determined as a function of time from the measured temperature profiles by calculating
the maximum value of dT,/dt. Results of such calculations show excellent agreement
with xf determined from video records for both wood and PMMA. Figure 3.2 shows
such agreement for PMMA where the solid lines correspond to the temperature meas-

urements and the dashed lines correspond to the video records.

 

 

 

 

 

 

 

 

 

800
Q 700— /
i,
‘3
B 600-
O .1
L
(I)
O- 500—
E
2
a) 400—
U
.9
S
m 300—

200 ' ' ' T ' ' ' T ' ' ' l ' ' fl 1 ' ' '

0 400 800 1200 1600 2000
Time (sec)
Figure 3.1 Measured surface temperatures during flame spread on PMMA
(U..=O.9m/s. Y,.=0.233).

60

50-
E -
U
V 40—
>2-
.. U YO
5 301 <m/s)
“T e 0.9 1.0
a) _ x 0.9 0.61
E 20 v 0.9 0.43
9 A 0.9 0.33
L *0— D 2.0 0.233

' o 1.5 0.233

as 0.9 0.233
0 o 0.6 0.233
' I - I ' I ' I ' I F
O 500 1000 1500 2000 2500 3000

Time (sec)
Figure 3.2 Flame-front history comparison between video records (dashed lines) and
surface temperature measurements (solid lines).

45

The pyrolysis front location, xp, is now found by defining a constant surface tempera-
ture, Tp, at which the solid begins to vigorously pyrolyze and contribute fuel to the
flame. For PMMA this is the vaporization temperature (643 K). Unfortunately, at
these temperatures the thermocouples often detach from the surface due to melting.
Thus, operator judgement is required in determining xp. To eliminate this difficulty,
At between xf and xp was consistently determined by (Tp - Tx)/(dT,/dt)mu. An exam-
ple of this method of determining Xf and x1) is illustrated in Figure 4.1. Figure 3.3
shows both the pyrolysis-flout location xp and the flame—front location xf histories for
most of the experiments (solid lines: xp, dashed lines: xf). Notice how close x.p and xf

are.

It is also important to note that the surface temperatures at xf are much higher
than ambient. They range from 533 K to 573 K for PMMA. This implies that most of
the temperatm'e rise has occurred in the preheat zone ahead of the flame. This obser-
vation is in sharp contrast with the previous assumption [Loh and Fernandez-Pello
1984, Annamalai and Sibulkin 1979] where the surface temperature is taken as

ambient until the arrival of the flame tip.

3.1.2. Flame Spread Rates

Once xf and XI, are determined as a function of time (Figure 3.3), the pyrolysis-
front speed Vp (defined as dx.p ldt) and the flame-front speed Vf (defined as de Idt) are
obtained from the slope of the least square fit lines. Data for only the first 0.5m were
used to minimize errors from an increase of the free stream velocity (U...) downstream
of the tunnel. In the first 0.5 m of the sample, both xp and xf increase nearly linearly
with time whereas in the last 0.15 m they exhibit slight acceleration. Figure 3.4 shows
VP and Vf plotted against the free stream velocity; it is seen that VI, and Vf increase
linearly with U...

46

nmmd
mnmd
nnmd
mmmd
00.0
med
0.0
0;

0.0
0.0
m;
0.N
0.0
0.0
0.0
0.0

XGBGBQUO

(220 :0 35.8505
32% one: 9:26 8:90: $8.25: 23 320-03% fin 95»...—

Aommv mEc.

000m 00.0N 00.0 P 00.0?

p b p -

 

 

 

 

 

(we) Jx 75> dx

47

NN

Annmarie <22“. .8 £8?

 

 

 

63.5.... 3.: 05 ca 38% 2:00-08“: 23 $32.3 05 we 8:8:800 1n 9:5...—
Am\Ev 362m; Eombm 00E
01m. 0; 0; a; N; 0; 0.0 0.0 ed
p _ . _ p _ _ _ b — . _ p _ . _ . F0.0
C> O
a> e
IN00
100.0
d 140.0

 

00.0

 

(S/LUO) 010.1 [303st

48

Figure 3.5 shows VI, and Vf plotted against the oxygen mass fraction (Y 0,,,.).
Note that several different measurements of VP and Vf for PMMA (from reflective to
blackened water-cooled bottom aluminum plate) are presented. It is once again

noteworthy that Vp and Vf are nearly equal (see Figures 3.4 and 3.5).

Figure 3.5 also shows that Vp depends upon the surface finish of the aluminum
plate directly under the bunting sample. The aluminum plate was used only for
PMMA since external radiation was not required. For the experiments on wood sam-
ples described in the next chapter, this plate was replaced with an infrared optical glass
window to allow external radiation. As is evident from Figure 3.5, the surface condi-
tion of this plate (and to a lesser degree, the two vertical sides of the tunnel that con-
tain the observation windows) considerably alters the flame spread rate because it
reflects the flame radiation back to the sample surface. This effect magnifies as Yo.
increases because the flame radiation increases. Measurements show that the flame
radiation at the plate surface increases linearly from nearly zero at Y0... = 0.2 to 3.5
W/cmz at Y0. = 1 (Figure 3.6). Thus, for a reflectivity as small as 0.1, the reflected
radiation becomes comparable with the external radiation used for the wood samples
discussed in chapter 4. The spread rate measured with a reflective aluminum foil for
Y0... = 0.6 is 2.7 times the spread rate for blackened and water-cooled aluminum plate
and 1.6 times the spread rate for a dull aluminum plate. Thus, the flame spread rate
depends upon the reflections inside the tunnel, and the result closest to the truth is that
of the blackened and water-cooled plate. For PMMA, these measurements show that
Vp ~ Yggf. This differs from the previous measurements for PMMA [Loh and
Fernandez-Per 1984] (dotted line in Figure 3.5), which show that vP ~ Y3... How-
ever, the present results involving a dull aluminum plate agree well with those of Loh
and Fernandez-Pello 1984; indicating that reflection of radiation inside the tunnel may
be the reason for disagreement (this corresponds to the systematic errors of the 2“‘1
kind described in chapter 2). Results for a water-cooled plate agree well with the
recent theoretical predictions for PMMA [Wichman and Agrawal 1991, Carrier et al.
1990], which are shown plotted by the dashed line in Figure 3.5.

49

Eamoudv <25 .8 Sues as: 5:6

Each» 8b 05 so 88% 88-08.“: 23 £523 05 do 8:3:88 Wm «Sufi

80> c0500: 30:: cmmxxo

o; No Ho eho

 

 

whim. 23233.2. m>_Hom_.Cwm
P b —y n — b _ h

 

. Cgv Become. om coEcoS) - u
$3: o_.ee|NeeceEe.._ e :3
Own—OOUKMBQB QMZMXUSm
23233.2 QmZMXUSm

.. . m:<|_n_ EDZ_§DI_< 4:50

)KOCIQ

[molmm

Tmolmv

 

molmo

 

(s/w) aim pDGJdS

50

F;

.302.le 5233 e8 .5th mng 5996 .555 ecu 883 .838

 

 

.8. on. .«o 383 05 3 cows—o8 2:3 @28on 05 do coco—ocean ed 0.53..—
80> c0500: mmoE cmmxxo
04 md md 50 md md To 00 md
. _ b r . _ b b p p L _ p _ . _ .X 0
I—
X

1N
rm.

 

 

<—

LUZ) N\ pDJHb UOIlDIpDJ GLUDIj
Z . .

51

The effect of the oxygen mass fraction on the flame spread rate can be further exam-
ined by looking at the chemisrry and the flame temperature. The following one step
chemical reaction can be used to describe the combustion process during flame spread:

CXHyO + (x+-§--;-)02——) xC02 + %H20

(til) (mam) (thcoz) (dingo)

+ [CO,THC,SOOT]
(that)
(Incomplete Combustion)
+ XAfAHu—XR) . (3.1)

(Heat Liberated)
where xA is the correction factor due to incompleteness of combustion (x =1 for com-
plete combustion), and am is the correction factor due to flame radiation losses (xR=0
for no flame radiation). If f is the stoichiometric fuel to oxygen ratio by mass, then

fro. gm of fuel requires Yo. gm of 02 and liberates [waAAHa-xnn J of heat.
This energy is utilized to

(i) raise the temperature of a unit mass of (02 , N2) mixture from T, to Tf:
[gar-Th]; '

(ii) raise the temperature of fYo, gms of fuel vapor from TI, to Tf: [CPCTf-TQwa];

(iii) provide for the heat required by the solid to produce fuel at the surface [QfYoa].

Thus, an energy balance would then lead to

waAa-«xnmn = cpcrf—T“) + gut-rpm... + QtYo, (3.2)

52

After rearranging the terms the flame temperature can be expressed as

fYoatxAAHa-xe-Ql—cpcrp—Te

(Tr—T?) = Cp (1 + We»)

 

(3.3)

where Q = L+Cp.(Tp-T..), where L is the latent heat of the solid.

For complete combustion and no flame radiation [xA(1-xR)=l]
Tf - Tp ~ Y0...

which leads to
vp ~ Y3“,
since vp ~ (Tr-T92.

However, as will be explained in section 3.1.3.2, the data show that xA decreases
and 7m increases with Yo... This makes the flame temperature less dependent on the
oxygen mass fiaction, and so also the pyrolysis-front spread rate. The results show
that Vp ~ Ygfi. Figure 3.7 shows 2 pictures taken of the flame during the flame spread
experiments (a) U..=O.9 m/s, Yo..=0.233, and (b) U..=O.9 m/s, Yo..=0.61. Notice the
difference in the flame brightness which indicates the change in flame radiation inten-

sity.

3.1.3. Species Production Rates

The species mass production and depletion rate histories are presented in Appen—
dix B for all the experiments performed. Figures 3.8 and 3.9, respectively, show the
species production rates for different free stream velocities and different oxygen mass
fractions plotted against the pyrolysis length, xp. Production rates of CO and total
unburned hydrocarbons are not presented because they are two orders of magnitude
smaller than the others; however, they are extremely important in fire research since

they were the primary reasons for many lost lives in building and other fires.

53

 

 

 

 

 

 

 

 

 

(b)

Figure 3.7 Picture of flame spreading on PMMA (a) U..=O.9m/s. Y°_--0.233
(b) U..=O.9m/s, Y°_=0.6l.

54

 

 

 

 

 

1E-01
ABE-02-
m
\ -
0‘
V4E—024
B .
O
(I
C
.9 .
*5 1E-021 . we" . , ,-
385—031 13'," _
‘o , -
O
L .
0- 45—03e
(D J 1,’
tn
0
2 - ' H20
-u- 02—dem.
-— CO
1E-O3 1 I l l n l7 l I
10 20 30 4O 50 60 708090100

Pyrolysis front Xp (cm)
Figure 3.8 Dependence of the species mass production rates on the pyrolysis-front
location for different wind speed (Y ,.=0.23; 1-U..=0.6m/s;
2-U..d).9tn/s; 3-U..=1.5m/s).

 

CDC

9.0
00
[LL

LlllL

_LLL J_Lllll

1

till

Mass Production Rate X102 (9/5)
8
O
J

 

 

 

 

10 U 40 I ' 80 100
Pyrolysis Front Xp (cm)

Figure 3.9 Dependence of the species mass [Induction rates on the pyrolysis-front

location for different oxygen mus fraction (U..=O.9m/s; l-Y...=O.233;

2-Y...=0.43; 3-Y...=l.0).

55

The data for small x9 is omitted since the measurements are affected by the igniter
flame and, hence, they are subjected to a large percentage error. Also, the soot pro-
duction rate was not measured even though a substantial amount of soot was formed,

especially during experiments at high Yo...

3.1.3.1. Burning zone behavior during flame spread

Species data were collected even after the flame had spread over the entire sam-
ple, i.e., in the boundary-layer burning zone. These data are presented only for
Yo..=0.43 for purposes of clarity (the rest of the data are in Appendix B). Vertical
straight lines are obtained (see Figure 3.9). because xp does not change during this
period since the pyrolysis front has already reached the end of the sample. The mass
production rate more than doubles during this period before achieving a steady-state
value for the case Yo..=0.43. The fact that the species production rate continues to
change substantially even after the pyrolysis front has reached the end of the sample
shows that the steady-state condition is not achieved in the boundary layer burning
zone. Vovelle et al (1987) showed that the mass loss rate kept increasing even after
the surface had started pyrolyzing as shown in Figure 1.1. Looking at this figure, the
surface temperature reaches steady state at about 300 sec, however the mass flux
reaches steady state at about 700 sec. Thus, the fact that the surface temperature in
the burning zone has become steady does not imply that the mass loss rate has become
steady.

The measured species mass rate can be expressed as

am
Ih(t) = w j m”(x,t) dx , (3.4)
0

where W is the width of the sample.

56

Taking the time derivative of the above equation gives

d1:1'l__(_t)= prj)a

lt-[m”(x, t)]dx + W m"(xp,t)-d- dip, (3.5)
0

dxp-V

where -— -

dt p = constant.

If a steady-state condition exists in the burning zone, (m”(x,t)=m”(x)), then

[dm/dxp]

rhll(xp) = W

(3.6)

For complete combustion, the production rates of C02 and H20 and the depletion
rates of 02 are directly related to the total fuel production rate between x = O and
x = xp. For the local mass flux to vary as x'05, the total fuel production rate at any
instant must vary as xp’z. Incompleteness of combustion will only serve to further
reduce this power. However, the data in Figure 3.9 shows that this power is greater
than 0.5 and closer to l; for PMMA at Y0... = 1.0 it is even greater than one. In addi-
tion, if xp=constant (end of the sample) then dxp/dt=0. Then, any changes in the mass

loss rate mm are due only to transient effects, with

. 0)
$de- = w x1 %-[m”(x,t)] dx . (3.7)
Consider now the experimental results of Figure 3.9. These results show that the
mass production rate keeps increasing even after the pyrolysis-front has reached the
end of the sample (or xp has become constant). This proves that the burning zone is
in a transient state during flame spread. This argument was further confirmed by con-
ducting flame spread experiments on 10 inch long PMMA samples. Video records of
these experiments showed that the flame length kept increasing even after the
pyrolysis-front had reached the end of the sample.

57

3.1.3.2. Incompleteness of combustion

Referring to Figure 3.9, we note that the production rate of H20 (tilHZO) seems to
follow that of C02 (taco) for all the experiments. The mass production rate of C02 is
always larger than that of H20. However, this is not true for the comparison between
the mass production rate of C02 and the mass depletion rate of 02 (rilez). For
Y0, = 0.23, rilco2 > onz, whereas for Y0, = 1.0, riico2 < ’hAOa The crossover
between rilco2 and film, can be seen at xp = 0.5 m for Y6” = 0.43. This indicates that
the combustion chemistry is changing as X? and Y0... are increased. Carbon in the fuel
is converted to unburned soot leading to lower mcoz. Reaction equation 3.1 implies

that:

[i] < [i]

mC02 comp. comb. InC02 incornp. comb.

This agrees with the physical observation that both larger flames and flames at higher
Yo... are brighter. Figures 3.10 and 3.11 show mega/saw: and taco/tum, respectively,
normalized with the corresponding mass based stoichiometric fi’action (i.e. a value of 1
corresponds to complete combustion), where PMMA is C15H40. The data show that
meg/mm: is slightly decreasing with x1) and that it drops significantly with Yo...
However, meg/map seems to be constant for all Yo... The data in Figure 3.8 shows
that the major chemical species mass rates increase with the wind speed, as expected.

since the heat transfer to the solid increases with U, as will be explained in detail in

Chapter 5.

58

 

 

m(:oz/moz
O
O)
l

 

 

 

.I vvvv
0.5- va'vvvvvvvvvv v
J VVvvvvvvvvvva
0.4“
0.2-‘
0.0 . I I I ' I ' I ' I '
10 20 30 40 50 50 70

Pyrolysis—Front Xp (cm)
Figure 3.10 Normalized rhea/mm: with the corresponding mass based

0.9 1.0
0.9 0.43
1.5 0.233
0.9 0.233
0.6 0.233

¥O$><

stoichiometric fraction (i.e. the value of 1 corresponds to complete

 

 

 

 

 

combustion).
1.6
1.44 or
O n .W' 2 fl
(£1 1.2" & .lv“ v‘.. ' p
\ 1 0- vIrv ’ “H- k V
N , vvvvvv ’v
o . v" v v
0 May v v p
E 0.8‘ AAAAAAAAAAAA 3:
0.6-i
0.4 I I ' l T I ' I I I I
10 20 30 4° 5° 60 70

Pyrolysis—Front Xp (cm)
Figure 3.11 Normalized Margo with the corresponding mass based

Ugo You,
(m/S)

0.9 1.0
0.9 0.43
1.5 0.233
0.9 0.233
0.6 0.233

X0894

stoichiometric fraction (i.e. the value of 1 corresponds to complete

combustion).

59

3.2. Discussion

It is interesting to note that there are several areas of agreement and disagreement
between these experimental results and the recent theoretical models developed for
non-charring materials [Wichman and Agrawal 1991, Carrier et al. 1990].

As predicted by previous theories[Wichman and Agrawal 1991, Carrier et al.
1990], both VI, and Vf are found to be linear with U, (Figure 3.4). However, these
theories also predict that usually Vf is significantly larger than Vp' which is in contrast
with the experimental results (Figures 3.4 and 3.5), which show that Vf = Vp' Like-
wise, Xf is found to be only slightly larger than xp (between 5 and 10%), regardless of
the free -stream velocity or the oxygen mass fraction. This discrepancy arises because
the theoretical models utilize the steady-state Emmons’ (1956) solution in the boundary
layer burning zone. As discussed earlier (Figure 3.9), the burning zone is unsteady in
the solid phase during the flame spread process. This leads to lower instantaneous fuel
mass production rates that result in smaller-than-predicted value of xf and Vf. The
existence of this unsteady bunting zone is further confirmed by the fact that the species
production rates vary roughly as Itp rather than as x35.

As is evident fi'om Figure 3.5, the theoretically predicted [Wichman and Agrawal
1991, Carrier et al. 1990] dependence of V1) on Yo... agrees well with that determined
experimentally. Increase in Y0, increases Vp primarily by increasing the heat flux
from the flame to the as-yet—unburned solid surface. This heat flux increases because
the flame temperature increases proportionally with Yo, in the absence of flame radia-
tion and incompleteness of combustion. This heat flux also decreases because of
increased shielding of the fuel surface from the flame by the evolved fuel mass flux.
As discussed earlier, the theoretical models overestimate the evolved fuel mass flux in
the burning zone by assuming steady state Emmons’ solution. This results in an
overestimation of the heat blockage factor. Fortunately, it is compensated by overes-

timating the flame temperature by neglecting flame radiation and incompleteness of

combustion.

CHAPTER FOUR

Flame Spread
over
C harring Solids

Previous experimental work on charting materials has been done on particle
boards in the upward (buoyancy-driven) flame spread mode [Saito et al. 1986, Kul-
karni and Fisher 1988, Quintiere et al. 1986]. However, studies in the forced convec-
tive mode have been primarily limited to non-chairing solids, such as PMMA
[Fernandez-Pello et al. 1981]. As discussed in Chapter 3, even for PMMA, uncertain-
ity exists regarding the prediction of flame spread rates under different environmental
conditions. The problem is further complicated for charrin g solids due to the forma-
tion of an insulating char layer. In addition, since wood requires preheating [Mekki et
al. 1990, Saito et al. 1986]. This problem is further complicated becauSe it introduces
two new parameters namely the external radiation and the preheat time. Atreya (1983)
studied the effect of the preheat time for certain incident heat fluxes on the horizontal
flame spread rate on wood. The results of his experimental work indicated that the
flame spread rate on wood increases with increasing either or both the external radia-
tion and the preheat time. His work also showed the dependency of the flame spread

rate on both the internal solid properties and the external environmental conditions.

60

61

This Chapter presents an experimental study of wind-aided flame spread over
wood slabs in the ceiling configuration. This study covers the three external conditions
that control the flame spread rate: wind speed (U..=0.2 to 1.5 m/s), oxygen concentra-
tion (Yo..=o.233 to 1.0), and external radiation (q"=0.5 to 1.3 W/cmz). The objectives
of this work are to (i) provide a physical understanding that will serve as the basis for
the development or refinement of theoretical models; and (ii) provide measurements for

chaning materials like wood.

4.1. Results

 

4.1.1. Temperature measurements

Figure 2.11 shows the schematic of the wind-aided flame spread in the tunnel
over a ceiling mounted sample of wood. The flame spread rate depends on how high
is the surface temperature at the end of the preheat (i.e. prior to ignition). For a certain
prescribed external radiation, both the surface temperature and the char layer thickness
increase in time. A high external radiation with a long preheat time will bring the
sample surface temperature close to its ignition temperature. This makes the flame
spread rate approach, at a critical point (q"(U,,, Yo... tww), that of the premixed-gas
flame propagation (flash through). In addition, the preheat time has to be chosen such
that the surface would undergo only a minor change due to external radiation. There-
fore, after a number of experimental trials, a 600 sec preheat time was chosen for low
external radiation (0.5 and 0.9 W/cmz) and 300 secs for high external radiation (1.3
W/cmz). Under these conditions, as shown in Figures 4.1 and 4.2, the surface tem-
perature nearly equilibrates with the surface heat loss; and its rate of change with time
becomes almost zero -i.e. prior to ignition. Thus, any changes in the surface tempera-
ture due to external radiation, during the flame spread process, can be ignored as
shown in Figures 4.1 and 4.2. These temperature histories also show that the external

radiation is uniform along the wood sample surface as discussed in section 2.1.4.

62

 

800

700~ T25
92

6004 t

500—

400~

Surface Temperature ( K)

  
 

 

 

.300 r .

)""_._‘.’_é_ -;
—T w
0 200

I ' I I
400 600 800 1000

 

Time (sec .)
Figure 4.1 Measured surface temperatures and their rate of change with time for wood
during flame spread (U..=0.6m/s. Y,.=0.233, Q"=05W/cm’).

600 .1-..“ —-- ___-.. --, _

 

500—

4004

EGO-i

Temperature (°C)
M
O
O
L

 

 

r v I I 1 ' I I I’ ' I r

100 200 300 400 500 600 700
Time (sec)

Figure 4.2 Measured surface temperatures for wood during flame spread

(U..=O.9m/s, Y,.=o.233, ('I"=l.3W/cm2).

63

The effect of wind speed on the flame spread rate was investigated for all these sets of
external radiation and preheat time. In studing the effect of the oxygen mass fraction
on the flame spread rate, however, a low level of external radiation (0.5 W/cmz) and a
large preheat time (600 secs) were used. This low external radiation was selected so
that a clear flame spread process is observed at high Y0... and to facilitate comparison

with PMMA (which burns without external radiation).

4.1.2. Pyrolysis and flame fronts

As previously mentioned in Chapter 3, visual observations led to the conclusion
that the peak rate of change in the surface temperature after ignition occurs at the
instant the flame tip arrives at the thermocouple location (Figure 4.1). Thus, the flame
tip location, x f, and the pyrolysis-front location, xp, may be determined as a function
of time from the measured temperature profiles as explained in section 3.1.1. The
methodology is illustrated on Figure 4.1. Results of such calculations agree exactly
with x f determined from video records. The pyrolysis temperature is taken as the
piloted ignition temperature (375 °C), at which the solid begins to vigorously pyrolize
and contribute fuel to the flame. Results of such calculations agree exactly with xf

determined form video records.

The three steps of preheating can now be distinguished during the flame spread
process. First the preheat by external radiation, then, after ignition, by the post-
combustion gases, and then by the diffusion flame extending downwind of the burning
zone. This extended flame is caused by the excess pyrolyzate. It is worth noting that
the surface temperature at the flame tip arrival ranges from 300 to 340°C, which is
higher by almost 120°C than at the conclusion of the preheating by the external radia-

tion. This indicates that there is major preheating by the hot gases ahead of the flame.

Figures 4.3 to 4.6 show both the pyrolysis-front and the flame-front histories,
after ignition, for different conditions. Both xP and xf have an almost exactly linear

dependency on time and are very close to each other in all cases.

 

Xp 8c Xf (cm)
(14
O
L r

N
O
l

O
l

 

 

 

O r r ‘ I '
600 700 800 900 1000

 

Time (sec)
Figure 4.3 Pyrolysis and flame front histories (Y o..=0.233, q"=O.SW/cm 2).

 

70
a 0.6
i C] 0.9
50— A 15

204

 

 

10 ' I T I ' I ' ”I ' I ' I T I ' I '
600 610 620 630 640 650 660 670 680 690

 

 

Time (see)
Figure 4.4 Pyrolysis and flame front histories (Y 0..=0.233, q"=0.9W/cm 2).

65

 

 

 

 

 

60
50—
A J
E
\o/ 40-
L.—
X
*5 304 0,,
XQ (m/S)
20— 9
‘ c: A 1.5
CI 0.9
O 0.5
10'I'I'I'I'T'F'I'I'I'
330 332 334 336 338 340 342 344 346 348 350

Time (sec)
Figure 4.5 Pyrolysis and flame front histories (Y,..=o.233. q"=l.3W/cm 2).

 

    
 

 

 

 

60
’0
50%
E
\0/ 40—
>;- .
(’8 3Oe Yo
O.
X
v 1.0
204 0 0.61
A 0.43
i' a 0.32
o 0.233
10 "'I"'I"'1"'I+j'
620 660 700 740 780 820

Time (sec)
Figure 4.6 Pyrolysis and flame front histories (U..=O.9m/s. Q"=0.5W/Cm 2).

66

The excess pyrolyzate (fuel that is n0t burned in the burning zone 0 < x < X?) causes
the flame front to be always ahead of the pyrolysis front. As the ambient oxygen mass
fraction increases, oxygen rich case, most of the fuel is consumed in the burning zone
causing the excess pyrolyzate to drop. Therefore the distance (Xf-Xp) should decrease
with increasing the ambient oxygen mass fraction Y0... The data shown in Figure 4.6

(as well as in Figure 3.4 for the case of PMMA) certainly agree with this hypothesis.

4.1.3. Spread rate

Once x f and x p are determined as a function of time, the pyrolysis-front speed
V p (defined as dxp/dt) and the flame-front speed V f (defined as dx f/dt) are obtained
from the slope of the least square fit line. Data for only the first 0.5 m was considered
to minimize errors due to changes in the free stream velocity. Figure 4.7 shows Vp
and V f plotted against the ambient oxygen mass fraction Y 0,, where both the free
stream velocity and the external radiation were kept constant at 0.9 m/s and 0.5
W/cm 2, respectively. For these environment condition, both V p and V f show a nearly

linear dependency on Y 0...

As discussed in Chapter 3 for the flame spread study on vaporizing solids, both
the decrease in combustion efficiency xA and the increase in the flame radiation cause
the flame temperature to be much less than that of the adiabatic case as the ambient
oxygen mass fi‘action increases. This is also true for charting solids. Brighter flames
(higher radiation) and soot deposition on the walls were observed as Yo“ was
increased. In fact, these two parameters are related since the brightness of the flame is

due to the hot soot particles. The data shows that Vp ~ YJJ.

67

.AnEo\3n.ourv £256an 5:25 ”8.: 5908

 

 

588 out 05 so 33% =8: 25: can $32.8 05 .o 8:85qu he 23E
000%
04 md 0.0 v.0 Nd

— _ _ _ _ . _ _ 00.0
Led
Imod

m; a

> D

 

 

no.0

/\

d

(S/w) JA 78>

68

Figure 4.8 shows that both Vp and V f sharply increase with the external radiation
for all the free stream velocities. Therefore, if the wind speed and the preheat time are
kept constant, increasing further the external radiation would result in a "flash-through"
at some critical point (q;(U,., Yo“, tpnhag). This very rapid flame spread would
occur when the surface temperature, at the end of the preheating -i.e. prior to ignition,
is at or above the ignition temperature of the solid (375°C). Figure 4.9 shows the
linear dependency of both Vp and Vf on U“, for all (1". Figures 4.7, 4.8, and 4.9 show
how close the flame spread rate Vf is to the pyrolysis-front spread rate Vp for all

experiments.

It is interesting to see the effect of the wind speed on the flame spread rate as
shown in Figure 4.9 where the flame spread rate for U..=0.2 m/s and q"=1.3 W/cm2 is
almost equal to that of U..=1.5 m/s and q"=o.5 W/cm 2. However, the effect of Y0,
can be seen when noting that Vp for Yo..=1.0 with q"=0.5 W/cm2 (Figure 4.7) is very
close to that of Yo..=0.233 and q"=0.9 W/cm2 (Figure 4.8). This shows that the flame
radiation for high Y 0.. may be thought of as external radiation. Stable flame spread
below 1.3 W/cm2 for 0.2 m/s was not possible.

4.1.4. Species production rates

The species production and depletion mass rate histories are presented in Appen-

dix C. A typical species production rate history during the three processes namely:
preheat, flame spread, and extinction, is shown in Figure 4.10 (U..=0.6 m/s, Yo“:
0.233, q"=1.3 W/cmz). Since in all the experiments performed, the total unburned
hydrocarbons (THC) and CO were negligible, only the major chemical species (pro-
duction of CO 2 and H 20, and depletion of O 2) mass rates are plotted versus the
pyrolysis-front xp and are presented in Figures 4.11 to 4.14 for all the conditions stu-
died. The data for xp<12 cm is not included because it is expected to have large
errors due to the ignition process. These figures are plotted in log-log scale to show

the relationship between the mass production rate and the pyrolysis-front.

69

 

 

 

 

   

 

12 . . . r . . . . 4 , . , . .
G—eU=1.5m/s
ia—aU=O.9m/s
IO—A—AUO _
v 00
A s
Q 8~ _
E _ .
8/
a. 6" _
> -i
025
‘0'- 44 _‘
> -I
2_. —4
04.,...,4..,...,.e
0.4 0.6 0.8 1.0 1.2 1.4
(1"(W/Cm2)

Figure 4.8 Dependence of the pyrolysis and flame front speeds on the external
radiation (Y...=0.233).

 

 

 

 

12
.J
104
Q 8-4

E
8,

0. SJ
> d
«=25
‘+— 4“
> .1

2—
0
0.0

 

Figure 4.9 Dependence of the pyrolysis and flame front speeds on the free stream
velocity (Y ...=0.233).

70

OI»

 

basin—ta £89....» £8538 Bee 8 82%
063 can??? do 88:9 m 05 mama—e roam: 83.. couoseoa «£0on

Aommv 0E;

 

 

8... use...

0

food

 

Iwod

IN P .6

T0 F .O
H cozoczxm emopwchw Second r

 

 

_ _ u _ e p . b p _ . b . l— .

 

(8/6) 9105 UOllOnpOJd ssaw

71

 

 

 

 

 

A 10.0
CD
\ 8.0‘
on
V 6.0-4
N
O
X 4.0'i
B
a?

2.0-3
C
.9
U
8
8 1.0“
Q_ 0.8“
3)) 0.6“ [3‘8 H20
0 0‘0 OZ‘dei.
E W CO

0.4 I I T I I .7 I I

10 20 40 60 80 100

Pyrolysis Front Xp (cm)
Figure 4.11 Dependence of the species mass production rates on the pyrolysis-front
location for different wind speeds (Y ..=0.233. q”=0.5W/cm 2).

 

 

 

 

 

20

A

m

\

3

N TO-l

O

.— 81

x

.9 6~

0

CK

c 4T

.9

23'

3

.0

e 2..

0.

[3'8 H20

0')

8 G—OOz-depl.

V—V CO
2 1 l l T 2 l
10 20 4O 60 80 100

Pyrolysis Front Xp (am)
“Inn 4.12 Dependence of the species mass production rates on the pyrolysis-front
location for different speeds (Y...=0.233, d"=0.9W/cm’).

72

 

 

 

 

 

1,7 :50
E
If . ‘. 0.9

(\l 20~ 1’ £933

0 o

X

.33

O

m 0.9
C

.9

73

3

.0

O

L.

Q.

5)) 4- 8—8 H20

g 9-9 02-depl.

3 V—V C07
l I I I
10 20 40 50 80 100

Pyrolysis Front Xp (cm)
Figure 4.13 Dependence of the species mass production rates on the pyrolysis-front
location for different wind speeds CY...=O.233, (1"==l.3W/cm 2).

 

100
80q

60% 0,
4o«

20—

 

2" B—E] H20
' " G‘Q 02-depl.
- V47 CO;
I I I —[ F
10 20 4O 60 80 100

Moss Production Rote XlO2 (g/s)

 

 

 

Pyrolysis Front Xp (cm)
Figure 4.14 Dependence of the species mass production rates on the pyrolysis-front
for different oxygen mass fraction (U..=O.9m/s. ti”=0.5W/cm 2).

73

The chemical species mass rate are related to the total quantities over the entire bum-

ing zone by

19.0)
m (t) = I ah (x,t)dx
x=0

with the local mass flux being higher near the pyrolysis front due to low char depth.

As the fire spreads, the mass rate increases roughly proportional to the pyrolysis-
front. When the pyrolysis-front reaches the end of the sample, the mass rate slightly
increases before reaching a plateau, then sharply drops (see Appendix C). The duration
of this plateau depends on the environment condition, it increases with increasing the
external radiation or the oxygen mass fraction. Therefore a thick char layer is
expected in these cases. The drop in the mass rate, however, is due to (i) the local
mass flux drop as shown in Figure 1.3, and (ii) the propagation of the extinction-front
(flame foot) behind the pyrolysis-front as observed during the experiments. However,
the flame does often flash back, as observed in the experiments, when some fuel has

occumulated enough to support a premixed flame upstream of the extinction-front.

The flame stand-off distance from the combustible surface in the ceiling
configuration mainly depends on the buoyancy, the wind speed, and the ambient oxy-
gen mass fraction. The buoyancy causes the flame (high temperature zone) to rise.
The wind speed controls the oxygen mass transfer rate to the flame. And the ambient
oxygen mass fraction controls the stoichiometric plane (or the flame) stand-off distance
from the surface and the flame temperature. The lower Y 0,, is, the higher the flame
stand-off distance and flame lenght are. Therefore, a combination of high wind speed
and high oxygen mass fraction would result into a short flame stand-off distance from

the surface, which in turn increases the net heat flux at the surface causing higher mass

flux.

The data shows that the mass production rates of the major chemical species
increase with U... Y0... and (1". This is expected since the flame approaches the sur-

face with increase in U... the flame temperature increases with Y0... and the pyrolysis-

74

front propagates deeper into the solid with increase in (1".

For complete combustion, the production rates of CO 2 and H 20 and the deple—
tion rate of O 2 should be related to the total fuel production rate between x=0 and
x=xp (the burning zone). For high Y 0... due to soot formation and deposition of heavy

hydrocarbons on the tunnel walls, the depletion rate of O 2 is not expected to follow

the production rates of CO 2.

As discussed in the previous chapter for PMMA, for the local mass flux to vary
as x435, the total fuel production rate at any instant must vary as x“. However, the
data shows that this power is about 0.4 at Yo..=0.233, and increases with Yo... up to 1
at Yoo.=l. The steady-state solid-phase can not be addressed in the charring solids like
wood due to the char build-up which continuously attenuates the production of the

pyrolysis products as shown in Figure 1.3.

Using the one step chemical reaction described in chapter three, the ratio of the

production rate of CO 2 to the depletion rate of O 2 is (equation 3.1)

tilC02 44x

mm: ’ 8(4x+y—2)

 

Where the chemical composition of the volatiles for poplar wood is shown in Table
4.1 taken from Atreya (1984). The above ratio should approach the value 1.25 for the
complete combustion case (Y 0.50333)- Figure 4.15 shows meg/mm? normalized
with its mass based stoichiometric fraction, plotted versus xp. Most of the room air
oxygen concentration experiments are close to 1 (complete combustion). However, as
Y 0.. increases, the normalized ratio {nah/mm2 drops to an almost 0.4 at 100% oxygen
environment. This indicates that the chemistry has changed as Yo. is increased. Car-
bon in the fuel is converted to unburned soot leading to lower mm: as discussed in
section 3.1.3.2. This agrees with the physical observation that both larger and flames
at higher Y0. are brighter. Figure 4.16 shows a picture of the flame taken during the
flame spread experiment (U..=O.9 m/s, Yo..=0.233, £1"=0.5 W/cmz).

75

 

 

 

 

 

 

 

 

 

 

1.2

. um You
1.0~ (m/s)

g g ”*9 . e 'o 80”” V 0.9 1.0

8 0'8 , u‘W- 0| 12°?" ' V 0.9 0.61

E ,2,” “U..-..‘ A 0.9 0.43

\ O 6“ V V v V O 0.9 0.32
8‘ VV"""" ao.90.233
o v ' v A 0.9 0.233
' ea 06 0.233
E04“ ”""H"" 00.60.2315
0 0.6 0.233
x 0.2 0.233
0’2 v 1.5 0.233
v 15 0.233
0 0.9 0.233

0-0 l 1 v r r l . , fire—T—fiefi *i

0 to 20 30 40 50 60 7O

PyrolySis Front XD (err)

Figure 4.15 Normalized moo/mm with the corresponding mass based
stoichiometric fraction (i.e., the value of 1 corresponds to complete
combustion).

Figure 4.16 Picture of a spreading flame on wood sample (U..=O.9 m/s. Yo..=0.233,

Q"=0.5W/cm2).

q
(W/cm’)

0.5

0.5
0.5
0.5
0.9
0.5
1.3
0.9
0.5
1.3
1.3
0.9
1.3

76

Table 4.1 Empirical formula of Poplar wood [Atreya 1984].

 

 

 

 

 

 

Char Empirical Lower Heat of
Yield Formula Combustion (KI/gm)
Poplar 0.33 (avg) C H O 19.33
1.66 2.43
Char - C H O 26.86
5.44 3.12
Volatile - C H O 15.62
0.99 2.31

 

 

 

 

 

Figure 4.17 shows, in chronological order, pictures of the flame as it spreads (U..=O.9
m/s, Yo..=0.61, £1"=0.5 W/cmz). The flame color in these pictures certainly indicates
the increase in the flame radiation with Yo... Figure 4.18 shows ducal/map, normal-
ized with its mass based stoichiometric fration 2 (according to Table 4.1), for all the

cases studied. The data shows that all the experiments fall close the value of l.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(4)

Figure 4.17 Instantaneous pictures taken of a spreading flame on a wood sample
(IJ..aO.9m/s. Yo..=0.61, q*=o.SW/cm’).

78

A8255

339:8 9 388:8 _ no 32> 05 6.5 SE oEoEoEuBfi

n. 2.0 No x
nd nmd 90.0
md nNd od D
3 Rd so a
nd nNd md Q
no no no a
n; mud ad 0
no as m... >
3 8.0 3 e
no and no e
0.0 n¢d ad I
no 5.0 no 9
no 3 no >

peoba Am};
..a so» a:

83.. age gaseous 2.. 55 one? 33oz

AEoV ax E8... m_m>_o._>n_
on on on om 9. on om

'P _ . _ p . _ . _ . _

 

 

 

§ a... 83E
S
0.0
flmd
i¢.o
rod am
r 0
Z
/
w
H
Z
O

 

 

CHAPTER FIVE

Convective and Radiative
Heat Transfer to the
Solid

During wind-aided flame spread over combustible solids, radiative and convective
energy transfer occur from the flame and the hot gases to the as-yet-unburned solid
surface. The rise of the solid surface temperature from ambient temperature far down-
stream of the flame to the ignition temperature at the pyrolysis front is due to (i) the
preheating downstream of the flame tip x>xf and (ii) the preheating underneath the
flame xp<x<xf (the overfire region). Therefore, the heating of the surface to its pyro-
lysis temperature in the overfire region depends on the surface temperature at the flame
tip arrival (x=xf). The higher is this temperature, lesser is the heating job needed from
the flame in the overfire region, and hence the shorter the distance (xf-xp). During the
flame spread experiments, the surface temperature at xf ranged from 573 K to 613 K
for wood and 533 K to 573 K for PMMA. It was noted from the experiments that this
temperature increases and the distance (Xf-Xp) decreases with Yo, , which supports the

above hypothesis. This latter point has also been reported by Ahmad and Faeth (197 8)
in their upward flame spread model.

79

80

Clearly preheating downstream of the flame tip cannot be neglected, it has to be
well understood for proper and realistic theoretical models, since in an analytical solu-
tion, the heat flux in the preheat zone is required to develop the flame spread model

[Femandez-Pello 1978]. Heat transfer underneath the flame is also equally important.

5.1. Formulation

 

A very interesting experimental observation leads to the development of a simple
model. It was observed in all the flame spread experiments in the ceiling
configuration, (that is for all free stream velocities and ambient oxygen mass fraction
used), that the flame lies just underneath the sample at an almost constant stand-off
distance all along the flame, from the attachment point to the flame tip. Gas-phase tem-
perature measurements (discussed in section 5.2.2) show that the flame stand-off dis-
tance changes by almost 2 mm from near the leading edge to near the flame tip when
xf=254 mm. Hence, the problem can be justifiably treated as that of a constant free-
stream velocity flow along a semi-infinite plate with an arbitrarily specified surface
temperature (Figure 5.1).

i

To. xf Tf

 

Figure 5.1 Schematic diagram of the model.

81

The surface temperature can then be considered as
T5 = TO. K < 0
T5 = T1: 0 < x S xf
Ts = F(x) xf < x

Thus, the heat flux from the wall surface in the preheat zone is [Kays and Crawford
1980]

II
a

dTs i=k
q = h(z.x) — + 2 h(zi.x) AT... (5.1)

dz i=1

05—.”

where h(z,x) is the local unit conductance from the single-step function solution,

defined as:
0.332 k Prm Re 1'2 3’ “’3
h(z,x) = g x " 1- :7;- (5.2)

 

There exists two discontinuous surface temperature: one at the leading edge and
another one at the flame tip location Xf. This will result in two terms in the above sum-
mation. In the integral term, however, the surface temperature gradient with down-
stream distance is zero for x<xf. For x>xf, the integral can only be solved if the sur-
face temperature is known. Since the resultant integral term in the preheat zone is only
a correction to the computed heat flux at the surface, and especially that dT/dz is only

significant near the flame tip, the temperature is assumed uniform (F(x)= a,,) in the rest

of the analysis.

Hence, the heat flux to the surface in the preheat zone (x > xf) takes the form

__ 0.332 kg Pr”3 Real/2 I z 3/ -1
q — x \Tf‘Tw) -1+ 1- ; (5.3)

 

01'

Nux 2 34 -1
W = 0.332 -1 + 1 - (x )- I] 3] (5.4)
ex

82

where Nux = q” x/(Tf - T...) kg and x*=x/xf is the related distance from the flame tip.

5.2. Results and discussion

5.2.1. Transient heat flux measurements

Figure 2.12 shows a schematic of a spreading flame on PMMA sample upsu'eam
of the ceramic detectors used in this study. The procedure to compute the heat
transfer modes from the transient measurements of the solid and gas phase temperature
is as follows. Inverse heat conduction calculations were used to obtain the incident
heat flux as a function of time [Beck et al. 1985]. The heat flux algorithm used in the
inverse heat conduction analysis is sequential, where few future temperatures, associ-
ated with future times, are used to compute the heat flux at every instant. The main
aspect of this algorithm is the use of the least squares criterion to estimate the heat
flux at the surface qm from the measured temperatures Y1, ...Yj , at times t‘“, rm”,
...,t‘“”H. The parameter r is the number of future time step and J is the number of
thermocouples. The criterion is to minimize S with respect to qm

K I +' l m+i-l 2
S = §§[ij r- _ Tj ] (5.5)
1: j:

where ij+H is the measured temperature at the j‘h sensor and at time t’m'i’l, and
iji'1 is the corresponding calculated temperature. The convective heat transfer rate

from the surface was computed from the temperature gradient at the solid-gas interface

as
tic.” = - k — (5.6)

where its is the gas thermal conductivity evaluated at the surface temperature. In the
transient flame spread experiments on PMMA, four thermocouples were maintained in
the gas-phase at some known distances from the surface. This small number of pOints

makes it difficult to rely with confidence on the curve fit results from which the

83

gradient is computed. Thus, the temperature gradient for these experiments is approxi-
mated as dT/dy = (Tg—Ts)/Ay, where 'I‘g is the gas temperature at distance Ay from the
surface. This is a good approximation since the thermocouple used is very close to the
surface (Ay<<8,, where St is the thermal boundary layer). First, the thermocouples
were constructed in such a way that measurements error due to heat conduction in the
thermocouple wires are reduced. Then, one thermocouple was placed about 1 mm from
the surface. An error of 0.1 mm on this distance causes a 10% error in evaluating the
heat convection. Therefore, the distance y has to be accurately found for every experi-
ment. At far distances from the flame, x/xf>>1, the radiation from the flame to the
surface located at distance x is negligible, then an energy balance at the surface leads
to err" = as".

the total and convection heat fluxes using all the data points N for x/xf>2

Hence the distance Ay can be well estimated using least square fit of

N 2
kg 2‘; ATi
Ay = N ‘= (5.7)
2 ATj (113"
j=l

 

This distance is then used to determine the convected heat flux during the whole
experiment.

The radiative heat flux from the surface was computed from the knowledge of the
surface and surrounding temperatures and the surface emissivity. Finally, the flame

radiation was computed by applying an energy balance at the solid- gas interface.

In order to compare and correlate the results of different conditions, the data has
to be plotted vs. x'=x/xf, which is the related distance fiom the flame tip. However, to
get the heat flux a function of x' for every experiment, the time dependent tempera-
tures and flame tip have to be curve fitted by a very low filter and outputcd with the
same time increment. Figure 5.2 shows such curve fitting for the gas-phas tempera-
ture. The different modes of heat transfer were then computed and are shown in Fig—
ures 5.3 (U..=O.9m/s, Yo..=0.233) and 5.4 (U..=O.9m/s, Yo..=l.0) and Appendix D.

84

ASE?» .Eommnx

. agate» .m>=o.cnre 5233 .5 32% 0:3: mats—o 2382—89

 

 

829mm» 3.5 256 05 new 3532: 05 no __SEQEOU «d 95$..—
Aoomv cc: 5
ON P 00 P om om ca ON 0
_ — p L _ _ . b . — F b O
icom
ioov
fix \. loom
x x
xxx . .x x
x .03 X -
x E l.

 

com

 

(Q) dine),

Q”(W/CN3)

¢'(W/cw3)

85

 

 

 

 

 

Figure 5.3 Heat flux measurements during flame spread on PMMA
(U..=O.9m/s,Yo..=O.233).
3.5 1
.4 3K cond
K
O conv
30 fl x [:1 (Cd
‘* ‘A re—rod
2 .5 48>:

 

 

 

X/Xf
Heat flux measurements during flame spread on PMMA
(U..=O.9m/s, Yo..=1.0).

Figure 5.4

86

In all the 21% oxygen experiments, the flame color was blue (see Figures 3.7 and
4.16), which indicates low flame radiation to the surface ahead of the flame. This is
confirmed by the results in Figure 5.3. The total heat conducted in the solid equals the
heat convected by the hot gases. Both radiation from the flame to the surface ahead of
the flame tip and the surface radiation to the surrounding are negligible. However, as
the oxygen mass fraction increases, the flame becomes brighter (see Figtnes 3.7 and.
4.17), which indicates a significant radiation to the surrounding. This observation is
confirmed by the results shown in Figure 5.4 for the case of 100% oxygen. As the
flame tip approaches, the radiation from the flame to the surface becomes more
significant. It was also observed that the deposition of soot on the walls increases with
increasing oxygen mass fraction. This introduces an error of the gas-phase temperature
measurements very close to the flame tip for high oxygen mass fraction, which in turn
under-estimates the computed heat convection. Again, this is only true for high oxygen

mass fraction and very close to the flame tip.

Since the energy lost by the flame to evaporate the fuel from the pyrolyzing sur-_
face [Cpflf-TproJ is not considered in the model (the flame is on the surface), then
the following adiabatic flame temperature is used in correlating the results

f Yo.
(Tr-T...) = —C;— (AH-Q) (5.8)

where Cp, the gas specific heat, is computed at the average temperature by the integral
7:

j Cpm dT = f Yo. (AH-Q) . (5.9)
T.

This makes the flame temperature go nearly (because Cp increases with Yo.) linear
with Yo... The flame temperature is then over—estimated because the real Tf is reduced
due to incompleteness of combustion and flame radiation. As discussed in both
Chapter 3 and Chapter 4, it was found that complete combustion holds only for
ambient air condition. As the oxygen mass fraction increases, the combustion process

becomes less efficient (xA decreases with Ya). In addition, the flame gets brighwr as

87

Yo. increases, (xR increases with Y”). Clearly, these two factors reduce the flame

temperature more significantly as You increases.

To approximate this incompleteness of combustion and radiation losses depen-
dence on Yo”, the convective heat flux measurements for different oxygen mass frac-
tion were normalized with (YOJO.233) “. This normalization makes the flame tempera-
ture adiabatic at ambient air condition, and less efficient as Yo. increases. The power
n was then approximated by using least square analysis on all the results. Expressing

the normalized convective heat flux measurements as in equation 5.4, the following

equation results

N
u" 1“ = (2 :6m (5.10)

Y
us 112 _2;
P’ R“ [0233

where C = 0.5, m = -2.3, and n = -0.25. The gas properties used in the above equa-
tions are evaluated at the film temperann'e Tm=(TrI-3T.)/4. The experimental results
normalized according to the left hand side of the above equation for all the conditions

tested are presented in Figure 5.5. In the same figure, the theoretical results of equa-
tion 5.4 are also presented.

The flame radiation, however, is treated as in the case of an expanding plane in
the flame spread direction with known temperature and emissivity and at a constant
stand-off distance from the surface (Figtue 5.6). The view factor is then determined
utilizing contour integration (Stoke’s theorem) and is

1 Xi" _ a
Fat-2 = '1; 31m 1[ ]

L\](xf--x)5+dz \Rxf—x): -I-dz

 

 

 

 

 

 

a -1 X -1 x_xf
+ tan -tan
V 82+d2 L ["1 a2+d2] [ Va2+d2]

 

 

+ x ‘1 a 5.11
W” [m] ‘ ’

88

 

Ln (Nux/PrmRe}’2(Yo../0.233)‘0'25)

 

 

U X
m/s cm

Theory

0.9 48
0.9 32
0.9 48
0.9 32
0.9 32
1.5 48
1.5 32
1.2 48
1.2 32
0.9 32
0.6 32

 

O$DEBI>DOQ<G+|

 

_J , ‘

0.0 0

Figure 5.5

      
 

' I ' I ' I ' I ' '
.2 0.4 0.6 0.8 1.0 1.2 1.4

Ln(X/Xf)
Nondimensional measured and predicted convective heat transfer during
flame spread on PMMA.

X Ceramic detectors

 
   
    
   
      

  

 

e5)

/

 

Water- 8 " . . ’ 0 ' ,
cooled plate ’ \ »
\

Radiometer

Figure 5.6

Schematic diagram of an expanding plane underneath the sample
surface.

1.0
0.61
0.43
0.43
0.31
0.233
0.233
0.233
0.233
0.233
0.233

89

where a = W/Z (W is the width of the sample) and d is the flame stand-off distance.

Using the adiabatic flame temperature, as in the convection analysis, and the
radiometer measurements taken at the bottom of the test section, the flame emissivity
is approximated. The flame radiation is then computed and is shown in Figure 5.7
with the experimental results. The predicted radiation is much higher than the experi-
mental data near the flame tip due to the adiabatic flame temperature asumption. Both
experimental and predicted results, however, show that (i) radiation can be neglected
near room air conditions (ii) and as the oxygen mass fraction increases, the radiation
becomes significant only underneath and very near the flame tip. For the case of
100% oxygen experiment, the radiation and convection are 24 and 76 percent of the

total energy transfer at the flame tip, respectively.

 

 

 

5 ‘
X 0.233
'1 O 0.33
C 0.43
4‘ + 0.61
V 1.0
A
N
E 3‘
U
\ .4
3
v
: 2‘
0..
1'1 v
v
+ v V
. + . ‘~-.___ h.“ ' V
0 ___, . .-_-.._¥ ‘——-—»_. h — .. -. .
1.00 1.02 1.04 1.06 1.08 1.10 1.12
X/Xf

Figure 5.7 Measured and predicted flame radiation for different Yo...

90

5.2.2. Steady-state heat flux measurements

To simulate the heat transfer from a diffusion flame in the overfire region (due to
excess pyrolyzate) during an actual flame spread. a long, steady and non-sooty flame
with low fuel flow rate and 19% free stream oxygen concentration was maintained in
the boundary layer over ceramic detectors for all experiments. Figure 2.13 shows a

schematic of the porous-metal burner diffusion flame used in this study.

Both the net heat flux into the solid and the surface radiation to the surrounding
are calculated as explained in the previous section and are presented in Figure 5.8 with
the total heat flux (q"m+q"md) (U..= 1.2m/s, xf=10 in, x= 7in). It is clear from this
figm'e that the gas-phase achieves steady-state relatively quickly. Hence, the gas-phase
thermocouple can be traversed at any instant after 300 seconds. The measured gas-
phase temperature profiles (after thermocouple radiation correction) are shown in Fig-
ure 5.9 for the case where U..=1.5 m/s and xf=10 in. The convective heat flux is then
computed by calculating the temperature gradient at the solid-gas interface, and using
air thermal conductivity evaluated at the surface temperatme. The combustion plumes
in Figure 5.9 show that the flame stand-off distance from the surface slightly decreases
in x: 2 mm in 152 mm distance. This certainly proves the validity of the above simple
model. Figure 5.9 also shows the flame temperature decreasing in x. This shows that
the presence of a cool wall inhibits complete combustion of the excess pyrolyzate by
quenching the reaction near the flame tip [Groff and Faeth 1978] in addition to the fact

that the rate of reaction in the flame decreases as the fuel concentration at the wall
decreases [Ahmad and Faeth 1978].

Temperature (°C)

1.4

91

 

 

 

q" (W/sz)

 

 

 

 

 

 

  
 

 

 

 

G-O rodiotion
A—A incident
0%) total
0.0“ 1‘10 1 r 1 r l v T 1 1 r I r i r l r
0 200 400 600 800 1000 1200 1400 1600 1800
Time (sec)
Figure 5.8 Net, total and surface radiation heat fluxes histories
(11.31% 1932546111, x=17.7cm).
1600
n/
- O l
/_
1200~ //8 \°\°
I
)4 o
/D
4 \
800-1 \o:
. 0* \0
°\
°\ .\
, °\ \ ..
400.4 Vf‘KN \ \ W 16in
'\\ \ OR o 11in
/ '\ \ \ C] 9in
\v\\o ‘ \O o 7in
§' 0 .
:34 \Q o x/O 3K 5m
0| 'r' I'r I""rx'*"
O 5 1O 15 20 25
Y (mm)
Figure 5.9 Measured

gas-phase temperature profiles of a methane diffusion flame

(U..=O.9m/s, Yo..=0.233, x,=25.4cm)

92

The flame radiation is then computed by performing an energy balance at the sur-
face. Figure 5. 10 shows the convective and flame radiative heat fluxes for all experi-
ments performed for different free stream velocities and flame lenghts. As expected,
for this non-sooty flame, convective heat transfer is dominant both ahead and under-
neath the flame. Radiation from these flames is very small and can be ignored for this
case. This is consistent with the conclusions drawn in the previous section for room air

oxygen concentration condition conditions.

Figure 5.10 shows that the convective heat transfer is decreasing in x both under-
neath and ahead of the flame. However, to get a better insight on the dependence of
this energy transfer on the variables studied in this work (U ., x, Xf), the experimental
results have to be correlated. Using the same correlation developed in the above sec-
tion, equation 5.10, the data ahead of the flame tip, where the formulation is valid, did .
fall together for all cases studied (Figure 5.11) and drops as x“? where p=2.2 (obtained
from besr fit), or it” ~ x‘2'7. This proves that the combination of the heat transfer"
parameters (Nux, Pr, Re.) are only a function of the related distance (x/xf) as shown in
equation 5.10. This clearly justifies the formulation. These results of methane
diffusion flame are in good agreement with those of flame spread on PMMA in the
previous section. Ahmad and Faeth (1979) found p=2.33 in their upward turbulent
natural convection fires. This is consistent with the distribution measured by Quintiere

et al. (1986) who found p=2.4 in their upward natural convection flame spread over
different combustible materials.

Eventhough the formulation is not valid underneath the flame, the data does
correlate when subjected to equation 5.10 and is approximately constant with x/xf (or *
{1” ~ x‘o's). This result is in good agreement with Emmons’ boundary-layer steady-
state solution. Therefore, the combination of these parameters (shown in equation
5.10) can be justifiably used both ahead and underneath the flame.

93

 

 

 

 

 

 

 

 

 

 

1.4
U Xf
1.2- cm/s in
1-0‘ +—+ 150 10
I?“ + H 150 10
E 0.82 W160 8
\ . v—v 150 6
5 ea 120 10
- 0.6e 949120 10
b <><> 120 8
0.44 a—a 120 6
J A--A 120 6
E83 90 1O
0.2e [3E] 90 8
‘ $69 90 6
0.0 T, _ .- . GD 90 6
0.0 0.5 1.0 1.5 2.0 2.5 3.0
X/Xf
Figure 5.10 Measured convective and radiative heat fluxes underneath and
downstream of steady porous-metal burner diffusion flames
(Y...=0.211).
s U Xf
A —1.0-: . : . 0 ° .5 U o cm/s in
§ 3’ ' ° 3 '1 . 2°
3 -1.4~ 'XD
6! . + 150 10
Q8 . o o: v 150 10
v V 150 8
g: ‘1 8: to; v 150 6
.3, j o 120 10
8‘ . v o 120 10
._. -2.2-‘ ug , O 120 8
Q” j , A 120 6
2 _ A 120 6
v -25- v 33 90 10
5 4 El 90 8
j 0 63 90 6
O
-30 ' l j I T 1 ' l ' I 7 i ' i r l ' 2 O 90 6
~08 —0.6 -0.4 —0.2 0.0 0.2 0.4 0.6 0.8 1.0
Ln(X/Xf)
Figure 5.11 Nondimensional measured convective heat flux underneath and downstream

of porous-metal burner diffusion flames (Y...=0.211).

94

5.2.3. Surface temperature

To obtain the sample’s surface temperature during the flame spread, the one
dimensional heat conduction in a semi-infinite solid is examined. The solid is initially

at ambient temperature, then experiences a transient heat flux at the surface
95 = Ts-T, = 0 y20 t=0
d." = 51”(9,.T...t) y=0 t>0

Since the heat flux applied at the surface in equation 5.10 can be expressed as (xFVft)

b
61” = 10.) -t- 0 < 3— < 1 (5.1221)
‘1 if
it” = for) 1< i < i (5.12b)
tf it

(where tf and tp are the times xf and xp reach x, respectively, and b=2.3), the problem
can be solved using the Green’s funcrion or Duhamel’s integral [Atreya 1983]. The
surface temperatme is then expressed as

6 = —-§tf)—I(tb) (513)
‘ 1:41:ka ’ '
where
t Tb
I b = — . .14
0.) loft-‘1‘“ (5 )

Ahead of the flame (0<t/tf<1), I(t,b) is reduced to a Beta function, and the surface

temperature solution becomes

fltf) tb+ll2
A t,” m

 

(5.15)

where

A = I‘(b+3/2)
I‘(b+1)

95

A7311? at b=2.3. The surface temperature ahead of the flame finally becomes (is terms

of all the parameters)
T84.” Yo... 0.25
Tf-T” 0.233

The PMMA surface temperature measurements of all cases are normalized as in

 

 

 

1/2
IL] [_L—‘ C k Prl/é] = 0'5 (16")"2'8 (5.16)
U” 198%

the left hand side of the above equation and presented in Figure 5.12 (a knowlege of
the flame spread rate is needed to find the surface temperature distribution). The right
hand side of the equation is also shown in solid line in the same figure. During the
sample ignition process, heat is transfered from the burner diffusion flame to the sam-
ple surface downstream as indicated by the gas-phase temperann'e rise presented in
Figure 5.2 at x=33cm and 1mm from the surface. This heat transfer causes the surface
temperature to rise above ambient and becomes dependent on the downstream distance
x as shown in Figure 3.1 near 100 sec (end of ignition). Therefore, the normalized
measured surface temperature would be higher near the leading edge than far down-
stream for small xf (or large x/xf), as indicated in Figtne 5.12. At the flame tip arrival,
however, the measured surface at different locations varies within 5 degrees for the
same experiment and is between 533 and 573 K for all the experiments conducted as
mentioned in seetion 3.1.1. Therefore, the normalized measured surface temperature is
expected to vary at the flame tip arrival as indicated in Figure 5.12. Hence, the
predicted surface temperature is expected to be lower than the measured for large X/Xf
for the first reason mentioned above. Near the flame tip, however, the predicted surface
temperature is expected to fall close to the measured data that corresponds to low oxy-

gen concentration experiments where flame radiation is negligible (since the model

only considers convection).

96

6038388 coats <22m c.3235 Ba 3582: 33655652 26 cam...

 

 

 

 

u_x\x
0 ¢ m. N F
23 0 ad x _ 9 t t . _ . _ _ cod
. . w . 9...... £2.8qu
ammo m mo x x... J. tie. ram“... 4
nnmd 2 ad + H
n86 m N; e .
name m N. e Iofio
ammo m m; o .
83 m f e .
6.2.0 2 m2 0 .
3.0 o ad < -
3.0 m ad a Iomd
5o m: We 4
3.0 m ad 0 w..,_
3.0 m we a
Ed 2 me e and
o; 9 me o
COED—thOU ll
5 Am};
6» x 3
3.0

 

 

97

Underneath the flame, however, I(t,b) is reduced to an incomplete Beta function

which may be approximated by the series

2 3 4
_.£.m£_£L_ b it- b 2.. a E
I(t’b) - tfb [2 b+1 [I 4(b+2) [t] 8(b+3) [ t] 64(b+4) [I] J (5°17)

This series converges very quickly where the 5‘h term is less than 1%. At the

 

 

 

 

pyrolysis-front arrival (t=tp), the H112 dependency on the pyrolysis temperature is in
good agreement with Quintiere et al. (1986) results.

5.2.4. Flame spread rate

The pyrolysis-front spread rate Vp can be deduced from equations 5.15 and 5.17

P: CP 1‘8] U” [TFT~]2 You T05

p c k rp-r” 0.233

= 0.251)2 1

VP 1; Fri/3

(5.18)

 

 

where D is the series inside the brackets of equation 5.17 evaluated at t=tp. The limits
of D are 1<D<2, where 1 corresponds to (since tf/tp=Vp/Vf) Vp=Vf (all the heating is
accomplished in the preheat zone) and D—-—)2 when Vf>vp (all the heating is accom-
plished under the flame). Since Vf is unknown, D is approximated by fitting equation
5.18 to the experimental results and found to be 1.3. Once again, the air properties are
evaluated at the film temperature (T rl-3T..)/4.

Equation 5.18 reveals the pyrolysis-front spread rate dependency on solid and gas
thermal properties and environment conditions. Figures 5.13 and 5.14 show experimen-
tal and predicted Vp dependency on U... and Y0.” respectively. The flame spread rate
is, therefore, linear with the free stream velocity and is proportional to Y}: for
PMMA as given by equation 5.18. This dependency is in agreement with the experi-
mental results where vp was found to be linear with U... and is proportional to Ygfi.
Loh and Fernandez-Pello (1984) found Y3 dependency on their floor configuration

assisted flow flame spread experiments. However, it was concluded that reflection

98

5E—O4

 

4E—O4~

3E-O4‘ '

vp (m/s)

25—04a ‘

 

 

 

1E-O4 T r j I I r I l I I I I I—I I I r r r I r
(15 (17 (19 lJ 1.3 L5

Ua. (m/S)

Figure 5.13 Measured and predicted pyrolysis-front spread rate dependence on
U. (Yo..=0.233).

 

4E—03

3E—03-

2E-03‘

vp (m/s)

1E-03~

 

 

 

 

0E+OO I l I T T I f r I I I l I I r l I
(12 (13 (14 (15 (16 (17 (18 (19 ‘L0 1.1

You,

Figure 5.14 Measured and predicted pyrolysis-front spread rate dependence on
Y... (U..=O.9mIs).

99

from the surroundings, in addition to the buoyant flame radiation, to the sm'face ahead
of the flame tip is the reason behind this higher dependency on Yo...

Equation 5.18 is a typical wind-aided flame spread formula as for those derived in
previous theories [W ichman and Agrawal 1991, Carrier et al. 1990] except for the
difference in the constant coefficient. Wichman and Agrawal (1991) who employed the
Oseen flow assumption, have a constant coefficient of 1 (see equation 1.3). Carrier et
al. (1990), however, who employed the boundary-layer assumption from the onset and
later invoked the Oseen flow approximation, have a coefficient of 0.25. The high
coefficient in Wichman and A grawal (1991) derivation explains their high predicted
ratio VWP. This coefficient is therefore dependent on the pyrolysis zone part of the

problem which determines the flame length and consequently the relation between VP
and Vf.

CHAPTER SIX

Conclusions

The focus of this study was to understand the wind-aided flame spread on wood
and PMMA. This included understanding and determening (i) the pyrolysis-front and
flame-front spread rates and (ii) the major chemical species mass production rates. In
the process, it was necessary to determine the the convective and radiative heat

transfer to the solid both ahead of and underneath the flame.

The conclusions of this work are summarized in the following two sections. Sec-
tion 6.1 summarizes the conclusions derived from the flame spread experiments on
chairing and non-charrin g solids. Section 6.2 summarizes the conclusions derived

from the results of the heat transfer study.

6.1 Flame Spread Experiments

The wind-aided flame spread experiments were conducted on both chairing
(poplar wood) and non-charring (PMMA) solids mounted in the ceiling configuration.
The pyrolysis front and the flame front spread rates were determined from the
pyrolysis-front and the flame tip histories, respectively. The mass production rates of
the major chemical species were derived from the transient species concentration meas-

urements during the experiments.

100

101

Several significant conclusions can be derived from the experimental results:

1-

The pyrolysis-front and the flame-front are much closer to each other than that
predicted by theoretical models. This is also true for the pyrolysis-front and
flame-front speeds regardless of the free stream velocity, external radiation
(applied only for wood), and/or the oxygen mass fraction for both wood and
PMMA. This is because unsteady conditions exist in the burning zone.

The pyrolysis-front and the flame-front spread rates vary linearly with the free
stream velocity for b0th wood and PMMA. The theoretical models for non-
charring materials predict the pyrolysis-front speed quite accurately despite the
solid phase assumption in the burning zone. This underscores the fact that the

flame spread rate depends primarely on local heating of the solid by the flame tip

in the adjacent preheat zone.

The pyrolysis-front and the flame-front spread rates sharply increase with the
external radiation for wood.

The pyrolysis-front and the flame-front spread rates vary with the free stream
oxygen mass fraction as Yolg.‘ for wood and Yolg.‘ for PMMA.

The species measurements show that the pyrolysis mass flux is roughly constant

both for wood and PMMA rather than vary as x'05 as predicted by Emmons
theory. The species measurements also show that the chemistry is changing as :1,

and Y0. are increased because of incompleteness of combustion.

6.2 Heat Transfer Experiments

The heat transfer to the preheat zone ahead of the pyrolysis front in wind-aided

flame spread has been addressed experimentally and theoretically. Even though the

model is simple, it provided the necessary heat transfer nondimentional parameters to
correlate the results b0th ahead of and underneath the flame.

102

Several key results emerge from this study:

1..

The convection is the dominant mode of heat transfer for near room air oxygen

concentration condition both ahead and underneath the flame.

The flame radiation becomes more significant underneath and near the flame tip

as the ambient oxygen mass fraction increases.

The steady-state convective heat flux measurements agree with those of the tran-
sient case ahead of the flame tip. This shows that the gas—phase reaches steady-

state rather fast (compared to the solid-phase)

Correlations have been developed for the convective heat transfer, the surface

temperature and the flame spread rate.

Ahead of the flame, the convection is transient (q” a :23) and goes as {2.8.
Underneath the flame, however, the convection is steady and goes as x‘m
(boundary-layer dependency).

The predicted pyrolysis-front spread rate varies linearly with U... and varies to the

1.5—power of Y0... These predictions are in agreement with the experimental
results.

Appendix A

Appendix A

Study of number and location
of thermocouples in a ceramic
solid for heat flux computations

The minimum error can be reached by using many thermocouples indepth and at
the surface in order to collect as much information as possible. However, extra infor-
mation is sometimes irrelevant and not worth the trouble of placing thermocouples in
the cast ceramic solid. Thus, the sufficient number of thermocouples needed for accu-

rate heat flux computation was investigated.

In addition, for a given number of thermocouples, the location of each thermo-
couple plays a role in the computed heat flux accuracy. The closer the thermocouple is
to the surface where heat flux is applied, the more sensitive it is to changes in heat

flux. Thus, the thermocouple locations were also investigated.

After casting a ceramic sample with indepth thermocouples, one needs to know
the exact thermocouple locations after the ceramic solid has dried out. The effect of

such error on heat flux results was also investigated. Error in solid properties is also

reported.

103

104

A.l. Experiment

Two thin resistance circular heaters were placed between two identical solid
cylinders of ceramic. Thermocouples were placed in the top solid as shown in Figure
A1 The heaters were set at 6.45 volts and 1.531 amps for more than 3 hours in order
to reach steady state temperatures in the solid. The power was then increased to 12.8
volts and 3.156 amps for 92 seconds and then turned off. This power step input is

shown in Figure A.2.

During this time, temperature data was collected from all the thermocouples, and
is presented in Figure A.3. As the data acquisition unit scanned over those 8 thermo-
couples, the timestep varied in time from 0.7 to 0.9 seconds for each thermocouple.
Therefore, for simplicity, a low filter was used to approximate the temperature data at
equal timestep of 1 second. A comparison of the temperature at x=0 before and after
smoothing is presented in Figure A.4.

Then the program CONTPC was run to compute the transient heat flux at the sur-

face. The number of future times used in all the computations was kept constant at a
value of 4.

A.2. Results and discussion

A.2.l. Usingonly one thermocouple indepth

The case of each thermocouple that was studied is presented in Figure A.5. As
the indepth distance of the thermocouple increases, the thermocouple becomes less
sensitive to heat flux changes at the surface, and so, the error in the heat flux, as seen

in Figure A.5, increases. The case of the deepest thermocouple, at x=2.7cm, has oscil-

lations with high magnitude. Therefore it is not reported.

Figure A.6 shows the residual of those cases presented in Figure A.5. Unlike the
heat flux; as the thermocouple location from the surface increases; the error between

the calculated and the actual temperature decreases.

105

 

100

 

m m _
h.

, w
m film

Experimental setup.

 

 

 

 

 

Aluminum plate

 

60 80

40

Time (sec)
Power step input history.

20.

 

mm - . -

.. 5r...”
32 000

1
I I
1

Figure A.l

 

—20
Figure A2

 

 

10
0.84
E one
04—

A. o}: ._o

o.2~
0.0
-40

106

 

80 Y I Y I V I

 

 

 

 

 

 

I l r T T
U 70-I _l
0
L.
:3
4.1
E
0.) 60" _l
O- W
E —/
0) " .4
l— .
50" "l
l 7
40 I I u I I I f I I r I I I I I
0 20 4O 60 80 100 120 1 4O 1 60

Time (s)
Figure A3 Temperature histories at the surface and in-depth

 

 

 

 

I I I I I I T I I I
‘+ .4
72-1 .4
- a
‘4 3 I
8 68-1 ..
3 d l
O " .
l...
3 64‘ d
1 _
E . .
Q)
I— 4
.1 .
60-1 ~
‘ o smoothed
' — real
56 I I I I I I I I r I fl r I W‘ I
0 20 40 60 80 100 120 140 160

Time (s)
Figure AA Comparison of measured and smoothed temperature data.

107

 

 

 

 

 

 

 

 

 

 

 

.1
_.
A d
(\l
E _
U
\ .1
3’»
v
x ..
2 .1
Li.
4.; 4
0 fi
Edi) ‘ ~ 5 v 1.14cm 1
I o 0.53cm
0'0 I 0.28cm d
o 0.0cm 4
-— ll TC
‘0-2 I I I I r I I I I I I I a I
0 20 40 60 80 1 00 1 20 1 40
Time (see)
Figure A.5 Computed heat flux history using one in-depth thermocouple
measurements.
0'6 T l l I l l I
d . ‘ -I
0.4-1 0 .1
4 ° 1
A '1
8 l
B
3 -1
.‘S.’
In . d
0
(I: 0 .4
.1 .0 . .4
_ e V 1 .1 4cm
OJH o 0.53cm
I I 0.28am 4
__0 6- o 0.0cm
' ' I I I I I I I I I I I I
O 20 40 60 80 1 00 120 1 40

Time (see)

Figure A.6 Temperatme residuals when using one in-depth thermocouple

measurements for heat flux calculations.

108

A.2.2. Usingtwo thermocouples indepth

Several cases were studied and are presented in Figure A.7. As the distance to the
first thermocouple from the surface increases, the error in the heat flux increases. The
residual (Y -T) of the cases in Figure A.7 are shown in Figures A81 and A.8.2 for the
first and second thermocouple, respectively. Figure A.8.l shows that a maximum error
of 0.6" C exists for the first thermocouple. However Figure A.8.2 shows a maximum
error of 2 ° C for the second one because the minimization of the root mean square is

heavily weighted on the first thermocouple.

A.2.3. Usingone thermocouple at the surface and others indepth

The thermocouple at the surface is more sensitive to changes in the heat flux than
the indepth ones. Thus, its effect dominates over the rest of thermocouples. Several

cases were studied on using one or two thermocouples indepth.

The cases of using one thermocouple indepth are shown in Figure A9. All the
results are very close, and a maximum error of 2% in the heat flux is seen for the case
of x=l. 14cm. The domination of the thermocouple at the surface, mentioned earlier,
can be observed in the case of having a thermocouple at x=1.l4cm. The addition of
the thermocouple at the surface brought the results from 10% oscillations to stable

results with low percent error.

The results of having two thermocouples indepth, in addition to the one at the

Surface, are very close, as shown in Figure A.10.

The cases of using 1, 2, and 3 thermocouples indepth with the one at the surface
are shown in Figure A.11. A maximum error of about 0.02W/cm2 (2%) is observed.
For comparison, the case of using all 7 thermocouples to compute the heat flux is

Pusented in Figures A.9, A.10, and A.11.

109

 

 

 

 

 

 

 

 

 

 

1.0 l I l l T I
. ...... I l
' u
A t.
(El 0.8“ _ , '1 ”J
\o ‘. .-
~;~ x " ‘
3x . x'.‘ 3'
x 0.6-I ‘: " “i
3 ‘u '0.
[1: xII ‘-
I l 8- .‘e' “
4-J I -
8 : .1 0". .53,2.7cm
I 0.4-4 3 . ‘3.“ 8.1.14cm -
-’ 7 .53cm
1 ll 2.7cm 4
0 .53cm
--- ll
0'2 I I ' l ' l f l ' a [c
0 20 4O 60 80 100 120 140
Time (sec)
Figure A.7 Computed heat flux history using 2 ill-depth thermocouple
measurements.
0.6 r f I I I TI T
'0
4 O I _‘
0.4-1. .. O "'
.3“
A J -- _s-II, -
a ’- 3»
0.2-1 I A ‘
'6 3’” l
1?, .0
ea —4
a)
QC .
. . I, an .53.2.7cm
-O.2- r: v .28.1.14cm ‘1
o .28..53cm .1
I .12.2.7em
—0.4 f . r t f r fl 0 .12.;53ctn
Time (see)
Figure A.8.l

Temperature residuals of the 1“ thermocouple when using
2 in-depth thermocouple measurements for heat flux calculations.

110

 

 

 

 

 

 

B l
B
3 '-4
39
(I)
a) 4
o:
-2-1 Vf .53.2.7cm -1
.28.1.14cm
. III .28,.53cm
o .12.2.7cm
“1 It .12,.53em
\J fl I T I 1 r Tr r I l T T I
0 20 40 60 80 100 120 140

Time (see)

Figure A.8.2 Temperature residuals of the 2'“l thermocouple when using
2 in—depth thermocouple measurements for heat flux calculations.

 

0.8—l

0.6-1

0.4-4

Heot Flux (W/cmZ)

 

 

 

0.2

 

I I I I r
0 20 4O 60 80 1 OO 1 20 1 40

Time ( sec)
Figure AS Computed heat flux history when using the thermocouple at x=0 and
another in-depth.

111

 

 

 

 

 

 

 

  

 

 

 

1.0 . , . , l I I I
‘ 1
A 0.84 _4
(\l
E 1 cl
0
; 0.6-i ..
'1 .4
’3‘
LT. 0.4% _.
+J
o . .
a)
I 0.24 _
l: 0.12cm. 1.1 cm
0 0.12cm. 0.2 cm
0.0 . , . , . , . , . f“. °" to .
O 20 40 60 80 100 120 140
Time (sec)
Figure A.10 Computed heat flux history when using the thermocouple at x=0 and
another 2 in-depth.
1.0 I I I I
1 .
1
A i
N _
E a
o
\ a
5 .
x —-I
3 01
LI.
‘6 .
al
I
III .12..53.2.7c
I .12.1.14cm
- o .28em
0.2 . - all TO
I ’ I I 1 r l T 1
0 10 20 30 4O 50

Time (see)

Figure A.11 Effect of the number of thermocouples used on the computed
heat flux.

112

A.3. Thermocouple locations

As mentioned in the previous section, the closer the thermocouples to the Stu-face
are, the better the results are. This is shown in Figures A.7 and A.8 for heat flux and

error, respectively.

A.4. Errors in the thermocouple location

For a given temperature history, an error in the thermocouple location would lead
to an error in the heat flux results. This is shown in Figures A.12 and A.13 for the
cases of two and one thermocouple indepth, respectively, with a location error of
2mm. Comparing these results with those of Figure A.7, it is seen that as the thermo—

couples are closer to the surface, the results show more sensitivity to location errors.

A.5 Errors in thermal properties

All the above results were computed by using published ceramic thermal proper-
ties. Just recently, the real properties of each ceramic block were computed by using
the parameter estimation code developed by Dr. J. Beck. For comparison, the heat
flux results of both cases are presented in Figure A.14 where all the thermocouples
were used to compute the heat flux at the surface. The error is more significant at the

beginning and the end of the step input.

A.6. Conclusions

 

- Besides the thermocouple at the Other boundary, it has been shoWn that the addi-

tion of another thermocouple gives approximate results if it is very close to the

surface where the heat flux is applied.

- Two thermocouples give even better results. For higher accuracy, one of the two

thermocouples must be at the surface.

- The location of the 2 "d thermocouple can be anywhere indepth as long as the l“

is very close to the surface.

113

 

 

 

 

 

 

   

 

 

r I l l 7 j if ' l
1 .6‘1 —1
A . I
N -
E 1.2. i
\ 1 -l
E . .
:5 0.8-4 3" 7 , 1 4
LI— 4 . ..O I I. .‘
+1 4 . I. . ' . -4
O .4 ' ' .
g 0 44 I: I O. _
. -< t. .' 0
hr .
: ~m\52.1.05cm
.73.2.7
0.0 I’ I V l I I I l I l o 1' r cm
0 20 4O 60 80 100 120 140
Time (see)
Figure A.12 Effect of 2mm location error of the in-depth thermocouple on
the heat flux results (using 2 thermocouples).
1 .2 1 l I r f T
. J
R?
E -
o
\
3 _
v
x
3 -1
LL J
.‘J
o a
a)
I o 1.1.4cm
a 0.53am -
. o 0.28cm
0 ,, ' —- 0!! TC
_ 0‘. 1 r I r I ‘r T f 1* f I
0 2O 4O 60 80 100 120 1 40
Time (sec)
Figure A.13 Effect of 2mm location error of the in-depth thermocouple on

the heat transfer results (using one thermocouple).

114

 

 

 

 

 

 

 

 

1' 0 I I T l l . l T l
f
p
0.8" . —J
’8?
E « : '
\U f .v .
g 0.6'* c ' m
V I 0'
x 1 $ ' '1
:3 , 0'
E: 0.4“J '0 ' '4
,4 . \
O 4 ' o .
0 ' .
I 0 .
0.2 : o ...
v ubliehed pro
4 o e ' oted pr“
— step input
0 0 T l T t T I T r r r r T
-20 0 20 4O 60 80 100 1 20 1 40
Time (sec)

Figure A.14 Effect of properties error on the computed heat flux (using all the
thermocouple measurements).

An error in the thermocouple location has more effect on those that are close to
the surface, since they are have higher sensitivity coefficients.

A maximum error of 2 0 C in the temperature was observed for the deepest ther-
mocouple, in case of using 2 thermocouples indepth.

Errors in' the solid properties have high effects on the results when sharp changes
happen in the heat flux applied at the surface. In addition, the computed thermal
pmperties of each ceramic solid give better results than the published ones.

Appendix B

Results offiame-spread
experiments on PMMA

(106

115

 

(105-

(104-

Moss Production Rate (9/5)

 

 

5E—O3

~4E-03

~3E-03

~2E-03

-1E-03

    

 

 

 

THC 8c CO Moss Production Rate (9/3)

 

 

 

 

0.00 “ , . , 0E+00
0 1000 2000 3000 4000
Time (see)
Figure B.Tl Species mass production rate: history during flame spread
on PMMA (U..=0.6m/s, Y...=0.233).
1E-O1 ,
fa? ..
\\ 55-02~
O) '1
v
e
*5 J
(I
C
:2 1E-021
0 i
3 .
8 55-034
L. a
0.
A
(D A
(I)
C 4 A H20
2 Cl 02-depl.
C
1 E-03 I I I r 6 I O? I I
10 20 4O 60 80 100
Pyrolysis—Front Xp (cm)
Figure B.Xpl Species mass production rates plotted against the pyrolysis-front

location for PMMA (U..=0.6m/s, Y,.=0.233).

116

 

 

 

 

 

 

 

 

 

0.10 4E—O3 Q
o 002 3
E 3 323““ 2
o» 008-1 ,6 “3C é:
v + C0 , ' + _ “ ”3E-03
B . . . g
o , . _ 2;;
05 0.06- . .- + g
C . 4' '0
g * . - -2E-03 g
‘5 . $3 0.
D 0.04- «5’ . . . 8
9 g" 0
OJ; _ - i it —1£-—03 5
0.02- . * ' O
(8 . f‘»; 0
E _ .
+ «Art 08
0.00 . e , f , . j . 05+00 (I)
0 500 1 000 1 500 2000 2500 1—
Time (see)
Figure B.T2 Species mass production rate history during flame spread
on PMMA (U..=O.9m/s, Yo..=0.233).
,0? 10.0 4 1' I l 1T m r t t _
\ 8.0+ _
3 .1 .1
6.0d _
(\l
o . ‘ ‘
§ 4.0‘ -i
m
4—3 " O "
o
0:
C 2.0 '3 _
.9 1 ° .
4" a
g 0“.“
‘O ‘A‘
8 1.01 a“
0- .: ‘ ‘ l -l
g 0.8 d ‘ ‘ A A H20
0 0.6‘ D 02-depl.
E 9 C09
| l l ' I I I l f
10 20 30 40 50 60 7O 80 90100

Pyrolysis—Front Xp (cm)
Figure 1!.sz Species mass production rates history during flame spread
on PMMA (U..=O.9m/s, Y...=0.233).

117

 

 

 

 

0.10 45-03 {

o 002 ~ 3

’(‘n‘ J D 02-depl. _ a)

\ V H20 , f *6
c» 0.081

v 1': 23° x115! -3I-:—0:s 0:

a) 1 . ,3! - c

*5 ‘ f .9

0‘ 0.064 ~ ' g

S " " 1’

2%; 4 .. ,' V g _2E*03 08-

'8 0.04- ‘, ‘1 I.“- J a)

O ‘ ‘ (D

L’ -I “ M. o

t x PIE-03 2

~ 0

8 0.02 )g? . 0

2 ~ MAéflfi %

0.00 ”‘W rs .fi , . , . . . 05+00 0

0 " m " 800 1200 1600 2000 5

Time (see)

Figure 8.13 Species mass production rates plotted against the pyrolysis-front
location for PMMA (U..=l.5m/s, Y...=0.233).

 

 

 

 

ft]? 1E-011

\ _ ..

01 8E 021

v -I

a) .

"5

[r 4E-02-

c i

.9

U

:3

.0

O

L

”- 1E-021

a 815-03-

: A H20
2 I U 02-depl.
4E-03 I ' I I T QICOI? I 1
10 20 3O 40 50 60 70 80 90100

Pyrolysis-Front Xp (cm)

Figure B.Xp3 Species mass production rates history during flame spread
on PMMA (U..=1.5m/s, Y...=0.233).

118

 

   

Mass Production Rate (9/3)

 

 

 

0.4 8E-03 {
v 002 A? - 3

J O 02-depl. o (D
1:] H20 3 b *5

0 THC a:

0'3“ x 00 : ”55‘03 c
.' . .7 _ g

I 'o

0.24 . —4E-03 9
a? - 0.

1’ °’

0‘)

, - - o

0.1- l" . ~2E—03 2
. - o

,wew ‘ o

0.0 " ' . T . I . , 1 1 . . . 0E+00 I
0 1 00 200 300 400 500 600 700 800 1—

Time (see)

Figure B.T4 Species mass production rates history during flame spread
on PMMA (U..=O.9m/s. YsO.43).

 

100.0 1 I ' I ' l ' l
80.01
60.0:

40.0-

lllllL

1

20.04

10.01
8.0-:
6.01
4.0-

4 llllllL

l

2.0-

11111

Moss Production Rote X102 (g/s)

 

 

 

 

PYronsis—Front Xp (cm)

Figure B.Xp4 Species mass production rates plotted against the pyrolysis-front
location for PMMA (U..=O.9m/s, Y...=0.43).

119

 

 

 

 

1.0 25—03 Q

0 C02 Q l- 3

”U? A 02-depl. Q)

\ U H20 it _ 45

3 01H x THC z; 0:

_ O

c 0.6 Q o . 0 8

.3 Q 0 —BE-04 9

.8 0.4— It 9 O 3}

0: .I X x o O P O

(I) X K O o 0 * F4E—04 E

0 020 xxx. Q Do .55” . 8
° Xix 683°° W i

2 x g 6 6 9 R 325
x* * 0° 5' ga-Uig** T

0.04 , . . . , 0 , . 0E+00 (i)

0 20 40 60 80 100 1 20 1 40 160 I—

Time (see)

Figure 3.13 Species mass production rates history during flame spread

 

 

 

 

on PMMA (U..=O.9m/s, Y...=l.0).
’0‘ 100.0. ' ' w ' ‘—
E 0.0: no ,. 5
v : 000° 0.. "
0° .O' ‘
65 40.0- 000° .... -‘
;2 ' 00000...." at 1
Duo .0. A““ d
0.) a o. a“
4...
O ”a: '... A‘ ‘
m 10.0“ 0. A‘ _I
80— ‘A .
C . J ‘0‘. _
a
'3 4 04 e“. ‘
O ' A‘ —
3 . ‘A‘ .
"U a
O
L
D.
W 1.01 A H20 -
‘8 0.81 D 02-depl. j
:2 63 C09
T l I t I l I I
10 40 80 100

Pyrolysis—Front Xp (cm)

Figure B.Xp5 Species mass production rates plotted against the pyrolysis-front
location for PMMA (U..=O.9m/s, Yo..=l.0).

Appendix C

Results of flame-spread
experiments on wood

Moss Production Rote X102 (g/s)

Moss Production Rate (9/8)

20

1

120

 

0.12
v 002
O 02-depl.
0.10- U H20
0 THC
1 % CO
0.084
0.06“
0.04-
«I
0.02‘
0.00 r 0E+00

 

 

 

' I I T I '
100 200 300 400 500 600 700

Time (sec)

Figure C.Tl Species mass production rates history during flame spread

on wood (U..=O.2mls, Y...=0.233, t'f'=l.3W/cm’).

 

v (:02

[1 H20

0 02-

depl.

 

 

10

Figure C.Xpl

 

l I T l
20 40 60 80 100
Pyrolysis Front Xp (cm)
Species mass production rates plotted against the pyrolysis-front
location for wood (U..=O.2m/s, Y...=0.233, é”=l.3W/cm’).

 

THC 8: CO Moss production Rate (9/8)

121

 

 

 

 

 

0.05 SSE-03
V 002
"m‘ 0 02-depl.
\\ C1 H20
3 0.04fi o THC /
'6 /¥ "2E-03
01 0.03- /
c -'
.9 . /:
.5 If I -\.’
g 0.020 0 3.1-Lg. ‘ _.
8 t :::- ' ~1E—03
0. 1 0
(0 ° “ .0
m 0.01—1 .- I ”I“, ”A:
0 % .‘A-fi'kl'.‘
:2 o f? 1f
.3:
(100 . ' .J- . 1 . I .r 0E+OO
0 200 ‘00 600 800 1000 1200
Time (see)

Figure C.T2 Species mass production rates history during flame spread
on wood (U..=0.6m/s, Y...=o.233. q"=o.swlcm2).

 

10.0
8.0n

6.04
4.0—

002
02-depl.

1304

2.0“

1.0“
0.8-4

0.6-1
O.4-+

 

0.2n

Moss Production Rote X102 (g/s)

 

 

 

0.1 T I i l
10 20 40 50 80 100

Pyrolysis Front Xp (cm)
Figure C.Xp2 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=O.6m/s, Y...=o.233, o"=o.SWIcm2).

THC & CO Moss Production Rate (9/5)

122

 

   

 

 

 

0.12 1 I . I . I ' I
V 002

A O 02-depl.
U) _ ...
\ 0.100 0 H20 g 15 02
3 1 an THC ,
Q) X CO '
.H 2
00: 0.08‘1 v
C i 9 ~8E-03
.9 0.06— a
‘g , 0 i .
U
9 0.04- 9!: L
0. . $0 —4I—:—03
a ’ -
O 0.02- 9 _
2 A

0.00 . , . 0E+00

0 200 400 600 800 1000

Time (see)
Figure C.T3 Species mass production rates history during flame spread
on wood (U..=0.6m/s, Y°.=0.233. (1"=0.9W/cm2).

 

 

 

 

U) V 002
F, O 02-depl.
V C] H20
N
104
9.. g;
X 7..
.‘L’ 6~
(is) 5d #0050
c 4—0 000099.09
0 0' 'V
‘8 v
v
8 2‘ V UDDDUDDDDCIP
0' 0000
U)
(D
U
2 1 1 I i r
10 20 40 50 80 100

Pyrolysis Front Xp (cm)
Figure C.Xp3 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=0.6m/s, Y,.=0.233, ti"=0.9W/cm2).

THC & CO Mass Production Rate (9/8)

123

 

   

 

 

 

”a?
0.20 -
_ v €02 * 2E 02 E»
ru? 0 02-depl. g." ‘ V
\ 016- D H20 1: - 3
3 . q o THC 9 ‘ ‘ZE—OZ CE
3 as CO , w ‘ C
O - o - 9
a: 012. v ~1E—02 T5
5 ‘ - 3
'8 . 3 9
' Q
§ 0.08:- ~8E-03 m
u ’ U)
0L- : ' °°°° ‘51.: g
3 0.04- “‘3 -4E-O3 o
o - é - 0
2 4 05 Q;
0.00” . , . , i , . . -OE+00 O
o 100 200 300 400 500 500 760 800 E

Time (see)
Figure C.T4 Species mass pmduction rates history during flame spread
on wood (U..=O.6m/s, Y...=o.233, q"=1.3wlcm2).

 

 

 

 

A 20
Q v (:02
on O 02-depl. V v
V ' o
D H20 V 8 ' o o
N v o o
O 3 °
; 10~ 3
3 5
.33 8o 0 9
V
(r a D D U D
c 5‘i U a a
O D a
'5 o D
U C]
5 4~
0
L.
Q
(I)
(n
O
2 2 I i T I
10 20 40 60 80 100

Pyrolysis Front Xp (cm)
Figure C.Xp4 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=0.6m/s, Yo..=0.233, ti"=l.3W/cm2).

124

 

 

 

 

 

 

 

 

 

0.12 7E-03
J V C02
’0? O (Dz-depl.
E 0.10% D ”20 M _
v T : THC ' om) _5E_03
0’ l-
*5 0.08-
0‘ l
c
.g 0.06— —3E—03
U i
3
. “o -
8 0.04~
0' ~1E-03
‘3 .
o 0.02“
2 “M *
W ,
0.00 . I I r —1E—03
600 700 800 900 1000
Time (sec)
Figure C.T5 Species mass production rates history during flame spread
on wood (U..=O.9m/s, Y...=0.233, Q"=0.SW/cm’).
10
’8‘ 9“ V C02
\ 84 O 02—depl.
3 7~ [3 H20 "
N 6_‘ '7'"
V
2 SJ 'vV'v oocpo
x "' 00°
4 _‘ ' V V ' 00
.2 v ' 000
O O o O 000
0: 3g 0 o
c
.9 U
"" D
g 2 4 000
U
o .30
L
0. 0°
C!
(CID, o
O 0 o
E 1 9 I I I j I I I I
10 20 30 4O 50 6O 7O 80 90100
Pyrolysis Front Xp (cm)
Figure C.Xp5 Species mass production rates plotted against the pyrolysisofront

location for wood (U..=O.9m/s, Y...=O.233, ti"=0.5W/cm’).

THC 8c CO Mass Production Rate (9/3)

125

 

 

 

 

 

 

 

 

 

1E-02
l v—v coz .
1; 0154 9—0 Cz-depl. f
4 B-El H O '
3‘ . 04> mic ,'. -8F_—03
m . H co l
*5 0.12— l'
01 ' ' -6E-O3
c d
.9 a
1;) 0.08~ L
‘O -l 4E-03
o .
L
Q .I
O
2
0.00 v . l . t OE+00
O 200 400 6 U 0 800 1000
Time (sec)
Figure C.T6 Species mass production rates history during flame spread
on wood (U..=O.9m/s, Y...=O.233, £1"=O.9W/cm2).
A 20
U) V C02
E O 02—depl. v
V D H20 ""'
N ' ' 0000
O ' ' ' 0°C
«— _4 v o
x 18d ' V V o O o O
.4 O
O: 7 v V o o 00
6“ o o 000
c: ‘,o
.9 5 —I a D D
6 c. 0 °
3 o
'0 4d 0 a
8 o D
O. 3.- 0 a
(I)
(D
o
2— 2
F l l i
10 20 4O 60 80 100
Pyrolysis Front Xp (cm)
Figure C.Xp6 Species mass production rates plotted against the pyrolysis-front

location for wood (U..=O.9m/s, Y...=0233,q"=o.9wIcm1).

THC & CO Moss Production Rotes (g/s)

126

 

0.3 1E—02
v 002
O 02-depl. -
~ U “20 “i. ~1E—02
O THC at II
3K CO fi'v " h
0.24 waxy -8E—03
0" P
0v
0'

Moss Production Rate (9/5)

 

 

~6E—03

~4E—03

”b b
: _ ° -2E—03

t

 

 

 

 

 

 

" ' . . I . r 0E+OO
0 1 00 200 300 400 500 600
Time (sec)
Figure C.T7 Species mass production rates history during flame spread
on wood (U..=O.9m/s, Y...=0.233, q"=1.3wlcm1).

A 30

fl *7 002
8 O 02—depl. 8

H a

O a

'- a

x a

a) a

*6 8 c1 0
O: 1 O J 5 o D

C 9 -4 U U

a . D

2 a

U ‘1 c1
8 6 —— 0

O 0

g .-
O. ‘J W 0

g} 4 1

o
E

3 T I I I
10 20 4O 60 80 100
Pyrolysis Front Xp (cm)
Figure C.Xp7 Species mass production rates plotted against the pyrolysis-front

location for wood (U..=O.9m/s, Yo..=0.233, d"=l.3W/cm2).

THC 8c CO Moss Production Rate (9/3)

127

 

 

 

 

 

0.12 ' 8E—03

J v—v coz .

’3 r

E 0.10a EHEJ H20 , l

o—o mc

.93 _

O 0.08“

0:

C .

.g 0.064 —4E-O3

U . .

3

U

E 0.04~ .
« ~2E-03

a .

O 0.02-

2 d

0.00 ,. ~— , . . . 05+00

0 200 400 600 800 1 000 1 200

THC & CO Mass Production Rate (9/8)

Time (sec)

Figure C.T8 Species mass production rates history during flame spread
on wood (U..=l.5m/s, Y...=0.233, Q”=;O.SW/cm’).

 

10.0
8.0-l

6.0-1 ' v v
0

4.0“ o a
2.0% D 0

1.0“

0.8“

O.6~ '3 H20

0 02—depl.
V CO

0'4 I ' i I I 1 l7 I l

10 20 30 40 50 60 70 80 90100

Moss Production Rote XlO2 (g/s)

 

 

 

Pyrolysis Front Xp (cm)
Figure C.Xp8 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=l.5m/s, Y...=o.233,d"=o.sw1cm2).

Moss Production Rate (9/3)

Moss Production Rote X102 (9/3)

128

 

«W002

0,16d 9—0 02-depl.

'lG-EIH20
«HTHC
qX—XCO

0.12-

 

 

I“ " ‘rr ‘

    
 
 
  

8E—03
~6E-O3

\”1—4E-03

 

0E+OO

 

 

 

 

 

TOO

THC 8c CO Moss Production Rate (9/8)

0.00   .,.1.
100 200 300 400 500 600 700 800 900
' Time (sec)
Figure C.'l‘9 Species mass production rates history during flame spread
on wood (U..=l.5m/s, Y...=o.233, q"=0.9wlcnt1).
20
V C02
0 02—depl.
[:1 H20
v ' 0
10 -< o
9 a , ; °
8 ._. ' o
74 ° .3
6'1 D D
_ a D 0
b " a
0 o
4_ D
D
:5 .4
2 T l I l
10 20 4O 60 80
Pyrolysis Front Xp (cm)
Figure C.Xp9 Species mass production rates plotted against the pyrolysis-front

location for wood (U..=l.5m/s, Y...=0233. d"=0.9W/cm2).

129

 

0.3

0.2-l

0.1q

Moss Production Rate (9/3)

 

0.0
O

 

1E—02
V—V C02
9-0 02-depl. '
- B—El H20 .

0—0 THC 3 ‘ ; “BE-03

H CO ' .. _
FEE-03
_4E-03
“25—03
"' OE+00

 

 

T I I l 1 I —U I I
1 00 200 300 400 500 600 700 800

Time (sec)

Figure C.T10 Species mass production rates history din-ing flame spread

50
40

30

20

Moss Production Rote X102 (9/5)

on wood (U..=1.5m/s, Y...=o.233. ('1"31.3WIcm’).

 

V C02
0 02—depl.
“ D H20

‘ .

 

 

 

l I T I

10 20 40 60 80 100
Pyrolysis Front Xp (cm)

Figure C.Xp10 Species mass production rates plotted against the pyrolysis—front

location for wood (U..=l.5m/s, Y...=0.233, d"=l.3W/em2).

THC 8: CO Mass Production Rate (9/5)

130

 

53
t3
1
x0004

Moss Production Rate (9/3)

 

.
I.
.
O ' ‘-
O
n" .
,
K
. I
Q: "
s.
wt
4
.0 ‘
I"' ~ .
x.‘
. e.
i o-
. a .

 

45-03
002
02-depl.
H20
THC
CO r3E—03

 

rza-os

~1E—O3

 

 

 

 

 

' r r l l m I t OE+OO
600 650 700 750 800 850
Time (sec)
Figure C.Tll Species mass production rates history during flame spread
(mvwndGLsOSmwtYmsailQEQJWMmfiL
20
r“? V 002
E O 02-depl.
V D H20
N ‘l —
(D O
.— 8‘
X
38
Q) 6‘ v'ggaz
*5 .
v 0
0: . 3 ° °
c 4“ ' 3°
' o
.9 ° 00‘3
+J DD
0 0°
3 0°
'0 0°
9. 2 - 0 °
0. D 0
D U
m
m
0
2 1 I I I I
10 20 4O 60 80 100
Pyrolysis Front Xp (cm)
Figure C.Xpll Species mass production rates plotted against the pyrolysis-front

 

lumumlRuummdaLFQSmfltYgs0334r=QSVMmm§.

THC & CO Moss Production Rate (9/3)

131

 

 

 

 

 

0.3 35-03
‘V 002
”a O 02-depl.
\\ < [3 H20 .
3 o THC
0) 316 CO
'6 Q2— -2E-03
0:
C
.9 . —
4.)
U
:3
.0 ..:';
9 0.1- . ~IE-03
Q. 9 .:§"' '
a)
tn ‘J‘ _
o .11"
0.0 i I I I 1 I I I I I T OE‘I‘OO
600 620 640 660 680 700 720 740

Time (s)
Figure C.T12 Species mass Induction rates history during flame spread
on wood (U..=O.9m/s. Y...=0.43, ti"=0.5W/cm2).

 

100
80 v 002

60 — O 02-depl.

Cl HO
40n 2

20‘

Moss Production Rote XiO2 (g/s)
5.5

 

 

 

I I I

10 20 40 60 80 100
Pyrolysis Front Xp (cm)

Figure C.Xp12 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=O.9m/s. Yo..=0.43, t'I”=0.5W/cm2).

THC & CO Mass Production Rate (9/5)

132

 

Moss Production Rate (9/8)

 

 

 

 

. I .7: . I l I v 'I I r I I OE+OO
600 620 640 660 680 700 720
Time (sec)
Figure C.T13 Species mass production rates history dm'ing flame spread
on wood (U..=O.9m/s, Y...=0.6l, ti"=0.5W/cm2).

 

100
V C02

0 02—depL
CI H20

45 07 CI:
0 O C
l 1 |

N
O
l

o
o
o
e
a
c

Moss Production Rote XiO2 (g/s)
ES
1

 

 

 

I i i I F
10 20 4O 60 80 100
. Pyrolysis Front Xp (cm)
Figure C.Xp 13 Species mass production rates plotted against the pyrolysis-front
location for wood (U..=O.9m/s, Yo..=0.6l,t'1"=0.5W/cm2).

THC & CO Mass Production Rate (9/3)

Moss Production Rate (9/5)

133

 

 

1.0 3E-03
V C02
‘ O 02-depl.
CI H20 I.
0-8‘ o THC OOOWO
X CO ° Q3000
‘ o q me W
000 ~2E-03
0°6" o W
o W
" 0 WV *
o a" If '
v m .—
0.4H 0° ”if. " vvvvvvvvvvvvvvvvv '
0° ' PIE—03
q o ' ‘
I v
0.2-i ............. .

 

 

 

 

 

 

 

 

 

 

THC 8c CO Mass Production Rate (9/3)

0.0 ' I r . OE-I-OO
600 640 660 680
Time (sec)
Figure C.Tl4 Species mass Iroduction rates history during flame spread
on wood (U..=O.9m/s, Y...=l.o, d"=0.5W/cm2).
A 100 _
< 80% v 002 o0o
O, 50— o 02-depl. o o 0 °°
V 4 U H20 0 0 v
N 40-1 o O ' ' v
Q o O o , v ' '
1'- 0 '
x V
8 20“ v ' ' ' 0000.300
0 v ' 0 o a
O: 101 o 0 0
s 81 . °
1.: 6—
(3) i
‘o 4‘
O
L
O.
U) 2 2
U)
o
2
1 T 1 l I I I I I
10 20 4O 60 80 100
Pyrolysis Front Xp (cm)
Figure C.Xpl4 Species mass production rates plotted against the pyrolysis-front

location for wood (U..=O.9mls, Yo..=l.0, d"=0.5W/cm2).

Appendix D

Results of heat transfer
experiments

q" (W600?)

134

 

 

 

0.5
X cond.
. O conv.
U re-roo.
O 4~ai A. rod.
Q
A ‘ E
N
g 034 a
U
\\. .
E
t 0.2-
U'
0.i~
“In
Dig."
t ’ 'I.. 'II II'u-uugulioll III II
0() “BEH!HWQ..~nfilUflfiSKIDOOKIIOIOENMMUDDG‘K‘Iflint.
. . T’ ' l ' FT '
l 2 3 4
X/xf

Figure D.l Heat flux measurements during flame spread on PMMA
(U..=0.6m/s, Y,.=0.233, x=33cm).

 

cond.
conv.
re-rod.
rod.

DDOBK

 

 

 

0:01 "Beleeoeaeuaag a a . . .

i 2 3 4 .5
X / Xf
Figure 0.2 Heat flux measurements during flame spread on PMMA

(U..=O.9m/s. Yo..=0.233, x=33cm).

 

0" (Wm?)

<1" (W/cm2)

0.0

135

 

 

Figure D3

   

a
2 3 4

X / Xf
Heat flux measurements during flame spread on PMMA
(U..=1.2m/s, Yo..=0.233, x=33).

 

 

 

Heat flux measurements during flame spread on PMMA
(U..=l.2m/s, Y,.=0.233, x=40cm).

 

 

136

 

q" (W/cm")

 

 

 

Figure D.5 Heat flux measurements during flame spread on PWA

('U.=1.Sm/s, Y...=0.233, x=33cm).

q" (W/cm’)

 

 

 

X/Xr

Heat flux measurements during flame spread on PMMA
(U..=O.9m/s. Yo..=0.33, x=33cm).

Figure 0.6

137

 

¢'(W/cn3)

 

 

 

X/Xr
Figure 0.7 Heat flux measurements during flame spread on PMMA
(U..=O.9m/s, Y°.=O.43, x=33cm).

 

q”(W/CW3>

 

   

 

to Its 20 :25 30 :15 4c» 4&5 50
X/Xf

Figure D.8 Heat flux measurements during flame spread on PMMA
(U..=O.9m/s, Yo..=0.61, 33cm).

("it‘ll/ C m 2y

138

 

 

   

 

5.5“” ’ " T
3K cond
1 C O conv
.).0 n 0 (Cd
n A re-rod

2 54%;;
2.04 E
1.5— '

J
1.0 II'l

*n
0.5—0CEJJ "u

40 ".""'..IIIIIIIIIIIII
0.0

1 2 3 4 5

X/Xr
Figure 0.9 Heat flux measurements during flame spread on PMMA

(U..=O.9m/s, Y...=1.0, x=33cm).

List of References

LIST OF REFERENCES

Abu-Zaid, M.Z., PhD Thesis, Michigan State University, East Lansing (1988).
Ahmad, T., and Faeth, G. M., Journal of Heat Transfer, 100, p 112 (1978).
Ahmad, T., and Faeth, G.M., Seventeenth Symp. anti.) Comb. (1979).
Annamalai, K., and Sibulkin, M., Comb. Sci. Techn.. 9, p.185 (1979).

Atreya, A., mo. Thesis, Harvard University, Cambridge, MA. (1983). Also published as a
National Bureau of Standards report, NBS—GCR-83-449, March 1984.

Beck, J.V., and Arnold, K.J., Parameter Estimation in Engineering and Science, John Wiley
and Sons (1977).

Beck, J.V., Blackwell, B., St. Clair, C.R., Inverse Heat Conduction, John Wiley and Sons
(1985).

Carrier, G.F., Fendell, FF... and Feldman, P.S., Comb. Sci. Techn., 23, p.41 (1980).
Carrier, G.F., Fendell, F.E., and Fink, 8., Comb. Sci. Techn.. 32. p.161 (1983).
Carrier, G.F., Fendell, F.E., Butler, (3., Cook, 8., and Morton, C., Comb. Sci. Techn.. (1990).

DiBlasi, C., Crescitelli, S., and Russo. 6., Proc. of the Eight Intl. Conf. on Comp. Methods in
Applied Sci. and Engg., Versailles, France, (1987).

DiBlasi, C., Crescitelli, S., and Russo, G., Comb. Flame, 72, p.205 (1988a).

DiBlasi, C., Crescitelli, S., Russo, (3., and Fernandez-Pello, A.C., Twenty-second Symp.(lntl.)
on Comb., Seattle, (1988b).

Emmons, H.W., Zeit. fur Angew. Math. and Mech. 36(2), p.60 (1956).
Emmons, H.W., Scientific American, 231(1), p.21 (1974).

Emmons, H.W., Ann. Rev. Fluid Mech., 12. p.223 (1980).

139

140

Fernandez-Pello, A.C., and Williams, F.A., Fifteenth Symp. (1nd.) on Comb., p.217 (1975).
Fernandez-Fella, A.C., Comb. Flame, 31, p.131 (1978).
Fernandez-Pello, A.C., and Mac. C.P., Comb. Sci. Tech., 26, p.147 (1981).

Fernandez-Pena, A.C., and Quintiere, 1.6., Fall meeting of Eastern Sect. of the Comb.
Institute, Atlantic City, December (1982).

Fernandez-Pello, A.C., and Hirano, T., Comb. Sci. Techn. 32, pl (1983).

Groff, E.G., and Faeth, G.M., Comb. Flame, 32, p.139 (1978).

Holman, J ., Experimental Methods for Engineers, McGraw-Hill Book Company (1984).
Kashiwagi, T., Ohlemiller, T.J., and Werner, K, Comb. Flame, 69, p. 331 (1987).

Kays, W.M., and Crawford, M.E., Convective Heat and Mass Transfer, McGraw-Hill Book
Company (1980).

Kulkami, AK, and Fisher, 8., Fall meeting of Eastern Sect. Comb. Institute, Clearwater
Beach, (1988).

Loh, H.T., and Fernandez-Penn, A.C., Twentieth Symp. (1nd.) on Comb. p 1575 (1984).
Markstein, G.H., and deRis, J., Fourteenth Symp. (1nd.) Comb., p. 1015 (1973).

Mekki,. K., Atreya, A., Agrawal, 8., and Mchman, 1.8., Twenty-third Symp. (1nd.) on Comb.,
(1990).

Nurbakhsh, 8., PhD Thesis, Michigan State University, East Lansing (1989).

Orloff, L., deRis, J., and Markstein, G.H., Fifteenth Symp. (Intl.) on Comb., p. 183 (1975).
Quintiere, J ., Design of Buildings for Fire Safety, ASTM STP 685, p. 139 (1980).
Quintiere, J., Harkleroad, M., and Hasemi, Y., Comb. Sci. Tech. 48, p. 191 (1986).

Saito, K., Quintiere, 1.6., and Williams, F.A., Proc. First Intl. Symp. on Fire Safety 8ci., p. 75
(1986).

Saito, K., Williams, F.A., Vlflchman, 1.8., and Quintiere, 1.6., National Heat Transfer Conf,
Pittsburg (1987).

Sibulkin, M., and Kim, J., Comb. Sci. Tech., 17, p.39 (1977).
Tewarson, A., and Pion, R.F., Comb. Flame, 26, p.85 (1976).

Vovelle, C., Akrich, R., and Delfau, J.L., Twentieth Symp. (1nd.) Comb., p. 1674 (1984).

141

Vovelle. C., Delfau, J.L., Reuillon, M., Bransier, J., Larqui, N., Comb. Sci. Tech., 53, p.187
(1987).

Wichman, 1.8, and Agrawal, 8., Comb. Flame, 83, p.127 (1991).