i: .
ﬁg,
. Luna. n .
! ... t .
Mm? mm?
a... 5.27:, .2 d...
; a? .3 «u.
Rﬂmummwmu. :LK. m%!
. ‘wxﬁe‘mﬁmﬁﬁmﬁmf .
x ..

.

.

, p
, Sm? ..
33.1..WWEWW ‘

. . wﬂm e.
. . .2?qu
‘ almﬁ _
.. r.
.mv. mu...
:WMP

.
.
b.

V a.
.u, . n.
. .mpnwwmmmﬁrc

.2. I.
it a . r.
, I

m: y
. :nﬁkfzrl L :r,
. , . ﬁ uvkécvhwﬁu. .
A . ‘ 7 . .. «My. .
V , A?»
. . . { . ‘
3m ,

L4:
. A.

5!;
“3'4

"I

x
l.

13».

.; 5w
. mama.“

1- ‘ .
.ﬁdﬂ. «.41.

3“.
)

 

 

a
it. .rll.‘ r.
u: l(\

 

 

 

*0
:E
w’y‘;

U)

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

This is to certify that the

thesis entitled

INSTABILITIES OF SPREADING DIFFUSION FLAMES IN
MICROGRAVITY AND THE DESIGN AND CONSTRUCTION OF A
HELE-SHAW APPARATUS THAT PRODUCES FLAMES IN THE NEAR
EXTINCTION LIMIT REGIME UNDER SIMULATED LOW
GRAVITY CONDITIONS

presented by

Lisa Marie Oravecz

has been accepted towards fulfillment
of the requirements for

MS degree in Whig

 

 

Major professor

Date 4/26/01

 

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

 

 

DATE DUE DATE DUE l DATE DUE
. APR 2 7 20
SST [111 l‘r2qﬁa A Mmll a®5 f1 A 107'- 'v‘jv
3* 'kx'c: :JJU'IIJI,

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIFIC/DateDuepes-p. 15

INSTABILITIES OF SPREADING DIFFUSION FLAMES 1N
MICROGRAVITY AND THE DESIGN AND CONSTRUCTION
OF A HELE-SHAW APPARATUS THAT PRODUCES
FLAMES IN THE NEAR EXTINCTION LIMIT REGHvIE
UNDER SIMULATED LOW GRAVITY CONDITIONS

By

Lisa Marie Oravecz

A THESIS

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

MASTER OF SCIENCE
Department of Mechanical Engineering

2001

INSTA
AN D '
TH Al

direct a}
“Flight ]
S Seton
Spreadir
thingui:
ﬂamdet:
flame Sp
[0 lack 0

Was idcn

ABSTRACT
INSTABILITIES OF SPREADING DIFFUSION FLAMES IN MICROGRAVITY
AND THE DESIGN AND CONSTRUCTION OF A HELE—SHAW APPARATUS
THAT PRODUCES FLAMES IN THE NEAR EXTINCTION LIMIT REGIME
UNDER SIMULATED Low GRAVITY CONDITIONS
By

Lisa Marie Oravecz

Microgravity diffusion ﬂame instabilities is a growing area of research due to the
direct applications to the safety of space ﬂight. Experiments, which are part of the
“Flight Deﬁnition Project”, named “ATHINA”, for NASA were conducted at the 2.2 and
5 Second Drop Towers, in which regions of oscillatory instabilities were observed.
Spreading ﬂame fronts fragmented into “ﬂamelets” which oscillated, recombined, or
extinguished. Flamelet oscillations were of order le. Blue intensity measurements of
ﬂamelets oscillated suggesting periodic variation of chemical processes. Microgravity
ﬂame spread rates of fronts and ﬂamelets were lower than normal gravity front rates due
to lack of buoyancy. Fuel surface energies were calculated, and a “critical energy ﬂux”
was identiﬁed relating number of ﬂamelets to losses at the sample surface.

A Hele-Shaw apparatus named the “MSU Flame Rig” (under ATHINA project)
was designed and constructed to produce diffusion ﬂame instabilities in a simulated
microgravity environment. Flame instabilities similar to NASA experiments and
unlimited test times indicated the device could be employed to determine thick fuels that
will be tested on the International Space Station. Unstable ﬂame behaviors unlike those
in the drop tower tests showed ﬂames to pulse, form blue cusps, and create ﬁngering

patterns. Flamelets and pulses oscillated at much higher frequencies than NASA tests.

I dedicate this thesis to:

My Husband, Ray for his love and support of me pursuing my life’s endeavors,
My Parents for showing me how to be brave and work hard,
And Myself for continuing to strive for my dreams.

iii

 

 

 

insight

OUISIZln

ofadxit
team FI
their Va:
igniter m

I
CODIbusz

the use 01

 

 

ACKNOWLEDGEMENTS

I would like to thank my advisor Dr. Indrek Wichman for his excellent guidance,
insight, and support, and also for giving me the opportunity to work on such an
outstanding project with NASA.

I would also like to thank Dr. Sandra Olson and Dr. Fletcher Miller for their hours
of advice, assistance, and support for this project. Thank you to the rest of the ATHINA
team, Frank Vergilii, Jim Bruewer, John Thomas, Bill Camperchioli, and Phil Stehno for
their valuable input on building the MSU Flame Rig particularly in the design of the
igniter and selection of diagnostic equipment.

I would also like to thank the NASA Glenn Research Center Microgravity
Combustion Branch for funding this project under the NASA Contract NCC3-662, and

the use of their drop tower facilities.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................................................... viii
LIST OF FIGURES ............................................................................ ix
NOMENCLATURE ........................................................................... xiii
INTRODUCTION .............................................................................. 1
CHAPTER 1
LITERATURE SURVEY ..................................................................... 7
1.1 OVERVIEW .................................................................. 7
1.2 DE RIS ANALYSIS OF SPREADING DIFFUSION FLAMES ...... 9
1.2.1 THEORY ............................................................. 9
1.2.2 THIN FUEL RESULTS ............................................ 1 1
1.2.3 SEMI-INFINITE MODEL RESULTS ........................... 12
1.3 OTHER WORKS OF DIFFUSION FLAME SPREAD IN
NORMAL GRAVITY .............................................. 13
1.4 DIFFUSION FLAME SPREAD IN MICROGRAVITY ............... 16
1.5 DIFFUSION FLAME INSTABILITIES ................................. 19
1.6 CANDLE FLAME IN MICROGRAVITY .............................. 22
1.7 IN STABILITIES IN SIMULATED MICROGRAVITY
ENVIRONMENTS ................................................. 24
1.7.1 LOW STRETCH FLAMES ....................................... 24
1.7.2 HELE-SHAW APPARATUS ..................................... 25
1.7.3 MICROGRAVITY FLAME SPREAD LIMITS IN HELE-
SHAW TYPE DEVICES .................................. 26
CHAPTER 2
EXPERIMENT ................................................................................ 29
2.1 OVERVIEW .................................................................. 29
2.2 NASA COMBUSTION TUNNEL RIG .................................. 30
2.2.1 NASA SAMPLES .................................................. 30
2.2.2 NASA TEST CONDITIONS AND FLOW SYSTEM ........ 33
2.2.3 NASA RIG ELECTRICAL SYSTEM AND
DIAGNOSTICS ............................................ 37
2.2.4 NASA DROP TOWER FACILITIES ........................... 39
2.2.5 NASA RIG TEST PROCEDURE ................................ 45
2.3 MSU FLAME RIG .......................................................... 48
2.3.1 MSU FLAME RIG MATERIALS ............................... 48
2.3.2 MSU FLAME RIG FLOW CONTROL AND
MEASUREMENT ......................................... 49
2.3.3 MSU FLAME RIG PLENUM AND CONTRACTION ...... 53

CH

 

RESULT

 

4.3

 

 

 

 

2.3.4 MSU FLAME RIG POROUS PLATE ........................... 55
2.3.5 FLOW IN THE MSU FLAME RIG TEST SECTION ........ 59
2.3.6 MSU SAMPLES ................................................... 60
2.3.7 MSU FLAME RIG SAMPLE HOLDER AND
SUBSTRATE .............................................. 61
2.3.8 MSU FLAME RIG IGNITER ..................................... 63
2.3.9 MSU FLAME RIG SPACERS .................................... 65
2.3.10 MSU FLAME RIG DIAGNOSTICS ............................. 68
2.3.11 MSU FLAME RIG TEST PROCEDURE ....................... 69
CHAPTER 3
CALCULATIONS ............................................................................ 72
3.1 OVERVIEW .................................................................. 72
3.2 NASA CALCULATIONS ................................................. 72
3.2.1 SURFACE ENERGY BALANCE DATA REDUCTION
CALCULATIONS ......................................... 72
3.2.2 TEMPERATURE DERIVATIVE CALCULATION .......... 76
3.2.3 PEAK TEMPERATURE ENERGY LOSS VERSUS
FLAMELETS CALCULATION ........................ 82
3.2.4 FLAME VELOCITIES ............................................. 82
3.3 MSU FLAME RIG CALCULATIONS ...................................... 83
3.3.1 SCALE ANALYSIS ................................................ 83
3.3.2 MASS FLOWRATE CALCULATIONS ........................ 90
3.3.3 VOLUMETRIC FLOWRATE CORRECTION AND
CONVERSION ............................................. 91
3.3.4 POROUS PLATE CALCULATIONS ........................... 93
3.3.5 PLENUM FILL TIME CALCULATIONS ..................... 96
3.3.6 ENTRANCE LENGTH AND BOUNDARY LAYER
CALCULATIONS ......................................... 97
CHAPTER 4
RESULTS AND DISCUSSION .............................................................. 106
4.1 OVERVIEW .................................................................. 106
4.2 NASA DROP TOWER RESULTS AND DISCUSSION ............. 106
4.2.1 NASA INSTABILITY REGIONS .............................. 106
4.2.2 NASA INSTABILITY CHARACTERISTICS ................ 112
4.2.3 NASA FLAMELET OSCILLATIONS ......................... 119
4.2.4 NASA FLAME SPREAD RATES .............................. 125
4.2.5 NASA SURFACE ENERGY BALANCES ..................... 127
4.2.6 NASA PEAK TEMPERATURE SURFACE ENERGY
BALANCES ................................................ 129
4.3 MSU FLAME RIG RESULTS AND DISCUSSION .................... 131
4.3.1 MSU FLAME RIG FLOWMETER VERIFICATION ......... 131
4.3.2 MSU FLAME RIG FLAME INSTABILITIES .................. 136
4.3.3 DISCUSSION OF MSU FLAME RIG VERSUS DROP
TOWERS .................................................... 151

vi

 

 

C HAP
C ONC

C HAP'
FLTL'}

BIBLIC

 

CHAPTER 5

CONCLUSIONS ............................................................................... 153
CHAPTER 6

FUTURE WORK .............................................................................. 1 5 5
BIBLIOGRAPHY ............................................................................. 158

vii

Table 1.

Table 2.

Table 3.

Table 4.

Table 5.

Table 6.

Table 7.

Table 8.

Table 9.

LIST OF TABLES

Constants of surface energy balance
equation.

Pressure ratio scaling values in the MSU
Flame Rig.

Ranges of volumetric ﬂow rates in MSU
Flame Rig.

Pressure calculations for determining
porous plate pressure drop.

Entrance length ranges in plane parallel
to tunnel side walls.

Entrance lengths in the plane parallel to
quartz plate.

Boundary layer thickness at the
beginning and end of sample.

Experimental results for drop tower tests
with aluminum foil substrate.

Experimental results for drop tower tests
with steel substrate.

viii

................................ 75

................................ 89

................................ 90

................................ 93

.............................. 100

.............................. 101

.............................. 104

.............................. 110

.............................. 111

Figure 1.

Figure 2.

Figure 3a.
Figure 3b.

Figure 4.

Figure 5.

Figure 6a.

Figure 6b.

Figure 7.
Figure 8.
Figure 9.

Figure 10.

Figure 11.

Figure 12a.
Figure 12b.
Figure 13a.

Figure 13b.

Figure 14.

LIST OF FIGURES

Images in this thesis are presented in color.

Diffusion ﬂame spreading over a solid

surface [2]. ............................ 9
NASA sample ﬁ'ont and side schematic. ............................ 31
Flat, linear ﬂame ﬁont in l-g. ........................... 32
Parabolic ﬂame front in l-g. ........................... 32
Flat, linear ﬂame ﬁ'ont in microgravity. ........................... 33
Flammability Map of Olson [24, 26]

with new extinction limit branch ........................... 34
Schematic of ﬂow system in Combustion

Tunnel Rig. ........................... 36
Picture of ﬂow system in Combustion

Tunnel Rig. ........................... 36
2.2 Second Drop Tower schematic. ........................... 41
Drag shield in 2.2 Second Drop Tower. ........................... 42
Zero-G or 5 Second Drop Tower facility. ........................... 43
Drag shield in the Zero-G Drop Tower. ........................... 44
Picture ofMSU Flame Rig. ........................... 48
Picture of MSU Flame Rig ﬂow system ........................... 50
Block diagram of MSU Flame Rig ﬂow

system ........................... 50
Uniform, ﬂat ﬂame front in MSU Flame

Rig. ........................... 52
Corrugated, unstable ﬂame {Tom in

MSU Flame Rig, ........................... 52
MSU Flame Rig plenum schematic. ........................... 54

ix

Figure 15.

Figure 16.

Figure 17.

Figure 18.

Figure 19.

MSU Flame Rig contraction in ﬂow
system.

Air path through porous plate in MSU
Flame Rig.

Double blocking plate system for porous
plate in MSU Flame Rig.

Schematic of MSU Flame Rig exhaust.

MSU Flame Rig sample holder system.

Figure 20a. Picture of MSU Flame Rig igniter

system.

Figure 20b. Schematic of MSU Flame Rig igniter

Figure 21.

Figure 22.

Figure 23.

Figure 24.

Figure 25.

Figure 26.

Figure 27.

Figure 28.

Figure 29.

Figure 30.

Figure 31.

system.
System of MSU Flame Rig spacers.

MSU Flame Rig camera mounting
system.

Fuel surface energy balance.

Schematic for scaling analysis of MSU
Flame Rig.

Calculation of test section velocities in
MSU Rig from rotarneter readings.

Pressure versus velocity behavior for
Mott porous plates.

Plenum ﬁll time calculations in MSU
Flame Rig.

Entrance length in plane parallel to
tunnel sides in MSU Flame Rig.

Momentum and mechanical energy
balances [54].

Boundary layers in plane parallel to
quartz plate in MSU Flame Rig.

Sample location and boundary layers in
MSU Flame Rig test section.

........................... 55

........................... 56

............. 57
........................... 59

........................... 62

........................... 91

........................... 95

........................... 96

........................... 98

........................... 99

........................... 102

........................... 104

 

 

Figure 1

Figure I
Figure I

Figure 3

Figure 3

Figure 3
Figure 3

Figure 3
Figure 3

Figure 3

Figure 3'

Figure 4i
FiEure 41

Figure 41
Figure 4:
Pi811m 42
Figure 43
FigUre 4.;
Figure 45

Fig” e 462

Figure 32. Typical drop tower test showing ﬂamelet
formation and interactions.

Figure 33a. Flamelet top view.
Figure 33b. Flamelet side view.

Figure 34. Side view of progression of ﬂamelet
extinction.

Figure 35. Typical spacing between three ﬂamelets.

Figure 36a. Test with close ﬂamelets.
Figure 36b. Combination of ﬂamelets.

Figure 37a. Spherical ﬂamelet.
Figure 37b. Oval ﬂamelet.

Figure 38. Typical test with faster side ﬂamelets.

Figure 39. Time sequence of right ﬂamelet
oscillations.

Figure 40a. Examples of time periods of major axis
ﬂamelet oscillations.
Figure 40b. Major axis of oval ﬂamelet.

Figure 41. Line proﬁle along which intensity
measurements were taken.

Figure 42a. Intensity in blue region over drop time
along line proﬁle in Figure 41.

Figure 42b. Intensity measurement for single frame
of drop with ﬂamelet along line proﬁle
in Figure 41 .

Figure 43. Flame speeds versus substrate distances
and air ﬂowrates.

Figure 44. Typical surface energy balances for 2
second drops.

Figure 45. Number of ﬂamelets versus peak energy
losses.

Figure 46a. Setup for verifying ﬂowmeter at
atmospheric conditions.

xi

........................... 124

........................... 126

........................... 128

........................... 130

........................... 133

Figure 46b.

Figure 47a.

Figure 47b.

Figure 48.

Figure 49.

Figure 50.

Figure 51.

Figure 52.

Figure 53.

Figure 54.

Figure 55.

Volume ﬂowrates of ﬂowmeter at
atmospheric conditions for two oriﬁces.

Setup for verifying ﬂowmeter at inline
pressure conditions.

Comparison of ﬂowmeter to calibrated
oriﬁce at inline conditions.

Typical progression of MSU Flame Rig
test.

Oscillatory motion of entire ﬂame ﬁ'ont
in MSU Flame Rig test.

Time sequence of uniform ﬂame front
changing into ﬂamelets in MSU Flame
Rig.

Periodic pulses traveﬂmg across ﬂame
front in MSU Flame Rig.

Periodic pulses that extinguish ﬂamelets
and ﬂame ﬁonts in MSU Flame Rig.

MSU Flame Rig test with unstable
ﬂamelets producing ﬁngering pattern.

MSU Flame Rig tests with stable cusp
ﬂames and unstable ﬂamelets.

Burn out of sample by thin ﬂame front in
MSU Flame Rig.

xii

........................... 133

........................... 135

........................... 135

........................... 137

........................... 139

........................... 140

........................... 141

........................... 143

........................... 145

........................... 146

........................... 149

As

{3‘00

wept-it:

8b

NOMENCLATURE

Pre-exponential factor for pyrolysis of fuel (8")

Speciﬁc heat of cellulosic fuel (J/(kg K))

Activation energy for fuel pyrolysis (kJ/mol)

Acceleration of gravity (9.8 m/sz)

MSU Rig test section height (cm)

MSU Rig test section length (cm)

Distance between fuel and substrate(cm)

Heat of Evaporation for cellulosic fuel (kJ/kg)

Total Pressure (psi)

Universal gas constant (8.314 J/(mol K))

Raleigh number based on test section height h

Reynolds number based on dimension x

Temperature of fuel surface (K)

Temperature of substrate backing (K)

Temperature of ambient surroundings (K)

Velocity of oxidizer in x direction (parallel to fuel sample) (cm/s)
Velocity of oxidizer in y direction (parallel to test section height) (cm/s)
Flame spread rate (cm/s)

Thermal diffusivity of air (mz/s)

Volume expansion coefﬁcient for air at constant pressure (K'l)

Substrate surface emissivity

xiii

Subscripts

air

Fuel surface emissivity

Thermal conductivity of air W/(m K)

Kinematic viscosity of air (mz/s)

Density of air (kg/m3)

Fuel surface density (kg/m3)

Stefan-Boltzmann constant (5.67 * 10'8 W/(rrr2 K‘»

Half thickness of cellulo sic fuel (in)

Ambient

Air

Backing or substrate
Flame

Test section height
Surface
Vaporization

lustrtﬁibmﬁg

xiv

INTRODUCTION

Combustion provides the majority of useﬁrl energy production in the world,
which makes it an important area Of research However, combustion has often been a
double-edged sword in its use by mankind. While combustion has produced efﬁcient
power plants, automobile engines, and gas and oil stoves to cook food and heat homes, it
has also produced destructive elements such as forest ﬁres and harmful pollution, which
are in most cases products of technological growth. Nonetheless, scientists are working
everyday to remedy the undesired effects of combustion while moving forward to better
society with its beneﬁts. In this respect, progress in combustion science is a worthwhile
endeavor. Therefore, research in this area must be conducted in a careful yet ambitious
manner.

Combustion involves many intricate processes that are coupled to each other.
Therefore, the combustion scientist must be proﬁcient in several areas; ﬂuid dynamics,
chemistry, heat transfer and thermodynamics. Such studies have produced more
complete and coherent ﬂame theories, in which ﬂames are categorized as either premixed
or diffusion ﬂames. The complicated nature of diffusion ﬂames has caused them to be
investigated less than premixed ﬂames, which have been studied extensively. In a
diffusion ﬂame, the fuel and oxidizer are not mixed prior to combustion, whereas in
premixed ﬂames the fuel and oxidizer have mixed before combusting. Detailed studies
of these ﬂames are crucial to understanding real world combustion scenarios because
diffusion ﬂames occur more frequently in both nature and technology than premixed
ﬂames. Most ﬂames we see are diffusion ﬂames: candles, matches, gas jet burners, jet

engines, and wood burning stoves. Some important investigations produced theories that

serve as the foundation for diffusion ﬂame research. Burke and Schumann [1] developed
a mathematical description of laminar jet ﬂames that agreed well with experiments, while
deRis [2] contributed a theory of laminar ﬂame spread over both thick and thin ﬁrels by
neglecting reaction kinetics.

With the advent of the space program over the last few decades, microgravity
ﬂames have become a signiﬁcant area of combustion research. The practical motivation
behind understanding this ﬁeld Of study is ﬁre safety in space vehicles like the shuttle and
Intermtional Space Station (ISS). In the study by Friedman [3], ﬁre safety in spacecraft
was investigated. Most of the prevention and defense tactics against ﬁres in space has
been established from experience of ﬁres in enclosed cabins like airliners. However,
more intensive studies focusing on the fundamental nature of microgravity ﬂames will
lead to a ﬁrmer basis for microgravity ﬁre safety strategies.

Fuel, oxidizer, and a heat or ignition source are the three factors required for a
ﬁre. Microgravity combustion safety is based on prevention of these factors by either
control or total removal. The control of ﬁre] is achieved by a systematic process of
determining those materials with low upward ﬂame spread in normal gravity. This
material testing on earth is actually a worst case scenario since microgravity ﬂames are
much less robust than l-g ﬂames due to the absence of buoyancy. Although many
nmterials produce high ﬂame spread rates, they are controlled by properly reserving them
in safe areas. Ignition is restricted by employing circuit breakers and electrical grounding,
and safeguarding against conditions that could lead to a spark. The oxidizer cannot be

controlled since it is vital for survival in space. For example, the space shuttle contains

21% oxygen at 1 atmosphere during normal operation, and 30% enriched sea level air at
72 kilo-Pascal reduced total pressure for Pre-Extra Vehicular Activities.

As larger space vehicles are constructed, the probability that the crew will detect a
ﬁre decreases. It is impossible to prevent ﬁres completely on the shuttle and space
station, however, NASA is continuously developing sensitive detectors and effective
extinguishing methods. Currently, the ﬁre detection and suppression system on the space
shuttle consists of ionization type smoke detectors that identify smoke in microgravity.
Non toxic ﬁre extinguishers containing C02 that will not exceed a dangerous
concentration level can suppress on board ﬁres.

Characterizing the nature of microgravity ﬁres for many materials under various
conditions is an equally important task that NASA strives to improve. Besides the
practical safety applications, quantitative explanations will lead to ﬁmdamental theories
of low gravity ﬂames. Furthermore, many of the underlying mechanisms of on-earth
ﬂames are masked by buoyancy, which causes a churning motion that increases the
mixing of the ﬁre] and oxidizer and produces a more vigorous ﬂame. In microgravity
however, processes that were quashed in l-g are now dominant mechanisms. An
understanding of these processes will give a more complete picture of what really occurs
in l-g ﬂames. Since approximately 85% of the United State’s energy come from
combustion, this knowledge is crucial. Any improvement in the efﬁciency in nuclear,
solar, or hydropower plants, automobiles, jet engines, or even outdoor grills could save
millions or perhaps billions of dollars.

Flames near extinction that are in either microgravity or on earth behave in

peculiar ways, and it is very important to determine the conditions under which these

ﬂames will survive or quench In fact, the consequences of not detecting and
extinguishing a ﬂame on board the shuttle or space station would be disastrous.
Microgravity diffusion ﬂames near extinction exhibit very unstable behaviors. Research
by Dietrich et al. [4] showed that candle ﬂames onboard the Space Shuttle and Mir
oscillated before extinguishing with periodic increases and decreases of ﬂame surface
area. Diffusion ﬂames examined by Olson [5] in a simulated microgravity environment
in low stretch conditions fragmented into cellular ﬂames that also oscillated and
meandered across the fuel surface prior to extinction. In another study by Olson et al.
[6], microgravity ﬂames smoldering over thin fuels produced ﬁngerlike ﬂames that
wandered over the fuel samples. The above examples indicate the elaborate nature of
microgravity ﬂame instabilities. Furthermore, because these ﬂames break apart into
smaller cellular ﬂames, which can travel across the fuel, detection Of them may be
difﬁcult. Knowledge of microgravity ﬂames in the “near extinction limit regime” is
limited because space exploration is still quite young. However, knowing the conditions
where these instabilities occur, and the underlying processes that prolong unstable ﬂames
that could undergo a transition into a dangerous ﬁre in low gravity is of the utmost
importance for space travel safety.

In general, ﬂame instabilities in microgravity are attributed to two mechanisms.
These are thermal diffusive and hydrodynamic instabilities. Thermal diffusive instability
processes occur when an imbalance in the mass and heat transfer in the ﬂame presents
itself as a bulge at various points along the ﬂame sheet. The hydrodynamic instability
occurs when differential thermal expansion of the gases across the ﬂame sheet shows

itself as a corrugation in the ﬂame front. A spreading diffusion ﬂame in microgravity at

 

 

the n
flame

displz

Anal}
Atmo:
ditfusi
diffusi.
endear
instabil
pmducc
Iberma
instabili
exiinctit
making
“The in
Space 3
CXperjmc
will row
Which 1'5
ﬂéime ins.
[hose f0un

R18 Will b

 

the near extinction limit becomes corrugated and later ﬁ'agments into smaller separate
ﬂames called ﬂamelets. These ﬂamelets my oscillate, recombine, or extinguish thus
displaying their unstable and complicated nature.

The overall goal of the NASA ﬂight project named “ATHINA” (stands for
Analysis of Thermodiffusive and Hydrodynamic Instabilities in Near-extinction
Atmospheres), which funded this thesis, is determining the extent to which the thermal
diffusive and hydrodynamic mechanisms contribute to causing instabilities of a spreading
diffusion ﬂame in microgravity. This thesis is the ﬁrst step in accomplishing this
endeavor. In the ﬁrst part of this thesis, the parameter ranges, in which diffusion ﬂame
instabilities of a wide thermally thin sample in a microgravity environment occur, will be
produced and experimentally investigated using the NASA Combustion Tunnel Rig.
Thermally thin ﬁrels are required in the drop tower to study microgravity ﬂame
instabilities because their spread rate is sufﬁciently fast to observe the near limit
extinction behavior. Thicker ﬁrels have much slower ﬂame spread rates than thin fuels,
making it very difﬁcult to observe instabilities in the two or ﬁve seconds of microgravity
time in the drop tower experiments. Even though most combustible materials on the
Space Shuttle and 188 are considered thick, thermally thin fuels provide valuable
experiments that can be used to understand thicker fuels. The second part of this thesis
will focus on the design, construction, and preliminary testing of the “MSU Flame Rig”,
which is a Hele-Shaw apparatus that simulates a low gravity environment and produces
ﬂame instabilities in the near extinction limit regime. The instabilities are similar to
those found in the NASA drop tests. The overall future primary use of the MSU Flame

Rig will be to examine and select thick fuels that will ﬂy on the International Space

 

SI

Ill:

ma
are
ex;

re 1;

Hit
con
defi
SCilI
of t

mus

Station under the ATHINA project. It will also be used to provide additional results to
the drop tower studies for thin fuel ﬂame instabilities. Since there is no limit to the testing
time of the MSU Flame Rig, long periods of simulated low gravity can be produced, so
long-time unstable ﬂame characteristics in this environment will be observed. The
success of the MSU Flame Rig will be important to the NASA ﬂight project because
many materials that are potential fuels for a dangerous ﬁre on the Space Shuttle or 188
are not thermally thin. This research is important for the space program because
experiments of this nature and the study of low gravity unstable ﬂame behavior is still a
relatively new area of study.

This project will be a collaborative eﬂon between NASA Glenn Research Center
Microgravity Combustion Branch and Michigan State University. This thesis is one
component of the larger “ﬂight deﬁnition” project called ATHINA for NASA. “Flight
deﬁnition” means that if certain scientiﬁc requirements are met, the experiment will be
scheduled to ﬂy on the International Space Station (188) in the future. Therefore, much
of this work is preliminary and will serve as a basis for some of the requirements that

must be met to ﬂy on the ISS.

CHAPTER 1
LITERATURE SURVEY

1.1 Overview

Flames are categorized as two types: premixed ﬂames and diffusion ﬂames. A
premixed ﬂame is one in which the fuel and oxidizer are premixed at the molecular level
before combustion, and a diffusion ﬂame is one in which the ﬁre] and oxidizer are not
mixed prior to combustion. In many systems, it is not necessarily advantageous to assert
this distinction because different regions of the ﬂame may have both diffusion and
premixed combustion. An example would be the internal combustion engine. From a
global perspective, the internal combustion engine contains a mixture of fuel vapor and
oxidizer that is sparked to produce combustion. However, on a microscopic scale, the
fuel vapor is really a gas comprised of many tiny droplets that are surrounded by a ﬂame
sheet. Therefore, the individual droplets can be categorized as diffusion ﬂames.

Premixed ﬂames have been investigated more extensively because, unlike
diffusion ﬂames, they can provide ﬁrndamental ﬂame characteristics such as burning
velocity or temperature. The burning velocity is more commonly referred to as ﬂame
speed and is the velocity at which the ﬂame travels into the unburned fuel/oxidizer
mixture. The fuel-to-oxidizer ratios as well as type of fuel and oxidizer are main factors
that affect these speeds, which can be important to combustion safety science. Examples
of premixed ﬂames are Bunsen burners and spark-ignition engines. In this thesis, the
ﬂames studied were diffusion ﬂames because the fuel is a solid.

Diffusion ﬂames lmve been studied less than premixed ﬂames due to their
complex nature. Most ﬂames are of this type in which the ﬁrel and oxidizer have not

formed a mixture prior to combustion. When ignition occurs, the fuel and oxidizer

diffuse toward each other and the reaction occurs in a zone between the reactants forming
a ﬂame sheet. Diffusion ﬂames are complex because they have no fundamental
characteristic that can be identiﬁed to them such as ﬂame velocity or mixture strength
although overall mass burning rates can be readily calculated. Because of this, scientists
in the early stages of diffusion ﬂame research concentrated on ﬂame height and shape
and their affect on soot formation, ﬂame stability, and burning rates. Since most ﬂames in
nature are diffusion ﬂames, scientists are continually studying them to improve engine
efﬁciencies, and ﬁre safety. Examples of diffusion ﬂames are gas jets, coarse oil sprays
or droplet combustion in diesel engines and gas turbines, candle ﬂames, and forest ﬁres.
Since this thesis involved ﬂames spreading over a solid fuel, this literature
review will begin with one of the ﬁrst monumental studies of ﬂame spread theory in
normal gravity. A brief history of some of the signiﬁcant works on diffusion ﬂame
spread follows in Section 1.3. Section 1.4 covers past research in microgravity diffusion
ﬂame spread, and Section 1.5 discusses ﬂame instability work in both normal gravity and
microgravity. Section 1.6 discusses the candle ﬂame in microgravity, and Section 1.7
describes studies that simulated diffusion ﬂames in microgravity. These last works in
Section 1.7 inspired the second part of this thesis, which involved the construction of an
apparatus that simulates the “near extinction limit regime” of spreading diffusion ﬂames
in low gravity. All sections will present the work in a chronological order to illustrate the

development of areas of research while showing how they are related to each other.

1.2 deRis Analysis of Spreading Diffusion Flames
1.2.1 Theory

In 1969, J.N. deRis [2] produced a classic study on the thermal theory of a
spreading diffusion ﬂame in an opposed ﬂow conﬁguration in which the oxidizer and
ﬂame travel in opposite directions shown in Figure l. deRis was the ﬁrst to provide an
analytical solution to the ﬂame spread problem in two dimensions for both a thin and a
semi-inﬁnite fuel bed. Because this was a landmark study and this thesis involves a

ﬂame spreading over a thin fuel, a more detailed description will be presented here.

Diffusion ﬂame

Air velocity Oxygen

 

    

Flame spread velocity
‘

Fuel vapor

 

y
Haj/[\H‘eat L“
Y \ \ X \ \ \ \ \

Unburnt fuel Vaporizing ﬁrel

 

Figure l. Diffusion ﬂame spreading over a solid surface [2].

In his theory, deRis assumed the following.
1. Heat Transfer

a. The constant air velocity affected the forward gas phase conduction.

b. Constant gas phase properties and a uniform velocity proﬁle were
assumed.

c. Downstream convection was accounted for in both the thin and semi-
inﬁnite cases, but radiation was only included for the semi-infmite case.

(1. The thin fuel was so thin that there was a uniform temperature distribution
across it, however, the thick fuel used the fully conductive solution for
temperature.

e. Gravitational buoyant effects were negligible.

2. Fuel Vaporization

a. The fuel was initially in the condensed solid phase and had not yet
vaporized.

b. When the ﬂame approached, fuel surface temperature increased sharply
causing the fuel surface to sublime. deRis assumed that the fuel vapor
pressure was strongly dependent on temperature.

c. The fuel does not vaporize until it reaches the vaporization temperature.

3. Mass Transfer

a. Diffusion and convection of fuel vapor, products, oxygen, and inert gases
were assumed.

b. Fuel mass transfer in the vertical direction off the fuel surface was by
diffusion, and the boundary conditions at the surface of the fuel were
linearized to give an approximation of perpendicular convection.

4. Gas-Phase Combustion

a. Combustion occurred only in the gas phase.

10

b. Fuel and oxygen reacted instantly upon mixing, indicating that the mass
transfer of the reactants rather than the chemical kinetics governed the rate
of combustion. This was referred to as “inﬁnitely fast kinetics” (Burke-
Schumann solutions). The ﬂame sheet was located directly between fuel
and oxidizer zones that did not overlap.

c. Schvab-Zeldovich diffusion ﬂame theory was used.

1.2.2 Thin Fuel Results

Using the above assumptions and fuel theory, deRis solved the thin fuel case ﬁrst
by employing a single global reaction equation, species, mass, and energy equations. He
obtained a set of solvable ordinary differential equations by taking the Fourier transform
of the governing partial differential equations, and produced the following relation

between the ﬂame spread rate and ﬂame temperature shown in Equation 1.

prpertfmp — T...) a- £20"; — Twp) (1)

In this equation, the constant properties pm, 1', 11, Too, Cw, TV.p are the fuel bed density,
fuel bed thickness, gas phase conductivity, far ﬁeld temperature, fuel bed speciﬁc heat,
and the ﬁre] bed vaporization temperature. The remaining variables Tr and V are the
adiabatic ﬂame temperature and ﬂame speed respectively. The left side of Equation 1
represents the amount of heat required for fuel vaporization, whereas the right side
represents the conductive heat that travels ahead of the ﬂame in the gas phase. Equation
1 is an important result illustrating that the ﬂame speed V depends on the adiabatic
stoichiometric ﬂame temperature with linearized mass transfer, Tr. Therefore, ﬂame

speed is independent of pressure and air velocity for a thermally thin fuel. However,

11

Equation 1 cannot be applied to lower oxygen concentration limits, when the ﬂame is
near extinction, because reaction kinetics become important and the thin ﬂame sheet
Burke Schumann approximation is no longer valid.
1.2.3 Semi-Infinite Model Results

The semi-infmite model governing equations consisted of the same gas phase
equations as in the thin fuel case as well as the full solid phase equations for the ﬁre] bed.
In the solid equations, both conduction and radiation were included. A system of
equations was formed with the gas phase and solid phase equations, and an eigenvalue
relation was formed ﬁ'om the ﬁrel bed relations. A solution was obtained by transforming
the governing equations to three simultaneous Wiener-Hopf integral equations that were

solved exactly. The result was a relation between the ﬂame temperature Ty and the ﬂame

 

 

spread rate V shown in Equation 2.
2
Tf —Tvap + 2R1F(2/1/pCpVall + 2R2 _ prpwle (2)
Twp - Too pCpVa(Tvap —T,°) IrpCpVa(Tvap —T°o) pCpra

Equation 2 contains the same constants as in Equation 1 as well as the constant values of
gas phase density p, gas phase speciﬁc heat Cp, radiation constants R., R2, 11, fuel bed
conduction in the y direction 11“,, and air velocity V.. As in the thin fuel model, Equation
2 gives a relation between the ﬂame velocity V and the adiabatic ﬂame temperature. The
semi-inﬁnite model, however, also includes a dependence on the air velocity V., which
was not obtained in the thin fuel case. The ﬁrst two terms on the left side are the forward
heat transfer in the gas phase, while the last term is the forward and downstream radiative
heat transfer. If radiation is neglected though (RI=R2=0), the thin fuel equation is

retained.

12

 

 

fuel
do:
eat

lie:

Vt?

ihi

tht‘
sir
er

()1

Ci
CC

bu

In deRis’ PhD thesis [7], the effect of radiation on the distance between ﬂame and
fuel, or standoff distance was calculated. For the semi-inﬁnite problem, the ﬂame is
close to the fuel, however, for no radiation, the ﬂame touches the surface, which may
cause extinction. Therefore, for thick fuels, the ﬂame spread rate is lower since more
heat is lost to the fuel surface.

The analysis by deRis showed that for thin fuels, the pressure and oxidizer
velocity do not affect the ﬂame spread rate when reaction kinetics are ignored. For
thicker fuels, an equation was derived that revealed the inﬂuences of radiation on ﬂame
spread velocity. The solution was approximate and exhibits physically realistic behavior.
1.3 Other Works of Diffusion Flame Spread in Normal Gravity

The work of deRis provided a basic understanding of opposed ﬂow ﬂame spread
theory that is still used today. The work by McAlevy and Magee [8-11], which was
simultaneous with deRis’ research, supported advances in ﬂame spread theory
experimentally. At the same time, Friedman [l2] wrote the ﬁrst published journal review
of ﬂame spread including the experimental and theoretical ﬂame spread ﬁndings of that
time. Together, these studies provide a foundation of diffusion ﬂame spread that has
helped to produce many of the ﬂame spread theories of today.

McAlevy and Magee’s work contributed experimental improvements to ﬂame
spread theory. They pioneered the use of small wind tunnels, which are now essential in
combustion laboratories, to measure ﬂame spread rates in an opposed oxidizer ﬂow
conﬁguration. They also were the earliest researchers to take advantage of the constant
burning properties of PMMA, which are used in many thick fuel ﬂame spread studies of

today. Moreover, McAlevy and Magee’s explanations are still used today such as the

13

need for understanding the ﬂame zone and aerodynamics of the ﬂame as well as the
Damkohler number (ratio of characteristic ﬂow time to characteristic chemical reaction
time) effects due to heat losses at the ﬂame tip.

In the studies by Sirignano [13-15], a model for ﬂame spread theory was
established which included solid fuel heat conduction ahead of the ﬂame and changes in
ﬁrel density. His analysis was similar to deRis, however, he produced a double branched
ﬂame spread eigenvalue. His contribution was important though because it addressed
some of the criticisms of deRis’ ﬂame spread theory.

Unlike deRis who used inﬁnite rate chemistry, Femandez-Pello and Williams
[16,17] employed ﬁnite rate kinetics in their ﬂame spread model. They also included
ﬁrst, a thorough investigation of the solid phase heat conduction ahead of the ﬂame and
second, the gas phase processes. In their analysis of a downward spreading ﬂame, they
included the effects of gravity throughout their calculations and solved the mass and
momentum gas phase equations. Fernandez-Pello and Williams also utilized one
irreversible Arrhenius reaction (F + vO —> P) with high activation energy to represent the
gas phase combustion, where F is fuel, 0 is oxidizer, and P is product. Overall, their
model was signiﬁcant in that it directly related ﬂame spread rate to forward solid heat
conduction.

Parker [18] and Hirano et a1. [19, 20] carried out considerable experimental
research on ﬂame spread. Parker measured the leading edge ﬂame temperatures in the
gas phase, which inﬂuenced ﬁrrther work on the gas phase velocity and temperature

ﬁelds near the ﬂame tip. Hirano et al. also took gas phase temperature measurements and

14

examined the aerodynamic similarities between the ﬂame leading edge and airfoils by
investigating a stationary ﬂame.

The ﬁrst thorough numerical investigation of ﬂame spread was by Frey and T’ien
[21]. One strong point in their study was the inclusion of ﬂame burnout a ﬁnite distance
downstream from the ﬂame tip. They generated a numerical model of steady-state ﬂame
spread over a thermally thin fuel in an opposed ﬂow conﬁguration. Their analysis was
two-dimensional and contained the elliptic, gas phase energy and species equations
including an overall one step chemical reaction and second order ﬁnite-rate Arrhenius
kinetics. In the unsteady solid phase, a ﬁrst order pyrolysis reaction was utilized as well
as conduction ahead of the ﬂame as a dominant factor affecting the ﬂame spread. They
looked at non-dimensional parameter effects on the ﬂame spread, and obtained good
qualitative agreement between experiment and the model for the changes of ﬂame spread
due to different ambient pressures, oxidizer mass ﬁactions, and opposed velocities. Their
results showed a relation between the ﬂame spread rate and Damkohler number. In
addition, they were able to produce the low reactivity ﬂame zone even near extinction.
This study was important because Frey and T’ien were the ﬁrst to numerically calculate
ﬂame spread extinction limits that until then had not been determined in other gas phase
theories.

Around the same time of Frey and T’ien’s work, Altenkirch et al. [22]
experimentally studied the effects of buoyancy on ﬂames spreading down thin solid fuels.
They concluded that if the ﬂame was assumed to be adiabatic, then the dimensionless

Spread rate correlated with the Damkohler number. They also reported that variations in

15

the buoyant force caused a difference in pressure, which affected ﬂame spread by
changing the chemical reacting rate in the ﬂame.

The above references are some of the major studies that established a basis for
modern day ﬂame spread theory. A detailed review is given by Wichman [23], which
presents the history and progression of opposed ﬂow ﬂame spread theory.

1.4 Diffusion Flame Spread in Microgravity

In the 1970’s NASA began major construction on the next phase of the space
program, prolonged human space ﬂight in the Space Shuttle. As a result, scientists at
NASA and abroad began rigorously studying the effects of microgravity on ﬁre. The
driving factor has been to preserve the safety of the astronauts in ﬂight. Microgravity
ﬂames are studied onboard the space shuttle and space stations, on airplanes that ﬂy in
parabolic arcs (KC-135) and in drop towers where experiments plunge to earth. The drop
towers supply between 2 to 10 seconds of microgravity in which the experiment is in free
fall, and the airplanes give approximately 20 to 30 seconds of 10'2 gravity level
conditions. These facilities have provided information to determine conditions and
parameter ranges that may yield useful results in space. They also can determine the
underlying ﬂame processes that are masked in l-g by buoyancy.

A very useﬁrl study in microgravity ﬂame spread was that of Olson et al. [24]. In
this research, a ﬂammability map for microgravity ﬂame spread over a thin solid fuel was
completed, which showed the effects of relative velocity between the ﬂame and opposed
ﬂow and the molar oxygen concentration of the environment around the ﬂame. The study
focused on the lower limit of opposed ﬂow velocity called “quenching” in which the

ﬂame extinguished due to lack of oxygen. At the opposite branch, called “blowoff’,

l6

which was already constructed from previous work, the Damkohler number is large and
the ﬂame extinguished because the rate of ﬂow was much larger than the rate of chemical
reaction. This study was directly applicable to space ﬂight safety because it showed that
the typical small gusts of air on the shuttle (6-10cm/s) can sustain a microgravity
diffusion ﬂame.

Bhattacharjee et al. [25] produced a theoretical model of a ﬂame spreading over a
thin fuel in a quiescent microgravity environment. In this analysis, the steady, two-
dimensional gas phase energy, species, and momentum equations were solved along with
energy and continuity in the solid phase to obtain a relation for the ﬂame spread rate.
Bhattacharjee et al. produced a dimensionless spread rate that correlated with the
Damkohler number, which agreed with previous studies. They also found that the spread
rates increased with increasing molar oxygen concentration in the quiescent environment.
Measured and computed rates disagreed, however, suggesting the need to include
radiation in the analysis.

Olson [26] further studied the microgravity ﬂame spread map for thin fuels,
discussed in [24], by describing three distinct regions of ﬂammability that are controlled
by different processes. In the low relative velocity region near the quenching branch, she
identiﬁed a new ﬂame control mechanism. In this near limit extinction regime, the ﬂame
was inﬂuenced predominantly by the decrease in oxygen transport due to the increase in
thickness of the boundary layers caused by the slow, Opposed ﬂow. The boundary layers
restricted the convective and diffusive transport of fuel and oxygen to the ﬂame tip. This
weakened the ﬂame, placing it closer to the quenching limit, where ﬁnite rate kinetics

became important.

17

The study of Bhattacharjee et a1. [27] compared experimental and theoretical
predictions of ﬂame spread rate, and total heat ﬂux from a microgravity ﬂame to the fuel
surface. The measurements agreed qualitatively with theory showing the importance of
including both surface and gas phase radiation in the model. An examination that
exhibited the effects of radiation was performed and showed that the surface radiation
decreased the heat ﬂux to the fuel by decreasing the ﬂame temperature, thereby slowing
ﬂame spread. Gas phase radiation increased the heat ﬂux to the surface, but was also lost
to the environment, which also decreased ﬂame temperature and spread rate.
Bhattacharjee et al. also showed that as the oxygen concentration is reduced, the heat
transfer due to radiation becomes a major factor in the survival of the ﬂame.

In 1993, Altenkirch et al. [28], developed a theoretical model and compared
results with drop tower and Space Shuttle experiments of a spreading diffusion ﬂame
when oxidizer velocity and ﬂame spread were of the same order of magnitude. In this
research, radiation was shown to be important at the near extinction limit. The effects Of
pressure on the gas phase radiation were also shown to be important.

Bhattacharjee et al. [29] studied the inﬂuences of pressure on a spreading
microgravity diffusion ﬂame experimentally and computationally. Their experiments
showed that the ﬂame spread rate increased with an increase in ambient pressure. These
results agreed with the computational model, which included gas phase radiation.
However, the experimental ﬁndings did not agree with the model that neglected gas
phase radiation, further supporting the inclusion of radiation in microgravity diffusion
ﬂames.

l8

1.5 Diﬂ'usion Flame Instabilities

This section will review some research on diffusion ﬂame instabilities. It
includes experiments in both normal gravity and microgravity. Most of the numerical
studies neglect gravity. Many studies have been done on diffusion ﬂame instabilities in
normal gravity, however, due to the relatively young age of the Space Shuttle program,
research of diffusion ﬂame instabilities in microgravity is limited. In addition, the
difficulty of observing microgravity ﬂames near extinction for prolonged time periods
has also restricted research. Most of the research presented here was not an examination
of ﬂame spread across thermally thin fuels. Only diffusion ﬂames were discussed that
displayed some of the same unstable behaviors seen in this thesis. This suggested that the
ﬂame instabilities in this study are a fundamental aspect of ﬂame behavior. The candle
ﬂame is considered in the next section in more detail since it closely resembles
qualitative experimental ﬁndings in this thesis. The studies are presented in chronological
order to give a sense of the subject development.

In the inverted wick experiments of Blackshear and Murty [30], ﬂame wrinkles
were seen when the fuel molecular weight increased. Orloff and deRis [31] attributed a
critical Raleigh number for observing cellular instability, which comprised two identical
internal rotating cells of fuel vapor located in a vortex.

W. Chan and J. T’ien [32] conducted an experimental study of spontaneous ﬂame
oscillations near extinction. In this work, which was inspired by Kirkby and Schmitz
[33] and Baliga and T’ien [34], high-speed motion picture and temperature measurements
were used to analyze extinction meclmnisms for wick, tube, and rod ﬂames in normal

gravity. Flame oscillations were local and multi—dimensional, and the unstable behavior

19

originated near the weakest point of ﬂame stabilization or the vicinity of lowest
reactivity. Due to the complexity of multi-dirnensional analysis, a simpliﬁed explanation
for the near limit behavior was proposed. The ﬁrst explanation discussed an increase in
temperature at certain points in the ﬂame, which caused an increase in fuel vaporization.
If the already disturbed ﬂame temperature was also a local maximum, then this increase
in pyrolysis enhanced the ﬂame oscillations. This placed the ﬂame closer to extinction
because it was balancing these processes to survive. The second explanation of Chan and
T’ien discussed a reduction in the concentration of the oxidizer causing the ﬂame to
quench locally due to locally weak chemical reactions. The reduction of oxygen caused
the ﬂame to extinguish, which decreased the temperature, and caused the local induced
velocity to decrease. At this point, the ﬂame tried to propagate into the quenched region,
so the ﬂame “ﬂashed” which appeared as an oscillation in these ﬂames.

Chao, Law and T’ien [35] numerically studied the structure and extinction
attributes of a burning ﬁre] droplet as a characteristic diffusion ﬂame. The interesting
ﬁnding was that radiative heat loss formed two critical extinction turning points. In
particular, smaller sized droplets extinguished due to the decrease in the chemical
reactions. For larger sized droplets, the heat loss by radiation promoted extinction and in
time also caused limited chemical kinetics.

Cellular structures resulting ﬁ‘om instabilities of diffusion slot jet ﬂames in
normal gravity were experimentally investigated by Chen et al. [36]. The unstable
cellular ﬂames were only seen near extinction, and ﬂame behavior was found to be

similar to premixed ﬂame instabilities. Cellular structures occurred for Lewis numbers

20

less than one (~0.8) of the more completely consumed reactant, which agreed with
Linan’s [37] criterion.

A very comprehensive numerical study was carried out by Kim [38,39] in which
he looked at a one-dimensional ﬂame and employed the thermodiffusive instability by
utilizing non-unity Lewis numbers. He neglected the hydrodynamic instability, but
contrary to other works, be accounted for ﬁnite rate chemical kinetics with the
thermodiffusive mechanism. In his analysis, a periodic wrinkling of the reaction sheet
and striped patterns were seen for ﬂames near extinction.

Cheatham and Matalon [40] conducted an investigation of the instability
mechanisms of a burning liquid fuel droplet near extinction by reducing the oxidizer.
The motivation of this study was to simulate the microgravity candle ﬂame, which was
spherical and showed unstable dynamics due to a decrease in the surrounding oxygen
concentration They noticed that reducing the Damkohler number manifested a gradual
rather than sudden transition to unstable conditions. Cheatham and Matalon further
showed that for the near extinction conditions of these ﬂames when the Damkohler
number or oxygen was reduced, radiative heat losses higher than a critical value would
produce unstable, oscillatory ﬂames. The ﬂames moved away ﬁ'om the fuel source when
the Damkohler number was diminished. The ﬂames, though, tended to return to the
undisturbed original location. Oscillatory motion started because radiative heat losses
below a critical value further reduced the ﬂame temperature and Damkohler number,
which again caused the ﬂame sheet to move outward because of weaker chemical
reactions. Oscillations increased in magnitude for large heat loss values or reduced

oxidizer values. In time the ﬂame quenched. Standoff distances were approximately 5 to

21

8mm for oscillation frequencies of order 1H2, which agreed with microgravity candle
ﬂame values.

Matalon [41] numerically researched the onset of the diffusive thermal instability
in a non-premixed ﬂame. He discovered that cellular instability occurred for a speciﬁc
range of Damkohler numbers when there was a loss of reactant from the reaction sheet.
Also, cellular instabilities transpired for fuel LeMs numbers less than one, and the
calculated cell wavelengths were of the same order as those in experiments.

Olson et al. [6] observed ﬁnger-er smoldering in experiments performed on the
Space Shuttle with thin burning cellulosic fuels. The ﬁnger smoldering fronts preferred
to move against the oxidizer ﬂow. Normal gravity smoldering eweriments did not
exhibit ﬁngering, possibly due to the large buoyant ﬂows available to the entire ﬂame
front. In microgravity, it was proposed that the very low ﬂow velocities did not permit
oxygen to be delivered on the sides of the smolder front. Therefore, smoldering in the
front was enhanced, which presented itselfas a bulge or ﬁnger.

1.6 Candle Flame in Microgravity

The most famous and extensively studied diffusion ﬂame in microgravity is the
candle ﬂame because it is a steady, non-propagating, diffusion ﬂame, which is quite
uncommon, but relatively easy to study. Microgravity candle ﬂame experiments were
conducted on board the Space Shuttle and Mir Space Station, and Dietrich, et al. [4],
compared results to a numerical model. Besides steady state behavior and numerical
simulation, unsteady ﬂame oscillations near extinction were observed and modeled.

The experiments on the Shuttle were conducted in a glovebox that allowed

oxidizer to diffuse into the candle environment. The Mir tests, though, used a container

22

with screens that allowed more oxygen to diffuse in due to the increase in open surface
area. AS a result, Mir candle ﬂames lasted longer than Shuttle ﬂames. Flames on both
the Shuttle and Mir began as yellow, sooty ﬂames that eventually changed into dim blue
hemispherical ﬂames. The blue hemisphere indicated the absence of the buoyancy and
the churning motion it causes, which would normally produce a yellow tear drop shape
on earth. In microgravity, less fuel vapor and oxygen mix than on earth, so a lower ﬂame
temperature and burning rate produce the blue color. The ﬂames burned steadily for time
periods ranging from lOOsec. to 45min. Before extinction, the candle ﬂames oscillated at
ﬁequencies of approximately le. The oscillations occurred when the base of the ﬂame
moved up and down, however, the top of the ﬂame remained stationary. This unsteady
behavior continued until the ﬂame extinguished. In the Shuttle experiments, the candle
ﬂame extinguished because inward diffusion of surrounding oxygen through the small
holes of the glovebox was not sufﬁcient to sustain combustion. However, in the Mir
experiments, the longer lasting ﬂames extinguished by fuel consumption. It was also
determined that the candle ﬂame can be considered “quasi-steady” because the duration
Of the ﬂames on Mir was much larger than the gas phase time scales.

The two dimensional axi- symmetric numerical model included ﬁnite rate
chemistry and radiation effects. The end of the wick was considered to be a small sphere
through which fuel evaporated. Numerical reaction rate contours compared well with
experimental ﬂame shapes in steady state for different oxygen mass fractions. Model
behaviors in the near extinction limit regime were brought about by sudden small
decreases in the oxidizer mass ﬁaction, Similar to the steady depletion of oxygen during

experiments. Oscillations in the model became larger in size until extinction, in which

23

the amplitude depended on the magnitude of oxidizer reduction. Frequency rates,
though, were approximately 0.2Hz., but this increased with decreasing ﬁrel Lewis
numbers. The model oscillations agreed qualitatively with the experiment in that the
motion of reaction contours was similar, nonetheless, the size of oscillation in the
experiment was greater. The microgravity candle ﬂame study is very important because
it looks at a diffusion ﬂame without the complex motion of buoyancy in both steady and
unsteady states experimentally and numerically.
1.7 Instabilities in Simulated Microgravity Environments

Experiments have been performed in normal gravity in which the gravitational
effect was “removed”. These types of experiments provide an economical means of
looking at ﬂame instabilities in pseudo-microgravity conditions. Coupled with drop
tower and in-ﬂight experiments, these simulated results can supplement ﬂame theory that
can be applied to characterizing combustion of different materials in space. The second
part of this thesis involved the construction of such an apparatus, which produces ﬂame
instabilities in a reduced buoyant atmosphere. The following experiments motivated this
endeavor.
1.7.1 Low Stretch Flames

In the doctoral thesis of Olson [5], low strain rates (~(g/R)m) yielded a simulation
of a microgravity ﬂame. This was achieved by varying curvature (radius) of the burning
PMMA cylinders. The diffusion ﬂames broke into smaller cellular ﬂames when the
radius of the PMMA was increased beyond a critical limit. The smaller ﬂames, herein
denoted as ﬂamelets, wandered across the surface for long periods of time on the order of

10min. F lamelets interacted with each other through their edges pulsating across the dark

24

non-ﬂaming portions. This dark channel was most likely ﬁrel rich due to the leakage
from the ﬂamelets. Prior to extinction, an entire ﬂamelet would oscillate.
1.7.2 Hele-Shaw Apparatus

Hele-Shaw apparatuses, which are wind tunnels where the test section is made of
two closely spaced parallel plates, have been typically used to observe inviscid ﬂow
patterns [42]. However, ﬂames spreading over thin solid fuels between the two parallel
plates in a Hele-Shaw apparatus will break up into smaller unstable ﬁngerlike ﬂames. If
the distance between the plates is below a “critical” value, buoyancy is suppressed,
thereby simulating a microgravity ﬂame. The critical spacing between the plates can be
estimated using the condition that the Raleigh number based on the spacing is greater
tlmn or equal to 1708 (Ra.I 2 1708) [43]. The Raleigh number is the product of the
Grashof number (ratio of buoyant to viscous forces) and the Prandtl number (ratio of
momentum and thermal diffusivities). For spacing that produces Raleigh numbers below
1708, the commonly known Benard cells or natural convective ﬂow patterns will not
occur. Therefore, the effects of buoyancy may be neglected. The ﬂame depends on
diffusion due to the absence of the bulk motion that would mix the fuel and oxidizer. In
the recent study by Zik et al. [44] a uniform smoldering front over a thin cellulosic ﬁrel in
an opposed ﬂow oxidizer situation became unstable and separated into ﬁngers when the
incoming oxygen ﬂow was decreased. Further reduction of the oxidizer ﬂow led to
splitting where only a few ﬁngers remained and eventually extinguished.

Zik et al. calculated that the two length scales, ﬁnger width and ﬁnger spacing,
were affected by the non-dimensional Peclet number (ratio between molecular convection

and diffusion). They further evaluated the critical Peclet number when the ﬁnger

25

spacing would be zero (a uniform ﬂame front). From the experiments, Zik et al.
concluded that the spacing between ﬁngers was determined by the contention of oxygen
by adjacent ﬁngers. Therefore, they assumed that the ﬁngering was a form of the thermal
diffusive instability mechanism. They also showed that ﬁnger width was linearly
dependent on the spacing between the parallel plates.

Zik et al. and Olson et al. [6] showed that ﬁngering instabilities could occur in the
form of smoldering in which the reaction is not in the form of a ﬂame and does not
appear as the typical yellow color. This instability is difﬁcult to detect and can undergo
transition into a ﬂame if oxygen concentration or heat losses are increased. Therefore, it
is important to know all of the conditions for their survival since they present a ﬁre
hazard onboard the space station.

1.7.3 Microgravity Flame Spread Limits in Hele-Shaw Type Devices

As mentioned earlier, determining the limiting conditions under which
combustion ﬁ'onts will be sustained and propagate over a solid material is a goal of the
space agencies. The Russian Space Agency (RSA) conducted an experimental study to
predict the worst case scenarios of ﬂames spreading over combustible solids [45]. They
based their experimental work on previous investigations that showed a ﬂame
extinguished if the oxidizer ﬂow rate was reduced below a critical value in a microgravity
environment. Therefore, a method of ﬁghting ﬁres in space could be based on
decelerating the fan ﬂow in space station compartments below the critical velocity if
combustion occurs. Coupled with this, a systematic selection of the proper materials,

which have a favorable limiting oxidizer ﬂow, would also be employed.

26

The RSA used three different experimental Hele-Shaw type devices that produce
microgravity conditions on earth:

1. A narrow horizontal chamber device containing two closely spaced parallel

plates, with samples burning on the bottom plate.

2. A vertical chamber device that also suppressed buoyancy by burning
cylindrical samples 1mm. away from a cylindrical quartz tube containing the
sample.

3. A horizontal chamber device with two closely spaced parallel plates with the
burning sample on the upper plate instead of the lower one.

Results from these facilities were compared to data Obtained during prolonged
space ﬂight onboard the Mir Space Station by using the Skorost facility. The Skorost
facility is a combustion ﬂow chamber that allows velocities from 2.5-20cm/s. Three
non-metallic fuels were burned. Data ﬁ'om the Skorost mini wind tunnel was used to test
the validity of the three ground apparatuses.

The RSA obtained limiting values of oxidizer velocity from all facilities, and
suggested that data ﬁ'om ground tests should be taken into account for determining
materials for space ﬂight. The RSA concluded that every effort had to be made to
decrease the oxygen concentration to the lowest allowable limit because doubling the
oxygen concentration lowered the minimum oxidizer velocity that would yield
combustion by four to ﬁve times. The RSA also concluded that further tests in “space”
had to be conducted in order to attain a critical Damkohler number that would aid in

deﬁning the combustion region of the materials.

27

The investigation performed by the RSA demonstrated the difficulty in choosing
non-hazardous materials for spaceﬂight. Currently, the only systematic procedure
involves determining the worst case scenario of ﬂame spread on earth. In the future,
Hele-Shaw devices coupled with drop tower and other ground-based experiments may

prove to be effective components in this process.

28

CHAPTER 2
EXPERIMENT

2.1 Overview

Experiments were conducted using the “Combustion Tunnel Rig” (CTR) at
NASA Glenn Research Center in Cleveland, Ohio, and the “MSU Flame Rig” at
Michigan State University in East Lansing, Michigan.

The CTR was a wind tunnel that was dropped in either the 2.2 Second Drop
Tower or 5 Second Drop tower at NASA that provide short periods of microgravity
conditions. Microgravity ﬂames were produced during the drops, by burning thin fuel
samples inside the CTR while low speed oxidizer ﬂowed across the fuel opposite to the
ﬂame propagation direction. Videos of the front and side views of the burning sample
were taken, and surface thermocouples recorded the temperature of the samples during
the drops. Results were analyzed using “TRACKER”, an image processing system
capable of tracking the ﬂame over time, which led to ﬂame speed measurements.
Thermocouple data was reduced using molliﬁcation and spline algorithms, which gave
temperature measurements over time. These thermocouple results were used for an
energy balance at the surface of the fuel sample. The overall use of the CTR in this thesis
was to locate the near extinction regimes of the thin fuels that can be difﬁcult to ﬁnd and
perform a preliminary analysis of these experimental results.

The “MSU Flame Rig” is a Hele-Shaw apparatus, which is a wind tunnel where
the top plate of the test section was placed at a very close distance to the ﬂoor of the test
section. Hele-Shaw devices are used to visualize inviscid and irrotational ﬂows in the two
dimensional plane parallel to the top plate, which occur due to the viscous ﬂow produced

between the sample and top plate. Another aspect of Hele-Shaw apparatuses was that

29

buoyant ﬂows caused by ﬂames inside the test section were suppressed by the close
proximity of the top plate to the bottom of the test section. The MSU Flame Rig utilized
this feature of the Hele-Shaw device to produce a simulated microgravity ﬂame near
extinction. The near extinction conditions resulted from the low air speeds (l-28cm/s)
produced in the test section, which were smaller than characteristic buoyant ﬂows
(approximately 30 cm/s). These velocities ensured that the burning samples on the
bottom plate did not extinguish due to blowout and that diffusive ﬂow was the main
process for mixing of fuel and oxidizer. The samples in the MSU Flame Rig were in the
same opposed ﬂow conﬁguration as the CTR, however fuels with different thickness can
be accommodated in the MSU Flame Rig. Distances between the sample and top plate
varied from 2-10mm. Most of the work in this portion of the thesis was devoted to the
successﬁrl design and fabrication of a device that burned thin samples that exhibited near
extinction limit behavior. Future results ﬁ'om this device will aid in the selection of thick
fuels that will ﬂy on the International Space Station (188), help to determine the
maximum sample width, and allow more time to qualitatively study ﬂame instability.
2.2 NASA Combustion Tunnel Rig
2.2.1 NASA Samples

A schematic of the experimental setup used in the CTR is shown in Figure 2. The
cellulosic, thin, solid ﬁrel used was Kimberly Clark1 brand Kimwipes 10.5cm. in width
and less than 10"in. thick A Kanthall brand hotwire tlmt was formed into many

sinusoidal curves ignited the sample. The sample was held straight and taut by gluing it

 

' Trade names or manufacturer’s names are used in this report for identiﬁcation purposes
only. This usage does not constitute an Ofﬁcial endorsement, either expressed or implied,
by Michigan State University or the National Aeronautics and Space Administration.

30

to the outer edges of the sample holder frame, which was either an aluminum

(conducting) holder or a polymeric (insulating) holder.

Front View Side View

 
   

Flame l===2======: ' e
Propagation
Direction

 

 

Nitrocellulose

S l HeatSink Insulated
am e -
1 l te P Backing Holder

Holder Sample

1“.— 105 cm—l

Figure 2. NASA sample front and side schematic.

Thicker fuels generated slower spread rates than thin ﬁrels. A fuel was considered
thermally thin if conduction into the thickness of the fuel (perpendicular to the ﬂow and
spread in Figure 2 side view) was negligible. In other words, the sample was not so thick
that it acted as a heat sink draining the ﬂame of its energy because gradients of
temperature across the thickness were inﬁnitesimal {(ATmpte mundATmaam-cxd}.
Thermally thin, cellulosic Kimwipes were chosen because the spread rates were fast
enough for ﬂames to show unstable behavior during the 2.25ec. and 5sec. drop periods.

Samples were ignited in normal gravity and then dropped after a one-dimensional

ﬂat ﬂame front was established like the one shown in Figure 3a. It was important that the

31

ﬂame ﬁont was linear as opposed to a parabolic ﬂame shape that lags near the sides of
the sample holder, like the one shown in Figure 3b. This is because for a parabolic ﬂame

front, the ﬂame instabilities might be attributed to edge effects instead of inherent ﬂame

dynamics.

   

Figure 3a. Flat, linear ﬂame ﬂoat in l-g. Figure 3b. Parabolic ﬂame ﬁont in l-g.

Several conditions assured the generation of a ﬂat spreading ﬂame front. The
wire was woven into dense sinusoids, so that better contact between the ﬁrel and wire was
achieved. Thus, uniform ignition across the entire sample was more likely instead of
patches of ﬂames travelling down the sample. An alumina ceramic bar held the igniter in
place, so that the wire would not sag when heated and cause a parabolic ﬂame front that
trailed at the sides. Also, a thin strip of nitrocellulose was glued across the top of the
sample that served as an extra source of initial heat generation for ignition. The thermal
expansion caused by exothermic heat release by the nitrocellulose did not complicate the
problem, since its effects were washed away by the opposed ﬂow before the drop

commenced. Lastly, the metal holder used in the beginning of testing was replaced with

32

an insulating holder because it caused the ﬂame front to be parabolic as it conducted too
much heat away at the sides of the ﬂame.

It was clear that the samples had to be ignited in normal gravity and then dropped
because it was nearly impossible to produce a ﬂat ﬂame in the near extinction regime
during the 2.2sec. or Ssec. drop time. If a front like the one shown in Figure 3a occurred,
the rig was dropped in the drop tower, and the ﬂame underwent transition into a ﬂat,

linear microgravity ﬂame front. An example is shown in Figure 4. Hence, if a ﬂame as in

 

Figure 4. Flat, linear ﬂame front in microgravity.

Figure 4 later corrugated, produced ﬂamelets, and progressed to near extinction, it was
not because of inﬂuences at the sides of the sample near the holder.
2.2.2 NASA Test Conditions and Flow System

As stated previously, a ﬂame in low gravity will enter into the near limit regime
and extinguish if one of the major components for ﬂame survival (fuel, oxidizer, or heat)
is reduced below a critical level. In the beginning of testing, the concentration of oxygen
was reduced to 17-18% in order to reach near limit behavior. Since it was difﬁcult to

produce a uniform ﬂat ﬂame front in normal gravity at such low oxygen percents, it was

33

obvious that a uniform ﬂame would not occur in low gravity. Hence, instead of lowering
the oxidizer mass ﬁaction, a metal or insulator backing that served as a heat sink was
placed a ﬁnite distance away ﬁ'om the sample. The backing provided a means with
which to vary the heat loss in the tests, so that the near extinction limit could be attained
while burning in air. This is best illustrated by observing the ﬂammability map of Olson

[24, 26] in Figure 5. In both normal gravity and microgravity, the burning sample was in

Flammability Map

 

MICROGRAVITY DATA
0 quiescent ﬂame spread
0 low speed forced convection
X ﬂame extinction

NORMAL GRAVITY DATA
I: high speed forced convection
A natural convection

Molar
O2 %

 

---newextinction1imitlino
duetoheatsinkbackinL

 

 

 

 

 

WWW
0-1 l 10 100 1000
characteristic relative velocity (cm/s)
Figure 5. Flammability Map of Olson [24, 26] with new extinction limit branch.

the ﬂame regime. For each drop, the ﬂame followed the green line from normal gravity
to microgravity. 1f conditions were favorable (low oxidizer ﬂows or low characteristic
relative velocities), the sample would reach the near extinction limit in microgravity,
Which corresponded to the quenching branch on the left. However, by employing the

heat sink backing, the quenching branch was in effect moved up and to the right as shown

34

by the portion of the red dashed line. This new displaced quenching branch enabled
higher oxygen concentration conditions such as air to be used. In summary, air was
preferable because steady downward spread in normal gravity was easily established
prior to each drop, and the backing (not the low oxygen concentration) enhanced
transition to instability when the drop occurred.

Care was taken to prevent the sample and backing from touching in order to avoid
heat conduction changing the ﬁrndamental processes of the burning sample, even though
it is considered a thermally thin fuel. At close distances around 5mm, the backing
quenched the ﬂame that would have otherwise existed on the sample back side (if there
were no backing, two symmetric ﬂames would spread across opposite sides Of the
sample).

Air velocities with uniform proﬁles from 1-8cm/s were produced in the test
section that contained the sample. A schematic and picture of the ﬂow system on board
the CTR are shown in Figures 6a and 6b. The tunnel reservoir stored up to 1800psia of
gas mixtures that contained less than 50% of Oz. The reservoir had to be ﬁlled before
each drop, so it would not entirely drain to zero gage pressure. This was especially
important for tests that required high velocities since the reservoir pressure decreased
more quickly than at low speeds. Instructions for ﬁlling the rig reservoir with air at
NASA are found in the Combustion Tunnel Rig (CTR) User’s Guide [46]. The pressure
regulator was adjusted to set the velocity in the test section, while the calibrated critical
ﬂow nozzle controlled the ﬂow and provided choked ﬂow velocities above Zem/s in the
test section. Mass ﬂow rates downstream of the oriﬁce remained choked as long as the

upstream pressure was twice as high as the downstream pressure. Calibrations of the

35

Sample Tunnel

‘ ‘ .. . ,3. .4‘ _._ LTOM '
. . .ttgitg _turml _eservoir
. i 3's? A" r g 9%}, |
“‘1

    
 

T vent

 

 

 

 

 

 

 

T
Lid hand
T “E ) valve
pressure
Iii; regulator) 6K)
F—el—E-
” ‘° solenoid mm

 

Hand Valve

 

 

 
   

 

Test
Section

 
 

.e ii‘éﬁ'r sm

 

t was» Qx~

(Red valve)

 

      

Pressure

 

  

 
 

Critical
Flow Oriﬁce

    

 

Figure 6b. Picture of ﬂow system in Combustion Tunnel Rig.

36

critical ﬂow oriﬁce were performed at NASA Glenn Research Center and can be found in
the CTR User’s Guide [46]. Porous plates in the plenum straightened the ﬂow before
entering the test section. Flow uniformity in the test section was veriﬁed by smoke
visualization Flow was exhausted to the atmosphere through a T-vent in the 2.2 Second
Drop Tower, and through a back-pressure regulator in the 5 Second Drop Tower. The
back-pressure regulator was needed because rigs were dropped in a vacuum in the 5
Second facility, so it maintained atmospheric pressure in the tunnel test section.

Prior to dropping, when the rig was secured at top of the drop tower, the test
section was ﬂushed with oxidizer ﬂow in order to establish uniform ﬂow conditions. The
amount of time for ﬁlling the test section varied with ﬂow magnitude, so for higher ﬂow
rates less time was needed. The ﬂush times are located in the CTR User’s Guide [46].
2.2.3 NASA Rig Electrical System and Diagnostics

The CRT was a self-sufﬁcient wind tunnel that operated in an isolated
environment for short periods of time. Three batteries on board the rig supplied power
to the electrical components. A model IV Tattletale Control System along with TTOOLS
computer software recorded the drop data and regulated different electrical processes on
the rig. Software programs that controlled data acquisition and rig electrical switches
were written in the BASIC language. Before each drop, programs were loaded and stored
in the on board computer, which ran the drop sequence. Hence, turning ﬂow on and off,
igniting the sample and other processes necessary for a successful drop did not have to be
done manually.

Several digital and analog channels permitted certain diagnostics on the rig. Four

pressure transducers at various points on the rig tabulated the pressure versus time for

37

each drop. These measurements were also needed for ﬁlling the rig reservoir with
oxidizer and setting the ﬂow rate. Therefore, the velocity through the test section was
monitored when ﬁlling and throughout the drop. Three portholes around the test section
chamber allowed video cameras or 16mm. color ﬁlm cameras for visualizing the front,
back, or side view of the tests. In this project, video cameras recorded the front and side
views of the burning sample. Type K thermocouples were incorporated into the rig
diagnostics. Up to three thermocouples recorded the temperature at the sample surface,
and either side of the heat sink backing. The temperature was measured at time intervals
tint could be preset by the CTR user. Thermocouples were sewn into the sample, so the
bead had thermal-physical contact with the fuel surface. The location of the
thermocouple on the sample was marked, so it could be seen on the video recording
during the real time of the drop. Since the researcher conducting the tests indicated to the
tunnel operator when to drop, this marker was essential to obtain the correct drop timing.
In this way, the researcher could determine when to drop the rig, so that the ﬂame would
spread over the thermocouple when it was in microgravity. Hence, it was assured that the
surface temperature of the microgravity ﬂame was measured regardless whether the
ﬂame or ﬂamelet was stable or unstable. In other words, it was not known prior to the
drop where along the ﬂame front breakup would occur. Consequently, it was not
completely certain beforehand whether the thermocouple would measure the temperature
of the ﬂame front, ﬂamelet, or the dead space between ﬂamelets. Besides sample surface
measurements, substrate temperatures were, recorded by gluing a thermocouple on the

front side of the backing for good contact.

38

2.2.4 NASA Drop Tower Facilities

NASA Glenn Research Center has two microgravity drop towers, a 2.2 Second
Drop Tower and 5 Second Drop Tower. The 2.2 Second Drop Tower has been used to
obtain preliminary data. This was because the turnover time was much faster, so as many
as 3 drops per day were conducted. Therefore, a large amount of information was
obtained in a smaller time period. The 5 Second Drop Tower is used for experiments that
are further along in their experimental programs. However, scientists have gained
information in the 2.2 Second Drop Tower that was sufﬁcient for their research.

Drop towers produce short periods of near zero gravity environments. The
following theory behind this is explained by considering the forces on objects located on
earth and in its vicinity. The ground provides a feeling of weight because of the
reactionary force that it exerts. When a person is falling toward the earth, though, there is
no feeling of their weight, hence they experience “weightlessness”. This condition is
called “free fall” or “zero gravity”. All objects falling towards earth experience the same
acceleration (g=9.8m/sz) in the absence of air resistance. Any object that moves ﬁ'eely
under the control of gravity regardless of initial conditions is said to be in a state of “free
fall”.

There are two ways to achieve zero gravity. The ﬁrst is to travel far enough away
from the earth’s surface, so that the force of gravity has no effect. This originates from

Newton’s equation of the force of attraction between two bodies of mass m1 and m;:

F = 01","; (3)

r

In Equation 3, G is the gravitational force and r is the distance between the two particles.

Therefore, if the distance between the mass (m) and the earth (m2) is very large

39

(r2>>Gm1m2) then the force between the two objects will be close to zero. If this is
applied to a person and the earth, then in order for the object to feel zero force or zero
gravity, the person would have to travel a distance far from the earth.

The second and more utilized way of achieving microgravity or ﬁ‘eefall was ﬁrst
explained in a thought experiment by Sir Isaac Newton. If a cannon is placed at the top
of a very high mountain and ﬁred, the cannonball will travel along a parabolic trajectory
and eventually hit the earth. However, if the cannonball is ﬁred with a large amount of
energy, it will reach a point where it will not return to the ground. In this state, the
cannonball is continuously falling not toward earth but around it. This condition of
continuous freefall is called orbit, in which the object never reaches the surface of the
earth. This is how astronauts feel weightlessness in space. The space shuttle travels with
sufﬁcient speed to reach earth’s orbit, in which the astronauts and all objects in the
shuttle fall together around the earth.

The drop tower facilities use the same principles to produce microgravity. The
experiments eventually hit ground, though, so the time of microgravity is not continuous
like in orbit, but much shorter. CTR experiments were placed in a capsule, which was
analogous to the space shuttle because they shielded the objects inside from the drag. In
this way, the objects inside the capsule ﬂoated (or dropped) at the same acceleration, thus
producing microgravity or weightlessness. Taller drop towers produce longer periods of
microgravity.

The smaller of the two drop towers at NASA produces 2.2 sec. of 10‘4 gravity

levels. A schematic of the tower is shown in Figure 7. Experiments in this tower fell

40

 

 

Figure 7. 2.2 Second Drop Tower schematic.

ﬁ'om the top ﬂoor below which there were various labs where researchers prepared
experiments and staff members built rigs and operated the tower for drops.

A picture of a typical drag shield in the 2.2 Second Tower is shown in Figure 8.
Each experiment was placed inside the drag shield on the ﬁfth ﬂoor. The drag shields in
this tower had a high weight to ﬁ'ontal area ratio, so drag coefficients were low. Drag
shields were up to 101.6cm. in height, 96.5cm. in width, and up to 158kg. (3501bs.) of
weight. The drag shield that contained the CTR experimental rig was lifted to the top of
the tower. After the CTR rig diagnostics and other parameters for the experiment were
prepared, the rig was dropped. The rig freely fell 20cm. within the drag shield, while the

entire drag shield and experiment fell 24.1m. into a lOft. airbag. The entire experimental

41

 

Figure 8. Drag shield in 2.2 Second Drop Tower.

package endured gravity levels between 15g. and 30g. upon deceleration.

A schematic of the ﬁve second drop or “Zero-G Tower” is shown in Figure 9. The
“Zero-Gravity Research Facility” was much more complex than the 2.2 Second Tower
because the size of the tower, which was actually a 155m. shaft into the ground, was
approximately as high as the Washington Monument. Another intricate feature of the
tower was that before each drop, the shaft was pumped to a vacuum. A larger staff than
in the 2.2 Second Tower was required to operate the vacuum hydraulic system. The
entire process usually lasted 45min. on a dry day and more than 1 1Ah. on humid days.

The vacuum was necessary because air drag would slow the acceleration of the capsule

42

   
    
 
    
 
 
    

  
 

20 Ton
Vacuum chamber crane
cap & experiment -.

loading area

   

 

 

 

Vacuum drop
chamber

 

'ottom of chamber1
155 meters below
ground level

 

Figure 9. Zero-G or 5 Second Drop Tower facility.

containing the experiment if it dropped 155m. in atmosphere. A large 20ton crane
hoisted the experiment that was no more than 455kg. (1003lbs.) at the top of the tower.
The experiment was again placed in an experimental capsule, which were
cylindrical and measured up to 3.4m. in length and 1m. in diameter. A typical test where
a capsule is being hoisted to the top of the tower is shown in Figure 10. These large
enclosed containers were needed to accommodate all of the electrical components that
operated the rig during the drop. There were no connections ﬁ'om the rig to the tower
once the experiment was dropped When the capsule was at the top of the tower and
sealed from atmosphere, an umbilical cable, connected to the top of the capsule, provided

electrical power and communication to and from the experiment before the drop. Seconds

43

 

Figure 10. Drag shield in the Zero-G Drop Tower.

before the drop, pins holding the capsule were removed, so that the capsule was hanging
from a support shaft less than lin. in diameter. A pneumatically controlled piston hit the
support shaft, shearing it and dropping the capsule. Therefore, a drop from rest was
achieved. During the drop, the electrical system, which contained video recorders and all
other electrical components, gathered and stored the data. Gravity levels during the drop
were 10'5g., producing 5.18sec. of microgravity.

At the bottom of the tower, the capsule decelerated in a pool of polystyrene
pellets. The radius of the cylindrical pool was larger than the capsule by inches.
Therefore, it was the viscous resistance of the polystyrene that slowed the capsule to a
halt. Within this 150millisec., the capsule deceleration was a ramp function during which

it experienced ﬁ'om 0g to 65g’s.

Due to the time to prepare for a drop in the Zero-Gravity Facility, no more than 2
drops per day were performed. Consequently, many researchers use this facility in the
ﬁnal experimental stages of their projects.

2.2.5 NASA Rig Test Procedure

In this section, an overall procedure is presented in order to give a notion of the
steps involved in conducting the drop tower experiments in this project. The following
steps illustrate the procedure that was followed to accomplish a drop. Since then,
improvements of the test method have been made. In particular, the nitrocellulose is no
longer used, and the igniter system is like the one that will be explained on the MSU

Flame Rig.

Sample Preparation

0 Glue nitrocellulose on cellulose fuel and glue sample to sample holder.

0 Form igniter wire, attach to wire leads, and weave sample through it.

0 Thread thermocouple in sample, mark it, and attach thermocouple to leads.
0 Screw hacking into place with thermocouple on front surface near sample.
0 Place sample holder in rig so it is secure and immobile.

0 Connect thermocouples leads to female ends at top of chamber. Make sure to
note which thermocouples correspond to their respective channels.

a Place top on test section and secure in place.

0 Fill rig reservoir with no more than 1800psia of gas (air, 02-N2 mixture) (See

CTR User’s Guide [46] for complete procedure of this step).

45

Set ﬂow velocity in test section by adjusting the pressure regulator to the desired
ﬂow. CTR User’s Guide contains velocity and pressure calibrations for this step.
(Complete procedure of this step is in CTR User’s Guide [46] ).

Make sure solenoid valve is off.

Cap T vent at top of chamber if leaving overnight.

Electrical Compgnents

Set igniter wire current to proper value, so that the wire will not break.
Calibrations for this step are in CTR User’s Guide. Igniter current controller
(ICC) is on side of the test section chamber, and switch must be ﬂipped to “set”
position. Also, igniter switch on Power Control Box (PCB) must be in the “on”
position. Make sure to switch ICC back to operate when ﬁnished. (See CTR
User’s Guide [46] ).

Check focus of video cameras by connecting to VCR and viewing on TV.
Reconnect cables to PCB.

Charge batteries with power box while the batteries are off.

On PCB make sure of the following: (See CTR User’s Guide)

3 Batteries are ON.

Igniter switches are OFF.

Manual Enable, Video Power, Aux 2, Tattletale Computer are ON.

Take caps off T vent.

Drop Procedure

Bring rig to 5“I ﬂoor and proceed with drag-shield packaging procedures. Tunnel

operator will help with this.

46

Connect ﬁber-optic cables that will transmit video data to VCR’s. Make sure
there are no tangles since ﬁber-optic cables damage easily.

Go to 8'” ﬂoor where rig will be hoisted.

Insert video tapes and record drop number for reference. Set videos to proper
channels.

Make sure red, green, and yellow switches on control panel are in off position.
This is very important since these control the ﬂow and igniter switches.

Open side of drag shield and connect computer cable to laptop computer.

Connect three cables at top of rig with the leads and alligator clips to either a red,
green, or yellow circuit. Make sure the two far cables have the same color code,
which means they are part of the same circuit. The third cable will indicate when
the rig is dropped.

Turn ON batteries and laptop computer.

Go to TTOOLS on computer and at the prompt load in proper program, which
corresponds to drop tower. Type run, then answer questions including ﬂow ﬂush
times, and unplug computer cable. Close drag shield.

Press play and record on VCR’s and announce the drop.

Flip gas solenoid circuit switch on control panel. This will start ﬂow ﬂushing into
chamber for a proper period of time before the igniter will automatically turn on.
While viewing video screen, decide when a ﬂat ﬂame front has developed before

reaching the thermocouple, and say “DROP”.

Post Drop Procedures

Collect Tapes and proceed to 5th ﬂoor to remove drag shield.

47

0 Bring rig to sample preparation area and connect commuter cable and battery
cables.
0 Open TTOOLS and type run to download pressure and thermocouple drop data.
0 Record drop data in blue log book, review drop tapes, and save drop ﬁle data
2.3 MSU Flame Rig
2.3.1 MSU Flame Rig Materiab
A picture of the MSU Flame Rig is shown in Figure 11. All materials for this
experimental apparatus were chosen with safety in mind. Over 90% of the materials
were either non-ﬂammable or had very high ignition temperatures. The very few
ﬂammable materials were placed at locations where ﬂames from the tests could not

possibly ignite them The plenum was constructed of aluminum plates, and the

Test Section

.ﬂ \
a

 

Figure 1 1. Picture of MSU Flame Rig.

48

contraction was a steel duct. The plenum also contained an aluminum pipe inside
through which the air entered. The porous plate inside the aluminum holder was
fabricated from stainless steel. The supporting plate and stilts of the test section were
aluminum. All spacers, which adjusted the height of the sample in the test section, were
made of either ground steel stock, aluminum bars, or steel shims. The sample tray was
aluminum with a mica piece to which the sample was glued. Pyrex windows, which
never experienced direct ﬂame, formed the side walls of the test section. A fused quartz
plate used to view the ﬂame was the only material that experienced direct ﬂame.
However, it was the purest form of glass that could be used having an operating
temperature (1250°C) and softening temperature range (1500-1680°C) well above the
temperatures of the ﬂames in the MSU Flame Rig. The plenum, contraction, and porous
plate holder were sealed using O-rings. Due to the difﬁculty of removing the quartz plate
after testing and not allowing any stresses on it, a soft, removable putty was used to seal
the test section, but it never came in close contact to the ﬂames.

A safe and durable ﬂow system was also constructed. Since the ﬂow source
produced shop air at 95-110psig, all ﬂow components used were rated to pressures well
above any operating values. Polyurethane tubing rated to lSOpsig connected all of the
components.

2.3.2 MSU Flame Rig Flow Control and Measurement

A photo and a block diagram of the MSU Flame Rig ﬂow system are shown in

Figures 12a and 12b. Oxidizer was supplied to a nozzle ﬁom the in-line shop air in the

Engineering Research Complex at between 95psig and llOpsig. The red arrows

49

 

Figure 12a. Picture of MSU Flame Rig ﬂow system.

 

To
Plenum

 

 

 

 

Figure 12b. Block diagram of MSU Flame Rig ﬂow system.

represent the air ﬂow direction. The air ﬂowed through a 5micron ﬁlter, which was
chosen because its size was approximately a thousand times less than any of the sonic
oriﬁces. Therefore the tiny oriﬁces would not clog because nothing larger than 1/1000 of
the oriﬁce size would pass through them. After the ﬁlter, the ﬂow entered a pressure
regulator that was used to set the ﬂow rate in the sonic oriﬁces.

The ﬂowmeter was placed directly after the pressure regulator to determine the
test section ﬂowrate and velocity. Since the ﬂowmeter was actually a rotarneter
calibrated at standard conditions (70°F and 14.7psig) with a range of S-SOscﬂr, it read a
lower volumetric ﬂowrate then the actual rate when it was placed in line and not
exhausting to atmosphere. This occurred because the density was substantially higher at
in-line pressures, so the volumetric ﬂowrate was lower for the same mass ﬂowrate, since
volumetric ﬂowrate equals mass ﬂowrate divided by the density. The actual volumetric
ﬂowrate was obtained using a correction factor, which will be explained in more detail in
Chapter 3 of this thesis.

After the rotameter, the ﬂow line split with a needle valve on one side and a
critical ﬂow oriﬁce on the other side. The needle valve was used to provide a surge of air
at the beginning of tests to establish a uniform ﬂame front. At this point in the test, the
ﬂow traveled predominantly through the needle valve because the diameter of the critical
ﬂow oriﬁce was hundreds of times smaller than the needle valve diameter. In other
words, the critical oriﬁce appeared as a closed port to the ﬂow, so it preferred to go
through the needle valve. When a uniform ﬂame front like the one in Figure 13a was
established, the needle valve was closed at a constant rate until either the needle valve

was barely open or the needle valve was closed and air ﬂowed only through the critical

51

ﬂow oriﬁce. This was one of the main factors that produced a corrugated ﬂame front like

the one in Figure 13b. The corrugated ﬂame front eventually led to ﬂame instabilities.

 

Figure 13b. Corrugated, unstable ﬂame front in MSU Flame Rig.

The critical ﬂow oriﬁces were used to produce the low mass ﬂowrates at which
ﬂame instabilities occurred. The oriﬁces were required because they allowed extreme
control of ﬂow on the order of centimeters per second because the ﬂow reached sonic
velocity at the throat. When the air ﬂowed only through the oriﬁces and not the needle

valve, the pressure regulator was used to control the ﬂow in the test section. However, to

52

ensure sonic ﬂow in the throat and consequently be certain of the mass ﬂowrate, the

compressed ﬂow relation in Equation 4 had to be satisﬁed.

P2 /1>l s .528 (4)

Here P. is the back-pressure or pressure of the regulator and P2 is the downstream
pressure. If the back-pressure is higher than approximately twice the downstream
pressure, then the ﬂow is choked at the nozzle throat. However, different ﬂowrates in the
test section were achieved by setting the back-pressure at values higher than twice the
downstream pressure. In these cases, the ﬂow was still sonic at the throat, but the mass
ﬂowrate at the throat changed because the density varied due to the different pressure in
the nozzle throat. One engineering setback of the nozzles, however, was that the ranges
of ﬂowrates for each nozzle were quite small, so the needle valve was utilized in
conjunction with them. Otherwise a complicated array of oriﬁces would be necessary,

and calibrations for these systems are often inaccurate.

2.3.3 MSU Flame Rig Plenum and Contraction

The critical ﬂow oriﬁce and needle valve were in two separate lines in parallel.
After the air ﬂowed through them, the two lines joined to form one ﬂow path again. This
ﬂow entered a 2in. diameter pipe that had three rows of 1/8in. diameter holes facing the
back of the plenum (see Figure 14). The size of the small holes was chosen to be
substantially less than the diameter of the pipe so that the ﬂow entered the plenum

through all of the pipe holes. In this way, the pipe appeared as a large plenum to the

53

 

2 inch diameter lenurn back wall

inlet pipe

 
  
   
 

 
 

 

 

 

 

 

inch inlet 1
. ortholes

 

 

Figure 14. MSU Flame Rig plenum schematic.

smaller holes. If the small holes were close to the size of the pipe diameter, then the air
may have only entered the plenum through the ﬁrst few holes of the pipe, thus producing
an irregular downstream ﬂow. The small holes in the pipe faced to the back wall of the
plenum to dissipate some of the entry ﬂow kinetic energy, so any individual jets of air
were less likely to affect the downstream ﬂow. The size of the plenum (1ft. x 13in. x
6in.) was larger than the test section to create a region with lower velocities than those in

the test section.

After the plenum, the lit. long by 13in. wide aluminum duct shown in Figure 15
contracted the ﬂow to the 2.3in. height of the entire test section. The contraction

contained a ramp that constricted the ﬂow the cross sectional area of the test section.

Both the plenum and contraction were welded along the side seams to prevent any
gas leakage. One-quarter inch plates with the duct openings in the middle were welded to
the ends of the plenum and contraction. These plates made it possible to facilitate
grooves for O-ring seals. The entire apparatus was constructed in separate pieces to

enable future additions or adjustments without reproducing the entire device.

54

“'V}.

.v

. Porous Plate Hoder

 

Figure 15. MSU Flame Rig contraction in ﬂow system.

2.3.4 MSU Flame Rig Porous Plate

The purpose of the porous plate was to ensure laminar test section ﬂow. In the
MSU Flame Rig, the .062in. stainless steel porous plates had a pore size of either
5microns or 10microns. In practice, the porous plate was the equivalent of many screens,
which in other experiments have been used to generate straight and uniform ﬂow. In the
MSU Flame Rig, the porous plate, shown in Figure 16, created “uniform” ﬂow by
making each ﬂow particle follow a tortuous path. In essence, the porous plate was a large
resistance that caused the upstream velocity distribution to “spread” out into a uniform

proﬁle downstream.

55

 

 

 

Figure 16. Air path through porous plate in MSU Flame Rig.

Since the plate had very small pores, the effect of the plate was to cause the ﬂow
lines to extend evenly over the entire face of the plate. Another way to describe this uses
energy balances. When the ﬂow travels through the plate, much of the kinetic energy
(due to velocity) is converted to heat energy due to the large viscous forces in the plate.
Therefore, mechanical energy ﬁ'om the ﬂow is dissipated, causing the velocity
distribution to become everywhere even. This was shown by Baines et al. [47] who
compared experimental velocity distributions of air through screens with Prandtl’s
theoretical expression for velocity modiﬁcation [48], shown in Equation 5.

Here it is the pressure drop coefﬁcient (k=AP/(pV2.,/2)), V0 is the mean velocity in the
free stream far from the porous plate, V“ is the velocity upstream of the plate, and V4 is
the velocity downstream of the plate. Equation 5 illustrates that the downstream velocity

depends highly on k, a measure of the energy change through the screen. Baines et al.

56

found that screens with very small pores created an even downstream velocity

distribution.

A ramp shown in Figure 17 was placed on the upstream side of the porous plate,

so the ﬂow was gradually guided into the porous plate. The ramp was connected to the

 

Top of
Contraction

 
  
   
 
  

Porous
ma

 

 

 

 

Air Flow
Directio

    
    

 

r ' r

 

‘1 — . ‘
Adjustable
[Floor of Contraction] “$33122" Spacers

 

 

Figure 17. Double blocking plate system for porous plate in MSU Flame Rig.

upstream “porous plate blocker” (ﬁrst blocker plate on upstream side of porous plate).
The ramp and porous plate blocker prevented mass ﬂow from being lost underneath the
test section. If the air was allowed to travel through the porous plate in the space below
the ﬂow opening, it would be nearly impossible to calculate correct ﬂowrates using the
methods described earlier because the ﬂowrates were calculated using the dimensions of
the test section. If the dimensions of the test section and spaces underneath were used,

only a very approximate estimate based on the area ratios of the two paths could be

57

determined at best. The ramp was sealed at the base and along the sides of the contraction
using duct tape. The ramp also directed the ﬂow gradually into the porous plate, to
prohibit separation in the contraction before the ﬁrst porous plate blocker. The longest
ramp possible was used so that the upstream cross sectional area changed slowly. The
ramp was connected to the ﬁrst porous plate blocker on the upstream side of the porous
plate. Three different upstream porous plate blockers were fabricated. Each plate had a
different test section height of 3mm, 6mm, and 9mm, so the plate openings would
match the test section heights and avoid pOSSible ﬂow separation from sharp comers or

steps.

On the downstream side of the porous plate was another aluminum plate to block
the ﬂow that would have otherwise entered under the test section. This double blocking
plate system shown in Figure 17 was needed becauseof the torturous path of the ﬂow. If
the second blocking plate was not present, the ﬂow could enter the porous plate through
the ﬂow opening and exit the porous plate near the bottom of it near the spacers. Three
aluminum plates were fabricated in order to allow an opening that was equivalent to the
test section area, to avoid separation due to changes in ﬂow area. The three plates were
made so that the sample tray rested on top of each one and produced the correct test
section height. In this way, there was no abrupt step down into the test section ﬁom the
porous plate. Before placing the aluminum tray in the test section, a piece of duct tape
shown in Figure 17, labeled “seal tape”, was attached to the sample tray and downstream
blocker plate. This prevented air from escaping below the test section, sealing it to obtain
proper ﬂowrates. Duct tape was believed sufficient on this low-pressure side. The side

edges and bottom of the plates were secured and sealed with putty.

58

2.3.5 Flow in the MSU Flame Rig Test Section

The ﬂow in the test section was highly viscous because the top plate was very
close to the sample. It was therefore concluded that the fully developed test section ﬂow
was the commonly known Poiseuille ﬂow in which the ﬂow formed a parabola of
constant negative curvature. Furthermore the Reynolds numbers in the test section were
computed to be very low in the laminar regime between 1.3 and 300, which can support

Poiseuille ﬂow. More of ﬂow in the test section will be discussed in Chapter 3.

The air and combustion products were expelled to atmospheric conditions in a

large exhaust plenum connected to the hood, shown in the schematic in Figure 18. The

 

 

Figure 18. Schematic of MSU Flame Rig exhaust.

dotted lines indicate parts of the system inside the exhaust plenum. The test section was

eFlitended to prevent any downstream disturbances from altering test section. The

59

extension was located inside the large plenum, so the researcher did not inhale
combustion products. In addition, the pressure in the test section was estimated to be

atmospheric because the hood fan was off during testing.

After testing, the hood ﬁn was turned on to dispose fumes, so velocities in the test
section during experiments were unaffected. If the fan were running during testing, then
it would have created a lower pressure and higher velocity in the test section than
conditions in which the system exhausted to atmosphere. In other words, the fan would
have added more of a pressure drop since there is a gain in energy across the ﬁn. Since
the ﬁn would have drawn the air and combustion products, the velocity in the test section
could not have been calculated, with the ﬁn operating during tests, using mass

conservation because the unsteady mass ﬂowrate in the test section would be unknown.
2.3.6 MSU Samples

The samples used in the MSU Flame Rig were Whatrnan2 brand pure cellulose
Grade 1 chromatography paper that were .18mm. thick and 200m. wide by 20cm. long.
Even though the MSU Flame Rig testing time is not limited to 25cc. or Ssec. like the
NASA Rigs, the pure cellulose ﬁlter paper was chosen to be the ﬁrst material tested for a

few reasons:

1) Pure cellulose was more likely to gasify ﬁ'om the solid state rather than melt

before it pyrolyzed. This was less complicated to study because fuels that

\

2Trade names or manufacturer’s names are used in this report for identiﬁcation purposes
Only. This usage does not constitute an official endorsement, either expressed or implied,
by Michigan State University or the National Aeronautics and Space Administration.

60

melt often undergo phase change (gasiﬁcation, bubbles) within the melted

material.

2) Thin ﬁrels on the order of paper thickness were considered thermally thin,

which meant that the thickness and thus one dimension of the fuel were

neglected.

3) A comparison between the NASA tests, which only utilized thermally thin
fuels, and MSU Flame Rig tests was desired. However, chromatography

paper was chosen over kimwipes because:

a) It was a research grade material, so many of its properties were

known.

b) Even though the chromatography paper was considered thermally
thin, it was thicker than kimwipes. Therefore, the ﬂame speeds
were slower than in the drop towers, and instabilities were

observed for longer time periods.

2.3.7 MSU Flame Rig Sample Holder and Substrate

The sample holder shown in Figure 19 consisted of a piece of mica and copper
backing attached to an aluminum ﬂame, which was attached to a large, aluminum plate
forming the ﬂoor of the test section. To ensure no obstacles were in the ﬂow, producing
separation and unfavorable ﬂow patterns, the sample was glued to the mica, so that it was

ﬂush with the aluminum tray. The mica served as an insulator in the same manner as the

61

front

- er
Substrate

Figure 19. MSU Flame Rig sample holder system

 

 

insulator holder of the NASA Combustion Rig in Section 2.2.1. Like the NASA holder,
the mica did not drain heat at the sides of the ﬂame, helping to produce a ﬂat, linear
ﬂame ﬂont. Like the NASA Combustion Rig, the MSU Flame Rig contained a copper
substrate below the sample. The distance ﬂom sample to substrate was adjustable. The
substrate covered the entire area of the burning sample and acted as a heat sink to
promote the onset of instabilities. The aluminum ﬂame sample holder, to which the mica
was glued, was attached to an aluminum tray, allowing the ﬂame to be adjusted. With
this design, thicker fuels up to ‘/4in can be tested in the MSU Flame Rig. For thicker
fuels, though, no backing is necessary. This is because the thickness of the fuel causes it
to draw heat to pyrolyze, so thicker ﬁrels act as their own substrate. The aluminum tray

was painted matte black in order to eliminate reﬂection ﬂom the metal in the video

footage.

62

2.3.8 MSU Flame Rig Igniter

The system igniter shown in Figure 20a provided a ﬂat, linear ﬂame ﬂont for
most of the tests. For clarity, Figure 20b shows a schematic of the igniter system. The
igniter system employed igniter wire holders made of spring steel. The same type of
igniter wire tlmt was used in the NASA tests (Kanthal) was pushed through a gold pin
holder that was crirnped at the end. The gold pin was similar to a dog screw in that it ﬁt
in the hole of the igniter holder, but did not go through it. It was crucial that the wire
touched the sample across the entire width of the sample to ensure contact between the
two. This was achieved by cutting the igniter wire shorter than the distance between the
two igniter holders. Therefore, the igniter holders were deﬂected inward (into the page)
towards the sample holder before current was passed through the wire. The inward
placement of the igniter wire holders also prevented the wire ﬂom sagging when it was
heated and losing contact with the sample. When sufficient current was applied to the
circuit, the wire glowed bright orange/yellow and became malleable, sinking towards the
copper substrate. The spring steel holders pulled the wire taut again and retained good
contact with the sample because there was already a load on the spring steel holders due
to their inward deﬂection. In this way, the holders kept the wire ﬂom sagging because
they acted as a spring and pulled the wire outwards causing it to become straight and

produce an even ﬂame ﬂont.

The igniter wire holder was attached to the sample ho lder in a way that insulated
it ﬂom the metal portion. Hence, the wire was not grounded and did not break when

electrical power was applied to it. The igniter holder was connected at the end to a red

63

‘ umrnurn tra
blocks vrew o
Insulator holder

d I niter wire Gold pin
wire holder

;" -Metal if": 1‘ i
j holder

for fuel holder and
substrate

  

 

 

Adjustment block
for fuel holder and
substrate

 
 
 

Figure 20b. Schematic of MSU Flame Rig igniter system.

Fiberglass insulator so the holder could be deﬂected inward. Insulator tape shown in
black, which insulated the rest of the igniter holder ﬂom the metal holder, was placed
along the metal holder. Wires connected the igniter wire holder leads to a voltage source.
Using Ohm’s law before each test, the proper amount of voltage was calculated to make
the wire glow orange/yellow and reach a temperature to ignite the sample. Usually

around 20-25volts and 3-5amps of current were utilized.

The space below the igniter wire holders was sealed so that no ﬂow could escape
ﬂom the test section. Also the wire holder slightly protruded into the cross sectional area
of the test section. This could have caused some ﬂow separation. However, our
apparatus was an opposed ﬂow conﬁguration, which meant that the ﬂame traveled in the
direction opposite of the ﬂow. Therefore, the ﬂame moved away ﬂom this area of
separation. Furthermore, the separation probably had little effect since the viscous forces
were so large in this direction. Moreover, an opposed ﬂow conﬁguration was desired so
that combustion products were washed away ﬂom the ﬂame instead of towards the ﬂame.
If the combustion products ﬂowed into the ﬂame, as in a co-ﬂow conﬁguration, they

would have inhibited the oxidizer ﬂom reaching the ﬂame.
2.3.9 MSU Flame Rig Spacers

A schematic ﬂom the side View of the MSU Flame Rig test section and spacers is
shown in Figure 21. The quartz plate lay on four posts, which are not shown, and were
precision ground to 2.336in. (accurate to 10"in.). The posts kept the quartz stationary
and were attached to a l/2irr. thick aluminum plate that supported the entire test section.

The aluminum tray that held the sample holder was placed on top of diagnostic spacers,

65

 

 

 

 

    

 

 

    

Aluminum Pogclruskplatc Spacer Supporting P1311e
0° er lates of test section

Fncasement

 

 

 

 

 

 

 

Figure 21. System of MSU Flame Rig spacers.

which were precision ground aluminum bars l.5in. thick. These bars allowed space for
any wires or equipment, such as igniter wires, thermocouple wires, heat ﬂux gauges, etc.,
that had to be attached to the sample or holder. The bars were placed so that the
aluminum sample tray did not sag anywhere, changing the distance between the sample
tray and quartz plate. Before each test, putty was placed along the side Pyrex windows.
When the sample tray was placed on the spacer bars, the putty stuck to the aluminum
tray. The test section was sealed because the putty prevented air ﬂom leaking along the
sides under the test section. Duct tape (labeled “Seal tape”) in Figure 21 was placed
along the bottom edge of the aluminum sample tray and connected to the downstream

porous plate blocker. This sealed the test section by preventing air leaking ﬂom the

66

porous plate below the test section. The side and bottom edges of the downstream porous

plate blocker were sealed with putty.

The series of spacers that changed the height of the sample were composed of
precision ground stock spacers accurate up to 104in. and aluminum shim stock spacers
accurate to 10% of their thickness. By this system, sample distances ﬂom 2m. to
10mm. were achieved by placing the correct spacers underneath the diagnostic spacers.
The system of spacers raised and lowered the diagnostic spacers and aluminum tray,
which changed the distance ﬂom sample to quartz. Therefore, the quartz plate was
always at a height of 2.336in. This setup also helped in preventing any expensive

machining of the plate.

The system of spacers were used to place the sample at very close distances to the
quartz plate, which in turn produced a pseudo microgravity ﬂame in this apparatus. The
close proximity of plate to sample prevented formation of the buoyant ﬂow pattern cells.
If the quartz were not present, the hot air and combustion products near the ﬂame would
have risen because of the ﬂuid’s lighter density and the bouyant force due to gravity
acting on it. After rising, the ﬂuid would have cooled ﬂom the change in temperature of
surrounding air, thus causing an increase in density of the ﬂuid. The higher density ﬂuid
would have descended towards the ﬂame, forming a continuous circular motion known as
Benard cells. The MSU Flame Rig did not allow this motion to transpire, so effects of
gravity were removed ﬂom the experiment, thereby simulating a microgravity ﬂame on

earth.

67

For the fully developed ﬂow in the test section, the air ﬂowed above and below
the sample. A downward ramp was formed that led the ﬂow below the sample so no
separation occurred because of the downward step. The airﬂow under the sample was
permitted for ease of future model analysis, so a ﬂow proﬁle between the sample and

substrate could be estimated.

2.3.10 MSU Flame Rig Diagnostics

Since the ﬁrst goal of this project was to produce ﬂame instabilities, the only

diagnostics utilized initially were video and digital cameras. Figure 22 shows the camera

~ Li t Shields

.1
h

a

‘\ l Test Section

 

Figure 22. MSU Flame Rig camera mounting system.

mounting system, which is on wheels. This apparatus placed the camera normal to the

test section so that distances would not be distorted when using either the digital or video

68

cameras. The mounting system allowed the user to slide over any part of the test section
and zoom in or out. Black poster board and material was draped around the camera
mount, so there was no light pollution in the picture. A red background light was used

sometimes to obtain a better picture of the ﬂame. For most tests, this light was off.

In the future, a recently purchased digital camcorder will give a large amount of
information when coupled with a software program called TRACKER. This software
program was obtained ﬂom NASA. It is a graphics program that is capable of measuring
distances, areas, color intensities and quantitative results for each ﬂame. Diagnostics that

will be used later are described in the Future Work Chapter.

2.3.11 MSU Flame Rig Test Procedure

The test procedure for the MSU Flame Rig is very similar to the NASA

Combustion Rig procedure.

Pre Test Preparation

0 Place the proper spacers in the test section to the correct test section height is

obtained.

0 Place putty and tape on downstream blocker plate, and press blocker against

porous plate.

0 Slide aluminum sample tray onto tape, and press ﬁrmly to seal. Place aluminum

tray on spacers and press ﬁrmly, so that the putty seals it along edges.

Measure four corners to ensure proper test section height.

69

0 Make sure using bubble tests that the ﬂow system is sealed. (This need not be

done every time.)

Sample Preparation

0 Attach igniter wire to igniter wire holders using gold pins.

0 Glue sample to mica using glue stick, and slide sample under igniter wire so that

it lies straight across top edge of sample.

0 Press and pull sides of sample to produce contact with the igniter wire. It is very

important that contact between paper and wire is optimal.

0 Slide sample tray into position on top of aluminum bars.

Quartz Plate Preparation

- Careﬁrlly place quartz plate on top of posts and slide against the porous plate

holder.

0 Using the soft putty, ﬁll all holes at points where the quartz and Pyrex side
windows meet. Also use putty to ﬁll where the quartz and porous plate holder

meet.

0 Connect igniter wire leads to voltage source.

Actual Testing

0 Center camera mount over sample.

70

Turn on the air source valve in the hood.

Set the pressure regulator to the proper pressure that corresponds to the desired

ﬂowrate produced by the oriﬁce at the end of the test time.

Open the needle valve so that the proper amount of volumetric ﬂow rate that

corresponds with the starting test section velocity is registered on the rotarneter.

Allow enough time to establish the proper pressure in the plenum, which will
produce the desired velocity in the test section. This should be calculated

beforehand.

Begin video taping.

Turn on power source that is already set to the proper voltage. Leave on for Ssec.

or until uniform ﬂame ﬂont occurs.

Once a one-dimensional ﬂame has formed, begin shutting needle valve at a
constant rate until ﬂame instabilities occur and develop over the sample. This
takes timing and practice. If done too slow the sample burns out, and if done too
fast the ﬂame extinguishes. Note that extinction of ﬂames may occur for either a
completely shut needle valve or one tint is partially opened. This will depend on

other experimental conditions.

Turn on hood fan to collect residual fumes.

71

CHAPTER 3
CALCULATIONS

3.1 Overview

Surface energy balances were performed for NASA drop tower tests that included
thermocouple measurements. Sections 3.2.1-3.2.3 show the method of data reduction to
obtain the energy calculations. Flame velocity calculations are in Section 3.2.4

MSU calculations begin in section 3.3.1 with a scaling analysis of the MSU
Flame Rig. Calculations were also performed before constructing the MSU apparatus so
that the correct volumetric ﬂow rate and test section velocity, lag time of test section
ﬂow, proper porous plate porosity, and sample placement in the test section could be
correctly determined. These are presented in sections 3.3.2 through 3.3.6. Due to the
nature of the MSU Flame Rig, it was difﬁcult to measure test section velocity proﬁles in
this preliminary stage of development. Therefore, many calculations were predictions of
ﬂow behavior involving conservative estimations.
3.2 NASA Calculations
3.2.1 Surface Energy Balance Data Reduction Calculations

An energy balance at the surface of the fuel was calculated using the experimental
data ﬂom NASA Rig drops. Surface energy values versus time during the entire drop
were plotted to determine the characteristic energy values of these ﬂames. This was done
in order to see if reasonable energy values could be obtained experimentally at the fuel
surface, since thin fuel microgravity ﬂame energies have been measured in the past.
Surﬁce energy values at the peak temperature of the drop versus the number of ﬂamelets
formed in that particular drop were plotted. A relationship was sought that exhibited a

“cutoff” energy where ﬂamelet formation occurred. Since the energy of the ﬂame often

72

determines its survival, a characteristic, or cutoff energy value was anticipated where the
ﬂame ﬂont could no longer sustain itself and would fragment into ﬂamelets. In this case,
there would be a deﬁnite distinction between peak energy values of tests with a large
number of ﬂamelets and tests with no ﬂamelets (a ﬂat ﬂont). The cutoff energy would
show a connection between them. Finding this cutoﬂ‘ energy may lead to distinguishing
values for the thin fuel ﬂames, which could be applied to space ﬂight safety.
Furthermore, the relationship between ﬂamelet formation and energy may lead to
important relationships that can explain ﬂamelet formation.

The surface energy balance illustrated in Figure 23 and shown in Equation 6 was

P-Gp(dT/dt)

 

0(Ts4'Tb4) [ss'lﬁh'l-l] MTril'bVL

 

Figure 23. Fuel surface energy balance.

 

a(T4—T4) (T —T) (if?) or
ass(T§—I£)+[l—SL+Aal-r SL b +Aspsr s Lv+Pscp§ = qinﬂamem)
—+ -1

5's 5b

73

used to calculate and plot the net energy losses at the fuel surface for each test. In the
above analysis, the energy equation was reduced to a two-dimensional system thereby
making the control surface a series of one-dimensional inlets and outlets. The ﬁrst term
in Equation 6 represents the radiation ﬂom the ﬁrel surﬁce to the surroundings, with T,
being the variable surface temperature. The second term was the net radiation between
the fuel and backing, with Tb being the variable backing temperature. The third term
represents the conduction ﬂom ﬁrel surface to backing, while the fourth is the pyrolysis
or heat of vaporization term [29], which represents the ﬂux of solid fuel being vaporized
per area. The vaporization term was originally cast as:

p s vf L, = vaporization term (7)

where

psvf = tit": mass ﬂux (8)

Equation 8 is the ﬂux of vaporized mass per area. However, using Equation 8 for the
mass ﬂux in the energy equation caused unrealistic non-physical numbers two orders
higher than real life measurements indicate. This was probably due to the fact that
Equation 8 was accounting for the entire thickness of the fuel sheet in the surface density,
and also postulating that the ﬂame speed is directly related to the quantity of the

vaporized mass. Instead, Equation 9 [29] was used to represent the vaporization term

—ES

—m~ ]
A s p S ze[ 5 LV = vaporization term (9)

because it gave realistic values in the energy equation. Furthermore, it included the fuel

half thickness in calculating the area density instead of the entire density as part of the

74

vaporization process, probably a more accurate depiction. Also, what made Equation 9 a
better candidate was that unlike Equation 8, Equation 9 contains the surface temperature
and activation energy. Thus the chemical processes and thermal properties of the ﬁrel
were accounted for as dominant forces in vaporizing the ﬁre] rather than simply the ﬂame
velocity.

The last term on the left side accounted for the time change in total energy. The
energy balance does not contain a total spatial change in enthalpy. It was hypothesized
that the energy would not change signiﬁcantly across a thin sample. The thickness of the
sample was so small that all thermal properties and variables in the energy equation were
estimated as constant across the thickness of the sample, so the fuel was considered to be
thermally thin ({AT(across sample)/AT(clmracteristic)}<<1). The incoming heat ﬂux
qin”g.m was the conduction and radiation ﬂom the ﬂame to fuel surface. Changes in the
strearnwise (ﬂow) direction were also neglected.

Table 1 deﬁnes the constants in Equation 6. From the relation in Equation 6, it

Table l. Constants of surface energy balance equation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

As, Pre-exponential factor, 5 x 1010 s" [27]
E, Activation energy for pyrolysis 139.6 kJ/kgL27]
0,, Fuel speciﬁc heat 1260 J/(kLK)

L Distance ﬂom fuel to substrate .67 cm and 1 cm
Lv Fuel latent heat of vaporization 368.5 kJ/kg

R Universal gas constant 8.314 J/(mol K)
T,o Ambient Temperature 30 °C

3., Substrate emissivity Screens=.l7, Cu=.03, Al=.04, Brass=.05
8s Fuel emissivity .95
7..., Conductivity of air @ atm Temp .0263 W/(m K)

pg Gas density at ambient 1.2 kg/m3

p, Fuel area surface density .0182 kg/m2

0 Stefan-Boltzmann constant 5.67 x 10‘8 W/(mz K‘)
1 Half thickness of fuel sheet .0004 in

 

 

 

 

 

75

was clear that the values needed to calculate the surface energy were the variable fuel
surﬁce temperature T,, the substrate temperature Tb, and the derivative of the surface
temperature with respect to time.

After a ﬂamelet regime was located using the 2 second drop tower, tests were
chosen in which the temperature of the ﬁre] surface and the substrate were recorded using
type K thermocouples. In addition, the ﬂame position was tracked over time to obtain the
ﬂame speed. Each thermocouple was located in the middle of the sample at an estimated
position where ﬂamelets would occur. However, sometimes the ﬂamelet did not pass
directly through the thermocouple because it formed elsewhere on the sample. Therefore
a smaller number of tests than originally planned produced the energy equation data. The
population of tests was sufficiently large to accomplish the goal of determining typical
energy curves for these tests.

3.2.2 Temperature Derivative Calculation
The derivative of the fuel surface temperature was calculated using a second

order, central in time, Finite Difference method shown in Equation 10.

£=T(r+l)—T(r—l) (10)
dt 2At
When this relation was used with the raw temperature data ﬂom the drops, it produced a

 

noisy, essentially incoherent derivative. The temperature data were reduced by
employing a routine ﬂom [49], [50] that “smoothed” the data. This algorithm produced a
curve for the temperature and derivative that did not include the inﬂuence of
measurement errors which rendered the derivative curve calculated ﬂom Equation 10

discontinuous.

76

Before the temperature values were smoothed and the derivative calculated,
additional temperature points were generated. Temperatures recorded before the drop
during the normal gravity ﬂame were taken at .05 sec. intervals, whereas temperatures
during the drop were taken at .Olsec. intervals. This change in time caused a spike or
discontinuity in both the molliﬁed temperature values and the time derivative. A linear
interpolation program generated four ﬁctitious temperature points between every original
.OSSec. interval, by calculating the slope of the line between each temperature point at the
.OSsec. interval. Then, using the .Olsec. interval the corresponding new temperature was
ascertained. This produced an even spacing of temperature values for smoothing the data.

The temperature values at equal time intervals were smoothed using a
molliﬁcation process that was based on weighting the data using a “blurring radius”, 8.
The molliﬁcation algorithm was written based on equations ﬂom [49], [50]. Equation 11

and 12 show the calculation for each molliﬁed point f(n) obtained ﬂom the data.

i =36
N?) = __§3§(i)r(n + i) (H)
where
E‘ ’2 i
p(i)=—1—e 52 (12)

(SJ;

The value Y(n+i) was the original value of the measured temperature at point (n+i). The
weighting ﬁmction p(i) was an exponential function which has a value based on the
distance ﬂom the original point. The negative sign in the exponential indicates that
temperature values Y(n+i) far ﬂom the original point did not contribute as strongly as

points near the temperature point being molliﬁed. The larger the value of i, the less effect

77

it had on the molliﬁed point. For example, for a blurring radius of 5=l, the summation
was ﬂom —3 to 3 in integer values (-3,-2,-1,0,1,2,3). The three points before and the
three points after the original temperature being molliﬁed were multiplied by the
weighting function calculated at each of those points and summed along with the value at
i=0. For this point, the exponential was equal to one, so the contribution to the

summation was just l/(Snl’z). In this case with a radius of 3, the ﬁrst molliﬁed point in a

list of data is shown in Equation 13. It was clear then that with a radius of 8=1 , the ﬁrst

f(1) = Eff: rosy-i + Y(—2)e_i + Y(—l)e—i + Y(O) + Y(ne—i + Yak—i + ray-i (13)
point that could be molliﬁed in the list of data was point number 4 because the previous 3
points were needed for smoothing. The last point that could be molliﬁed for this radius
was the n-3 point since the three points in ﬂont of the molliﬁed point were required. This
will be discussed in following paragraphs.

Some points needed to be considered when choosing an effective blurring radius.
If the radius chosen is too small, the molliﬁed data resembles the noise and measurement
errors in the original data. However, if the radius chosen is too large, the molliﬁed data
do not represent the true data and can possibly be offset in some places. For example,
peak terrrperatures may be lower due to a large radius which includes many points that
would “weigh” the data down. These behaviors have been observed in all types of data
that have used this molliﬁcation routine. The best method to insure proper molliﬁcation
was to begin with a low blurring radius, and then increase the radius and compare

molliﬁed groups of data to each other and the original data. This ensured that peak

78

temperatures would not be in large disagreement with the original data. Consequently, a
continuous curve for the temperature that still represented the original data was produced.
This study began with a radius of one and increased it to employ a radius of ten for
calculation of the smoothed temperatures.

For the blurring radius of 10 used in this work, a signiﬁcant amount of
temperature data at the beginning and end of the test were not mollified. This was
because the summation occurred ﬂom i = -30 to i = 30, so smoothing could not begin
until the 31” temperature point, and it had to end at the (n-30) point. No points before the
30" or after the (n-30) points could be molliﬁed because the summation included the 30
values before and after the smoothed temperature point. To try to mollify points before
the 30m one when no data existed before them would cause the molliﬁcation to be
weighted toward the side where points existed, giving an incorrect value.

In order to make the data not molliﬁed appropriate for calculation of the
derivative, the data at the beginning and end of the test was smoothed using a cubic
spline, matching the ends to the smoothed temperature data. A computer program was
developed to calculate a cubic polynomial that ﬁt the data at the beginning and end of
each set of temperature data. This was likewise done for the 30 end temperature points of
each drop. From this a smooth proﬁle of temperature points at the desired time values
necessary for differentiation were calculated.

In the spline method, the following equation represents the data when solved for

a(t-to)3+b(t-to)2+0(t-to)+d=T(t) (14)
the constants a,b,c,d. In Equation 14 to was chosen so that time zero occurred where the

spline curve was connected to the molliﬁed data. So for a blurring radius of 10, (t-to) = O

79

at points 30 and at n-30. These times were only used in this section of calculation. Once
the spline was computed, the original times were used for the derivative and all other
plots. The constants c and d were automatically determined by the boundary conditions
between the spline curve and the molliﬁed data. These boundary conditions insured that
the curve was smooth and continuous where it joined the molliﬁed temperatures. The
smooth part was accomplished by enforcing continuity of temperature at the ﬁrst
molliﬁed temperature value. This criterion solved the constant (1. At time (Mo) = 0 the
value of the polynomial equaled the ﬁrst molliﬁed temperature. Continuity was achieved
by requiring a continuous derivative across the spline curve and molliﬁed data. This

meant that Equation 15 was satisﬁed at the end of the spline or at the point where the

 

spline joined the smooth data.
dT T(i+l)—T(i—1)
— t—t = = I5
(1! ( o) 0 2A ( )

This criterion solved for c. For example, in the spline at the beginning of the test T(i+l)
was the second molliﬁed point, while T(i-l) was the second to the last raw temperature

point used in the spline algorithm.

The constants a and b were solved by generating a matrix using all the raw
temperature values applying Equation 14 along with c and d. This produced a rmtrix of

the form:

[f]=[A][x] (16)

which is further illustrated in Equation 17. The variables a and b, which are actually

constants in this system of equations, were solved using the least squares method. From

80

__ 3 2 "
[T(t-to)-d-C(t-to)l-[(f-to) (t-Io) ] [b] (17)
the above method, an equation for temperature was used to create a smooth curve, and

thus temperature values at the appropriate times were calculated for differentiation.

After the temperatures were molliﬁed and the spline curves were used to obtain
temperatures at either end of the smoothed data, two columns for time and corresponding
temperature were constructed. The time period was over the entire drop. The
molliﬁcation and spline data reduction processes produced a temperature curve that,
when graphed, was smooth and continuous, so the derivative was now determined using a

second order central in time scheme shown before in Equation 10.

£_ T(i+1)—T(i—1)
dt 2At

 

(10)

Returning to the surface energy balance in Equation 6, the derivative in Equation 10 was

 

-E
4 T4 a;
0'03 - ) (T —T) JRT 6T _n
03s(7:r4‘7:6)+[1 1 b +lair—"S‘T'b—“i"15:10.97 S Lv‘l'PscpE = qinﬂame(6)
+ 4]

2‘— 2'—
used for computirfg th: time dependent term. The two radiation terms and conduction
terms used the molliﬁed fuel surface temperature that corresponded to the time and value
of the derivative at that time. It was also concluded in testing that the temperature of the
backing, which was recorded throughout the drop at a location directly behind the fuel
thermocouple, did not change signiﬁcantly ﬂom its initial (atmospheric) temperature.
Even when the ﬂame passed over it, the temperature of the substrate barely changed ﬂom

Tm. Therefore, in tests where Tb was not recorded, it was estimated to be the average

temperature of the substrate during the drop, which was usually very near Tan“.

81

3.2.3 Peak Temperature Energy Loss versus Flamelets Calculation

In order to accomplish the second goal and compare energy values of different
drops to determine if a “cutoff” energy existed, one characteristic temperature was
chosen to substitute into Equation 6, so that there was only one energy value representing
each test. The peak fuel surface temperature for each test was used because typically in
past studies, this has been the temperature that dictated the instability of the ﬂame at the
leading edge. Choosing this temperature also caused the change in enthalpy term (dT/dt
term) to vanish since the slope of the line at the peak temperature was zero. Therefore,

the energy equation was reduced to:

4 4 [-55]
GT —T T —T RT
(sp b)+l-(Sp b)+AspsTe spL

i L- arr L
(8‘ + as 1)

oz.(T.‘},—T£)+ . = 41./1.... (I8)

 

01'

" H

4 net losses = q in ﬂame (19)
The energy losses due to radiation, conduction and heat of combustion for each test at the
peak temperature, TSP, were determined. Plotting these values for each test versus the

number of ﬂamelets for each test produced a relationship that yielded a critical energy

ﬂux value.
3.2.4 Flame Velocities

Flame velocities in normal and microgravity conditions were determined using
the TRACKER software at NASA Glenn Research Center. These spread rates were used

to compare ﬂamelet velocities to microgravity normal gravity ﬂame ﬂont velocities.

82

They were also used in the energy equation to determine the mass ﬂux in the vaporization
term. This term was, however, later calculated using a different expression. Video of each
drop was digitized at NASA into thirty ﬂames per second, then each ﬂame of the drop
was read into TRACKER A ruler was located in one of the drop ﬂames to determine a
length scale for the drops. For each ﬂame, the tip of the ﬂame or ﬂamelets was tracked,
in which the software recorded the x and y positions of the ﬂame that were later
converted to actual length units. The y position of the ﬂame was used in the analysis,
since the ﬂame traveled downward and not sideways toward the sample edges. Vertical
ﬂame position versus time was plotted in Excel and the slope of the line was calculated

for the drops, which was the ﬂame speed. Flamelets in microgravity were also tracked.
3.3 MSU Flame Rig Calculations

3.3.1 Scale Analysis

Figure 24 shows the schematic used to perform a scaling analysis in order to

 

 

 

Top piate T°°
y‘ 1 v
L-ex h L.“
U—' Heat source
BSttom plate

 

—~ .. t»

Figure 24. Schematic for sealing analysis of MSU Flame Rig.

83

check the assumption that the effects of buoyancy would be suppressed for test section
heights, h, under 1.1cm. It was also desired to ﬁnd the approximate upper limit of the

test section height, h, which would suppress buoyant motion.

First, the governing equations of mass, momentum, and energy were deﬁned in

Equations (20-23) for the system. Species equations and chemical reaction terms were

Mass Conservation Qu— + 2V— = 0 (20)
536 5y
2 2
x-MomentumConservation u—+v§£ = ——L-a£+ _6__u+6_u (21)
6x By P 6x 6x2 5y?-
2 2
y - Momentum Conservation u g + vi = ——1-§£ + V 31 + 53—V— — g (22)
6x 6y P 6y 6x2 2
2 2
Energy Conservation u 6— + v91 = a 6_g_'_ + g (23)
6y ax ay

neglected because it was assumed that the bottom of the test section was a very hot plate,

representing the ﬂame, with temperature much greater than ambient. This would serve as

the source of heat in this estimation.

In the MSU Flame Rig the characteristic test section height, h, is much smaller
than the test section length, L, so (h/L <<l). Therefore, second derivatives with respect
to x, (62/ 6x2 terms), could be neglected since they would be much smaller than second
derivatives with respect to y. Also, the Boussinesq approximation [43] was employed in

the y-momentum equation where the change in pressure in the y direction is dictated by

the hydrostatic pressure gradient. Combining this with the gravity force term and using

the equation of state, the governing equations become:

Mass Conservation Qu— + 23 = O (20)
5x 5y
2
x - Momentum Conservation u Qu- + v 92 = __1__6£ + V 21 (24)
6x 6y P 6x ay2
2
y - Momentum Conservation uQ—Z + v 2v— = gﬂ(Ts — T 00) + v -a—Y— (25)
6x 6y ay2
2
Energy Conservation u 531 + v 6_T = a Q—Z (26)
6x 6y 5y?-

where (T,-T,,)=AT . Starting with mass conservation, Equation 20, and noting that the
characteristic velocity in the x direction is U, the characteristic x value is L, and the

characteristic y value is h, the velocity in the y direction scales as:

Uh
~— 27
v 1. t)

From the original fact that (h/L)<<l, the y direction velocity, v, is neglected, and the x-

momentum equation becomes:

6n 1 or 62a
V—

 

_ = ___ + 28
6x P 6x ayz ( )
inertia = pressure + friction
which scales to
U 2 AP vU
_ , x , _ (29)
L pL hi

However, for very low ﬂows, as in the MSU Flame Rig, the pressure and very large

viscous forces would most likely dominate the inertia term in this equation due to the

85

close proximity of the plates and low Reynolds numbers. Therefore, the x-momentum
equation renders a pressure versus ﬂiction balance, so the change in pressure in the x

direction scales as:

L vU
up, ~——Ph2 (30)
Likewise, the y momentum equation is:
62V

gravity = friction
The y-momentum equation therefore is a balance between gravity and viscous forces.
The ﬂiction term was not neglected, even though the y velocity v is presumably
negligible. This is because both the y-velocity and characteristic y are small, so the
second derivative of the y velocity with respect to y may not necessarily be small.

Equation 31 scales to:

w
EMT ~ h—2 (32)
gravity = friction

Now focusing on the energy equation, which when reduced, scales as:

UAT ~ aAT

 

33
L h, ( 2
convection ~ conduction
From this, a scale for U is determined as:
U ~ £2 (34)
h

86

By inserting this in the x-momentum equation, the pressure change in the x-direction is

approximated as:

~ pLZav
h4

 

My: (35)

Using the y-momentum equation and realizing that the pressure depends on the
hydrostatic gradient caused by the temperature difference, an effective pressure change in

the y-direction is approximated in Equation 36.

My ~ pkg/3M (36)
This is because the original pressure term in the y-momentum equation was approximated
using the Boussinesq relation. Comparing the pressure changes in the x and y directions

gives:

APy N 8347775
AP x asz

 

(3 7)

Grouping terms together into non-dimensional units, Equation 37 is actually the Raleigh

number based on the height of the test section times the ratio of (h/L)2.

 

AP 2
y ~ Rah (5) (38)
APx L

Equation 38 is the key to suggesting that buoyancy may be neglected in the MSU Flame
Rig. Since (h/L)<<l, the pressure ratio is very small. Then pressure changes in the y-
direction are very small compared to pressure changes in the x-direction. Pressure
changes in the x direction over the entire test section length were approximated to be
much smaller than the pressure changes required to produce test section velocities. This

is because there are no obstacles, such as screens or oriﬁces, in the test section to induce

87

large pressure drops. Therefore, for the higher test section velocities like 30 cm/s, the
velocity in the test section is assumed as this value over the whole length of the test
section, so the change in x-velocity throughout the test section is believed to be very
small. The change in pressure in the y direction is taken over the y-direction
characteristic length or the test section height. In this case, if the pressure change in the
y-direction were much less than x-direction pressure changes, then the y-direction
velocities would be much smaller in magnitude than the x-direction velocities. This
meant that for (h./L)<<1 and for low ﬂow rates even as high as 30 cm/s, the y-direction
velocities are much smaller than buoyant velocities. As a consequence, it was assumed
that buoyant cells did not form because the characteristic velocities that support them
were not present. In ﬁct, as mentioned earlier, in theory, there is a critical Raleigh
number below which buoyant cells may not form between two long parallel plates. For
Raleigh numbers equal to or greater than 1708 [43], the commonly known Benard cells

or buoyant motion will occur due to gravity and changes in temperature and density.

Equation 38 was calculated for various test section heights, in order to determine
the upper limit where the test section height is too large and traces of buoyant motion
would begin. In Table 2, several test section heights and two temperature differences
were used to calculate the Raleigh number and pressure ratio. The test section length or
x-direction scale, L, is 30in. (76cm). The temperature difference between the ﬂame and
ambient was assumed as 600°C and 700°C. For both temperatures, the pressure ratio
seemed to be much less than unity for all test section heights except 1.2cm. At 1.2cm.,
the Raleigh number was still below the critical number for convection, however, any test

at this height would be suspect to have gravitational effects. This is because at a 12an

88

Table 2. Pressure ratio scaling values in the MSU Flame Rig.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

T,=600°c
Test section height(cm) h/L Rah Ra.,(h/L)2

0.2 0.003 6.9 4.8*10’5
0.5 0.006 109 0.004
0.8 0.01 446 0.049
1.0 0.013 873 0.15
1.2 0.016 1508 0.37

r.=700°c
0.2 0.003 5.8 4.0*10"
0.5 0.006 91 0.004
0.8 0.01 374 0.041
1.0 0.013 730 0.12
1.2 0.016 1261 0.31

 

 

 

test section height, Rah(h/L)2 was not much less than one, and Rah was much closer to the
critical Raleigh number than other heights. The critical Raleigh number (1708) is not
necessarily a clear cutoff. Many numbers of this nature are approximate, in that buoyant
motion may develop at values near 1708, and it may not be forming complete Benard
cells. Therefore, it was assumed that the effects of buoyancy in the MSU Flame Rig were
neglected for test section heights equal to or below 1cm, since both the scaling pressure

ratio was much less than one and the Ra.I was below the critical value for buoyant cells.

89

3.3.2 Mass Flowrate Calculations

The desired volumetric ﬂowrates shown in Table 3 for the velocity and test
section height ranges for the MSU Flame Rig were calculated by multiplying the desired
test section average velocities by the cross sectional areas of the test section. The largest

ﬂowrate, which corresponded to the largest cross sectional area times the largest velocity,

Table 3. Ranges of volumetric ﬂow rates in MSU Flame Rig.

 

 

 

 

Ranges Test Test Section Test Section Cross Volumetric
Section Heights Sectiorml Areas Flowrates in Test
Velocities (cm) (ftz) Section at STP
(cm/s) (113/11,)
Smallest 1 0.2 0.007 0.840
Largest 40 1.0 0.035 167

 

 

 

 

 

 

 

 

was 167ft3/hr. The smallest ﬂowrate, which corresponded to the smallest cross sectional
area times the smallest velocity, was 0.84ﬁ3/hr. A range of 1-40crn/s was the test section
velocity range because higher velocities than those in NASA tests were needed to
Produce the instabilities. This was because the top plate would also drain some of the
heat from the ﬂame. It was anticipated, though, that instabilities would occur below
400Ill/S. This upper limit was an estimation of maximum ﬂow velocity needed to ignite
the satnple and produce an initial ﬂat even ﬂame ﬂont. Therefore, to be safe 40cm/s was
used for all ﬂow calculations. The above range of ﬂowrates in Table 3 was required for

certain calculations to choose the correct ﬂow oriﬁces and porous plates.

90

3.3.3 Volumetric Flowrate Correction and Conversion

For the MSU Flame Rig, it was estimated that instabilities would probably occur
between 1 and 30cm/s. However, the range of velocities that would produce unstable
ﬂames was not exactly known for each test section height. The only way to obtain this,
was to conduct several preliminary tests for each plate height, and record the air speeds

that produced the near extinction regime.

In order to pinpoint the range of velocities where instabilities occurred, a
rotameter calibrated to standard conditions was utilized. Since the pressure of the air was
known directly after the pressure regulator, the rotameter was placed at that point and
volumetric air ﬂowrate ranges were recorded for unstable ﬂame conditions. From the
volumetric ﬂowrate, and assuming the entire MSU Rig system was sealed, the velocity

could be indirectly obtained in the test section. Figure 25 illustrates this in a series of

 

1 Convert observed volume P
ﬂowrate to actual volume {2:} Q = (41.11
ﬂowrate at standard conditions. ac Qobs Pas

2 Convert actual volume ﬂowrate

P
to equivalent volume ﬂowrate erq = Qac —‘-"'—
at standard conditions. Pam

3 Calculate mass ﬂowrate by .
multiplying converted ﬂowrate r32} m = erq
times atmospheric density.

4 Calculate test section velocity

by dividing mass ﬂow by _ .
atmopheric density and test “1'. V1931 80° — m KM)

 

 

 

 

 
 
      

 

 

Figure 25. Calculation of test section velocities in MSU Rig ﬂom rotameter readings.

91

steps rather than numbered equations for ease of presentation. Since the rotameter was
calibrated to atmospheric conditions, the conversion factor shown at the ﬁrst step in
Figure 25 was used to correct for the observed volume ﬂowrate. The rotameter that was
placed in-line at much higher pressures would read much lower volumetric ﬂowrates
since volume ﬂowrate = mass ﬂowrate/ density. At high pressures, the densities would
also be high, thereby producing a much larger denominator than atmosphere, and
consequently a lower volume ﬂowrate. The second conversion factor at the second step
in Figure 25 converted the observed volume ﬂowrate to an equivalent volume ﬂowrate at
standard conditions. Therefore, the mass ﬂowrate could be calculated ﬂom step 3 by
multiplying the corrected volume ﬂowrate times atmospheric density. By mass
conservation, the correct test section bulk velocity was determined by dividing the mass
ﬂowrate by the density and test section area in the fourth step. In the correction and
conversion factors in steps 1 and 2, instead of pressure values, densities are usually used.
It was assumed, though, that air was an ideal gas and that there was little temperature
ﬂuctuation in the line. Also in this conversion, P,c has usually been taken to be the outlet
pressure of the rotameter. However, in this experiment, P” was taken to be the pressure
indicated by the pressure regulator, which was placed directly before the rotameter. This
was because the rotameter was placed at a very short distance away ﬂom the pressure
regulator, and the pressure drop across the rotameter (supplied by the manufacturer) was
negligible. The correction and conversion factors ﬂom steps 1 and 2 in Figure 25 were
well known results of rotameter application and found in references by [51], [52], and

[53].

92

3.3.4 Porous Plate Calculations

The porosity of the plates was chosen based on ﬂow estimations and calculations,
so that they would produce straight and laminar ﬂow. First the minimum pressure
change allowable that would produce laminar ﬂow was calculated by estimating the
maximum jet velocity in the plenum and calculating the dynamic pressure associated with

it. Table 4 shows these results in the ﬁrst row. Since the maximum velocity that would

Table 4. Pressure calculations for determining porous plate pressure drop.

 

 

 

 

Density p Upstream Dynamic Pressure Pressure Drop AP
of Porous Plate V2 p V2 for Jet Velocities Across Porous Plate

V=40 cm/s 100 times 1/2 9 V2

1.2 kg/m3 0.096 Pa (1.4 r 10:5 psi) 9.6 Pa (.0014 psi)

7.6 kg/m3 0.61 Pa (8.8 * 10'5psi) 61 Pa (.0088 psi)

 

 

 

 

produce a stable ﬂont was assumed to be 40cm/s, then to be safe, this was chosen as the
maximum velocity that would occur in the contraction or plenum. Furthermore, this
maximum test section velocity was a very conservative number because the areas of the
contraction and plenum were much larger than any test section area, so the velocities
would be less than the test section velocity for the same mass ﬂowrate. The porous plate
was chosen so that the pressure drop across the plate was much larger than any dynamic
pressure that occurred due to velocities upstream of the plate. In this way, undesirable
ﬂow patterns ﬂom these jets of air due to separation were dissipated in the plate because

its pressure drop was so large. A large pressure drop was considered to be two orders of

93

magnitude higher or one hundred times larger than the dynamic pressure of any jets
behind the plate. This gave a value of 9.6Pa. or 0.0014psi. From this calculation, the
pressure drop across the porous plate helped to guarantee that any effects of the ﬂow
upstream of the plate did not create turbulent or undesirable ﬂow downstream of the

plate.

Different upstream densities of the ﬂow were also taken into consideration in
order to calculate the minimum pressure drop more accurately. From previous
calculations it was determined that the largest density in the polyurethane tubing before
the plenum of the system was 7.6kg/m3. This value was obtained ﬂom the ideal gas law
at ambient temperature and tubing pressure known ﬂom the pressure regulator.
Therefore, the maximum possible density in the tubing would be less than this while the
dynamic pressure was calculated using 7.6kg/m3 to be conservative. Using this value, the
pressure drop across the plate was still very small at 0.0088psi. From this analysis, it was
evident that the smallest porous plate pressure drop allowed that would still drive the

ﬂow through the plate and minimize turbulent ﬂow downstream was 14.7088psia.

The Mott Corporation3 manufactured the porous plates and provided pressure and
velocity behaviors in each plate for different pore sizes in the plate. Figure 26 shows this
pressure and velocity data utilized to determine the porosity of the plates. The pressure
drop corresponded to the pressure drop across the plate and the airﬂow corresponded to

the velocity inside the plate. To decide the appropriate porosity, the maximum and the

 

3 Trade names or rmnufacturer’s names are used in this report for identiﬁcation purposes
only. This usage does not constitute an official endorsement, either expressed or implied,
by Michigan State University or the National Aeronautics and Space Administration.

94

Mott Corp.

10

“—10 micron
1 —5 micron

0.1

0.01

0 Pressure Drop (PSI)

 

.001
Air Flow (CFM/FTAZ)

Figure 26. Pressure versus velocity behavior for Mott porous plates.

minimum velocity of the airﬂow directly after the plate was used. This was actually the
test section average velocity. The velocity at the exit of the plates was chosen as an
estimate because, in reality, the velocity of the air inside the plate was unknown. The
minimum entry velocity in the test section was lcm/s (1.97ft/min) and the maximum was
40cm/s (79ft/min). From the Mott data, it was apparent that the 5micron and lOmicron
size porosity plates were desirable because for the smallest velocity value (2ft/min), both
plates gave a minimum plenum pressure drop above the 0.0088psi value calculated
previously. From Figure 26, the lOmicron plate showed a pressure drop of 0.01psi for
1.97ft/min velocity. Also, the highest pressure drop across the plates which corresponded
to the largest velocity value (79ft/min) was approximately 2psi., which was possible in

our system because the supply pressure of air was approximately 80psig.

95

3.3.5 Plenum Fill Time Calculations

The longest and shortest amounts of time required to ﬁll the plenum were
calculated. After these time periods, the plenum would be sufﬁciently pressurized, so
that the correct velocity and therefore the desired ﬂow conditions were established in the
test section. At the beginning of each test, the largest ﬂowrate for that test section area
was used so that the sample was ﬂushed with a large quantity of oxidizer for ignition.
Therefore to determine the plenum/contraction ﬁll time, the quantities in Figure 27 were

calculated. For ease of presentation, the calculations are presented in a ﬁgure rather than

 

 

 

 

1 Use largest and smallest P .
plenum pressures to calculate p p l = M
plenum density. Use ambient RT ambient
temperature.

2 Calculate mass of air in plen m . = V
at corresponding max and min at r Pp] pl
plenum pressures.

3 Calculate plenum ﬁll time by .
dividing corresponding max . __ m
and minimum ﬂowrates by ﬁll ”me " m
mass in plenum. Pl

 

Figure 27. Plenum ﬁll time calculations in MSU Flame Rig.

in the text. Before the ﬁrst step, pressure limits of the plenum that corresponded to the
largest velocities for the minimum and maximum test section areas were determined. For
these calculations, ﬂow was assumed to accumulate in the plenum/contraction and not
ﬂow through the porous plate. In reality, some ﬂow was probably ﬂowing through the
porous plate before the driving pressure in the plenum contraction was established. The

maximum and minimum pressure limits of the plenum were determined by adding

96

atmospheric pressure (14.7psi) to the high and low pressure drops (.4psi and lpsi) at
30cm/s (60ft/min) across the porous plate ﬂom the previous section. Using the pressure
limits, the density in the plenum/contraction volume was calculated ﬂom the ideal gas
law (step 1). From this density, the mass of air inside the plenum/contraction volume was
calculated for the corresponding densities (step 2). Lastly, using the largest and smallest
mass ﬂowrates that corresponded to each of the pressure values that were originally used,
the time to ﬁll the plenum/contraction volume was computed by dividing the mass

ﬂowrates by the mass of air in the plenum/contraction volume (step 3).

Using the above analysis, it was assumed that test section velocities greater than
30cm/s would be required to ignite the sample. At this low-end ignition limit of pumping
up the plenum, the longest ﬁll time was calculated using the above equations and Mott
pressure data. For 30cm/s and all test section heights, the longest plenum ﬁll time was
estimated to be less than lOsec. Therefore, it was concluded that this was the worst case
scenario and the longest time period before steady ﬂow conditions were established. For
most tests, the plenum ﬁll time would be even less than this because initial velocities
would greater tlﬁn 300m/s. Before testing however, ignition usually did not occur within

lOsec. of turning on the ﬂow supply due to additional test preparations.
3.3.6 Entrance Length and Boundary Layer Calculations

The entrance lengths in the plane parallel to the Pyrex sides for all test section
heights and corresponding velocities were calculated in this section using the analysis
from Arpaci [54]. In the MSU Flame Rig, the entrance length in the test section was the

distance ﬂom the porous plate after which the velocity proﬁle no longer changes. The

97

concept of entrance length in the above mentioned plane in this apparatus is shown in

Figure 28. The boundary layer was the region in which the ﬂow velocity varied due to

Porous elocity Quartz
Plate

 

Figure 28. Entrance length in plane parallel to tunnel sides in MSU Flame Rig.

the viscous forces of the wall. Outside of the boundary layer was the ﬂeestream velocity
or core ﬂow, which was constant. When ﬂow entered the duct, the boundary layers
increased until they reached a maximum value and met. After this distance, fully

developed ﬂow in the form of the Poiseulle parabolic proﬁle occurred.

The analysis began with the control volume denoted as CV in Figure 29, where
the pressure acts on all points of the ﬂow, not just the boundary layer. From the balance

in Figure 29, the momentum equation for this control volume is:

d n n dp Bu
—]uudy=-——-v{—] (39)
de de 6y y=0

98

 

 

 

For Momzenturn Balance: For Mechanical Energy Balance:
A= ingot A= 7......

0 o

5”“ 2 +2; 5”“? 0 4x 0
C= -91: 32. _ __cjg 9:91

[(1) ¢h+ ay JuayJa‘x C— [(1) ¢h+ ay aya‘x

 

 

 

 

 

Figure 29. Momentum and mechanical energy balances [54].

Where it is the velocity dependent on x and y, p is the pressure, p is the air density, and v

is the kinematic viscosity of the air. The mechanical energy equation for the control

volume is:
d n u2 ndp na‘t'yx
—lPu—dy=-l—udy+l—u05’ (40)
4bro 2 odx 0 By

The following ﬁrst order proﬁle for velocity was assumed.

2
u _ z _ z .
_U__ a) (a) , OSySé' (41)
l ; §Sy=n

99

Along with the dimensionless variables:

 

Ezg , (7:50 , f=(%)Re-l (42)

where U is the velocity in the core ﬂow, 8 is the boundary layer thickness, and Re is the
Reynolds number based on the half channel width n Equations 41 and 42 were
substituted into Equations 39 and 40, which were then combined to yield the following
relation between the dimensionless boundary layer thickness and the entrance length
75+l§331n(1—-§)—%1n(1—5)+%§1:i& =10§= ”(ii—(7:?) (43)
located in the Reynolds number. The entrance 3length occurs where the dimensionless
boundary layer thickness (8 tilde) is equal to one, when the boundary layers merge. Using
a value of unity however, causes an inﬁnite value of the natural logarithm, so the
boundary layer was instead taken to be very close to unity. Substituting 8 tilde equal to
0.99 into Equation 43 gave the entrance length x, which was within the dimensionless

variable x tilde.
Table 5 shows the calculations of the entrance lengths in the plane parallel to the

Table 5. Entrance length ranges in plane parallel to tunnel side walls.

 

 

 

 

 

 

 

 

 

 

Velocity U0 V2 channel width Reynolds # Entrance length
(cm/s) (cm) (cm)
1 0.1 0.67 0.02
40 0.1 26 0.66
1 0.5 3.3 0.41
40 0.5 133 17

 

100

tunnel side-walls for the maximum and minimum test section velocities and heights. The
largest entrance length that was calculated for all the MSU Flame Rig ﬂow conditions
was approximately 17 cm (6.7 in). This was calculated using the largest possible velocity
that would be used in the largest test section height, which would produce the largest
possible Reynolds number. The sample was not placed at a downstream distance ﬂom
the porous plate smaller than the entrance length because the boundary layers would still
be developing. If it were, then the velocity proﬁle would be changing over the sample
and therefore possibly introducing more variables affecting ﬂame instabilities. In the
MSU Flame Rig, 2.3in. additional distance was allowed, so the beginning of the sample
was 23cm. (9in.) ﬂom the porous plate location. By doing this, ﬂow over the sample
would always be fully developed, so that an approximate average velocity was estimated

over the entire sample.

Entrance lengths were also calculated in the plane parallel to the quartz plate
using Equation 43. In this dimension, the half channel width is 16.5cm. (6.5in.) because
the width of the quartz plate or test section is 33cm. (13in). From the results shown in
Table 6, it was evident that boundary layers never merged since entrance lengths were so

much larger than the length of the test section which was 84cm. (33 in). Therefore, the

Table 6. Entrance lengths in the plane parallel to quartz plate.

 

 

 

 

 

Velocity Uo V2 channel width Reynolds # Entrance length
(cm/s) (cm) (cm)
1 16.5 104 460
30 16.5 3300 13830
40 16.5 4400 18440

 

 

 

 

 

101

sample or at least part of the sample was always located in the core ﬂow. The boundary
layer thickness on each side of the test section along the Pyrex sides was computed to
determine where the sample was located in relation to the boundary layer. Before
proceeding, it is important to point out that every effort was made to place most of the
sample in the core ﬂow in this dimension. However, due to the high cost of a large

quartz plate, 33cm. (13in) was the maximum width that could be purchased.

The boundary layers shown in Figure 30 along the sides of the Pyrex windows

were computed using the Blasius equation [43].

 

_ (44)

g _ 4.92
x Reg/2

Here x is the location of the beginning or end of the sample in the test section, 8 is the

Top View of Test Section
plate holder

 

x —> side

 

Length of Test Section

Figure 30. Boundary layers in plane parallel to quartz plate in MSU Flame Rig.

boundary layer thickness, and Rex is the Reynolds number with respect to the location of

the sample and test section velocities. Table 7 shows the boundary layer thickness at the

102

beginning and end of the sample shown in Figure 31. The beginning and end of the
sample were chosen as locations to measure the boundary layer, to ascertain how the

boundary layer changed along the sample and if the sample was located in the core ﬂow.

Most of the boundary layer values indicated that the sample was in the ﬂeestream
velocity in this dimension, but the maximum value of the boundary layer for all the MSU
Flame Rig conditions at lcm/s test section velocity was 12.5cm. (4.9m). This value was
obviously for the lowest test section velocity and at the end of the test section. For the
same velocity at the beginning of the test section, the boundary layer was 9.2cm. (3.6in.)
meaning the boundary layer grew 3.3cm (1.3in.) on each side along the sample. In the
MSU Flame Rig, the sample was placed in the middle of the Aluminum tray at a distance
of 6.4cm. (2.5in.) ﬂom the sides of the Pyrex walls shown in Figure 31. From Table 7, it
was clear tint the sample was located in the core ﬂow for all test section velocities
greater than or equal to 4cm/s. At this velocity, the boundary layer at the end of the
sample was located at the very edge of the sample. Also ﬂom the table, at 2cm/s, at the
beginning of the sample, the boundary layer was at the edge, however, it grew to include
the edges of the sample downstream. Therefore, at 2cm/s, the sample edges were within

the edge of the boundary layers.

Since we were studying ﬂame instabilities, which are a function of velocity, the
boundary layers for velocities lower than 4cm/s should be taken into consideration. If

instabilities arose at the sides of the sample for this range of velocities, then explanations

must include the eﬁects of the changing velocity within the boundary layers. However,

103

 

Table 7. Boundary layer thickness at the beginning and end of sample.

 

 

 

 

 

 

 

 

 

 

 

Test Boundary layer thickness in (cm) Boundary layer thickness in
Section at distance from beginning of (cm) at distance ﬂom
Velocity in test section to beginning of beginning of test section to
(cm/s) sample (9 in or 23cm) end of sample (17 in or 43cm)
1 9.2 cm 12.5 cm
2 6.5 cm 8.8 cm
4 4.6 cm 6.3 cm
10 2.9 cm 4.0 cm
30 1.7 cm 2.3 cm
40 1.4 cm 2.0 cm
Porous Plate

Pyrex side

   

        

23 cm (9 in)

43cm (17 in)

Pyrex side

Boundary layers

 

Figure 31. Sample location and boundary layers in MSU Flame Rig test section.

 

an argument can be made to show that this may be inconsequential. It was explained
before that the MSU Flame Rig was really a Hele-Shaw apparatus. In this type of
experiment, the viscous forces were so dominant in one dimension, that the plane of the
other two dimensions would exhibit “inviscid” ﬂow behavior. In ﬁct, Hele-Shaw
devices are used to observe inviscid and irrotational ﬂow patterns (“Potential Flow”) in
the dimension for which the preceding boundary layers were being determined (Figure
31). A short sequence of calculations can show that ﬂow is irrotational [42] in this top
view plane in Figure 31. Therefore in this plane, the ﬂow can be described using
“Potential Flow Theory”, so the viscous forces in this plane are presumed to be
negligible. Consequently, the boundary layer would have had little effect on the ﬂow
across the sample in this plane. Explanations of the instabilities at the edges of the
sample for the suspect velocities will include all parameters that can destabilize the ﬂame

ﬂont.

105

CHAPTER 4
RESULTS AND DISCUSSION

4.1 Overview

Drop tower results provided regions of ﬂame instability, which were
characterized by oscillating ﬂamelets. Oscillation ﬂequencies and behaviors were
measured in size and ﬂamelet intensity. Flame spread rates were calculated using
Tracker. Select tests also included thermocouple measurements throughout the drop on
the fuel surface and substrate. Thermocouple data in conjunction with ﬂame spread
results were applied to a ﬁrel surface energy balance. Two types of energy balances were
plotted: one energy balance over the entire test, and a second energy balance using the
maximum temperature so tests could be compared. All results were preliminary, but
beneﬁcial for pinpointing required future measurements to quantitatively explain
instability mechanisms. Drop tower tests indicated the necessary upgrades to the sample
holder and diagnostics for the MSU Flame Rig and the International Space Station
speciﬁcations.

The MSU Flame Rig was successfully completed, and preliminary ﬂow
measurements were recorded. An effective ignition system consistently formed ﬂat,
ﬂame ﬂonts in the test section. Flame instabilities were produced ﬂom a stable ﬂame
ﬂont, and video data were recorded.

4.2 NASA Drop Tower Results and Discussion
4.2.1 NASA Instability Regions

Initial groundwork testing performed without any substrates attempted to produce

ﬂame instabilities by reducing the oxygen concentration to the normal gravity extinction

limits, which are l7-20%. Stable, one-dimensional ﬂame ﬂonts were not produced at

106

such low values, so an increase in heat loss was pursued instead. The heat loss parameter
was introduced by placing a metal backing within 1cm. away ﬂom the back surface of the
fuel. In most tests, a stable ﬂat, ﬂame ﬂont began in normal gravity, continued through
transition to microgravity, and ﬁnally broke into separate ﬂamelets that oscillated,
recombined, or extinguished. A ﬂame ﬂont resembling the original ﬂame ﬂont often
resulted when ﬂamelets occasionally recombined.

A pictorial story line of a typical test is shown in Figure 32 on the following page.
Most photographs of experimental tests in this thesis were mnipulated using image-
processing software so the ﬂames and ﬂamelets could be more easily viewed. This was
done solely because many of the ﬂames were dimmer than the video after being digitized
into separate ﬂames. If the images had been printed after being digitized, the geometry
and colors of the ﬂamelets would be indiscernible. The relative colors and intensities
between ﬂames or ﬂamelets in the same picture were not changed. For a speciﬁc ﬂame,
the color or brightness of one ﬂame was not improved without changing the other ﬂames
in the picture by the same amount. Therefore, the original contrasting features between
ﬂamelets was retained. Furthermore, the geometry of the ﬂames and ﬂamelets were not
changed ﬂom the source video. Time zero is taken to be the instant a uniform normal
gravity ﬂame ﬂont was formed.

The ﬂame spread rate direction is downward on all top views in this thesis, such
as the ones shown in Figure 32. The transition to microgravity was denoted by the ﬂame
color change ﬂom yellow to blue, suggesting the decrease in oxidizer supplied to the
ﬂame due to the lack of buoyant motion. Transition duration depended on the velocity of

the oxidizer. Soot formation was reduced in microgravity. The small bright spots were

107

 

t=0 sec t=3.46 sec

 

t=l.43 sec t=4.76 sec

      

t=2.27 see

Figure 32. Typical drop tower test showing ﬂamelet formation and interactions.

t=4.99 sec

108

not ﬂames but char pieces, regions of smoldering that usually did not survive.
Instabilities occurred when the smooth ﬂame ﬂont became corrugated and broke apart
into separate ﬂames. The substrate acted as a heat sink by absorbing heat ﬂom the
backside of the sample. Without the heat sink, the ﬂames would have otherwise
continued as a ﬂat ﬂame ﬂont until the test was over, or all of the fuel was consumed. In
essence, the backings allowed higher oxygen concentrations and higher opposed ﬂow
velocities. Air was used for all tests because the fuel was guaranteed to ignite for the
smallest ﬂow velocities (lcm/s) in normal gravity, or beginning test conditions. Opposed
ﬂow velocities dispensed any residual buoyancy that accumulated near the ﬂame in the
pre-drop normal gravity conditions. The air ﬂow also assured sufficient spread rates tlurt
would undergo transition into ﬂamelets. In tests with the substrate, the low opposed ﬂow
velocities were ample to produce a ﬂat ﬂame ﬂont in normal gravity, but high enough to
sustain the microgravity ﬂames and avoid blowoff (ﬂow velocity is much larger than
chemical reactions for ﬂame to survive). In ﬁct, the opposed ﬂow rates were chosen so
that the ﬂame would continue regardless of the heat loss and also produce instabilities
within the 2.2sec. drop time.

Aluminum foil was the ﬁrst substrate, which helped attain the ﬁrst goal of
producing an observable region of instabilities. The results of this substrate are presented
in Table 8.

An observable region of ﬂamelets existed for velocities between 2cm/s and 4cm/s.
For tests below 2cm/s of ﬂow, the microgravity ﬂame or ﬂamelets always extinguished,

and those above 4cm/s always generated a smooth ﬂame ﬂont that did not break apart.

109

Table 8. Experimental results for drop tower tests with aluminum foil substrate.

 

 

 

 

 

Air Velocity Number of tests Tests with Tests that
(cm/s) conducted ﬂamelets extinguished
2 2 1 1
3 3 1 1
4 6 4 0
5 3 O 0

 

 

 

 

 

 

Above 4cm/s the ﬂame remained stable because enough oxidizer was delivered to burn
all of the fuel across the entire sample and thus keep the ﬂont ﬂom corrugating. Below 2
crn/s, an insufﬁcient amount of oxygen was supplied to the ﬂame ﬂoat and ﬂamelets,
which made them extinguish. Most ﬂamelet development was seen at the 4cm/s velocity
where four out of the six tests showed ﬂamelets and none extinguished. For the lower air
ﬂow rates, the ﬂamelets were most likely not receiving enough oxygen to survive, so they
quenched. The rmin shortcoming of the aluminum foil backing was that it did not remain
rigid during the tests. Instead, heating and cooling due to the ﬂame would cause the foil
to deform. This changed the foil/sample distance at different points along the sample,
causing an uneven distribution of the heat sink. This irregularity in the substrate may
have caused the onset of instabilities in local regions of the fuel. If one portion of the foil
were wrinkled, placing it closer to the sample, then at that point more heat may be drawn
ﬂom the ﬂame. This could cause the ﬂame to break apart at that point. In order to
control the substrate distance ﬂom the sample, thicker, more rigid substrates such as
steel, copper, and brass were later utilized.

A thicker steel backing produced a larger region where ﬂame instabilities were
observed because the conductivity of steel was four to ten times smaller than aluminum.

The results are shown in Table 9. To account for the thicker heat sink, the steel was

110

Table 9. Experimental results for drop tower tests with steel substrate.

 

 

 

 

 

 

 

Air Velocity Number of tests Tests with Tests that
(cm/s) conducted ﬂamelets extinguished
1 3 2 1
1.5 2 2 0
2 4 3 0
3 3 3 0
4 3 2 0
5 1 0 0

 

 

 

 

 

 

placed further ﬂom the sample than the aluminum foil. This reduced the heat being
absorbed ﬂom the ﬂame, and possibly aided in producing a larger instability range. In
these tests, ﬂamelets were seen at tunnel velocities between lcm/s and 4cm/s.

Similar ranges for the copper and brass substrates were produced. However, only
the beginning two substrate results are shown because the ﬁrst goal of showing an
observable region of microgravity ﬂame instabilities was accomplished. Some solid
backings were painted black to change the emissivity of the substrate. These results will
be discussed in a later section. It was discovered that if steel screens were used as
backings, which allowed air to diffuse ﬂom the back of the sample, a ﬂatter beginning
ﬂame ﬂont resulted. The screens, though, were no longer used because a better igniter
system like in the MSU Flame Rig was installed which produced excellent initial ﬂame
ﬂonts. Also, the diffusing air towards the back of the sample added another complication
for future models.

From these tests, the fact that different substrates produced similar instability
ranges, supported regarding these instabilities as a fundamental attribute of diffusion

ﬂames.

111

4.2.2 NASA Instability Characteristics

The top and side camera views, like those shown in Figures 33a and 33b,

4— llamc spread

ﬂame spread

leading cduc

g

 

Figure 33a. Flamelet top view. Figure 33b. Flamelet side view.

illustrated that the ﬂamelets were three-dimensional entities. The top view shows oval or
circular ﬂamelets, and side view shows the familiar crescent shape. Flame spread
direction for all side views was to the left, or in the direction the leading edge pointed.
Flame spread for all top view ﬂames was always downward. The two views indicated
that any numerical computer simulation of such ﬂames must be three-dimensional. For
ﬂames close to extinction, the leading edge of the ﬂarnelet lifted vertically away ﬂom the
paper. A time sequence of this is shown on the next page in Figure 34. As the drop
progressed, the ﬂame became thinner denoting possibly a decrease in chemical reaction
intensity. The leading edge lifted more and moved back towards the tail until it ﬁnally
disappeared. Therefore, it was evident that the ﬂame needed more oxygen because it was
“reaching” away ﬂom the fuel surface. Standoff distances (the distance ﬂom ﬂame to

fuel) of these ﬂames were larger than stable ﬂames, indicating they were fuel rich and

112

 

t=l.47 sec t=l.67 see

Figure 34. Side view of progression of ﬂamelet extinction.

113

needed more oxidizer. Stable ﬂamelets that lasted throughout the drops had leading edges
that were closer to the sample than unstable ﬂamelets. Even though the nature of the
ﬂamelets suggested that an entire three-dimensional simulation is necessary, valuable
information was discovered from two-dimensional analyses of experimental results that
will be discussed in later sections.

The side view also showed that sample to substrate distance affected the ﬂamelets
and ﬂame fronts. For substrate distances close to the sample (around 5mm.), only the top
side of the ﬂame spread along the sample because enough heat was removed that the
ﬂame quenched on that rear sample side. Examples of this are shown in the previous
Figures 33b and 34. For substrate distances between 0.5-O.8cm., the ﬂame partially
wrapped around the leading edge of the sample. In other words, the usual crescent
shaped ﬂame spread on the top; however, at the leading edge, the ﬂame wrapped around
to the backside of the sample near the substrate. On the backside, though, only part of the
crescent slmpe was seen. Substrate distances around lcm. were far enough away from
the sample that another ﬂame which mirrored the top ﬂame formed, so two crescent
shaped ﬂames spread on both sides of the fuel.

For the 10.5cm. sample width that was used in these tests, usually no more than
three ﬂamelets would form. Sometimes only one or two ﬂamelets occurred, however,
when three were produced as in Figure 35, they were ﬁ'equently evenly spaced across the
sample. This spacing suggested that ﬂamelets could not survive very close to each other
due to the fuel and oxygen transport regions around the ﬂamelets. If the ﬂames were too
close to each other, one of them would consume the oxidizer inﬂux near the second

ﬂame, depleting it of a necessary component. This would cause the other ﬂamelet to

114

 

Figure 35. Typical spacing between three ﬂamelets.

extinguish. In fact, this is exactly what happened in certain tests. In some drops like the

one shown in Figure 36a and 36b, the ﬂame front corrugated and separated into ﬂamelets

   

Figure 36a. Test with close ﬂamelets. Figure 36b. Combination of ﬂamelets.

that were quite close to one another. Usually, one of the ﬂamelets was larger than the
other one. As time progressed, the larger ﬂamelet grew in surface area and seemed to
swallow the smaller ﬂamelet. In another view, the smaller ﬂamelet extinguished.
Sometimes the recombination of the two produced a mini ﬂame ﬁont that extended

partially across the surface. The mini ﬂame front in Figure 36b was the result of the

115

combination of the right two ﬂamelets shown in Figure 36a. In other drops, the larger
ﬂamelet increased or remained the same size. In all of these cases, it was clear that a
deﬁnite region around the ﬂamelets existed where the fuel and oxygen transport was
crucial to the survival of the ﬂamelet. It was concluded that the even spacing in some
tests was a semi-stable conﬁguration in which sufﬁcient fuel and oxidizer enabled the
ﬂamelets to exist.

Flamelets near extinction would oscillate. From the top view, the ﬂamelet would
grow in size, shrink, and then grow again. This motion was quantiﬁed and will be
discussed in greater detail in the next section.

Flamelets were either circular or oval ﬁ'om the top view. Examples are shown in

Figure 37a and Figure 37b. The more stable ﬂamelets were oval, with the major axis

‘%

ﬂame
smear

ﬂame
sprezu

   

Figure 37a. Spherical ﬂamelet. Figure 37b. Oval ﬂamelet.

along the width of the sample and the minor axis along the direction of ﬂame spread.

They resembled mini ﬂame fronts on a smaller scale. Circular ﬂamelets were closer to

extinguishing. Flamelets that would shrink to quenching were always circular, and this

116

rapid decrease in size was seen at the lowest air velocities. This suggested that stability
depended highly on oxygen delivery to the ﬂamelet. The fact that the shape was circular
near extinction also insinuated that the ﬂamelets near extinction might be trying to create
a “lowest energy conﬁguration” to survive. Many processes in nature tend toward the
least energy possible, and an example is the raindrop, which is spherical in shape because
this is the geometrically lowest attainable energy-state.

All ﬂames and ﬂamelets in microgravity were blue in color implying lower
temperatures than ﬂames in normal gravity which were yellow from the high soot
production. The lower temperatures occurred because the absence of buoyancy made the
diffusion of fuel vapor and oxidizer the primary source of mixing, so the ﬂame burned
less vigorously in microgravity than normal gravity. As seen from the previous ﬁgures,
the intensity of blue was different in various regions of the ﬂamelets. Oval ﬂamelets had
bright blue leading edges with more diffuse backs. This was enhanced for more stable
oval ﬂamelets because these ﬂamelets would not oscillate, but spread down the fuel. The
leading edge and sides of the ﬂamelets were bright blue tracing out the shape of a bowl,
whereas the far middle and back portions were less luminous. Circular ﬂamelets near
quenching had a more homogenous distribution of the color intensity. One possible
explanation for these differences in intensity is that the more stable oval ﬂamelets were
attempting to become a fuel front due to an adequate amount of ﬁlel and oxidizer. The
leading edge of this mini ﬂame front was preventing three dimensional oxidizer ﬂow
from traveling to the back of the ﬂame so the chemical reactions were weaker indicated
by the decrease in color. The intensity of circular ﬂamelets was more uniform because

oxidizer could reach the back portion of the ﬂame by traveling along the sides of the

H7

ﬂamelet. Therefore, chemical reactions in the back of the ﬂame would be more intense
in the circular rather than oval ﬂamelets. Furthermore, circular ﬂamelets were near
extinction and attempting to achieve a lower energy limit. These are all incipient
rationalizations for which attempts at quantiﬁcation will be made in future testing.

A very interesting characteristic that was also seen in Dartﬁre testing of thick
ﬁiels by Olson [55], was that the side ﬂamelets near the insulating holder would
sometimes propagate ahead of the middle ﬂamelet or horizontal line of the ﬂame front.
This phenomenon was mainly seen in the longer time period testing in the Zero-G drop

Tower. An example is shown in Figure 38. One possibility for this behavior was that the

ﬂame spread

1

..\~“..“

1" w d' ._

 

Figure 38. Typical test with faster side ﬂamelets.

side of the ﬂamelet close to the edge of the holder contained more oxygen than the side
near another ﬂamelet. On the sides, there were no adjacent ﬂamelets to compete with for
oxygen, so the three-dimensional oxygen transport mechanism was correspondingly

magniﬁed. Another explanation was that the holder, which was not a perfect insulator,

ll8

may have been conducting heat ahead of the ﬂamelet feeding the ﬂame with energy, and
thereby increasing the ﬂame speed. Future investigations will address quantifying and
supporting the above proposed explanations.

In the 5 Second Drop Tower, ﬂame fronts corrugated and broke apart into
ﬂamelets that oscillated, recombined, or extinguished. However, air velocities in the 5
Second Tower in which ﬂamelets were observed were larger (4-10cm/s) than in the 2.2
Second Tower (l-Scm/s). This may have been caused by the back-pressure regulator
used to regulate airﬂow in the test section in the 5 Second Tower or by the fact that the
gravity levels in this facility are lower. The same ﬂame phenomena, though, were seen in
the tests in the Zero-G tower as in the 2.2 Second Tower.

4.2.3 NASA Flamelet Oscillations

In many tests, ﬂamelets exhibited oscillatory motion in size from the top view.
Flamelet oscillations always began as an increase in the surface area, then a decrease, and
repetition of this action until the drop ended. A time sequence of this is shown in Figure
39 on the next page. In the time sequence, the right ﬂamelet formed at 2.25 seconds,
shrank to its smallest size at 3.36 seconds, then grew larger at the end of the drop. This
periodic change in size indicated the beginning of oscillatory behavior, which most likely
would have continued if the drop time were longer. In some cases, the ﬂamelet would
shrink to quenching after a couple of oscillations. For most tests, though, the short drop
time would end before the ﬂamelets extinguished. Thus, it was not possible to determine
if the oscillations would persist indeﬁnitely, which supported the longer duration of

testing needed in space ﬂight.

119

 

t = 2.47 sec t = 4.03 sec

 

t=3.14sec t=4.22 sec

 

t=3.36 sec t=4.55 see

Figure 39. Time sequence of right ﬂamelet oscillations.

120

Oscillations were much more prominent along the major axis of the ﬂamelet,
parallel to the width of the sample. Minor axis movement was much less extreme.
Therefore, oscillation frequencies were measured using the major axis. Examples of
oscillations along the major axis are shown in the plots in Figure 40a. The line along
which the major axis was measured for the ﬁrst plot in Figure 40a is shown in Figure
40b. The results in Figure 40a had identical air ﬂow velocities. However, the ﬂamelets
measured were located at different parts of the sample, so frequencies were not affected
by position on the sample. Time zero was the instant that the ﬂamelet formed ﬁ'om the
ﬂame from within the 2.23ec. drop interval. Frequencies were always on the order of
le., the same frequency as the microgravity candle ﬂame experiments, [4]. This
suggests that the unstable mechanisms are fuel independent, and that the present and
future related studies can be applied to fundamental knowledge of the near extinction
limit region. One limitation in these tests was that no more than 2 oscillations were seen
in the 2.2 Second Tower.

The oscillations indicated that the gas phase processes were unsteady. A potential
interpretation is that the ﬂamelet increased its surface area to enhance diffusional oxygen
delivery, then shrunk to preheat the ﬁle] and sustain the chemical kinetics. The ﬂamelet
continued this balance mechanism until it recombined with another or could no longer
endure and extinguished. As the ﬂame moved outward, the chemical kinetic times were
longer compared to ﬂow times so the Damkohler number became smaller. Therefore,
since the mixing of the fuel and oxidizer was faster than the reaction, one or both of the
reactants leaked through the reaction zone. This may have caused a thermal diffusive

instability.

121

 

 

 

 

1.8 ,
~ - . :»- .t
Major Axrs . '
(cm) 16 ' ‘ u.
.‘ . ‘ ‘ .0.
O. ‘ 0.. Q I
14 ' ’o”b? . .. .9.i.
. . . 9.? ,.
1-2 _ ; f°°0'o i Steel backing
; } Zeta/s .
1.0 - 9 . : airﬂgm
0 .2 .4 .6 .8 1.0 1.2 1.4 1.6
Major 5 's ”I“? > Steellbaclcmg
(cm) ' ‘ i . 'i"“2?"‘.“/s
15 : ' X ' . airﬂow
' {e : : . I i
1.1 . , . .+“.f..... . 0.0 g‘..,. 1
V . o". ...1‘:... ..
o _ 0°
0.7 .. ‘0“.

 

o .2 .4 .6 .8 1.01.2 1.4 1.6 1.8
Time(sec)

Figure 40a. Examples of time periods of major axis ﬂamelet oscillations.

 

Major Axis

Figure 40b. Major axis of oval ﬂamelet.

122

Oscillations were never seen at the lowest oxidizer velocity with the aluminum
foil backing, but were seen at the same ﬂow rates with the steel backing. This showed
that the stability of the ﬂamelets was directly related to the backing material, distance
between backing and sample, and oxidizer velocity. Future tests will strive to determine
these relationships in more depth.

Tests conducted in the 5 Second Drop Tower were in general agreement with the
2.2 Second Tower. Oscillations were observed and occurred along the major axis, and
the substrate backing greatly affected the heat loss and thus instabilities. Oscillation
frequencies were the same order of magnitude as 2.2 Second Tower tests. In this tower,
some ﬂamelets continued to oscillate during the entire drop without extinguishing while
others quenched or recombined to form a new ﬂame front.

In some of the 5 Second Drop Tower experiments, ﬂamelet intensity in the blue
region was tracked over time and showed an oscillatory nature. Figure 41 shows a Me

for a typical test in which this behavior occurred. In Figure 42a, the change in blue

r. ...

,_ Intensity
measured
along lme

 

Figure 41. Line proﬁle along which intensity measurements were taken.

123

B
Integselty

  

3 2
Distance (pixels)
50 pixels=lcm

3 4
Time (see)

Figure 42a. Intensity in blue region over drop time along line proﬁle in Figure 41.

180
160
140
Blue 120

Intensity 100
80

60
40
20

0

 

73 98 126 158 191 225 258 290 320
Line length (pixels)

Figure 42b. Intensity measurement for single frame of drop with ﬂamelet along line
proﬁle in Figure 41.

124

intensity along the white line proﬁle shown in Figure 41 was measured over the time of
the drop. For each frame of the video, the intensity in the blue region was measured at
every pixel along the line. An example of a single frame measurement is in Figure 42b.
This plot represents the blue intensity measurement along the line for a frame in the drop
when the ﬂamelet shown in Figure 41 was located along the line. The “hump” in Figure
42b represents the ﬂamelet intensity. The steep increase in the blue intensity on the left
denotes the intense blue leading edge of the ﬂame, while the more gradual decrease on
the right indicates the more diffuse back. Figure 423 was obtained by “stacking” the
consecutive frames of intensity measurements (like in Figure 42b) for the entire drop
time.

From Figure 42a, it was evident tlmt the blue intensity of the ﬂamelet oscillated
throughout the drop. Sometimes, the intensity was measured in the dead space between
ﬂamelets during the drop since they would recombine with others or extinguish. The
oscillatory nature, however, was not due to the absence of a ﬂame along the line. For
example, the oscillations in Figure 42a were not because the ﬂamelet extinguished during
part of the drop. In Figure 41, the ﬂamelet was present along the line for the entire drop.
The oscillatory results indicated a periodic ﬂuctuation in the chemical processes of the
ﬂamelets, not just the size. Since it was evident that energy processes may be oscillating,
these measurements pointed out the need for quantitative temperature measurements.
4.2.4 NASA Flame Spread Rates

Flame spread rates in microgravity were calculated by tracking the leading edge
of the ﬂame or ﬂamelet over the entire drop. All plots showed a linear curve, for which

the slope was the spread rate. Figure 43 displays the relation between ﬂame spread rate,

125

opposed air velocity and backing distance for both microgravity and normal gravity

ﬂames. The backing distance is denoted by the size of the bubble radius, so a larger

 

bubble radius = substrate distance
smallest radius: 0.673 cm

 

 

 

 

 

 

 

 

 

0.813cm
1.00cm
largest radius: 1.68cm
1.40 -———————1———J
1.20 - ‘l-g '
1.00 -* .“8
Spread 030 - Oﬂamelets
Rate
(cm/s) 0.60 -
0.40 -
0.20 -
0.00 T i r
0 2 4 6 8

Flow Rate (cm/s)

Figure 43. Flame speeds versus substrate distances and air ﬂowrates.

radius means that the backing was further from than sample than for a smaller bubble
radius. The purpose of this plot was to check that the ﬂame spread rates for ﬂame fronts
and ﬂamelets were plausible and to study any relationships that may exist between ﬂame
spread rate and substrate distance.

Normal gravity spread rates did not change due to the opposed oxidizer ﬂow or
the change in backing distance. This was consistent with the opposed ﬂow velocities
being smaller than buoyant ﬂows, which dominate the spread rates at low ﬂow in normal

gravity. The buoyant velocities created vigorously burning ﬂames that were much hotter

126

than microgravity ﬂames, so the substrate was not an effective heat sink that altered the
spread rates. Nearly all microgravity ﬂame ﬁonts and ﬂamelet spread rates were much
lower than the normal gravity spread rates due to the weaker burning rates and lower
temperatures in microgravity from lack of buoyant mixing. Only one microgravity ﬂame
lmd a large spread rate near normal gravity values. This occurred for the largest substrate
distance, so the substrate was totally ineffective.

Figure 43 shows that the substrate distance inﬂuenced the spread rates for
microgravity ﬂame fronts and ﬂamelets more than the oxidizer ﬂow rate. For most
oxidizer ﬂowrates, approximately the same range of microgravity ﬂame velocities was
produced. However, for each individual ﬂowrate, the microgravity ﬂamelet and ﬂame
front spread rates varied. Therefore, it was concluded that backing distance was a
primary parameter that affected microgravity ﬂame spread rates and ﬂame behavior.

4.2.5 NASA Surface Energy Balances

Energy balances at a single point on the surface of the fuel were calculated
throughout the entire drop. From the calculations presented in Sections 3.2.1-3.2.3, the
surface energy losses due to radiation, conduction, and vaporization, plus the total
enthalpy change in time throughout the drop were plotted. The sum of these quantities

equaled the unknown heat ﬂux coming in ﬁom the ﬂame. In particular, Equation 6 was

-ES

0(Ts4 4:) (TS —Tb) RT;

[ _1— + L _1] l"lair —L"""""’
8S 81,

calculated for each time step in the drop and graphed. Examples of these energy balance

 

[ J 6T
“szf-To‘iﬂ +AsPsTe Lv+Pscp57 = qinﬂame (6)

curves are shown in Figure 44. In the plot, the large increase in energy represents ﬂame

127

—Al back, 1cm, 3 ﬂamelets
10 —Alback,lcm,2ﬂamelets
9 —Screen back, .6cm, 3 ﬂamelets

8 —Cu back, .6 cm, I ﬂamelet
7

Surface 6
Energy 5

(W /cm2)4
3

 

2

1

0

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Time (see)

Figure 44. Typical surface energy balances for 2 second drops.

passage over the thermocouple (point on the surface where the data was recorded). At
the beginning of the tests, the temperature of the fuel surface was room temperature. As
the ﬂame approached the thermocouple, the temperature of the surface increased in a
very short period of time indicating the vaporization of fuel. Many tests like those in
Figure 44 Ind energy values at the end of the drop larger than zero. This was most likely
due to the fact that most drops ended while ﬂamelets were still present. Therefore, the tail
end of the ﬂame would have still been traveling over the thermocouple, causing
temperatures and consequently energy values exceeding atmospheric values.

Most drops had energy values between lW/cm2 and IOW/cmz, which was

consistent with previous studies of microgravity ﬂames such as [27]. In Figure 44,

128

surface energies were compared for drops with different, substrates, substrate distances,
and numbers of ﬂamelets. Time zero was the beginning of the drop, and one ﬂamelet
denotes a uniform ﬂame ﬁ'ont that did not break apart. In Figure 44, the drop with three
ﬂamelets showed the surface to have larger amounts of energy losses than drops with
fewer ﬁamelets. This agreed with observations that tests with more ﬂamelets were
unstable compared to a stable uniform ﬂame front because they were closer to extinction.
It was also concluded that radiation terms dominated the energy balance, which agreed
with current near limit extinction theory of ﬂames. When compared with other energy
terms, such as conduction, vaporization, etc., in all but one test radiation accounted for
over eighty percent of the total energy contribution.

The above conclusions were the ﬁrst step in assessing the validity of the
measurements and preliminary calculations. In the future, a detailed gas phase energy
balance will be attempted, so that a more comprehensive energy analysis can be
completed. This will lead to a better understanding of the thermal difﬁisive (Lewis
number effects) or hydrodynamic contributions to ﬂame instabilities.

4.2.6 NASA Peak Temperature Surface Energy Balances

Surface energy values at the peak temperatures for tests versus the number of
ﬂamelets formed were analyzed. The goal was to see if a relationship between ﬂamelet
formation and energy existed. The peak temperature was chosen because in theory, it
was the temperature that would most likely dictate the instability of the ﬂame at the

leading edge. In the energy balance from equation 18, the conduction ﬁom paper to

4 4 ;E

4 0(T -T) (T ’Tb) [RT] ..

0350:; —Tb)+(:i+p'i—bﬁ+lair‘_spL_’+AspsTe sp Lv = qnetlosses (18)
SS 5b

129

backing, radiation and reradiation from backing to surface, and vaporization terms were
summed. The contribution from the change in total energy (time dependent term) was
zero since at the peak temperature, the slope was zero. Therefore, the energy balance was
really the maximum of net losses ﬁ'om the surface. These net losses equaled the incoming
radiation and conduction from the ﬂame.

The number of ﬂamelets formed versus the peak energy losses for each test are

shown in Figure 45. Since ﬂamelets did not always pass over the thermocouple or

 

4.....-rr..4.........--.
3- I II I a
Number
of 2- I -
Flamelets
1L I I I -
0 IIAIJJIILJLIAAAIIALLI

 

 

 

o 2 4 6 8 10 "12
Surface Energy Losses (W/cm2)

Figure 45. Number of ﬂamelets versus peak energy losses.

thermocouples would break, not all drops were included in Figure 45. In the plot on the
vertical axis, one ﬂamelet corresponded to a single ﬂame front. It was anticipated that

the losses at the peak temperatures would be closely related to any ﬂame instabilities. As

130

 

expected, a region between 3W/cm2 and 6W/cm2 of losses occurred at ﬂamelet
formation. However, the plot also implied that in tests with more ﬂamelets, the ﬂamelets
had greater net losses than a single ﬂame front that stretched across the sample. The two
outlying points showed this: the test with the ﬂame front, with losses of l.5W/cm2, and
the test with three ﬂamelets, with losses around lOW/cmz. This was expected because
ﬂamelets were more unstable compared to uniform ﬂame fronts, most likely because of
large energy losses. The interesting feature of this plot is the qualitative feature that
exhibited a critical net loss of energy at which ﬂamelet formation occurred. There was a
general critical region of losses between 3 and 6 W/cm2 common to all tests, regardless of
the number of ﬂamelets. In this region, the large energy losses caused by the substrate
made it difﬁcult for the ﬂame to survive, so the ﬂame front had to fragment into smaller
ﬂames in order to exist. It is important to note that Figure 45 shows the beginnings of a
qualitative trend, and more tests are needed to deﬁne a critical net loss quantity. Thus,
this will be the basis for measuring surface energy losses in ﬁiture tests as well as
searching for gas phase energy connections with number of ﬂamelets.
4.3 MSU Flame Rig Results and Discussion
4.3.1 MSU Flame Rig Flowmeter Veriﬁcation

The in-line ﬂowmeter, or rotameter, readings and correction and conversion
factors explained in Chapter 3.3.3 were veriﬁed. The ﬂowmeter readings were ultimately
used for calculating the bulk test section velocity using mass conservation. This was

accomplished using the calibrated sonic oriﬁces for low ﬂow velocities. The oriﬁces

came with a calibration sheet that gave the volumetric ﬂowrate at standard conditions for

131

different upstream pressures. The calibrations were NIST (National Institutes of
Standards and Technology) standard, so the given ﬂowrates were precise.

Before determining if it was feasible to use the ﬂowmeter in-line and correct for
density changes, the ﬂowmeter was validated under the environment that it was
calibrated which was standard conditions. To do this, the pressure regulator, oriﬁce, and
ﬂowmeter were placed in a line connected to the 95psig air supply. A block diagram of
the setup is shown in Figure 46a. The losses of the tubing were considered negligible
since the tube lengths were less than one foot. The ﬂowmeter exhausted to atmosphere.
Since the tubing was short and the ﬂowmeter had negligible pressure losses, it was
assumed that the downstream side of the oriﬁce was at atmosphere. If the ﬂowmeter
were accurate, then the measurements from this setup would match the volumetric ﬂow
measurements of the calibrated oriﬁce, since both were at standard conditions.

The ﬂowmeter readings at atmospheric conditions compared well to the calibrated
oriﬁce readings shown in Figure 46b. Two oriﬁce sizes were chosen to test the
ﬂowmeter. Since the ﬂowmeter and oriﬁce calibrations were at standard conditions, it
was only necessary to compare volumetric instead of mass ﬂowrates because the
densities were the same. From Figure 46b, it was clear that the ﬂowmeter worked
properly. Volume ﬂowrates for the meter working at stande conditions were very close
to the calibrated values from the oriﬁce. The ﬂowmeter readings were slightly higher
than the oriﬁce readings. The average error between the ﬂowmeter and calibrated values
for both oriﬁces was 6.5%. This was most likely because the resolution of the ﬂowmeter

was smaller by one factor of ten than the calibrated values of the oriﬁces.

132

Checking Validity of Flowmeter
at Standard Conditions
High
Pressure
Source

 

Flowmeter mun

Pressure % I
Regulator

Figure 46a. Setup for verifying ﬂowmeter at atmospheric conditions.

 

 

 

 

 

 

 

 

——Orifice .0079" diameter measured
+Orifice .0079” diameter calibrated
Ari-Orifice .011" diameter measured

10 +Orifice .011" diameter calibrated :' " '

 

 

 

 

Volume Flow8 -

@ 6 3

Standard _'

Conditions _
(SCFH) 4

 

 

 

E
2 L
L
t

nan-I-aall..-.l-..nl-..nl....I-L-_LLIIL

0 u.
10 20 3O 4O 50 60 70 80 90
Upstream Pressure (PSIG)

 

Figure 46b. Volume ﬂowrates of ﬂowmeter at atmospheric conditions for two oriﬁces.

133

The next step was to test the validity of placing the ﬂowmeter in-line and using
the correction and conversion factors in Chapter 3.3.3 to calculate the correct mass
ﬂowrate. In this setup shown in Figure 47a, the ﬂowmeter was being tested for its
operation in-line, so it was placed before the oriﬁce at pressures indicated by the pressure
regulator. At pressures higher than atmospheric, the ﬂowmeter had to be corrected for
the change in density of the air moving through it by multiplying the ﬂowrate times
(P.un/P.c)m. This gave the actual ﬂowrate at the in-line pressure. The actual ﬂowrates
were then converted to standard conditions (atmospheric) by multiplying the actual
ﬂowrate times (Pm/Pm). Therefore, the corrected and converted ﬂowrates of the in-line
ﬂowmeter could be compared to the volumetric ﬂowrates given by the calibrated oriﬁce
because both were at standard conditions. If it functioned properly, then the corrected and
converted volumetric ﬂowrates would be the same or close to calibrated oriﬁce ﬂowrates
by mass conservation. The result of comparing the ﬂowmeter with the previous used
calibrated oriﬁce with 0.011in. diameter is shown in Figure 47b. The smaller diameter
oriﬁce was not used because the ﬂowmeter inline values were below the ﬂowmeter scale
and could not be determined.

From Figure 47b, the ﬂowmeter measurements that were corrected for inline
pressures and converted to atmopheric conditions were close to the calibrated values of
the oriﬁce. The corrected and converted meter values were generally lower than the
calibrated values, and were not as accurate as measurements taken when the ﬂowmeter
was at atmospheric conditions. Larger deviations ﬁ'om calibrated values were seen as the
pressure increased. However, the average error between measurements was 10%.

Therefore, it was concluded that the ﬂowmeter could be used to estimate the bulk

134

Checking Validity of Flowmeter
at High Pressure Conditions

  
  

 

 

Flowmeter *

  
   

Pressure
Re ator

 

 

 

Figure 47a. Setup for verifying ﬂowmeter at inline pressure conditions.

 

 

 

10 +Oriﬁce .01 1" diameter measured .
9 +Oriﬁce .01 1" diameter calibrated E
8
Volume Flow 7 3
@ 3
Standard 6 1
Conditions 5 i
SCFH i
( ) 4 -:
3
2

 

lllllll

 

20 3O 40 50 6O 7O 80
Upstream Pressure (PSIG)

Figure 47b. Comparison of ﬁowmeter to calibrated oriﬁce at inline conditions.

135

velocity ﬂow in the test section within 10%. This was acceptable because in the
preliminary testing encompassing this thesis, approximate ranges of velocities where
instabilities occurred were desired. After these ranges are conﬁrmed, more sophisticated
means of measuring test section velocity will be utilized. Also, to reduce error, tests were
run at pressures of 60psig and below.
4.3.2 MSU Flame Rig Flame Instabilities

Preliminary testing in the MSU Flame Rig made it evident that many more
dynamic instability behaviors may be seen in a zero gravity environment. The MSU
Flame Rig produced ﬂame instabilities similar to drop tower experiments, and additional
unstable behaviors not seen in NASA tests. Initial ﬂame ﬁ'onts and ﬂame instabilities
required test section velocities above lOcm/s compared to NASA tests. Generally,
instabilities were seen from 5-28cm/s, using predominantly the needle valve. As the ﬂow
velocity decreased below 28cm/s, the ﬂame front corrugated and broke into smaller blue
ﬂames or yellow ﬂamelets. A time sequence of a typical test in the MSU Flame Rig is
shown in Figure 48. Times are located in the lower right corner of each frame. Total test
times during which ﬂame instabilities formed and developed were much longer than
NASA tests. Most tests lasted l-5min. Higher ﬂowrates were needed in the MSU Flame
Rig than in the NASA Rig in order to establish a ﬂat ﬂame front because there was no
buoyant motion to feed the ﬂame and the top plate of the test section was probably
removing heat away from the ﬂame. The top plate basically was forcing the ﬂame into
an enclosed area with a limited supply of oxygen, most likely not sufﬁcient enough for
ignition. Therefore, test section velocities as high as 30 to 40 cm/s were employed at

ignition.

136

 

Figure 48. Typical progression of MSU Flame Rig test.

137

After a uniform ﬂame ﬂont was generated across the sample, the ﬂow was
steadily and slowly decreased. In these tests, both heat loss and diminished oxygen
supply produced the instabilities. As the oxidizer ﬂow was reduced, distinct behaviors
would occur. In some tests, the entire ﬂame ﬂont slowly oscillated tracing out a
sinusoidal motion like a standing wave produced by a string with ﬁxed nodes. The ﬂame
ﬂont would then turn ﬂom yellow and sooty to a very thin and blue ﬂont that was on the
order of a few millimeters in thickness. A time sequence of this is shown in Figure 49.
Less than thirty ﬂames per second were digitized ﬂom the MSU tests because more than
this would give huge quantities of information when images were downloaded.
Therefore, ﬂames with the same time stamps are not identical ﬂames, and they denote a
change of the ﬂame within one second. In Figure 49, the beginning ﬂont became
corrugated like a sine wave at t=20 seconds. The ﬂame ﬂont resembled a standing wave
by oscillating throughout the test until it became a thin blue ﬂont. The “wave” motion
can be seen when the middle of the ﬂame that started concave with bulges on either side,
would periodically protrude forward. When this happened, the ﬂame bulges on the sides
became ﬂatter, and then it repeated, while thinning and becoming blue. The motion of
the concave middle changing to ﬂattening sides is shown ﬂom t=20 to 21sec., again ﬂom
t=2l to 27sec., and again ﬂom t=28 to 40sec. The ﬂames that show the most prominent
concave middle with bulging sides were the first ﬂame at t=208ec., the second ﬂame at

t=213ec., the ﬁrst ﬂame at t=29sec. At t=403ec., the ﬂame is totally ﬁat.

138

COUCH.

mid dl

_) '

1—20 see

F28 see

I 40 see

 

Figure 49. Oscillatory motion of entire ﬂame ﬂont in MSU Flame Rig test.

139

 

After the ﬂame ﬂont became thin and blue, small ﬂames behind it sometimes
remained. These would later turn into ﬂamelets that generated a corrugated ﬂame ﬂont
with many smaller sine waves. Some ﬂamelets survived while others extinguished. An

example is in Figure 50.

 

Figure 50. Time sequence of a uniform ﬂame ﬂont changing into ﬂamelets in MSU
Flame Rig.

In other tests, very fast pulses traveled periodically across (parallel to) the initial
ﬂame ﬂont or across already formed ﬂamelets. As the ﬂame ﬂont advanced, there were

also ﬂashes of light moving across it. Figure 51 shows a time sequence of a test with

140

pulse 1

pulse 3

pulse 1
ﬂ

pulse '3

ﬂ

pulse 3

ﬂ
pulse 2

'e
g
‘.

*4
iw-‘U

 

Figure 51. Periodic pulses travelling across ﬂame ﬂont in MSU Flame Rig.

l4]

pulses moving across the ﬂame. The pulses were in the region of the photos with bright
lines. The lines appeared because the pulses were moving so fast that the camera could
not capture them clearly. The pulses had a very fast ﬂequency much larger than the drop
tower ﬂamelet oscillation ﬂequencies, with one pulse usually proceeding across the ﬂame
in under a second. Therefore, the ﬂames in Figure 51 were labeled ﬂom a to j to signify
their order. Pulses usually originated at the middle of the sample and moved outwards
disappearing at the edges. This was most noticeable ﬂom ﬂames c to g. The bright pulse
started in the middle in ﬂame, and progressed upward in the picture to the sample edge.
The pulse reached the edge in ﬂame f as it is located under the last cusp at the top of the
picture. In ﬂame g, it was still present, but had lost intensity, and had disappeared in
ﬂame h. A third pulse commenced again in ﬂame i, but was so fast that it disappeared in
the last ﬂame, j.

The tests with uniform ﬂame ﬂonts that exhibited sinusoidal curves, like in Figure
49, resembled the tests with the fast pulses like in Figure 51 because the sinusoidal
motion was like a very slow pulse across the ﬂame ﬂont. The trailing yellow sooty part
of the ﬂame in Figure 49 also exhibited sinusoidal behavior, moving like a slow pulse
across the ﬂame. In experiments with pulses, the ﬂame ﬂont was usually yellow and
sooty. Sometimes, both mechanisms would be present, as tests would begin with pulses
and then progress to the slow sinusoidal motion, turning into the thin blue ﬂame ﬂont.

The pulses moving across ﬂamelets would sometimes cause the ﬂamelets to
extinguish. An example of this is shown in Figure 52. Again, the ﬂames were labeled in
order ﬂom a to j. In ﬂame a, there were mini ﬂame ﬂonts or blue cusp ﬂames burning

over the sample. The ﬁrst pulse, shown by the bright spot in ﬂame b, occurred in the

142

 

. f
puke 1 g I pulse '.-
h .. T: h

pulse 3
pulse 2
i? ‘—
1 4—

pulse 3

‘ pulse 2.

~‘6.

1:4—

pulse3

~ 1 llarnelel
I ﬂamelel

’g ‘—pulse 1

 

Figure 52. Periodic pulses that extinguish ﬂamelets and ﬂame ﬂonts in MSU Flame Rig.

143

middle of the sample. As the pulse moved downward in the photo to the edge of the
sample, shown in ﬂames c through f; it extinguished the blue cusp ﬂames, present in
ﬂame a. In ﬂame g, two new pulses formed in the middle of the sample. Pulses 2 and 3
moved opposite each other to the edges of the sample. From this motion, pulse 2, which
went upward in the photo, extinguished the remaining blue cusps on the upper side of the
sample. However, after the two pulses had moved outwards, two ﬂamelets remained and
continued to burn over the sample. This quenching phenomenon resembled more the
hydrodynamic instability, in that the motion of the pulse extinguished the ﬂamelets or
smaller blue cusp ﬂames, like blowing out a candle. In NASA tests, when ﬂames moved
near each other, they usually combined. The larger ﬂamelet would swallow the smaller
and become bigger still. However, in the MSU Flame Rig, it was as though the force of
the pulse created a blowout condition for the ﬂamelet, in which the ﬂow times were much
larger than the chemical processes, so the ﬂamelet could not sustain itself. Pulses rarely
ever moved across the beginning uniform thin blue ﬂame ﬂont.

Aﬂer the ﬂame pulsed or became thin and blue, it corrugated as oxidizer was
reduced and broke apart into smaller ﬂames that resembled the drop tower ﬂamelets. The
ﬂamelet produced ﬁngering patterns very similar to the smoldering combustion ﬁngers
seen by Zik [44]. An example of a test in which this action transpired is in Figure 53.

The ﬂamelets in the MSU Flame Rig test oscillated, however, their ﬂequencies
were very fast compared to the drop tower ﬂamelet ﬂequencies. The oscillations were
also a periodic increase and decrease in ﬂame let surface area, so it is likely that very
similar controlling processes as in the drop towers ﬂamelets are present in the MSU

Flame Rig. The MSU ﬂamelets were yellow, not blue like the drop tower tests. Some-

144

 

Figure 53. MSU Flame Rig test with unstable ﬂamelets producing ﬁngering pattern.

145

times the ﬂamelets would extinguish, or if they were close enough, would recombine into
a larger ﬂame. Oﬂen, they did not recombine, implying that adjacent ﬂamelets were
competing for oxygen. This would affect the mass transfer of reactants, suggesting that
the ﬁngering was a thermal diffusive instability.

A very interesting phenomenon that was also seen, were blue crescent ﬂames.

Examples ﬂom different tests are illustrated in Figure 54. When the ﬂame ﬂont became

 

Figure 54. MSU Flame Rig tests with stable cusp ﬂames and unstable ﬂamelets.

corrugated, tiny blue ﬂame ﬂonts that were very thin and cusp shaped would form across
the sample and produce the ﬁngering pattern. In many cases, ﬂamelets and blue cusps
were present. All four tests in Figure 54 contained ﬂamelets, which were the smaller

yellow circular ﬂames. Test 1 contained three ﬂamelets, one at the top and two at the

146

bottom of the picture; Test 2 had one at the bottom; Test 3 showed ﬂamelets across the
whole sample, some behind the cusps; Test 4 had four at the bottom.

The cusp ﬂames looked nearly identical to the stable blue oval ﬂame lets with
bright leading edges as seen in the drop tower tests. The difference though, was that the
MSU tests did not have diffuse backs, only a leading edge or cusp. This may have been a
result of the quartz plate prohibiting ﬂow over the top of the ﬂame. Therefore, there was
no three dimensional oxygen-transport behind the ﬂame. This would have been due to
° the ﬂame ﬁlling most of the cross sectional area and preventing oxidizer ﬂow over itself.
The MSU Flame Rig blue cusps always had oval crescent ﬂonts, not circular ones like
some of the NASA ﬂamelets. The oval shape of the cusps made them appear as mini-
ﬂame ﬂonts like the oval ﬂamelets in the NASA tests. The leading edge of the cusps
acted again as a barrier to three-dimensional transport of oxygen that traveled ﬂom the
sides to the back of the ﬂame. This was because the ﬂame ﬂont was not spherical, so the
oxidizer could not be uniformly distributed to the back of the ﬂame, so there were little or
no chemical reactions in that area. The close proximity of the plate worked in
conjunction with this to prevent oxidizer ﬂom ﬂowing over the top of the ﬂame. Thus
there were no reactions behind the cusps compared to the diffuse backs of the ﬂamelets in
the drop towers.

Like the drop tower tests, the blue crescent ﬂames were much more stable than
the oscillating ﬂamelets because they did not oscillate. In fact the ﬂamelets would often
oscillate very fast, and then ﬂame let would transform into a bright blue cusp. It appeared
as though the ﬂamelet oscillations provided the mechanism for gaining a more stable

state. In some cases though, oscillations would result in extinction. One possible

147

 

ermlanation for the pulses across the ﬂame ﬂont and sinusoidal nature was that the
reduction of the oxidizer velocity was changing the time scales of the gas phase
processes. As a result, it was possible that there existed different modes of ﬂequency in
ﬂamelet instability behaviors. In the drop tower testing, le. may be the dominant
ﬂequency seen at the low time scales produced by the low velocities. In the MSU Flame
Rig, there are higher oxidizer velocities, so it is possible that the dominant ﬂequency of
oscillation is much higher and the lower ﬂequency behavior is present but can not be
detected by simply looking at the video.

More interesting was the fact that the ﬂamelets sometimes traveled along the
underside of the fuel sample. The MSU Flame Rig was arranged so that oxidizer ﬂowed
across the top and bottom of the ﬁiel sample. It was anticipated that the underside of the
ﬂame near the substrate would not be present like in the NASA tests. Instead, the
ﬂamelets were sometimes traveling closer to under the fuel than on the top of the sample.
One explanation could be that the the] char would curl under and obstruct the ﬂow,
producing an accumulation of oxidizer under the sample, and making a more favorable
ﬂame environment in that area. Future testing will employ ashless ﬁlter paper to try to
remedy this. During some tests, the ﬂarnelets under the sample appeared as though they
had extinguished, so all that was left was a ﬁnger. The ﬂamelet would either extinguish
or transform into a stable blue cusp that would appear along the ﬁnger on the top side of
the sample. Since the cusps were more stable than ﬂamelets, it is possible that they were
able to survive above the sample where there may have been less oxidizer.

In some tests, the ﬂat yellow ﬂame ﬂont would turn into the thin blue ﬂont,

which would continue to the end of the sample without corrugating. The ﬂame ﬂont was

148

 

 

parabolic instead and did not break apart. An example of this is shown in Figure 55 with

time stamps in the lower right corner. When the ﬂame reached the edge of the sample, it

 

Figure 55. Burn out of sample by thin ﬂame ﬂont in MSU Flame Rig.

was still thin and blue. If there were any ﬂamelets behind the sample, oscillations began,
as quick ﬂashes of light. However, as soon as the fuel was totally consumed in the
middle of the sample, the ﬂont turned ﬂom blue to sooty yellow as the remaining sides of
the ﬂame consumed the rest of the fuel. In these cases, the ﬂame was acting like a barrier
to the incoming oxidizer, so the ﬂame was weak and blue. The oxygen was only being
received ﬂom the direction of the leading edge of the ﬂame. There was most likely no

ﬂow behind the ﬂame ﬂont, so there was no multi-dimensional feeding of oxygen to the

149

ﬂame. The ﬂame, which stretched across the entire sample and was almost the width of
the test section, resembled a solid object and also ﬁlled most of the cross sectional area of
the test section. Therefore, the ﬂow was forced to either become part of the chemical
reaction or to pass around the sides of the ﬂame. A solid object in this type of ﬂow ﬁeld
would produce considerable drag behind it, causing many vortices and non-uniformity in
the ﬂow. It is possible that this was not sufﬁcient to feed oxidizer behind the ﬂame to
turn it yellow and sooty. When the ﬂame ﬂont split in two after the fuel burned out, a
rush of oxygen ﬂowed through the middle of the sample, which produced a more
vigorously burning ﬂame, indicated by the yellow color. This behavior helped to support
the previous explanations that the stable blue cusps in the MSU Flame Rig and oval
ﬂamelets in the drop towers acted as barriers to the opposed oxidizer ﬂow.

Flames in the MSU Flame Rig displayed many different instability processes.
Different behaviors hinted at the particular type of instability mechanism. The oscillating
ﬂamelets pointed to a thermodiffusive instability because it is well known that this
mechanism promotes bulges in ﬂame ﬂonts, due to imbalance in heat and mass diffusion.
In the MSU Flame Rig, the bulges were exaggerated to the point that ﬂamelets produced
ﬁngering patterns. Zik et al. [44] concluded that the smoldering ﬁngering patterns were
predominantly caused by a thermodiffusive mechanism because the ﬁnger ﬂames
competed for oxygen causing diffusion ﬁelds that were directed at the ﬂont and sides of
the ﬂame. This appeared to be likely in the MSU Flame Rig because a ﬁrll-ﬂedged ﬁnger
ﬂamelet rarely combined with an adjacent ﬂamelet. However, the pulsing action of the
ﬂames implied a hydrodynamic mechanism because they were so powerful, that they

sometimes extinguished the weak ﬂames in the vicinity. The pulsing action was not seen

150

 

 

in the NASA tests, in which the ﬂow was constant. The MSU Flame Rig induced
unstable ﬂames using both heat loss and reduced ﬂow, so it was likely that both
instability mechanisms existed. Further quantitative testing is needed in the future.

4.3.3 Discussion of MSU Flame Rig versus Drop Towers.

The similarity of ﬂame instabilities seen in the MSU Flame Rig to those in the
NASA drop towers show that it is possible to use a Hele-Shaw apparatus as a supplement
to studying microgravity ﬂames. This does not mean that Hele—Shaw devices can or
should replace drop towers. As with all experiments, there are deﬁnite unwanted side
effects. The Hele-Shaw apparatus is not in a true zero gravity environment. It only
simulates it, at the expense of having a large heat sink very near the top of the burning
sample. This could be affecting the heat ﬂux and energy processes of the ﬂame.
Furthermore, a Hele-Shaw apparatus will always produce very viscous dominated ﬂow
due to the close proximity of the two plates. In many cases, researchers would like to
have a uniform ﬂow proﬁle across their samples. To obtain any resemblance of bulk
plug ﬂow, higher velocities are needed. However, this would undermine the effect of the
tunnel since ﬂows well above buoyancy driven ﬂows are required because the purpose of
the apparatus is to simulate ﬂows that are not larger than buoyant velocities in any
direction. Despite these limits, there are deﬁnitely advantages of the MSU Flame Rig.
Most importantly, unlike the drop towers, test times are unlimited. Consequently,
instabilities of thicker ﬁrels can be tested, since it is very difﬁcult to produce them in the
drop towers due to slow spread rates. Also, the oxidizer velocity can be varied in situ,
and very wide samples of many widths may be incorporated. The fact that the ﬂame

instabilities in the MSU Flame Rig resemble those in the drop towers supports the use of

151

an. 1.).I I.-.‘.\“” Is. _‘I'

 

this device for microgravity experiments. In effect, coupled with drop tower or airplane
experiments, the MSU Flame Rig can be very helpful in gaining more information about

thin and thick fuel instabilities under simulated microgravity conditions.

 

152

 

CHAPTER 5
CONCLUSIONS

Instability regions of ﬂames in microgravity spreading over a thin solid fuel were
found for different substrate backing materials. The unstable nature was shown when a
ﬂat ﬂame ﬂont would corrugate and then break into ﬂamelets. The ﬂamelets oscillated,
recombined, or extinguished, and always traveled in the direction of the opposite oxidizer
ﬂow. Oscillations were on the order of le., which was comparable to microgravity
candle results. Line proﬁle measurements showed oscillatory, unstable behavior of the
intensity of the ﬂamelets in the blue region. Flamelet spread rates for different substrate
distances were usually less than microgravity ﬂame spread rates. Substrate spacing
affected microgravity ﬂame spread rates and also determined the existence of ﬂames on
the backside of the sample. An energy balance at the fuel surface during the drop
produced ﬂux values that were of the same order as previous studies. More importantly,
when peak surface energy ﬂux values were compared with number of ﬂamelets for each
drop, an area where there was a “cutoff” ﬂux was evident. This critical region may
provide some indication of the minimum energy of the ﬂamelets. These preliminary tests
and measurements in the NASA drop towers revealed an instability regime of
microgravity ﬂames that can survive at ﬂow rates near those on the Space Station and
Space Shuttle. Such ﬂames could possibly compromise the safety of the people onboard
these vehicles. In effect, the need for further studies to determine the underlying
processes of unstable microgravity ﬂame behavior was demonstrated.

Using ﬂuid dynamic principles, a Hele-Shaw device called the MSU Flame Rig

was successfully constructed which was capable of producing low opposed ﬂow

153

 

velocities, variable test section heights, and a burning sample on the ﬂoor of the test
section in order to simulate ﬂame spread in a low buoyant atmosphere. The ﬂowmeter
readings of the device were veriﬁed so that approximate bulk ﬂow velocities could be
determined in the test section. The wind tunnel incorporates various fuel widths, and a
substrate backing for thin samples, whose distance was variable. A successful igniter
system was fabricated that produced ﬂat ﬂame ﬂonts like NASA tests. These ﬂame
ﬂonts became corrugated and broke into ﬂamelets similar to drop tower results when the
ﬂowrate was decreased. Flameﬂonts were blue, with yellow ﬂamelets. Additional
dynamic processes were present in the MSU Flame Rig, as the very thin ﬂame ﬂont
would often oscillate as a whole, like a standing wave or sinusoidal curve. The ﬂame
ﬂont also broke into tiny ﬂames, which oscillated at ﬂequencies higher than ﬂamelet
ﬂequencies in NASA tests. F ingerlike smoldering also occurred analogous to previous
works. Generally, many different ﬂequencies and unstable behaviors in the MSU Flame
Rig were present for ﬂames in the near extinction limit regime. Observations suggested
that both the thermal diffusive and hydrodynamic mechanisms affected ﬂame
instabilities. The MSU Flame Rig appears to be a very valuable tool for studying ﬂame
spread rates in a reduced buoyant atmosphere. It will aid in the selection of thick fuels to
be tested on the International Space Station. The MSU Flame Rig showed that weak
ﬂames can survive under low ﬂow conditions common to many everyday environments
as well as those on the Space Station. Further studies of the ﬂame mechanisms in the
MSU Flame Rig may give some indication of ﬂame behavior in zero gravity
environments, helping to ensure the safety of persons living and working in enclosed

microgravity environments.

154

 

 

CHAPTER 6
FUTURE WORK

This thesis was a preliminary portion of a “Flight Deﬁnition” project called
ATHINA for NASA Glenn Research Center Microgravity Combustion Branch and MSU.
Work is in progress to incorporate additional diagnostics in the Combustion Tunnel Rig
(CTR) so more knowledge of the ﬂame instability nature will be gained. In particular,
the igniter system that inspired the MSU Flame Rig design has been added to the CTR.
A gas phase energy balance will supplement the solid fuel balance to determine the
energy processes controlling ﬂamelet formation. Particle Image Velocimetry (PIV)
measurements are being tested in normal gravity, leading to the development of a system
on the CTR. The PIV will help to determine the multidimensional ﬂow ﬁelds of the
ﬂame. Heat ﬂux sensors are also being tested to measure ﬂame heat ﬂuxes. Fuels
similar to the ﬁlter paper will be tested. Drop tower testing will help to determine the
conditions and necessary equipment for tests on the International Space Station (188).
Airplane (KC135) experiments that provide approximately twenty seconds of
microgravity will help determine selection of thicker fuels that can be tested on the 188.

For the MSU Flame Rig, a deﬁnite range of instability conditions for four
diﬂemnt thin fuels will be determined. Thermocouple measurements will give fuel
surface and substrate temperatures for energy balance calculations. A hotwire placed
near the test section opening will be calibrated to the ﬂowmeter, so unsteady ﬂow rates
can be measured. The hotwire will also indicate the ﬂuctuation intensities in the test
section. Due to low Reynolds numbers, it is unlikely that any genuine turbulence exists
there. More sophisticated ﬂowmeters will be incorporated in the apparatus, as well as

pressure transducers. Microphones or other devices capable of measuring the pulse

155

 

 

ﬂequencies and ﬂamelet oscillations will be tested at different locations in the test
section. Since the ranges of velocities where instabilities occur has been obtained, the
next step will be to set the velocity at one value (as in the NASA tests) and try to produce
a ﬂame ﬂont that corrugates into unstable ﬁngering patterns. Then the NASA tests and
MSU tests can be more closely compared. Most importantly, ﬁrels thicker than paper
will be tested and instability ranges recorded. Computer simulations using CFD software

will be attempted to model a simpliﬁed system with comparisons to experimentation.

156

 

fl»—

5‘5.»

2!. .k {'Iﬁim 'L-Y‘h

 

 

I Ii

‘I'..

BIBLIOGRAPHY

 

157

 

10.

11.

12.

BIBLIOGRAPHY

Burke, S. P., and Schumann, T. E. W., 1928, “Diffusion Flames,” Industrial &
Engineering Chemistry, Vol. 20(10), pp. 998-1004.

deRis, J. N., 1969, “Spread of a Laminar Diffusion Flame,” Twelfth Symposium
(International) on Combustion, The Combustion Institute, pp. 241-252.

Friedman, R., 1993, “Fire Safety Practices in the Shuttle and the Space Station
Freedom,” Proceedings of the Second International Microgravity Combustion
Workshop, N93-20204, Cleveland, OH, Sept. 15-17, pp.213-225.

Dietrich, D. L., Ross, H. D., Shu, Y., Chang, P., and T’ien, J. S., 2000, “Candle
Flames in Non-Buoyant Atmospheres,” Combust. Sci. and Tech.,Vol. 156, pp. 1-24.

Olson, S. L., 1997, “Buoyant Low Stretch Stagnation Point Diffusion Flames over a
Solid Fuel,” Ph.D. Thesis, Case Western Reserve University.

Olson, S. L., Baum, H. R., and Kashiwagi, T., 1998, “Finger-Like Smoldering over
Thin Cellulo sic Sheets in Microgravity, ”Twenty-Seventh Symposium (International)
on Combustion, The Combustion Institute, pp. 2525-2533.

deRis, J. N., 1968, “The Spread of a Diffusion Flame over a Combustible Surface,”
Ph.D. Thesis, Harvard University.

McAlevy, R. F., and Magee, R. S., 1969, “The Mechanisms of Flame Spreading over
the Surface of Igniting Condensed-Phase Materials,” Twelfth Symposium
(International) on Combustion, The Combustion Institute, pp. 215-227.

McAlevy, R. F., Magee, R S., and Wrubel, J. A., 1967, “Flame Spreading at
Elevated Pressures over the Surface of Igniting Solid Propellants in Oxygen/Inert
Environments,” Meeting of the Western States Section of the Combustion Institute.

Lastrina, F. A., Magee, R. S., and McAlevy, R. F., 1971, “Flame Spread over Fuel
Beds: Solid-Phase—Energy Considerations,” Thirteenth Symposium (International)
on Combustion, The Combustion Institute, pp.935-948.

Magee, R S., and McAlevy, R. F., 1971, “The Mechanism of Flame Spread,” J. Fire
Flammability, Vol. 2, pp. 271-297.

Friedman, R., 1968, “A Survey of Knowledge About Idealized Fire Spread over
Surfaces,” Fire Res. Abstr. Rev., Vol. 10, pp. 1-8.

158

 

 

 

l3.

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

25.

Sirignano, W. A., 1972, “A Critical Discussion of Theories of Flame Spread Across
Solid and Liquid Fuels,” Combust. Sci. T echnol., Vol. 6, pp. 95-105.

Sirignano, W. A., 1974, “Theory of Flame Spread Above Solids,” Acta Astronaut,
Vol. 1, pp. 1285-1299.

Feng, C. C., and Sirignano, W. A., 1977, “Further Calculations Based Upon a
Theory of Flame Spread Across Solid Fuels,” Combust. Flame, Vol. 29, pp. 247-
263.

Fernandez-Pello, A. C., and Williams, F. A., 1975, “Laminar Flame Spread over
PMMA Surfaces,” Fifteenth Symposium (International) on Combustion, The
Combustion Institute, pp. 217-231.

Fernandez-Pello, A. C., and Williams, F. A., 1977, “A Theory of Laminar Flame
Spread over Flat Surfaces of Solid Combustibles,” Combust. Flame, Vol. 28, pp.
251-277.

Parker, W., 1972, “Flame Spread Model for Cellulosic Materials,” J. Fire
Flammability, Vol. 3, pp.259-269.

Hirano, T., and Kinoshita, M., 1975, “Gas Velocity and Temperature Proﬁles of a
Diffusion Flame Stabilized in the Stream over a Liquid Fuel,” Fifteenth symposium
(International) on Combustion, The Combustion Institute, pp. 379-387.

Hirano, T., and Kanno, Y., 1973, “Aerodynamic and Thermal Structures of the
Laminar Boundary Layer over a Flat Plate with a Diffusion Flame,” Fourteenth
Symposium (International) on Combustion, The Combustion Institute, pp. 391-398.

Frey, A. E., and T’ien, J. S., 1979, “A Theory of Flame Spread over a Solid Fuel
Including Finite-Rate Chemical Kinetics,” Combust. Flame, Vol. 36, pp. 263-289.

Altenkirch, R. A., Eichhom, R., and Shang, P. C., 1980, “Buoyancy Effects on
Flames Spreading Down Thermally Thin Fuels,” Combust. Flame, Vol. 37, pp. 71-
83.

Wichman, I. S., 1992, “Theory of Opposed-F low Flame Spread,” Prog. Energy
Combust. Sci, Vol. 18, pp. 553-593.

Olson, S. L., Ferkul, P. V., T’ien, J. S., 1988, “Near Limit Flame Spread over a Thin
Solid Fuel in Microgravity,” T wenty-Second Symposium (International) on

Combustion, The Combustion Institute, pp. 1213-1222.
Bhattacharjee, S., Altenkirch, R. A., Srikantaiah, N., and Vedha-Nayagam, M., 1990,

“A Theoretical Description of Flame Spreading over Solid Combustibles in a
Quiescent Environment at Zero Gravity,” Combust. Sci. and Tech, Vol. 69, pp. 1-15.

159

 

 

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

Olson, S. L., 1991, “Mechanisms of Microgravity Flame Spread over a Thin Solid
Fuel: Oxygen and Opposed Flow Effects,” Combust. Sci. and Tech, Vol. 76, pp.
233-249.

Bhattacharjee, S., Altenkirch, R A., Olson, S. L., Sotos, R G., 1991, “Heat Transfer
to a Thin Solid Combustible in Flame Spreading at Microgravity,” Transactions of
the ASAE (Journal of Heat Transfer), Vol. 113, pp. 670-676.

Altenkirch, R. A., Bhattacharjee, 8., Olson, S. L., Sacksteder, K., 1993, “Opposed-
Flow Flame Spreading in Reduced Gravity,” Proceedings of the Second
International Microgravity Combustion Workshop, N93 -20206, Cleveland, OH,
Sept. 15-17, pp.237-243.

Bhattacharjee, S., Altenkirch, R. A., Sacksteder, K., 1996, “The Effect of Ambient
Pressure on Flame Spread over Thin Cellulosic Fuel in a Quiescent, Microgravity
Environment,” Transactions of the ASM (J. Heat Transfer), Vol. 118, pp. 181-190.

Blackshear, P. L., and Murty, K. A., 1967, “Some Effects of Size, Orientation, and
Fuel Molecular Weight on the Burning of Fuel Soaked Wicks,” Eleventh Symposium
(International) on Combustion, The Combustion Institute, pp. 545-552.

Orloff, L., and deRis, J., 1971, “Modeling of Ceiling Fires,” Thirteenth Symposium
(International) on Combustion, the Combustion Institute, pp. 979-992.

Chan, W. Y., and T’ien, J. S., 1978, “An Experiment on Spontaneous Flame
Oscillation Prior to Extinction,” Combust. Sci. and Tech, Vol. 18, pp. 139-143.

Kirkby, L. L., and Schmitz, R. A., 1966, “An Analytical Study of the Stability of a
Laminar Diffusion Flame,” Combust. Flame, Vol. 10, pg. 205.

Baliga, B. R., and T’ien, J. S., 1975, “Unsteady Effects on the Low Pressure
Extinction Limit of Solid Propellants,”AIAA J., Vol. 13(12), pp. 1653-1656.

Chao, B. H., Law, C. K., and T’ien, J. S., 1990, “Structure and Extinction of
Diffusion Flames with Flame Radiation,” Twenty- Third Symposium (International)
on Combustion, The Combustion Institute, pp. 523-531.

Chen, R. H., Bradley Mitchell, G., Ronney, P. D., 1992, “Diffusive-Thermal
Instability and Flame Extinction in Nonpremixed Combustion,” Twenty-Fourth
Symposium (International) on Combustion, The Combustion Institute, pp. 213-221.

Linan, A., 1974, “The Asymptotic Structure of Counterﬂow Diffusion Flames for
Large Activation Energies,” Acta Astronaut, Vol. 1, pp. 1007-1039.

160

 

 

38

39.

40.

41.

42.

43.

45.

46.

47.

48.

49.

50.

51.

Kim, J. S., 1996, “Diffusional-Thermal Instability of Diffusion Flames in the
Premixed-Flame Regime,” Combust. Sci. and Tech, Vol. 118, pg. 27-48.

Kim, J. S., 1997, “Linear Analysis of Diffusional-Thermal Instability in Diffusion
Flames with Lewis Numbers Close to Unity,” Combust. Theory & Modeling, Vol. 1,
pp. 13-40.

Cheatham, S., and Matalon, M., 1996, “Near-Limit Oscillations of Spherical
Diffusion Flames,” AIAA J., Vol. 34(7), pp.1403-1409.

Matalon, M., 1999, “Instabilities in Non-Premixed Combustion,” Thirty-Seventh
AIAA Aerospace Sciences Meeting and Exhibit, AIAA 99-0584, Reno, NV, Jan. 11-
14.

Potter, M. C., and Foss, J. F., 1982, Fluid Mechanics Great Lakes Press, Inc.,
Okemos, MI, pp. 392-393.

 

Bejan, A, 1995, Convection Heat Transfer, Second Edition, Wiley-Interscience,
New York, pp. 252-256, pp.48-49, pp.159-161.

Zik, O., Olami, Z., Moses, E., 1998, “Fingering Instability in Combustion,” Phys.
Rev. Lett., Vol. 81, pg. 3868-3871.

Ivanov, A. V., 1997, “A Study of Material’s Combustion Processes in
Microgravity,” Russian Space Agency Keldysh Research Center, Technical Report
No. 2625, on treaty No. 920/9-5208/95:400-I/34-95 project TM-9.

Olson, S. L., 1997, Combustion Tunnel User’s Guide, NASA Glenn Research Center.

Baines, W. D., and Peterson, E. G., 1951, “An Investigation of Flow Through
Screens,” Transactions of the ASME, Vol. 73, pp. 467-480.

Prandtl, L., 1932, “Herstellung einwandﬁeier Luﬁstrome (Windkanale),” Handbuch
der Experimentalphysik, Vol. 4, part. 2, pg. 73, (NA CA Technical Memorandum No.
726).

McMasters, R. L. IV, 1997, “Modeling Heat Transfer for Parameter Estimation in
Flash Diffusivity Experiments,” Ph.D. Thesis, Michigan State University.

Murio, D., 1993, The Molliﬁcation Method and the Numerical Solutions to 111 Posed
Problems Wiley-Interscience, New York.

 

Omega Engineering, Inc., 1995, The Flow and Level Handboog Vol. 29, Omega
Engineering, Inc., Stamford, CT.

 

161

52.

53.

54.

55.

Miller, R. W., 1989, Flow Measurement Engineering Handbook, Second Edition,
McGraw Hill, Inc., New York, pp.(14-46)-(14-58).

The ASME Special Research Committee on Fluid Meters, 1971, Fluid Meters Thjei_r
Theory and ApplicaLion. Report of the American Society of Mechanical Engineers
Research Committee, Sixth Edition, editor Howard Bean, The American Society of
Mechanical Engineers, New York, pp. 81-89.

Arpaci, V. S., and Larsen, P. S., 1984, Convection Heat Transfer, Prentice-Hall, Inc.,
New Jersey, pp. 146-150.

Wichman, I. S., and Olson, S. L., 1998, “Investigation of Diffusion Flame Tip
Thermodiffusive and Hydrodynamic Instability Under Microgravity Conditions,”
Research Proposal submitted in response to NASA Research Announcement NRA-
97-HEDS-01, Microgravity Combustion Science: Research and Flight Experiment
Opportunities. Accepted and funded Jan. 1999 under contract number NCC3-662,
Project Monitor Dr. Fletcher Miller.

162

 

 

muwujﬂlﬂtjﬁlﬁlﬂfw1|