.al .I... 32...:
W23. 3-H}! ‘us. fizz“. k1
‘ t. . a “In? 3| V
3 8..
t9?‘€i.o¢5§. r
ire-51:12 V.
. «I'lfilltn :35!
.3. £3.14: ha. a.”
. in???

. £~
caiillnkllz.iv ‘
fiuL‘ .493. dud“?
P... \'
..§)I.!II!I II

.i I. .\. .3... I:
Stagnaaiutsauitel
\ssznttfip..asilu.£.11§|lo

1.1.1.833

 

{‘13.}.i33i
11.... i 351.112.
. it. tiiir .1

 

it"s

 

 

main
’1

,7-le C)

 

LlF?Q A RY
Michigan State
University .

,—

 

 

 

This is to certify that the
dissertation entitled

COMBUSTION AND CHEMICAL KINETICS STUDY OF JET
FUEL, BIOGAS AND SOLID FUEL IN DIFFUSION FLAMES

Doctoral

presented by

Saeid Jahangirian

has been accepted towards fulfillment
of the requirements for the

degree in Mechanical Engineering

 

 

 

Major Projéssor’s Signature

{0 lfléloq

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

 

 

 

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 DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K'IProj/Acc&Pres/CIRCIDaIeDue.indd

 

COMBUSTION AND CHEMICAL KINETICS STUDY OF JET
FUEL, BIOGAS AND SOLID FUEL IN DIFFUSION FLAMES

By

Saeid Jahangirian

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Mechanical Engineering

2009

ABSTRACT

COMBUSTION AND CHEMICAL KINETICS STUDY OF JET FUEL, BIOGAS AND
SOLID FUEL IN DIFFUSION FLAMES

By

Saeid J ahangirian

This study is organized into three parts focusing on jet fuel combustion, biogas

combustion, and solid fuel microgravity combustion.

In Part I, combustion and chemical kinetics of high molecular weight hydrocarbon
blends at the onset of soot formation are studied. The chemical structure of ethylene
counterflow diffusion flames doped with trace amounts of jet fuel or two promising jet
fuel surrogates is analyzed by gas sampling via quartz microprobes and GC/MS. A
dataset for the pyrolysis, oxidation and sooting behavior of jet fuel in diffusion flames is
provided. The critical fuel decomposition products and soot precursors, such as acetylene,
benzene and toluene, are compared to evaluate surrogate formulations. The data for C7-
C12 alkanes are consistent with typical decomposition of large alkanes with both
surrogates (6 and 2-component) showing good qualitative agreement with jet fuel in their
pyrolysis trends. The acetylene profiles present a unique multimodal behavior. Good
agreement between jet fire] and the surrogates is found with respect to critical soot

precursors such as benzene and toluene.

In Part II, combustion and kinetics of biogas, which is a viable alternative gas

turbine fuel, is modeled in premixed and non-premixed configurations. A modeling study

is conducted on blends of CH4 and C02 simulating biogas from digestion plants or

landfills to compare predictions for four non-sooting counterflow diffusion flames and to

examine their thermal and chemical structure. In addition to evaluation of thermal
influences of biogas C02, the chemical influences of C02 are quantified because C02

dilution through chemical effects is shown to reduce soot precursors, emissions of NOx

and greenhouse gases even without flame temperature reduction.

In Part III, diffusion flames spreading near solid fuel surfaces are investigated
numerically and analytically in 2-D domains for both unconfined and confined
environments. A model of surface-attached solid fuel flames with weak convection is
constructed. Flame spread over thin fuels is studied in a confined geometry because of its
implications for fire safety in normal gravity. It is demonstrated that the buoyancy is
suppressed in the MSU Narrow Channel Apparatus, which produces test conditions that

can simulate conditions achieved in actual microgravity.

Copyright by
SAEID JAHANGIRIAN

2009

DEDICATION

CI»)! (It/Cd»

This book is dedicated to

my mother, Houra, my father, Hassan, and my wife, Tayebeh.

ACKNOWLEDGEMENTS

I express my sincere gratitude to my advisors, Dr. Indrek Wichman and Dr.
Abraham Engeda, at Michigan State University for their continuous support, friendship
and encouragement during my graduate studies. I gratefully thank Dr. Alessandro Gomez
at Yale University, where a valuable part of this dissertation was conducted during a

sabbatical leave from Michigan.

I am very grateful for having an exceptional doctoral committee and would like to
sincerely thank Dr. Charles Petty and Dr. Farhad Jaberi, for their continual

encouragement, and their valuable comments and criticisms.

I would like to gratefirlly acknowledge the financial support of AF OSR (Grant
#FA9550-O6-1-0018, Dr. Julian Tishkofi", Program Manager) and the supply of jet fuel
from Dr. James (Tim) Edwards (Wright-Patterson Air Force Base). I also wish to express
gratitude to the Microgravity Combustion Branch at NASA Glenn Research Center for
funding this project under contract #NCC3-662. Financial support of the MSU

Turbomachinery Laboratory for bio gas research is greatly appreciated.

I would like to thank Dr. Sandra Olson and Dr. Fletcher Miller of the NASA
Microgravity Combustion Division for their technical advice and their ongoing
encouragement. Discussions with Dr. Charles McEnally in the Department of Chemical

Engineering at Yale University are also acknowledged.

My journey through these years was much easier with the friendship of many

friends and colleagues especially Mehdi Abarham.

vi

Finally and most importantly, my deepest thanks to my wife, Tayebeh, for all her
support and patience during the years I worked on my PhD. research. It was her
encouragement, love, and passion that gave me strength and confidence to look for the

best, pass all obstacles, and plan for reaching to great horizons.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................... xi
LIST OF FIGURES .................................................................................. xii
NOMENCLATURE ................................................................................ xix
CHAPTER 1 INTRODUCTION ................................................................................... 1
PART 1: EXPERIMENTAL STUDY OF JET FUEL COMBUSTION ...................... 4

CHAPTER 2 BACKGROUND AND INTRODUCTION TO JET FUEL AND JET

FUEL SURROGATE COMBUSTION ............................................................................... 5
2.1. JET FUEL SURROGATES .......................................................................................... 6
2.2. SURROGATES FOR ALTERNATIVE JET FUELS .......................................................... 8
2.3. KINETIC MODELING OF JET FUEL SURROGATES ..................................................... 9
2.4. REDUCTION OF SURROGATE MECHANISMS ................................................... 10
2.5. GAS CHROMATOGRAPHY AND MASS SPECTROMETRY (GC/MS) ........................ 11
2.6. EXPERIMENTAL STUDY OF JET FUEL COMBUSTION .............................................. 12

CHAPTER 3 EXPERIMENTAL PROCEDURE FOR DETAILED CHEMICAL

ANALYSIS OF JET FUEL DOPED FLAMES ................................................................ 15
3.1. THEORY OF COUNTERLFOW FLAMES ................................................................... 15
3.2. EXPERIMENTAL SETUP ........................................................................................ 18

3.2.1 . The electrospray ............................................................................................ 22
3.3. CRITERION FOR FLAME COMPARISON .................................................................. 23

CHAPTER 4 EXPERIMENTAL STUDY OF ETHYLENE COUNTERFLOW
DIFFUSION FLAMES DOPED WITH JET FUEL AND ITS SURROGATES UNDER

INCIPIENTLY SOOTING CONDITIONS ...................................................................... 30
4.1. SIMILARITIES IN TEMPERATURE AND SPECIES PROFILES AMONG FLAMES .......... 30
4.2. C7-C15 ALKANES .............................................................................................. 35
4.3. C7-C11 ALKENES ............................................................................................... 41
4.4. C3-C6 ALKANES, ALKENES AND DIENES ............................................................ 42

4.4.1. C5-C6 Alkenes .............................................................................................. 42
4.4.2. Quantification of dienes ................................................................................. 43
4.4.3. C3 alkanes and alkenes .................................................................................. 44
4.5. C2 HYDROCARBONS ........................................................................................... 46
4.6. AROMATICS ........................................................................................................ 49
4.6.1. Benzene and toluene ...................................................................................... 52
4.6.2. Effects of component chemical class ............................................................. 56

viii

4.6.3. Threshold sooting index (TSI) ....................................................................... 58
4.6.4. Possible role of the antistatic additive ........................................................... 60

PART H: BIOGAS COMBUSTION AND CHEMICAL KINETICS. . . . . . . ...62

CHAPTER 5 BACKGROUND AND INTRODUCTION TO BIOGAS

COMBUSTION RESEARCH ........................................................................................... 63

5.1. BIOGAS PRODUCTION AND BENEFITS ................................................................. 64

5.1.1. Biogas composition and upgrading ............................................................... 64

5.1.2. Biogas benefits .............................................................................................. 69

5.2. GLOBAL UTILIZATION OF BIOGAS ....................................................................... 71

5.3. BIOGAS POTENTIAL As AN ALTERNATIVE GAS TURBINE FUEL ............................. 72
5.3.1. Fuel-flexible gas turbines utilizing biogas ..................................................... 72 .

5.3.2. Biogas-driven micro-gas-turbines (MGT) ..................................................... 73

5.3.3. Premixed turbulent biogas combustion ......................................................... 75

5.4. DISCUSSION: BIOGAS RESEARCH OBJECTIVES ..................................................... 77

CHAPTER 6 CHEMICAL KINETICS STUDY OF SIMULATED BIOGAS IN

PREMD<ED COMBUSTION ........................................................................................... 80
6.1. BIOGAS AS A DILUTED HYDROCARBON BLEND .................................................... 80
6.2. GAS PHASE CHEMICAL KINETICS MECHANISMS ................................................... 82

6.2.1. Standard-state thermodynamic properties in CHEMKIN ............................. 82
6.2.2. Detailed, reduced and skeletal reaction mechanisms .................................... 84
6.3. EVALUATION OF REACTION MECHANISMS ........................................................... 88
6.4. CARBONE DIOXIDE DILUTED METHANE/AIR COUNTERFLOW FLAMES .................. 93
6.5. WATER VAPOR AND NITROGEN INFLUENCES IN BIOGAS COMBUSTION ................ 98
6.6. BIOGAS PREMIXED FLAMES: RESULTS AND DISCUSSION ................................... 101
6.6.1. Equilibrium calculations for biogas/air combustion .................................... 101
6.6.2. Chemical kinetic modeling of biogas ignition ............................................. 103
6.6.3. Laminar flame speed of biogas/air mixtures ............................................... 109

CHAPTER 7 THERMAL AND CHEMICAL STRUCTURE OF BIOGAS

COUNTERF LOW DIFFUSION FLAMES .................................................................... 114
7.1. NUMERICAL APPROACH ................................................................................... 117
7.1.1. Criterion for biogas flame comparison ........................................................ 119

7.2. COMPARISON OF REACTION MECHANISMS FOR FLAME I ................................... 121
7.3. INFLUENCES OF C02 CONTENT OF BIOGAS; COMPARISON OF FLAMES 1-4 ........ 125
7.3.1. Temperature and velocity ............................................................................ 125
7.3.2. Major species, radicals and intermediate species ........................................ 128
7.3.3. Soot precursors ............................................................................................ 133
7.3.4. NOx emissions ............................................................................................. 135
7.3.5. Greenhouse gas emissions ........................................................................... 138

PART III: FLAME SPREAD OVER SOLID FUELS IN MICROGRAVITY. . . . . . . . ....l43

ix

CHAPTER 8 BACKGROUND AND INTRODUCTION TO FLAME SPREAD IN
MICROGRAVITY CONDITIONS ................................................................................. 144

CHAPTER 9 NUMERICAL AND ANALYTICAL STUDY OF DIFFUSION
FLAMES SPREADING NEAR SOLID FUEL SURFACES IN MICROGRAVII Y 150

9.1. MODEL PROBLEM ............................................................................................. 152
9.2. 2-D THEORETICAL MODEL ................................................................................ 153

9.2.1. Analytical solution of VZZ = O .................................................................. 154

9.2.2. Numerical and analytical solution of 8(6Z / 6x) + d (OZ / 6y) = V2 Z .156
9.3. DISCUSSION: 2-D THEORETICAL MODEL ........................................................... 157
9.4. FLAME-SURFACE INTERACTION OVER SINGLE- AND MULTIPLE SOLID FUEL
SEGMENTS IN A CHANNEL CROSS FLOW ....................................................................... 161
9.5. DISCUSSION: SIMULATION FOR ONE FUEL SEGMENT ......................................... 164
9.6. DISCUSSION: SIMULATION FOR MULTIPLE FUEL SEGMENTS .............................. 168

CHAPTER 10 FLAME AND FLAMELET SPREAD OVER THIN FUELS IN

SIMULATED MICROGRAVITY CONDITIONS ......................................................... 174
10.1. THE NARROW CHANNEL APPARATUS (N CA) FOR NORMAL GRAVITY
EXPERIMENTS ................................................. 177
10.2. FLAME AND FLAMELET SPREAD IN THE NCA .................................................... 179
10.3. VELOCITY SCALING ANALYSIS IN NARROW CHANNEL APPARATUS .................. 181

10.3.1. Results and discussion: Velocity scaling analysis in NCA ..................... 184

CHAPTER 1 1 CONCLUSIONS ............................. ' .................................................... 1 92
1 1.1 . CONCLUSIONS: JET FUEL COMBUSTION ............................................................. 192
1 1.2. RECOMMENDATIONS AND FUTURE WORK: JET FUEL COMBUSTION ................... 194
11.3. CONCLUSIONS: BIOGAS COMBUSTION AND KINETICS ........................................ 195
1 1.4. RECOMMENDATIONS AND FUTURE WORK: BIOGAS COMBUSTION AND KINETICS

199
1 1.5. CONCLUSIONS: FLAME SPREAD IN MICROGRAVITY .......................................... 200

11.6. RECOMMENDATIONS AND FUTURE WORK: FLAME SPREAD IN MICROGRAVITY 202

APPENDIX A ................................................................................... 204
APPENDIX B ................................................................................... 205
REFERENCES .................................................................................. 207

LIST OF TABLES

Table 3-1. Boundary conditions for the non-sooting methane flames doped with jet fuel

and Utah/Y ale surrogate fiom the previous study of Bufferand et a1. [36] at Yale ........... 25
Table 3-2. Boundary conditions for the ethylene diflusion flames under incipiently
sooting conditions. The jet fuel was provided by WPAF B (POSF No. 465 8). ................. 27
Table 3-3. Jet fuel surrogate mixtures; values are reported in molar %. ........................... 28

Table 4-1. Concentrations of each chemical class in the jet firel and two surrogates.
Values are in vol%. ............................................................................................................ 57

Table 4-2. TSI, aromatic content, H/C ratio, and cetane number for jet fuel and surrogate
mixtures. ............................................................................................................................ 59

Table 6-1. Selected compositions for simulated biogas used in the computational study.94

Table 6-2. Ignition delay times of biogas mixture 6 fOr different equivalence ratios at l
atrn and 1000 K ................................................................................................................ 106

Table 6-3. Ignition delay times of biogas mixture 6 at various pressures and fixed
equivalence ratio and initial temperature ......................................................................... 107

Table 6-4. Comparison of calculated laminar flame Speed for methane with the data in the

literature. .......................................................................................................................... 109
Table 7-1. Boundary conditions of Flames 1-4 ............................................................... 120
Table 7-2. C02 exhaust and inlet mass fluxes, and EIC02 for Flames 1-4* ................... 140

xi

LIST OF FIGURES

Images in this dissertation are presented in color.

Figure 3-1. Schematic of a counterflow burner with a diffusion flame. .......................... 16
Figure 3-2. Schematic of the experimental setup. ............................................................. 19

Figure 3- 3. Schematic and photograph of the electrospray unit for liquid fuel dispersion.
- _ ..................................... 22

 

 

 

Figure 3-4. Photograph of the flame at the onset of sooting. The thickness of the sooting
zone with orange luminosity underneath the blue zone is about 0.6 mm .......................... 29

Figure 4-1. Temperature profiles. Symbols: Flames A, C and A* with microprobe in.
Line: Flame A in the absence of microprobe. ................................................................... 31

Figure 4—2. Major species profiles: Flame B (full black symbols), Flame C (open
symbols), Flame D (red symbols) and Flame A (blue symbols). . ....... - - 34

 

 

Figure 4-3. Methane mole fraction profiles: Flame B (full black symbols), Flame C (open
symbols), Flame D (red symbols) ...................................................................................... 35

Figure 4-4. Comparison of C7-C15 Alkanes; a) Flame B, b) Flames C (open symbols)
and Flame D (red symbols), Flame A (blue symbols) and Flame A* (+). ........................ 36

Figure 4-5. Utah/Y ale surrogate chromatogram presenting distinct peaks in
correspondence with nitrogen (leftmost peak) and the 6 components of the surrogate:
from left to right, methylcyclohexane, iso-octane, m-xylene, tetralin, dodecane and
tetradecane. ........................................................................................................................ 38

Figure 4—6. Typical jet fuel MS chromatograrn, as measured from a gaseous sample
extracted near the burner mouth. The arrows denote compounds that were identified and
quantified. .......................................................................................................................... 39

Figure 4-7. Partial carbon mole fraction profiles for Flame B (full symbols), Flame C
(Open symbols) and Flame D (red symbols). Contributions from ethylene, C0, C02,
acetylene and ethane are not considered ............................................................................ 40

Figure 4-8. Comparison of l-octene and l-heptene profiles: Flame C (open symbols),
Flame D (red symbols). None of these species was detected in Flame B. ........................ 41

Figure 4-9. Comparison Of C5-C6 alkenes profiles: Flame B (full black symbols), Flame
C (Open symbols), Flame D (red symbols) and Flame A (blue symbols). ........................ 43

xii

Figure 4-10. Comparison of cyclopentadiene profiles: Flame B (full black symbols),
Flame C (open symbols), Flame D (red symbols) and F lame A (blue symbols). ............. 44

Figure 4-11. Comparison of profiles of a) l-propene and b) propane: Flame B (full black
symbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+). .................................................................................................................... 45

Figure 4-12. Profiles of molar fractions of a) ethane and b) acetylene: Flame B (full black
symbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+). .................................................................................................................... 47

Figure 4-13. Comparison of profiles of m-xylene, tetralin, and 1,2,4-trimethylbenzene
with compressed (a) and expanded (b) ordinate scales: Flame B (full black symbols),
Flame C (open symbols) and Flame D (red symbols). ...................................................... 51

Figure 4-14. Global pathways of naphthalene formation. Isopropyl benzenes and species
with similar structure are rare in the fuels under study. ............................... - --53

 

Figure 4-15. Comparison of profiles of benzene mole fractions: Flame B (full black
symbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+). .................................................................................................................... 54

Figure 4-16. Comparison of profiles of toluene mole fractions: Flame B (full black
symbols), Flame C (open symbols) and Flame D (red symbols). ..................................... 55

Figure 5-1. Landfill gas characterization by methane content (%) and technologies gas
operating ranges ................................................................................................................. 67

Figure 5-2. LHV’S of alternative gaseous firels compared to natural gas adopted fiom
[81] ..................................................................................................................................... 68

Figure 6-1. Distribution of the scaled reaction rates for San Diego mechanism reactions at
Tad of stoichiometric methane/air mixture (Tad= 2224 K) ................................................. 87

Figure 6-2. Temperature and major species profiles for (23%CH4-77%N2)/Air flame at 1
atm, a =56 S'l (GRI-Mech 3.0 mechanism). ...................................................................... 89

Figure 6-3. Computed temperature profiles for (23%CH4-77%N2)/Air flame at 1 atm,
with GRI-Mech 3.0 mechanism: a) VF=V0=45 cm/s (line, computed), b) Vp=Vo=25.5
cm/s (line, computed) and c) VF=V0=45 cm/s (solid squares, experiments of Sung et al.
[135]) ................................................................................................................................. 90

Figure 6-4. Comparison of computed profiles of temperature near the peak point for

(23%CH4-77%N2)/Air flame at l atrn (a =56 S']) with GRI-Mech 3.0, reduced GRI-Mech
1.2, and San Diego mechanisms. ....................................................................................... 92

xiii

Figure 6-5. Comparison of computed profiles of H radicals near the peak point for
(23%CH4-77%N2)/Air flame at l atrn (a =56 s") with GRI-Mech 3.0, reduced GRI-Mech
1.2, and San Diego mechanisms. ....................................................................................... 92

Figure 6-6. Temperature and major Species profiles for four compositions of C02 diluted
methane flames (P=1 atm, VF=V0=45 cm/S, GRI-Mech 3.0 mechanism) ........................ 95

Figure 6-7. Profiles of H radical mole fi'actions for four compositions of C02 diluted
methane flames in thermal mixing layer (P=1 atm, VF=V0=45 cm/S, GRI-Mech 3.0
mechanism) ........................................................................................................................ 97

Figure 6-8. Profiles of CH4 mole fraction for four compositions of C02 diluted methane
flames in thermal mixing layer (P=1 atm, VF=V0=45 cm/s, GRI-Mech 3.0 mechanism).
........................................................................................................................................... 97

Figure 6-9. Profiles of scaled C02 for four compositions of C02 diluted methane flames
in a counterflow configuration (P=1 atm, VF=V0=45 cm/s, GRI-Mech 3.0 mechanism).
- ................................................. ......... 98

 

Figure 6-10. Temperature, major species profiles (a) and minor species profiles (b) for
biogas mixture 5 (P=l atm, VF=V0=45 cm/s, GRI-Mech 3.0 mechanism). ..................... 99

Figure 6-11. Comparison of NO and H radical mole fi'actionsz biogas mixture 1 (black
lines); biogas mixture 5 (blue lines). ............................................................................... 100

Figure 6-12. Adiabatic flame temperature of CH4 and a typical biogas (mixture 6) in lean
to rich equivalence ratios. ................................................................................................ 102

Figure 6-13. Equilibrium mole fractions of H, OH, NO and CO for CH4 and a typical
biogas (mixture 6) in lean to rich equivalence ratios ....................................................... 103

Figure 6-14. Ignition delay times of biogas mixture 6 at 1 atrn and <p=1 (The peak of the
OH profile criterion was employed). ............................................................................... 106

Figure 6-15. Ignition delay times of biogas mixture 6 at various equivalence ratios with
P=I atm, and Tinifial=1000 K. .......................................................................................... 107

Figure 6-16. Ignition delay times and peak gas temperatures of biogas mixture 6 at
various pressures, Ip=l, T=1000 K. ................................................................................. 108

Figure 6-17. Effect of C02 concentration on the laminar flame Speed of CH4/C02
mixtures simulating biogas. ............................................................................................. 111

xiv

Figure 6-18. Effect of C02 concentration on the concentration of H radicals and
downstream temperature for CH4/C02 mixtures simulating biogas. .............................. 111

Figure 6-19. Relative effect of N2 and C02 concentration on the laminar flame speeds of
CH4/C02/ N2 mixtures with constant CH4 concentration, [CH4]=0.6. ........................... 112

Figure 7-1. Schematic of the counterflow burner setup for the chemical kinetics study
(L=14.1 mm; SP: stagnation plain). A nitrogen shroud is usually used in experiments. 118

Figure 7-2. Profiles of major species (02, CH4, C02, and CO) for Flame 1. Comparison
between a) solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism; c) dotted
lines: reduced GRI 1.2 mechanism; (1) symbols: experiments of Bufferand et a1. [36]. .122

Figure 7-3. Profiles of C2 species (blue: C2H2, red: C2H4, black: C2H6) for Flame 1.
Comparison between a) solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism;
c) dotted lines: reduced GRI 1.2 mechanism; (1) symbols: experiments of Bufferand et al.
[36] The fuel supply in experiments contained about 120 ppm of ethane. ..................... 123

Figure 7-4. Profiles of H, OH, and CH3 radicals for Flame 1. Comparison between a)
solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism; c) dotted lines: reduced
GRI 1.2 mechanism. ........................................................................................................ 124

Figure 7-5. a) Comparison of temperature and axial velocity profiles for Flames 1-4; b)
Magnified version of temperature profiles near the peaks. The stagnation plane for Flame
1 is at 2 ~ 7.3 mm. All flames are on the fuel side of the SP ........................................... 126

Figure 7-6. Comparison of heat release profiles (signatures) for Flames 1-4 in the reaction
zone .................................................................................................................................. 127

Figure 7-7. Comparison of the mole fractions Of C02, H20 and CO for Flames 14. Solid
lines for CO and C02 profiles in this Figure and Figure 7-2 are the same. .................... 127

Figure 7-8. a) Comparison of the mole fractions of H, O and OH radicals for Flames 1-4;
b) Magnified version of Figure (a) Showing the mole fractions of H radicals. The

percentages Of C02 in the fuel stream have been written. Note the quantitative definitions
of “chemical effect” and “thermal effect”, which are employed (without explicit
demarcation) in all subsequent figures. ........................................................................... 131

Figure 7-9. Comparison of the mole fractions of a) CH3 radicals and CHZO; b) CH
radicals for Flames 1-4. The percentages of C02 in the fuel stream have been written. 132

Figure 7-10. Comparison of the mole fractions of acetylene for Flames 1-4. The decay of
CH4 profiles is also demonstrated in this Figure. ............................................................ 134

XV

Figure 7-11. Comparison of the mole fractions of N radicals and NO (in ppm) for Flames
1-4. The percentages of C02 in the fuel stream have been written. ............................... 136

Figure 7-12. Comparison of the mole fractions of N02 and N20 for Flames 1-4. The
percentages of C02 in the fuel stream have been written ................................................ 137

Figure 7-13. Tradeoff between scaled flame temperatures, scaled NO, and scaled mass
flow rates for Flames 1-4. ................................................................................................ 142

Figure 9-1. Boundary conditions for a single fuel segment to solve mixture fraction

equationVZZ = 0 . The “fuel inflow segment” is shown in red. The rightmost boundary
is for downstream (x-mo) and the topmostboundary is for transverse (y—mo). ............. 155

Figure 9-2. Comparison of numerical and exact analytical solutions of the equation

V2Z = 0 for Zf = 0.1-0.9. (Solid line: numerical solution, Dashed line: exact analytical
solution). .......................................................................................................................... 158

Figure 9-3. Comparison of approximate analytical solution and numerical solution of

e(az /6x) + d(az/ay) = V22 for e = 0.1 and d = 0. (Solid line: numerical solution;
Dashed line: lowest order analytical solution given by Eq. (9-9)). ........ 159

 

Figure 9-4. Contours of constant Zf from the numerical solution of

e(0Z/6x)+ d(0Z/5y) = V2Z for e = 0.05, d = 0.05. Red dashed lines represent the
approximate analytical solution for Zf': 0.5. ................................................................... 160

Figure 9-5. Contours of constant Zr from the numerical solution of

e(6Z/ 6x) + d (BZ/ 6y) = V2 Z for e = 0.8, d = 0.2. Red dashed lines represent the
approximate analytical solution for Zf= 0.5 .................................................................... 161
Figure 9-6. Schematic of the upstream section (12 cm x 10 cm) of computational domain
with three 1 cm firel inlet Slots and 240 x 150 grids. Grid compression is applied in the y
direction. .......................................................................................................................... 162

Figure 9-7. Contours of a) temperature and b) oxygen mass fraction in fuel rich
conditions (¢= 1.7). ........................................................................................................ 165

Figure 9-8 Contours Of a) temperature and b) oxygen mass fraction in stoichiometric
conditions (¢= 1). ........................................................................................................... 166

Figure 9-9 Contours of a) temperature and b) oxygen mass fraction in fuel lean conditions
(q) = 0.5) ........................................................................................................................... 167

xvi

Figure 9-10. Y0 and Y F contours near the leading edge (dotted line: oxidizer; solid line:
firel) .................................................................................................................................. 169

Figure 9-11. Temperature contours upstream of the flame for the case with two fuel inlet
segments (p = I). ............................................................................................................. 170

Figure 9-12. Temperature contours upstream of the flame for the case with three fuel
inlet segments (p = I) ...................................................................................................... 171

Figure 9-13 Reaction rate contours for the case with three fuel inlet segments (p = I).
......................................................................................................................................... 172

Figure 9-14. Product dispersion plot Showing broadening of the C02 mass fraction (o =
1). Three fuel inlet segments are employed (Figure is not to scale) ................................ 172

Figure 10-1. A single flamelet (~6 mm Wide) propagates steadily in an air flow of 0.5
cm/s in the NASA NCA. When the air flow is abruptly increased to 50 cm/S (100x), the
flamelet grows 200% in less than 10 seconds. (Each photograph is 2 see after the
preceding one) [155] ........................................................................................................ 176

Figure 10-2. Schematic of the Narrow Channel Apparatus (NCA). a) Flow enters the left
side through holes in the tube, and is distributed in the plenum section, straightened
through screens and honeycomb, and then enters the test section. The flow exhausts from
the right plenum. b) Side view of test section, showing the thin fuel suspended in the
center of the narrow channel. .......................................................................................... 178

Figure 10-3. Flamelet tracks on the filter paper in the normal gravity NCA, illuminated
with green LEDS. The uniform flame consumes the entire sample (black solid area at the
left Side of the image) and then breaks up into ~10 flamelets as the flow is reduced. The
unburned paper between flarnelets appears green, while the flamelet tracks are black. The
flamelets are the luminous tips at the ends of the branching pattern [155]. The falmelet
burned fraction is approximately 0.6. .............................................................................. 180

Figure 10-4. Definition of the various velocities that occur in the NCA. The sample is
placed between top (quartz) and bottom walls. Inflow is left to right, while the flamelet
moves right to left. In this figure, gravity vector points downward; buoyant velocity is
upward. ............................................................................................................................ 181

Figure 10-5. Schematic of geometry of numerical simulation used for velocity scale
analysis. The channel height used was h=5 mm, and the hot spot was also 5 mm centered
at x = O (the vertical (y) axis is located at x = 0). Flow velocity (5, 10, 20 cm/s) and
gravity level (0g, 1 g) were varied. ................................................................................... 182

Figure 10-6. Numerical results for vertical velocity at x=0 for isothermal and adiabatic
top wall boundary conditions under two inflow (forced flow) velocities: Vfl0w= 5 cm/s
and Vflow= 20 cm/s. .......................................................................................................... 183

xvii

Figure 10-7. Numerical predictions of the vertical velocity component along the line y =
(1/3)h for three values of the inflow velocity at both normal and zero gravity. .............. 185

Figure 10-8. Maximum velocity for three values of the inflow velocity at y = (I/3)h and y
= (2/3)h for zero gravity conditions. ............................................................................... 186

Figure 10-9. Numerical results of vertical velocity component for the six cases along a
vertical slice centered above the hot spot at x = 0. .......................................................... 187

Figure 10-10. Numerical results of the buoyant velocity, calculated as the quantity

[Vy(1g) - Vy(0g)] along horizontal slice y=(I/3)h. Inset shows the trend in maximum
buoyant flow as a function of forced flow. ...................................................................... 188

Figure A-l. Photograph of the experimental setup at Yale University. Right: the
counterflow burner setup; center: the GC/MS setup; left: the data acquisition system. .204

Figure A-2. The counterflow burner setup at Yale University. ....................................... 204

Figure B-l . Flammability map with opposed flow velocity and heat loss as axes, showing
good agreement between Zero Gravity Research Facility tests and normal gravity tests in
the NCA, both in the ‘u’=upward facing orientation and ‘d’=downward facing orientation
[180, 181]. The flamelet regime is a near-limit zone. 0g data with 0.02 mrn'1 Spacing is
from [185]. ....................................................................................................................... 206

xviii

Abbreviations

NG

RNG

LNG

BG

GT

GHG

PIV

LDV

LIF

DLN

GC

MS

PAH

TSI

CN

NOMENCLATURE

Natural gas

Renewable natural gas
Liquefied natural gas
Biogas

Gas turbine

Green house gas

Particle image velocimetry
Laser doppler velocimetry
Laser induced florescence
Dry low NOX

Gas chromatography

Mass spectroscopy
Reaction rate

Polycyclic aromatic hydrocarbons
Threshold sooting index
Cetane number
Fischer-Tropsch

Carbon dioxide emission index

Anaerobic digestion

xix

LHV
EGR
SD
SP
NCA
ZGRF

MCH

Symbols

CP

Lower heating value

Exhaust gas recirculation

San Diego detailed reaction mechanism
Stagnation plane

Narrow Channel Apparatus

Zero Gravity Research Facility at NASA
Methylcyclohexane

Trimethyl benzene

Global Equivalence ratio, q)=(Air/Fuel)s,o,-c;/(Air/FueDacmal
Constant pressure specific heat, kJ/kg-K
Enthalpy, kJ/kg

Entropy, kJ/kg-K

Laminar flame speed, crn/S

Swirl number

Turbulent flame speed, cm/S

Flame thickness, mm

Strain rate, l/s

Dokahler number

Turbulent Karlovitz number

Thermal diffusivity, m2/s

Flow velocity, cm/s

XX

Vblow
Vbuoy
Vdiff
Vflow
Vtherm

v, Vy

I)

Mixture density, kg/m3

Mole fraction

Mass fraction

Gas constant, kJ/kmol-K or radius, m
Pressure, atm

Temperature, K

Stoichiometric mixture fraction
Burner separation, mm

Radial coordinate, mm

Axial distance in counterflowgburner, mm
Gravity level, 1 g=9.8 m/SZ, 0g = 0 m/S2
Gap spacing, cm

Diffusion coefficient, m2/s

Viscosity, 1112/8

Fuel surface blowing velocity, cm/s
Buoyancy-induced velocity, cm/s
Diffusive velocity, cm/s

Forced flow velocity, cm/s

Thermal expansion velocity, cm/s
Velocity in the y direction, cm/S

Heat flux, W/cmz, 0 = adiabatic

pi, 3.14159.....

xxi

u
Superscripts

0

Subscripts
k, i
F
0
FF

00

ad

is

min

mix

Velocity in the X direction, cm/s

Standard-state

Species (components) in a mixture
Fuel

Oxidizer

Fuel fiaction in fuel inlet (surface)
Oxidizer fraction in oxidizer inlet
Flame

Adiabatic

Ignition

Far field value

Cross flow

Jet flow

Maximum value

Minimum value

Surface, solid

Mixture

xxii

Chapter 1 Introduction

Combustion research can be broadly classified into two categories: simple fluid
mechanics, but complex kinetics, including real fuels such as blends of large
hydrocarbons; or, the reverse, complex fluid mechanics, but simple chemical kinetics.
Fundamental studies on both categories are necessary before starting to face the greater

challenge of complex fluid mechanics and complex chemical kinetics [1].

This thesis is organized into three parts (Parts I—III) focusing, respectively, on jet
fuel combustion, biogas combustion, and solid fuel combustion in microgravity
conditions. It focuses mostly on diffusion flames and ranges fiom studies on complex
chemistry with simple fluid mechanics (as in Part I) to problems involving complex
patterns of heat and flow transfer modeled with a simple underlying chemistry (as in Part
III). In Parts I to 111, a broad range of fuels are studied, which are, respectively, in liquid,

gaseous, or solid state.

In Part I, combustion of jet fuel and jet fuel surrogates, which are blends of large
molecular weight fuels with complex kinetics, is experimentally studied. The Objective is
to study the structure of jet fuel doped diffusion flames at the onset of soot formation. In
Part II, a combustion and chemical kinetic modeling is performed for biogas, which is
produced from anaerobic digestion of biodegradable materials and is presently a viable
alternative fuel (e. g., for gas turbines). Finally, in Part HI, flame—surface interactions and

combustion Of solid fuels in simulated microgravity conditions are numerically and

analytically investigated. Thin biomass samples (e.g., thin cellulosic fuels) are the

principal solid fuels used.

The dissertation is organized SO that this short introduction is followed
consecutively by Parts I to 111. Each part starts with a chapter in which a review of
previous work, the motivation of research, and an introduction to the methodology of
research conduct are all thoroughly presented. The second and third chapters of each part
define the metrics of the problem after which results are presented and discussed.
Conclusions in Chapter 11 include all concluding remarks as well as recommendations
for future work pertaining to Parts I to 111. Chapter 11 is followed by appendices.

References for all parts are listed in 0. .

Research in Part I, was conducted at Yale Center for Combustion Studies under the
Air Force Office of Scientific Research (AFOSR) jet fuel surrogate program. A
background of the research on jet fuel and its surrogates is presented in Chapter 2. In
Chapter 3, the experimental procedure followed for the detailed chemical analysis of
simple or complex fuels is explained. In Chapter 4, a well-defined baseline ethylene
diffusion flame under incipiently sooting conditions is perturbed with the addition of
either jet fuel or two jet fuel surrogates. Attempts are made to validate the surrogate
formulations with respect to the flame structure and to provide an experimental database
for the pyrolysis and oxidation behavior of jet fuel at the onset Of sooting in diffusion

flames.

Research in Part II on biogas combustion is motivated by the worldwide interest in
alternative fuels, especially for stationary power generation. Particularly, it is of interest

to the US. Departments of Energy and Agriculture as well as many State Governments.

In Chapter 5, the latest progress and challenges in biogas research are discussed. Also,
biogas benefits, production methods and compositions are explained. In Chapter 6, a
chemical kinetics study is performed for simulated biogas blends in premixed flames.
First, a laminar counterflow diffusion flame configuration is chosen to evaluate some
detailed and reduced mechanisms in regard to their applicability for the biogas chemical
kinetics study. Gas phase equilibrium calculations are then performed, followed by
ignition delay time and laminar flame speed calculations. In Chapter 7, thermal and

chemical structure of biogas counterflow diffusion flames is examined. A detailed kinetic

modeling of C02 diluted methane blends, which are typical biogas blends, is conducted.

In addition, thermal and chemical influences of C02 content of biogas on flame

temperature and emissions of NOx, soot and green house gases are evaluated.

Microgravity flame spread over solid fuels, investigated in Part III, is conducted in
collaboration with NASA Glenn Research Center under the NASA microgravity
program. Chapter 8 explains the motivation to investigate diffusion flames near solid fuel
surfaces in either an “Open” or a “confined” enviromnent. In Chapter 9, an analytical and
numerical model of surface-attached solid fuel flames with weak convection is
constructed. The ultimate purpose of this study is to establish a model to examine flame
structure, heat transfer, and reaction rate in flames attached to solid fuel surfaces,
particularly surfaces in which multiple flames (or flamelets) may exist. In Chapter 10,
flame and flamelet spread over thin solid fuels in simulated microgravity conditions is
studied. Numerical analyses are compared with two sets of experiments involving flame
spread in a Narrow Channel Apparatus (NCA, available at MSU and NASA) in normal

gravity, and the others taking place in actual microgravity in the NASA drop facilities.

Part I

Experimental Study of Jet Fuel
Combustion

Chapter 2 Background and Introduction to Jet Fuel
and Jet Fuel Surrogate Combustion

Transportation fuels, including jet firels, constitute a Significant share of the world’s
energy consumption. Common jet fuels include Jet A-l, Jet A, and JP-8. They comprise
hundreds of aromatic compounds and aliphatic components, such as straight chain
paraffins, branched chain paraffins, cycloparaffins, and alkenes [2]. The increasing need

for energy security as well as imposing stringent emission regulations for NOx, CO,
particulates, and green house gases such as C02 necessitates: 1) Improved chemical

modeling of conventional aviation fuels such as JP-8; 2) Development and
commercialization of alternative transportation fuels like Fischer-Tropsch (F-T) jet firels.
Depending on the source of the parent crude and the refinery process, the jet fuel
composition may vary significantly. The future fuel supply will become more and more
diversified and burning a broad range of fuels as well as reduction of pollutant (e. g. soot)
formation will pose new challenges to the implementation of their combustion. This trend
will necessitate fundamental studies in well-defined and well-controlled environments to
establish, among other aspects, the chemical kinetic behavior of these complex fuel

blends.

Characterization and simulation of jet fuel chemical kinetics and transport is only
practical by identifying surrogate mixtures having a relatively small number of
components. The surrogate physical and chemical properties should capture essential
features of real fuels in prototypical combustion conditions. The chemical community

selects and characterizes surrogate fuel formulations. Normal paraffins, iso-paraffins,

cyclO-paraffins, Olefinic species, single ring aromatics, and multi-ring aromatics are
typical constituents for a surrogate mixture. Colket et al. [3] proposed a road map for
future development of surrogate fuels, which resulted from discussions at a number of
meetings of a surrogate fuel working group. Developed surrogate formulations would
need experimental validation with respect to the real jet fuels. Surrogate mixtures have
been defined and tested in many experimental conditions and configurations, including
flow/stirred reactors, shock tubes, premixed flames, pool fires, and counterflow diffusion
flames. A comprehensive review was presented by Dagaut et al. [4]. Parallel attempts are
made to compile semi-detailed kinetic mechanisms for a suggested surrogate. Each
kinetic modeling should also be validated with experimental data of both the surrogate
and the real jet fuel. Once the surrogate formulation and the semi-detailed kinetic
mechanism are validated, the next step will be a systematic reduction of the surrogate
kinetic mechanism. Reduced mechanisms can potentially be utilized in CFD codes with a

Significantly lower computational cost.

2.1. Jet fuel surrogates

In the late 1980’s, Wood et a1. [5] formulated a 14-component J P-4 surrogate based
on its compound class composition and distillation curves. Subsequent efforts by Schulz
et al. [6] led to a 12-component jet fuel surrogate. Feasibility, simplicity, fuel class
similarity, and cost are essential criteria that guided subsequent work aimed at decreasing
the number of components to produce more manageable formulations. Violi et al. [7]
reported a six-component Utah surrogate designed to match the volatility of jet fuel and
its overall sooting behavior based on smoke point tests. Five-component surrogates e. g.

the Drexel surrogate [8] or the UtahNioli Surrogate #3 [7], four-component surrogates

e.g. the surrogate in [9], and three-component surrogates e.g. the UC San Diego/Milan
surrogate [10] have also been reported. More recently, the number of components has
further decreased with the minimization effort culminating in the two-component Aachen
surrogate [1 l]. Single-component surrogates, though initially contemplated, are now
generally thought to lack the necessary flexibility to match the jet firel performance in a
sufficiently broad parameter space. In recent work in pressurized flow reactors, Natelson
et al. [12] experimentally studied jet fuel and a three-component jet fuel surrogate
suggested by the surrogate fuels working group at pressures as high as 0.8 MPa. The
three-component surrogate showed higher reactivity than jet fuel suggesting that an
improvement is possible by adding iso-paraffins. Experiments by Holley et al. [13] in a
counterflow non-premixed configuration found that the six—component Utah surrogate [7]
increased ignition propensity and resistance to extinction compared to jet fuel. This
behavior was attributed to mismatched transport properties. Vasu et al. [14] measured
ignition delay times of jet fuels in a shock tube and compared them with predictions of
some current kinetic mechanisms [15, 16]. The five-component Utah surrogate (V ioli
Surrogate #3 [7]), when used with the Milan mechanism [15], revealed the closest
agreement in ignition delay times, especially in capturing the high-temperature trend.
This work has continued on individual components of the surrogate mixture, n-dodecane
(n-C12H26) and methylcyclohexane (MCH), with shock-tube experiments on the former
[17] and with measurerrients of OH time-histories of oxidation behind reflected Shocks

for the latter [1 8].

2.2. Surrogates for alternative jet fuels

Alternative jet fuels can be derived from shale, coal, biomass, or natural gas.
Fischer-Tropsch (F-T) jet fuels are derived fi'om synthesis gas (CO+H2) made fi'om

natural gas using the Fischer-Tropsch process [19]. These firels can significantly reduce
particulate matter (PM) emissions. The complex F-T jet fuels are drawing more and more
interest worldwide and in the US. The US. Department of Defense and the Department
of Energy are jointly working through programs such as the Clean Fuels Initiative to
develop F-T jet fuels and to assess their national security benefits and weigh them against

cost and availability concerns [20].

JP-8 mixed with synthetic jet fuels (synjet) makes an aviation fuel blend e.g. F-T
Jet A-1 (S-8). Since the F-T jet fuels consist mainly of paraffins [21], which in case of S-
8 are alkanes with one or two attached methyl groups [22], a surrogate for these fuels
should likely include iso-paraffin as a major component. The reason for this extent of
isomerization is to meet the jet fuel density and freeze point specifications [19]. Recently,
Natelson et al. [23] experimentally studied the pre-ignition and autoignition behavior of
the F-T jet fuels and possible surrogates to understand the influences of compositional
differences on the chemistry of pre-ignition and autoignition. Kahandawala et al. [24]
investigated J P-8, an F-T synthetic jet fuel, and a one-component surrogate synjet fuel (2-
methylheptane) in a Shock tube and showed that both surrogate and actual synjet, and JP-
8 fuels had similar ignition delays. Mawid [25] suggested a two-component S-8 F-T

surrogate, which showed a similar ignition behavior to a JP-8 surrogate.

2.3. Kinetic modeling of jet fuel surrogates

Jet fuel surrogate mechanisms consist of hundreds of species and thousands of
reactions (e.g. Complete Ranzi mechanism: 310 species and 8335 reactions; Zhang
mechanism: 208 species and 1087 reactions [16]; Mawid mechanism: 226 species and
3230 reactions [26]). These kinetic mechanisms for various surrogate blends have been

computationally modeled in both premixed and non-premixed configurations.

Ignition delay times and laminar burning velocities of jet fuel surrogates have been
calculated using FlameMaster [27] or CHEMKIN [28] codes. Honnet et al. [1 l] modeled
a 2-component J P-8 surrogate (the Aachen surrogate) with FlameMaster. Vasu et al. [14]
used CHEMKIN to find ignition delay times of 5— and 6- component jet fire] surrogates
(Violi surrogates [7]). Dagaut et al. [29] used PREMIX [30] and PSR [31] CHEMKIN-

based codes to model their 3-component jet fuel surrogate.

Counterflow diffusion flame codes such as OPPDIF [32] and FlameMaster [27] are
well established and are widely used to study jet fuel surrogates [10, 11, 33, 34]. Highly
optimized libraries and Optically thin radiation models have been employed to evaluate
the thermodynamic and transport properties, the chemistry, and the radiative flux [33].
When large chemical mechanisms are to be modeled, many modifications are necessary
to the original codes. Many combustion mixtures of our interest include large jet fuel
hydrocarbons or their surrogate components. These contain molecules of disparate
molecular weight. Also, in the case of sooting flames, multi-ring PAH’S can form heavy
reaction intermediates. Relevant transport mechanisms and coefficients would be affected
by such a “molecular weight contrast” [35]. All diffusivities (v, a, D) in the computer

code should be revisited to accommodate for such phenomena. For example, the code

modified by the Smooke group at Yale [36] incorporated a modification of Fickian

transport models and an inclusion of the Ludwig-Soret effect for heavy molecules.

To my knowledge, modeling the sooting or incipiently sooting surrogate-doped
ethylene flames with the well-tested mechanisms like the Ranzi mechanism is yet to
agree with experiments [3 7]. Also, the detailed chemical structure of counterflow

diffusion flames has not been studied with simple jet firel surrogates such as the Aachen

surrogate.

2.4. Reduction of surrogate mechanisms

Several leading research groups are working on compilation and reduction of jet
fuel reaction mechanisms. Among all reaction mechanisms studied by the Hanson group
[14], the Ranzi mechanism [15] showed the closest agreement with the experimental
ignition delay data. This kinetic mechanism is an example of successfirl mechanism
compilation. It was compiled using existing hierarchically-constructed models for alkanes
(rt-dodecane, iso-octane, and n-tetradecane) and simple aromatics extended to account for
the presence of tetralin and methylcyclohexane [33]. The Smooke group at Yale has used
this mechanism in the non-sooting counterflow diffusion flames [36] and is attempting to

reduce it.

Recently, a comprehensive discussion on the development of reduced mechanisms
was presented by Lu et al. [38]. Various skeletal reduction methodologies have been
employed to remove unimportant species and reactions from a detailed mechanism:
directed relation graphs (DRG), DRG with error propagation (DRGEP), DRG-aided

sensitivity analysis (DRGASA), computational singular perturbations (CSP),

10

optimization, principal component analysis, detailed reduction, J acobian analysis, and
sensitivity analysis [3 8, 39]. One of the fastest skeletal reduction algorithms is DRG [40,
41], which have been recently used by Lu and Law for the reduction of n-heptane [39,
40] and iso-octane [40], and by Pepiot-Desjardins and Pitsch [42] for iso-octane (using
DRG-based techniques). The skeletal reduction is followed by lumping and time-scale
analysis. A novel lumping method was recently suggested by Pepiot—Desjardins and
Pitsch [43] and used for n-heptane and iso-octane. Time-scale analysis methods may

include QSSA, PEA, ILDM, and CSP [3 8].

2.5. Gas Chromatography and Mass Spectrometry (GC/MS)

Flame sampling involving the physical extraction of gaseous samples by a probe
for subsequent chemical analysis provides a wealth of data about the underlying flame
chemistry. Flame sampling is virtually indispensable for a detailed chemical kinetics
study of high molecular weight hydrocarbons such as jet fuel or its surrogates. A GC/MS
is used in this work for the chemical analysis. The GC/MS instrument is a complex piece
of equipment composed of various parts; the operation of these parts allows achieving

separation and identification/quantification of different species in a gas mixture [44].

The term gas chromato graph is often used to generically refer to the group of
components (e. g., oven, chromatographic columns, etc.) operating to achieve species
separation. The GC operates on gas mixtures, although analysis of liquid samples can
also be performed by introducing the sample through appropriate injectors, after which
the liquid is vaporized. Chromatographic separation of selected Species in the gas mixture
takes place within chromatographic columns. Identification and quantification of species

as they emerge from the columns is performed by means of dedicated detectors. The

11

time, under a given and fixed set of conditions, that each component of the sample is
retained in the chromatographic column is called the retention time. The different species
can be identified based on their retention time that is typically determined on the basis of
the columns standard chromatograrns. The separation of the mixture is recorded as a
series of peaks. The area under each of these peaks is uniquely proportional to the
concentration of that component in the original mixture. Integration of the area can be
used for quantification provided that accurate calibration curves are available for the

component.

The Mass Spectrometer (MS) allows for both identification and quantification of
species. Because of its operational principle, the MS does not always require species
separation. Other detectors (e. g. Flame IonizatiOn Detector (F ID) and Thermal
Conductivity Detector (T CD)) are generally only used to quantify absolute or relative
amounts of separated compounds. A software (Agilent Chernstation) is used to run

analyses and analyze the data extracted from the GC/MS.

2.6. Experimental study of jet fuel combustion

In an earlier study at Yale [33], good agreement was reported for extinction strain
rate and temperature profiles between jet fuel and a Six-component surrogate in non-
sooting counterflow diffusion flames. This contribution began in collaboration with the
research groups at University of Utah and University of Milan, with the Utah group
establishing the surrogate mixture formulation and the Milan group developing the
chemical kinetic model for this surrogate including 221 species and 5032 reactions [33].
The formulation used in [33] and in all subsequent work at Yale departs slightly from the

Utah surrogate in that the %vol composition of the six components (Surrogate #1) in

12

Violi et al. [7] was reinterpreted as molar fractions, which resulted in relative average
changes of 27% in the component molar compositions. We shall refer to this surrogate as
the Utah/Y ale surrogate. More detailed investigations followed in the Yale laboratory,
with the chemical analysis of the structure of a methane counterflow diffusion flame
perturbed with thousands of ppm of either jet fuel or the six-component jet fuel surrogate
[34, 45] in highly diluted and non-sooting flames, including detailed one-dimensional
modeling using the Milan mechanism. The surrogate captured the general jet fuel
behavior reasonably well except for ethylene and small aromatics such as benzene and
toluene. The discrepancy in aromatics is cause for concern, if confirmed under sooting

conditions, since aromatics are critical precursors to soot.

The sooting behavior of jet fuels, especially at take-off, is an issue in most aero-
turbines for which non-premixed configurations are preferred. Therefore, any surrogate
formulation needs to be validated with respect to this performance. Beyond global
combustion properties, such as extinction, ignition, and flame speed of jet fuels, detailed
probing ensures validation of surrogate composition in sooting conditions and also
provides details on the structure of jet fuel sooting flames [46]. Gas sampling is
problematic in the presence of large amounts of soot because of the inevitable clogging of
the microprobe orifice. This problem can be circumvented if conditions of incipient
sooting are chosen, in which the soot loading is kept at the minimum necessary to discern
a faint blackbody luminosity in the flame. These are the conditions chosen in the present

study.

A comprehensive investigation is conducted on the detailed flame structure of

ethylene counterflow diffusion flames perturbed by trace amounts (2000 ppm, molar) of

13

jet fuel and two surrogates via gas sampling and chemical analysis. The ultimate goal is
to ascertain if the surrogate formulations are reasonably successful in mimicking the
performance of jet fuel, especially with respect to the behavior of small aromatics for
which discrepancies were observed under non-sooting conditions [36, 46]. In addition to
attempting to validate the surrogate formulations with respect to the flame structure, the
objective is also to provide a database for the pyrolysis and oxidation behavior of jet fuel
in a diffusion flame at the onset of soot formation for other investigators to use in parallel
research efforts. Two- surrogates are examined: the six component Utah/Y ale surrogate
used in all previous work in the Yale laboratory [33, 34, 45], and a two component
Aachen surrogate, that has been reported to mimic not only conditions of extinction and
autoignition, but also to match the soot volume fi'action behavior, especially under
relatively high strain rates [11]. Semi-detailed chemical kinetic mechanisms are available
for both surrogates [11, 47, 48]. The study of the Aachen surrogate is conducted in
collaboration with the research groups at University of California San Diego and RWTH

Aachen.

l4

Chapter 3 Experimental Procedure for Detailed
Chemical Analysis of Jet Fuel Doped Flames

In this chapter, details of the experimental procedure for the chemical analysis of
jet fuel and its surrogates are explained. The experimental work included in this thesis
was performed at the Yale Center for Combustion Studies at Yale University. The work
presented in this thesis is the latest progress made in the experimental part of the jet fuel
combustion research (started few years ago and supported by the AFOSR and ARO) at
the Yale Center. The details of the experimental setup which includes a counterflow
burner equipped with a semi-automated GC/MS are presented. Criterion for comparison
of various ethylene diffusion flames either as baseline flames or flames doped with jet
fuel and its surrogates are elaborated. The adjustments in boundary conditions are also
discussed. Photographs of the experimental setup, used in this work, are presented in

Appendix A.

3.1. Theory of counterlfow flames

Counterflow (opposed-flow) burner configurations have been extensively used in
the study of non-premixed as well as premixed flames. The configuration has many
desirable characteristics for experimental and computational studies. It is symmetric and
the resulting flame is planer. Thus, detailed experimental and computational investigation
01' the flame structure can be conducted in the direction normal to it along the centerline.

30th sides of the flow can be seeded with particles for LDV or PIV mapping of the flow

field. A schematic of a counterflow diffusion flame, which can be generated by

15

impinging a uniform oxidizer jet against a uniform fuel jet is Shown in Figure 3-1. A

photograph of a counterflow burner, used in this work, is presented in Appendix A.

In the counterflow configuration, buoyancy effect is small and the resulting flame
is quite steady because the flow is usually dominated by forced convection [49]. In the
laboratory, stability can be further facilitated by adding a guard flow (also called Shroud
flow) of inert gas and adjusting its flow velocity. This shroud flow also isolates the
reactant jets from the environment. However, the Shroud flow may have a drawback: it
sometimes moves the flame. As a result, when computational profiles are to be compared
with the experimental profiles a shift, which depends on the velocity Of the shroud flow,

may be indispensible to match profiles.

 

 

 

 

 

Oxidizer
I
A
Flame
Ci) .
________________________________________ Stagnation
L r F surface
Inert guard
v flOW
Fuel

Figure 3-1. Schematic of a counterflow burner with a diffusion flame.

According to Law [49], “the inverse of the velocity gradient, 1/a, represents a
characteristic flow time, which, when compared to the characteristic reaction time, yields
the system Damkohler number”. Because the flow velocity along the centerline near the
stagnation region varies linearly with distance, the flow can be characterized by a single
parameter, namely its velocity gradient a, which constitutes the local strain rate. In the
laboratory, aerodynarnically shaped nozzles usually generate uniform flows. The strain

rate, a, can be approximated for the counterflow as [50]

(3-1)

.0 :[ F—zfil

(3-2)

“F V-Fl :z—El

where a, V and p are strain rate, velocity and density, respectively, subscripts 0 and F

denote fuel and oxidizer and L is the distance between nozzles (or burner separation).

Seshadri and Williams [50] defined a density corrected strain rate, a, for the

counterflow flame as

_2L__Q_[l+ +VF__\/E_] (3_3)
VOJF;

where V0 and VF are the flow velocities normal to the stagnation plane (SP) at the

oxidizer and fuel boundaries, respectively, and p0 and pp are the mixture densities at the

oxidizer and fuel boundaries, respectively. Density corrected strain rates defined by Eq.

(3-3) are used in this part.

17

The stoichiometric mixture fraction, 2f, is defined as

1
_1+sYFF/Y00’

 

2f (34)

where s is the stoichiometric mass ratio of oxygen to fuel, YFF and Y00 are the feed

stream mass factions of the fuel (regardless of the chemical composition) and oxygen,
respectively.

Highly turbulent counterflow flames have been recently proposed as a very useful
benchmark of complexity intermediate between laminar flames and practical systems [1].
By operating in a turbulent Reynolds number regime of relevance to practical systems
such as gas turbines and internal combustion engines, they retain the interaction of
turbulence and chemistry of such environments, but offer several advantages from both a

diagnostic and a computational viewpoint.

3.2. Experimental setup

Figure 3-2 shows a schematic of the experimental setup. It consists of a counter-
flow burner [46], including a nitrogen shroud that shields the flame from room drafts and
ensures burning in the controlled atmosphere that is determined by the composition of the
feed streams. Carefully designed convergent sections allow for a top-hat velocity profile
at each burner mouth. The inner diameter of the fuel and oxidizer outlets is 12.5 mm and
the burner separation (L) is 14.1 mm. Slightly nitrogen-diluted air is used as the oxidizer
while the fuel is nitrogen-diluted ethylene doped with either jet fire] or the surrogates. To
ensure complete vaporization of the dopant liquid, an electrospray operated in the multi-

jet mode [51] disperses the liquid fuel in the preheated fuel/nitrogen stream. This

18

approach provides flexibility in flow rates without compromising the stability of the
flame. To prevent condensation downstream of the electrospray unit, PID controllers
keep the fuel line at 430K, which is well above the dew point of the dopant/C2H4/N2

mixtures. Gas samples are extracted from the flame through a microprobe, consisting of a

 

 

 

.l Nitrogen
GC Oven Sample I .l
@‘ I .’ Air
470 K .,
.I. ~2
e "

 

 

 

 

     
 
  

 

Heated lines

 

 

 

 

 

Carboxen Nitrogen
4 K
HP1 20 Electrospray
Multi-Position Valves
5000V
It
Jet Fuel

Figure 3-2. Schematic of the experimental setup.

small silica probe with an outer diameter of 340 um and an inner diameter of 170 um.

Details of the different microprobes used previously can be found in [34, 45].

The chemical analysis is performed by a gas chromatograph (Agilent 6890A)
equipped with mass spectrometer (MSD, Agilent 5973N), thermal conductivity (TCD),

and flame ionization (FID) detectors. The instrument is capable of quantifying complex

hydrocarbon mixtures, C0, C02, 02 and N2. It uses two capillary columns, a Supelco

Carboxen and an Agilent HP-l, connected to the PID and MSD, respectively. In addition,
the TCD measures non-hydrocarbon stable gases separated by means of a third column
(Alltech, Packed Molecular Sieve). Because of its much wider linear range, this detector

is better suited than the MSD for the analysis of gases present as large fractions of the gas

sample and/or in greatly varying amounts e. g. N2 and 02. A homemade nickel-based

catalytic converter (Methanizer) allows for FID quantification of CO and C02 upon their
conversion into methane in the presence of hydrogen. The system can separate and
quantify H2, N2, 02, C0, C02, light gaseous hydrocarbons and higher hydrocarbons up to

at least C14.

Species are identified during the GC/MS data post-processing by both the column
retention time and the molecule-specific mass spectrum. GC/MS analysis produces a
wealth of information, but has one main drawback: it takes a very long time to perform a
flame scan. At the small liquid flow rates of interest for jet fuel (e. g. 1.6 ml/hr), a syringe
pump is used to deliver the fuel. Preserving a steady flame over the analysis time, which
is on the order of one day, would be challenging: it would require repeated flame shut-
offs for reloading of the syringes. Thermal transients would affect boundary conditions,
and other inevitable consequences such as sampling probe distortions might cause
reproducibility problems that affect the self-consistency of the data. To sidestep these
problems, a semi-automated chemical analysis method is employed that consists of
sampling the gas and storing it in a battery of sampling loops using two pneumatic-
actuated injection valves and two multiposition valves, as Shown in Figure 3-2 [12, 34,

45, 52, 53]. One of the pneumatic injection valves is responsible for the TCD analysis

20

(top left in the figure) that is executed in real time, since it requires about two minutes per
data point. The other valve controls the PH) and the MS. Since these two instruments
require a much longer time to complete a measurement, the samples are stored in two sets
of 16 sample loops using two multi-position valves (V alco®) and analyzed overnight by a
computer-automated sequence. An optimized time-temperature program for the
chromatographic columns keeps the total analysis time at a minimum and ensures that

aging of samples does not change the sample concentrations significantly.

Temperature measurements are performed using flame-welded, silica coated Pt-
10%Rh/Pt thermocouples with a typical junction diameter and a wire diameter measuring
on the order of 70 um and 50 pm, respectively. Standard corrections for radiative losses
are applied. In some temperature measurements, a gas sample probe is also placed within
the flame to assess the probe’s intrusiveness. AS a result, any spatial offset between
temperature profiles with and without the probe would be mostly due to the sampling
probe intrusiveness. Temperatures of both feed streams at the burner outlets are measured

by a K-type thermocouple.
Accuracy in the GC/MS analysis and reproducibility of the data were ensured by
analyzing gas mixtures of known composition (Scotty®) and repeated sampling at the

same position in the flame. Standard calibration gases (Scotty®) are used for calibration

of light gaseous species. Aliquots Of liquid hydrocarbons dissolved in acetone are
injected into the GC/MS for the calibration of heavy liquid hydrocarbons. The total error

is estimated at 7% for light species and 12% to 15% for the heavier ones.

21

3.2.1. The electrospray

It was mentioned earlier that an electrospray operated in the multi-jet mode
disperses the liquid firel in the preheated fuel/nitrogen stream. Figure 3-3 shows a
schematic and photograph of the electrospray unit. For small flow rates (up to 10 mL/h),
a tapered needle is used to disperse fuel. The ground electrode consists of a ring
positioned 2 cm downstream the tip of the needle. A voltage between 4 and 6 kV is
applied between the two electrodes. The Spray is often run in the multi-jet mode,
depending on the flow rate and voltage. To monitor the electrospray behavior and ensure
that the liquid is well-dispersed into vapor without condensation, a transparent Pyrex

electrospray chamber (evaporator) is used. This electrospray chamber is heated by

 

 

Ring Connected to
Ground

Needle

    
 

Insulated
Pyrex
Evaporator

Heating Tape

 

 

I]
l

Liquid Fuel from .

 

 

 

 

 

* -

 

Syringe Pump

Figure 3-3. Schematic and photograph of the electrospray unit for liquid fuel dispersion.

22

heating tapes covered by an insulation layer. A thermocouple is positioned inside the
chamber, very close to its wall, to monitor its temperature that is kept at around 160 °C to

avoid condensation. The electrospray must be operated with particular attention. Two
copper tubes (not shown in the Figure), painted in black to avoid glare, are mounted
perpendicularly to the Pyrex evaporator to realize small “openings” in the insulation layer
and enable occasional visualization of the spray behavior using a microscope. Sparks
may cause misbehavior of the spray. They may also damage the temperature PID
controllers that, at the very least, will require resetting. Beside the visualization of the
electrospray with microscope fi'om the black Opening, current can be checked. The
current through the voltage amplifier should be zero Since no shortcut is desired while a

steady multi-j et spray is present.

3.3. Criterion for flame comparison

Diffusion flames “doped” with trace amounts of prevaporized liquid fuel have
shown many computational and experimental advantages. The advantages of doping a
baseline flame were mentioned in the previous studies at Yale [34, 45]. This approach
was first employed in non-premixed flames by Harnins et a1. [54] and has been more
extensively used by McEnally and Pfefferle [55, 56]. This approach minimizes the
potential for vapor condensation,“ Since the condensable species is at very small
concentrations. The temperature-time history can be easily adjusted to be identical for all
flames, so that this counterflow flame environment can be regarded as a flow reactor in

which the residence time and the temperature profile are adjusted by varying the strain

 

t
Evidence in preliminary experiments at Yale had showed that, probably because of condensation effects
in the unheated probe, the species with largest molecular weights were underestimated.

23

rate and the feed stream composition, respectively. The temporal history is identical since
the flames have essentially the same mass averaged velocities at the boundaries and
identical temperature profiles, which leads to the same evolution of the velocity profile
via gas expansion. Furthermore, critical, non-chemical variables, such as temperature and
velocity, as well as probe-induced perturbations, can be evaluated once and for all in the

baseline flame.

The counterflow configuration is chosen, as in previous studies at Yale, as the work
horse for a systematic study of firels with complex chemical kinetics and their coupling
with transport in the simplest possible fluid dynamic environment: a one-dimensional
laminar flow. It is amenable to detailed computational models that are now routinely
applied in most combustion laboratories, at least for simple firels. Since our goal is to
focus on conditions yielding soot formation and since soot is an issue primarily in non-
premixed flame environments [56], a non-premixed counterflow flame is appropriate for

our study. For such a flame, the density corrected strain rate, a, was defined by Eq. (3-3)

and the stoichiometric mixture fraction, zf, was defined by Eq. (3 -4).
The previous study at Yale [3 6] focused on a highly diluted methane baseline flame
doped with 1000 ppm of jet fuel or the Utah/Y ale surrogate under non-sooting conditions,

with Tmz 2012 K, a= 144 3'1, and 2;: 0.76. For comparison purposes, the boundary

conditions for these non-sooting methane flames studied by Bufferand et al. [36] are

presented in Table 3-1. Here we use an ethylene baseline flame, Since C2H4 has a greater

soot propensity than CH4, SO that its boundary conditions can be adjusted with ease for it

to be at the onset of soot formation. TO that end, the temperature/time history needs to

24

Table 3-1. Boundary conditions for the non-sooting methane flames doped with jet fire]
and Utah/Y ale surrogate from the previous study of Bufferand et al. [36] at Yale.

 

 

 

 

 

 

 

 

 

 

 

Flame C
Flame A Flame B l ,1,
Baseline Jet Fuel U ale
Surrogate
Molar Composition
N2 0.897 0.902 0.902
CH4 0.103 0.097 0.097
C2-CS alkane impurities 232 ppm 218 ppm 218 ppm
Jet fire] (C11H21) 992 ppm
Methyl-Cyclohexane 200 ppm
Fuel ISO-Octane 100 ppm
Side m—Xylene 150 ppm
Tetraline 50 ppm
Dodecane 300 ppm
Tetradecane 200 ppm
Mass Flux (g/(cm2.min)) ' 2.80 2.97 2.97
Temperature (K) 379
Molar Composition
N2 0.227
Oxidizer 02 0.773
Side 2
Mass Flux (g/(cm .min)) 3.19 l 3.42 l 3.42
Temperature (K) i 340 5
Strain Rate (s“) 134 | 144 | 144
zf 0.76

 

 

 

 

 

favor soot formation, with higher temperatures and lower residence times [57-59], and
the stoichiometric mixture fraction, zf, would have to decrease sharply in comparison

with our previous studies to values less than 0.5 to ensure that the flame locates itself on

the oxidizer side of the stagnation plane and oxidation of soot precursors is avoided [60].

A value of zf= 0.18 was chosen for all flames. The stoichiometric mass ratio of oxidizer

25

to fire], 3, for the baseline flame is 3.42, whereas Ypp and Y00 are 0.27 and 0.2,
respectively.

Table 3-2 specifies the overall mean strain rate and the boundary conditions (mole
firactions, total mass flux and outlet temperatures of both fuel and oxidizer streams) for
the five flames under consideration: the baseline ethylene flame, Flame A; the
ethylene+jet fuel flame, Flame B, the ethylene+Utah/Y ale surrogate flame, Flame C; the

ethylene+Aachen surrogate flame, Flame D; and a fifth flame, Flame A*. The latter is a

modified baseline ethylene flame whose total fire] carbon flux matches the doped flames.
The total carbon flow rate is 3.81><10'2 mol/rnin and 3.96><10'2 mol/min for Flame A and
Flame C, respectively. Flame A* is established by modifying the boundary conditions of
Flame A to ensure the same zf, temperature profile, and carbon molar flow rate as Flame
C. To avoid clutter, the only data that will be presented for Flame A* is the

concentrations of a few minor species such as aromatics. A chemical formula of C11H21

and density of 0.81 gr/cm3 were used for jet fuel [61]. The jet fire] was provided by

Wright-Patterson Air Force Base (POSF No. 4658). TO account for the considerable
variability in the composition of jet fuel from different refineries, an “average” jet fire]
was synthesized by mixing together 5 Jet A fuels from different U.S. manufacturers. The
composition of that blend in vol% is: 55.2% paraffins (n- and i-), 17.2%
monocycloparaffins, 12.7% alkyl benzenes, 7.8% dicycloparaffins and 4.9% indans and

Tetralin. The balance, z 2%, is in naphthalenes and trycylcoparaffins.

26

Table 3-2. Boundary conditions for the ethylene diflirsion flames under incipiently
sooting conditions. The jet fire] was provided by WPAF B (POSF No. 4658).

 

 

 

 

 

 

 

 

 

 

 

 

 

Flame A Flame A* Flame B Flame C Flame D
. Equal Utah/Y ale Aachen
Baseline Jet Fuel
Carbon Surrogate Surrogate
Molar Composition
N2 0.7278 0.7280 0.7340 0.7339 0.7339
C2H4 0.2722 0.2719 0.2641 0.2641 0.2641
C2 (Ethane) 637 ppm 636 ppm 618 ppm 618 ppm 618 ppm
impurities
Jet Fuel (C11H21) 1953 ppm
Methylcyclohexane 394 ppm
g ISO-Octane 197 ppm
E73 m-Xylene 295 ppm
'2' Tetralin 98 ppm
In n-Dodecane 591 ppm
n-Tetradecane 394 ppm
Total= 1970 ppm
1,2,4-trimethylbenzene 450 ppm
n-Decane 1520 ppm
Tome 1970 ppm
Mass Flux (g/(cm2.min)) 1.619 1.683 1.684 1.683 1.682
Temperature (K) 407 407 407 407 407
‘9 Molar Composition
.3 N2 0.8070 0.8070 0.8070 0.8070 0.8070
i. 02 0.1843 0.1843 0.1843 0.1843 0.1843
5‘
:5 Mass Flux (g/(cm2.min)) 1.891 1.925 1.925 1.925 1.925
0
Temperature (K) 370 370 370 370 370
Strain Rate (6‘) 89.9 85.2 92.3 92.3 92.3
zf 0.18 0.18 0.18 0.18 0.18
Carbon mole fraction % 54.45 54.39 54.97 54.85 54.74
Carbon flow rate
- 3.81 3.958 . . .
(mol/min x 10 2) 3 967 3 958 3 951
V0 (cm/s) 33.17 31.03 33.77 33.77 33.77
VF (cm/s) 32.17 31.13 33.16 33.16 33.16

 

27

 

Table 3-3. Jet fuel surrogate mixtures; values are reported in molar %.

 

 

 

 

 

 

 

 

 

Utah/Y ale Surrogate Aachen Surrogate

iso-octane 10

methylcyclohexane (MCH) 20

m-xylene l 5

n-dodecane 30

n-tetradecane 20

tetralin 5

1,2,4-trimethylbenzene (TMB) 22.8
n-decane 77.2

 

 

 

 

 

The Utah/Y ale surrogate is the six-component blend of well-known hydrocarbons
used in the previous works at Yale [33, 45], while the Aachen surrogate is a two-
component surrogate as used in [11]. Compositions of these surrogates are given in Table
3-3. The surrogate components are made by mixing high purity hydrocarbons (Sigma-
Aldrich and Fluka 2 99% purity). Trace amounts of an ethane impurity in the C2H4

supply were revealed by chromatographic analysis at about 620 ppm. The ethane

impurity of each flame has been listed in Table 3-2.

A photograph of Flame C, the baseline flame doped with the Utah/Y ale surrogate,
is Shown in Figure 3-4. A layer of orange luminosity is visibleundemeath the flame
chemiluminescence. The dopant concentration is chosen based on two considerations: (1)
it should be sufficiently small so as not to change the overall flame structure and to
preserve incipient sooting conditions, such that clogging of the microprobe orifice is

avoided without resorting to special probes with attendant deterioration of the spatial

28

resolution [62]; and (2) it should be sufficiently large so that the jet filel/surrogate
contribution to the production of critical Species, such as soot precursors (e.g., small
aromatics) can be discriminated from the contribution of the baseline flame. On the basis
of these considerations, the ethylene flame was doped with 2000 ppm of either jet fuel or
its surrogates. However, the addition of even such a small amount Of these fuels increases
the peak temperature by 20-30 K and the flame location shifts slightly toward the
oxidizer side. To preserve the temperature-time history and ensure comparable Arrhenius
kinetics among all of the flames, the temperature profile needs to be the same in all of
them. Also, since the mixture fraction is a single-valued complementary error function of
the axial position, fixing its value at 2f= 0.18 ensures that the flame position is unaltered
by the perturbation. To maintain the same temperature profiles as in the baseline flame,
we increased the inert mole fraction in the fuel stream, which leads to a small change in 2f
and a firrther shift of the flame towards the oxidizer side. We compensated for this shift
by a small increase in the oxidizer flow rate to move the stagnation flame in the opposite

direction, towards the fuel Side.

 

Figure 3-4. Photograph of the flame at the onset of sooting. The thickness of the sooting
zone with orange luminosity underneath the blue zone is about 0.6 mm.

29

Chapter 4 Experimental Study of Ethylene
Counterflow Diffusion Flames Doped with Jet Fuel and
its Surrogates under Incipiently Sooting Conditions

In this chapter, results Of the chemical analysis of jet fire] and its surrogates are
presented. An ethylene counterflow diffusion flame doped with 2000 ppm on a molar
basis of jet fire], Utah/Y ale surrogate or Aachen surrogate is studied under incipiently
sooting conditions. As discussed earlier, the doped flames have identical stoichiometric
mixture fractions and strain rates, resulting in a well-defined and fixed temperature/time
history for all of the flames. Five flames are under consideration: the baseline ethylene
flame, Flame A; the ethyleneI-jet fuel flame, FlameB, the ethylene+Utah/Y ale surrogate
flame, Flame C; the ethylene+Aachen surrogate flame, Flame D; and a fifth flame, Flame
A*, which is a modified baseline ethylene flame whose total fire] carbon flux matches the
Utah/Y ale surrogate doped flame. Profiles of temperature, critical fire] decomposition
products, and soot precursors are compared. These profiles include major species, C7-
C15 alkanes and alkenes, C3-C6 alkanes and alkenes, C2 hydrocarbons, and some
aromatic species. By comparing and contrasting the detailed structure of these flames, the

surrogates’ pyrolysis, oxidation, and most importantly the sooting behavior are examined.

4.1. Similarities in Temperature and Species Profiles among Flames

Figure 4-1 presents a comparison between the temperature profiles of Flame A
(baseline), Flame C (doped with the Utah/Y ale surrogate), and Flame A* (baseline with
the same carbon mole flux as Flame C). Note that the temperature profiles are

indistinguishable among the three flames, which means that our strategy of establishing

30

virtually identical time-temperature histories is successfill. Since the overall jet fuel heat
release behavior was indirectly validated with the Utah/Y ale surrogate in [33] and with
the Aachen surrogate in [l 1] by examining the extinction behavior and the temperature
profiles of such flames, the same concentration of jet fuel and of the Aachen surrogate

should lead to identical temperature histories.

 

2400

 

 

 

 

K I Flame A -Baseline
- 0 Flame C -Utah/Y ale surrogate
2100 ’_ + Flame A*-Equal carbon
- Flame A -without microprobe
A 1800 :
é ..
o i-
g 1500 _
g .
o. 1200
E I
or
I...
900
600 Z
300

0123456789101112
Distance from fuel inlet (mm)

Figure 4-1. Temperature profiles. Symbols: Flames A, C and A* with microprobe in.
Line: Flame A in the absence of microprobe.

The intrusiveness of any physical probe inserted into the flame is a common
criticism raised to probe sampling. In addition to introducing a heat sink into the
combustion environments, the probe may perturb the fluid dynamics of the flame [36].
McEnally et a]. [55, 56] have comprehensively reviewed such pitfalls of flame sampling.
In situ analysis, primarily by optical techniques, has the advantage of being essentially

nonobstructive. However, it is usually less general, often applying to at most only a few

31

of the species present in the flame [63]. To assess the probe perturbations, an additional
temperature profile is shown for the baseline flame in the absence of the sampling
microprobe. Comparison between the temperature profiles confirms an overall probe-
induced perturbation that results in a 0.5 mm Shift towards the fuel side, consistent with
similar conclusions fiom OH planar laser induced fluorescence [3 6]. This shift accounts
for roughly 7 % of the physical domain where chemistry is playing a role (the region

between 2 and 9 mm from the fuel inlet), is therefore considered a modest shift.

Figure 4-2 compares some major species profiles among Flames A to D. Unless

Specifically noted in the subsequent figures, no experimental data of Flame A* are shown

to avoid excessive cluttering of figures. Full-blue symbols are used for the baseline
ethylene flame (Flame A), full-black symbols for the jet fire] doped flame (Flame B),

open symbols for the Utah/Y ale surrogate counterpart (Flame C), full-red symbols for the
Aachen surrogate counterpart (F lame D), and + for the modified baseline flame (Flame

A“) in this and all subsequent figures. Our goal was to perturb a well-defined baseline
ethylene flame with the addition of trace amounts of j et fire] or surrogates. No distinction

is observed in the profiles of the five flames with respect to CO and C02. 02 was only
measured in Flame A and is shown in the figure to illustrate the overall nonpremixed
fl 31116 structure. The profiles of C2H4 are also identical, except at the fuel side boundary
Where the mole fraction has necessarily changed between the baseline flames (Flames A
mid A*) and the doped flames (Flames B, C, and D) because of the flow rate adjustments
‘liSCussed in Section 3.3. As elaborated before, the temperature profiles (Fig. 3) of these
11 allies are virtually the same. The absence of significant changes in the concentrations of

CO and C02 and Of temperature because of the presence of either jet fuel or the

32

surrogates indicates that these species and the heat release are produced predominantly by
the oxidation of 021-14. The blue flame in Figure 3-4 locates itself at approximately z=6.2-
6.4 mm. The thickness of the orange sooting zone is about 0.6 mm. The asymmetry in the
temperature profile is due to endothermic pyrolysis reactions between 2 = 5 and 7 mm.
Detailed modeling of ethylene nonpremixed flames confirms the existence of such a zone
[64]. The maximum C02 mole fraction is slightly larger than that of the methane non-
sooting flames in our previous experiments [36], whereas the peak of CO mole traction is
nearly 1.5 times larger than that of the non-sooting flames. This difference is attributable

to the nitrogen mole fraction in the fuel stream being smaller in the ethylene flames.

A drawback with probing of non-premixed flames in general is that the layer where
interesting products are abundant is narrow and comparable to the spatial resolution Of
regular probes. The CO profiles of our sooting flames are wider than in [36]. One of the

reasons for this behavior is that the strain rate of these flames is smaller than the strain

rate of the non-sooting flames (the flame thickness scales inversely witth ). Having a
Wider region where chemical reactions are occurring makes the probing easier and more

rel iable, especially since we want to track the formation of aromatics.

33

 

 

0.3 _

r . '02

i . ‘ CO
0.25 ' ' I O

”a .cfl ' i I C02

0.2 :

0.15 :

Mole fraction

0.1 :

0.05 :—

 

 

 

 

012 345 6 78 910
Distance from fuel inlet (mm)

Figure 4-2. Major species profiles: Flame B (full black symbols), Flame C (open
symbols), Flame D (red symbols) and‘F lame A (blue symbols).

Figure 4-3 compares the methane mole fractions among the five flames. Methane in
the present study, unlike [36], is a minor species produced by pyrolysis and oxidation of
the ethylene, and of the jet fuel or surrogates in concentrations below 100 ppm. The
De=aks of the methane profiles for all flames are located at 225.4 mm. The doped flames

Ila-\/e virtually the same methane profiles, whereas the peaks of methane mole fraction for
the baseline flames (Flames A and A*) are 15% smaller. The similarity of the methane
lDt‘ofiles, combined with the similarities in major species and temperature that were
DI‘eviously discussed, lends credence to our use of these flames as a sort of flow reactor

vl"’i~'th virtually identical conditions, except for the imposed perturbation.

34

 

 

 

 

0.001 _
i .80
_ . a.
0.0008 _- 34.. .90.
O + 0

_ (y .2

c ’ 0* .9

,9 0.0006 - a, +.

'5 " e

as c °

it: - o

2 -. " ,

o - e

2 0.0004 _ .3 Q
_ S 0
.. . 0

0.0002 - +
. e.
- g g
M“ 2M
0 .I....I...lrnr.in..l.|.l _
8 9 10

O 1 2 3 4 5 6 7
Distance from fuel inlet (mm)

Figure 4-3. Methane mole fraction profiles: Flame B (full black symbols), Flame C (Open
symbols), Flame D (red symbols).

4-2- C7—C15 Alkanes

Figure 4-4 shows the profiles of the C7-C15 alkanes in the doped flames, B, C and
D - Only one set of data is Shown for the jet fuel doped Flame B to avoid cluttering of the
"figure. The iso-octane concentration in Flame B is about 0.5 ppm and it is not Shown.
MQst alkanes in the jet fuel-doped Flame B (Figure 4-4a) disappear because of pyrolysis
at 2:5 mm, except for decane, undecane, and tridecane that decay at zz5.2 mm, and
quecane that disappears at 235.4 mm. In the Utah/Yale surrogate-doped Flame C, the

di Sappearance of surrogate iso and normal paraffins is delayed to nearly z=5.4-5.5 mm.

35

  
 
 
 
 
 
 
 
   

 

 

 

 

 

 

heptane
octane
nonane
decane
. undecane
A dodecane
E tridecane
& tetradecane
v pentadecane
.5 MCH
*5
E
E
o
E
0 1 2 3 4 5 6 7 8 9 10
Distance from fuel inlet (mm)
(a)
1600
I v dodecane
_ . O
1400 - e . s . o tetradecane
5 ’ , a MCH
1200 f ’ D isooctane
g : ° 0 decane
3 1000 _- e
g :
'4: +- O
8 800C .
2 VV 0
400 ES $ E E v
85 vv 0
200 up l>l> I>I> ta. D9586: 0
DD
0 L I l I 1 DD

 

 

012 3 4 5 6 7 8 910
Distance from fuel inlet (mm)

(b)

Figure 4-4. Comparison of C7-C15 Alkanes; a) F lame B, b) Flames C (open symbols)
and Flame D (red symbols), Flame A (blue symbols) and Flame A* (+).

36

In the Aachen surrogate-doped Flame C that has only one alkane (decane), its
disappearance occurs at $35.6 mm. The data presented here are consistent with typical

decomposition of large alkanes. The observed profiles are reflective of the chemistry of
jet fire] or the surrogates, since the oxidation of C2H4 does not yield any of these large

alkanes, as also confirmed by preliminary simulations of the baseline flame with the

semi —detailed Milan mechanism. Jet fire] and both surrogates Show a reasonably good

agreerrlent in the pyrolysis trends of large alkanes.

Whereas the chromatograms of the surrogates are clean with distinct peaks (as
shown in Figure 4-5) and the quantification of their components poses no difficulties, for
complex fire] blends such as jet fuel that contain a large number of alkenes and
alkYIBenzenes having different isomers, the chromatograms are subject to interference.
n13 quantification of these isomers is often affected by overlapping peaks fiom other
rnolecules. AS a result, the chromatograrn, shown in Figure 4-6, contains a large “grassy”
IDEIQ1<ground and is very hard to analyze. The reason for the modest presence of alkanes in
the jet fuel dOped F lame B (Figure 4-4a), despite the fact that they are known to be major
coll'iponents of jet fuel, is because the total alkane component is spread over many

di fi‘erent individual alkanes. The chromatograrn was measured for a gaseous sample
exhumed near the burner mouth, that is, before any significant chemistry had taken place.
The final number of species we were able to measure is 33, and another 20 were correctly
identified but their quantification lacks the necessary accuracy. There is a plethora of
8th aller, unidentified peaks and a pedestal on which the peaks are superimposed. The
Dre sence of the pedestal prevents us from performing an accurate integration for these

peaks. The dominant peaks in Figure 4-6 are associated with C9-C15 alkanes, whereas

37

only a few of the smaller peaks, associated with smaller alkanes and aromatics, are
marked. The quantified alkanes in Figure 4-6 are regarded as tracers for a broader group

of large alkanes in jet fire]. Similarly, the quantified small aromatics are indicators of a
larger number of similar molecules. Therefore, the present data provide a general picture
of the chemical evolution in the pyrolysis and oxidation of jet fuel in a sooting diffusion
flame, with the caveat that the comparison with surrogates can only be qualitative for C7-

C15 alkanes or alkenes upstream of the flame zone.

TOTAL ION

 

 

 

 

 

 

 

 

LL—UL, L, L L.

1'
TIME [min] 5 10 15 20

 

 

Figure 4-5. Utah/Y ale surrogate chromato gram presenting distinct peaks in
Correspondence with nitrogen (leftmost peak) and the 6 components of the surrogate:
fiom left to right, methylcyclohexane, iso-octane, m-xylene, tetralin, dodecane and
tetradecane.

38

undecane
<--

dodecane
<--

Response
(Total ion)

4— decane

<— tetradecane

4.... 1,2,4—trimethylbenzene
<— tetralin

‘— nonane

 

 

 

 

4— pentadecane

<—- octane
«— m-xylene

‘— methylcyclohexane

4—toluene

! “T ‘ Time (min)

4— benzene
4- heptane

 

ryvvv—v— vvvvvvvv

V6 10 '14' ' I 18 22

Figure 4-6. T ypica] jet fire] MS chromatograrn, as measured from a gaseous sample
exJil‘acted near the burner mouth. The arrows denote compounds that were identified and
quantified.

A comparative analysis Of the total carbon count helps us to evaluate the GC/MS ‘
performance with respect to large hydrocarbons. For our sooting Flames B, C and D a

colIilparison of the total carbon counts from the liquid dopants is shown in Figure 4-7. The

t
otal carbon mole fraction excludes the contributions from major products CO and C02,

tld their primary source, C2H4. Unlike a Similar figure in [36], it also excludes the

QBliltributions from acetylene since, as shown in Section 4.5 below, the production of
ateetylene is mostly attributed to the ethylene portion of the fire]. Therefore, the
ch"Iflparison in Figure 4-7 Should account for all of the measured carbon in the parent
1iqtlid fuel dopant and their intermediate pyrolysis products. We measured approximately

6 ‘4 times as much carbon at the fire] boundary in the surrogate-doped flames as in the jet

39

fuel-doped flame. Yet, the actual carbon must be nearly the same at this location since the

overall molecular weight and flow rates of the injected vapor are comparable in the three

 

 

 

20000 .
- $.- 43 e FlameB
Tia - 2“}: .13 A FlameC
. I FlameD
‘A
615000? A
:‘I u
e i A
E - a
§10000- '
E a
19 A
8 f
5000- 4";
l: 0 e e e e e e .A‘A-
i. A
0 '''''''' IlullllllllllIJHLIIHW.

 

Distance from fuel inlet (mm)

Figure 4-7. Partial carbon mole fraction profiles for Flame B (full symbols), Flame C

(Open symbols) and Flame D (red symbols). Contributions fiom ethylene, C0, C02,
acetylene and ethane are not considered.

fleltnes. The data for Flame B show a flat profile hovering around 3100 carbon ppm up to
2 “'E 3.8 mm. For Flames C and D, the total carbon count has an initial plateau around
1 9 ,000 ppm up to z z 3.3 mm, after which it monotonically decreases as CO and C02 are
t\Ql‘med. This discrepancy suggests that we properly quantified only roughly 15% of the
tQta] carbon introduced as jet fire] vapor, whereas the fractions of carbon recovered from
the Utah/Y ale surrogate and the Aachen surrogate are roughly 93% and 92%,

reES-pectively. After z z 5.4 mm, all three profiles follow a similar quantitative trend. At

40

about 2 z 7.7 mm, the oxidation of dopants is completed and no carbon persists in

hydrocarbons in these flames.

4.3- C7-C11 Alkenes

In the non-sooting methane flames doped with jet fuel, small quantities of l-decene
and l-undecene were found immediately after the concentration drop of their parent
alkanes [36]. None of these two olefins were detected in Flames A-D. Analysis of the C7-
C 1 1 Olefins in the jet fuel doped Flame B is extremely difficult because of the
oVerlapping spectra of multiple isomers. The largest olefins detected (and quantified) in
the doped flames were l-Octene and l-heptene. These olefinic Species were not detected

in F lame B. Figure 4-8 illustrates the profiles of l-heptene and l-octene for the Aachen

 

 

 

2.5
[ I t-heptene
: I 1-octene I
2 - I
S : .
o.
E 1.5 l: .
:8 r e
E - '
:03 1 - e O
s ' ° - °
- e
O O
0.5 :- .
i I
W-
03 4 5 6 7

Distance from fuel inlet (mm)

Figure 4-8. Comparison of l-octene and 1-heptene profiles: Flame C (Open symbols),
Flame D (red symbols). None of these species was detected in Flame B.

41

 

and Utah/Y ale surrogate doped flames (Flames C and D). The peak of l-heptene and 1-
octene appear after the concentration drop of the parent alkanes (see Figure 4-4). This is
consi stent with the well-known decomposition pathways for alkanes: H-atom abstraction

or unimolecular dissociation to alkyl radicals, followed by beta scission of the alkyl

radicals to olefins and additional alkyl radicals [57].

A difference between the surrogates with respect to the formation of olefins is
Observed in the presence of large olefins. AS seen in Figure 4-8, in the Aachen surrogate
Flatne D, about 2 ppm of l-heptene and 1.5 ppm of l-octene were quantified near the
Peak of the C5-C6 olefins (224.9 mm shown in Figure 4-9). On the other hand, in the
Utall/Yale surrogate Flame C, only l-octene was detected, but peaks for cyclohexene and

rnetl'lylcylohexene were observed. Figure 4-9 shows the profile of cyclohexene for Flame
C-

4'4- C3-C6 Alkanes, alkenes and dienes

4'4 - l - C5—C6 Alkenes

Figure 4-9 illustrates how C5 and C6 olefins, such as l-pentene and l-hexene, are
folTried as larger alkanes decompose in Flames A-D. There was no detectable l-hexene in
file baseline ethylene Flame A. Smaller concentrations of 2-hexene and 3-hexene, on the
order of 1 ppm, were identified in some of the doped flames. The profiles of l-pentene

and 1-hexene Show peaks at z: 4.9 mm. Comparison of Figure 4-9 with the profiles of
alkanes in Figure 4-4 confirms that these olefins are formed as the fuel alkanes are
Q1§composing Flames B, C and D Show good agreement in their l-pentene profiles, with

Rigrlificantly larger concentrations as compared to the baseline Flame A. The agreement

42

with respect to ]- hexene on the other hand is relatively poor, especially for Flame C. The
Utah/Y ale surrogate-doped flame has nearly 2.7 times as much l-hexene as the jet fuel-
doped flame, while the Aachen surrogate-doped Flame D has a peak l-hexene value

almost 50% larger than that of jet fire] doped flame.

 

 

 

20
L o t-pentene
L o t-hexene
: a A cyclohexene
15 - ElEll
E ’ I
c. I.
5.1- I
.8 i 'E 0'6
‘5' 10 ' o
e - . ii 0
‘3 ' 'i
E — D Q ’ I
5 l— O . ‘ Cl
. a .8
_ u 4.53,,
D
- I. 2003484 goal?
0 b I I AAA -

0 1 2 3 4 5 6 7 8 9 10
Distance from fuel inlet (mm)

F igure 4-9. Comparison of C5-C6 alkenes profiles: Flame B (full black symbols), Flame
C (open symbols), Flame D (red symbols) and Flame A (blue symbols).

4-4.2. Quantification of dienes

The overlapping spectra of multiple isomers make the analysis of larger olefins in
the jet fuel doped flames extremely difficult. Similar challenges are posed by the
ghantification of dienes. The only diene that was detected and quantified in the jet fuel
aQt‘rd other flames is cyclopentadiene (C5H6), which is shown in Figure 4-10. Oxidation of
chylene by itself produces cyclopentadiene in the order of 4 ppm. Both surrogates

“Verpredict cyclopentadiene although they Show a Similar trend with respect to the

43

location and magnitude of the peak. The Utah/Yale surrogate doped Flame C produces
nearly 1.5 times more cyclopentadiene than the jet fuel doped Flame B. The

cyclopentadiene peaks appear further downstream as compared to the peaks of C3-C8

 

 

 

olefins.
18_
: o
16" O.
o
14: 0
A : '
E12". 0
e s -
.510? .
it? 8'-
2 E o
0 6b 0
2 I 00
4:- {\o
: // ‘
2; . / O \\
: C 002*". O .\
: O.'.’ O \ b
onW—mm
8

3 4 5 6 7
Distance from fuel inlet (mm)

Figure 4-10. Comparison of cyclopentadiene profiles: Flame B (full black symbols),
Flame C (open symbols), Flame D (red symbols) and Flame A (blue symbols).

4~4.3. C3 alkanes and alkenes

Figure 4-11a shows the profiles of l-propene for Flames A, A*, B, C, and D. The
baseline ethylene flames (Flames A and A*) produce similar propene mole fractions, but
the doped flames produce noticeably different propene profiles. The propene mole
fiaction is significantly larger than other alkenes, except obviously for the fuel species
chylene, and the peak mole fraction appears slightly after the peak of larger alkenes. The

bl-opene mole fraction peak for all three doped flames occurs at z 225.2 mm with the

highest magnitude for jet fuel (Flame B) and the lowest for the Aachen surrogate (Flame
D). The difference between propene mole fraction profiles of jet fuel/surrogate—doped

flames and baseline flames can be attributed to the beta-scission of large alkyl radicals all

 

 

 

250
200 :- .
. .-
A D
s - “a
53: 150 - '56-:
c h
.g .
8 r It! I?)
e: i .0 '
2 100 f _- a
O
2 ID If." g
_ It] I i
50 - 0....- C;
I.+I I’.'
: lug: . '
0 WWW

Distance from fuel inlet (mm)

 

 

(a)
25
A
L A
20 F €A ‘
’E‘ L AAA
Q r A A
9; 15 b A
c .. A AA
.9
E . *e
E 10': ‘ +
'5
_ A
5 ‘ AA
b M ”A
- A M
04

 

012345678916
Distance from fuel inlet (mm)

(b)
Figure 4-11. Comparison of profiles of a) l-propene and b) propane: Flame B (full black

Symbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+).

45

 

the way down to the smallest alkenes. We note that the ability of both surrogates to
predict the jet fuel behavior with respect to the formation of alkenes gets better for
Smaller alkenes, probably because many more pathways can produce the smaller alkenes
and so the specific details of the fuel alkane structure are less important. It is noteworthy
to compare the ordinate of the profiles of C7-C8, C5-C6, and C3 alkenes. The ordinates

Show at least one order of magnitude difference.

Figure 4-11b shows profiles of propane; no butane, pentane, or hexane were
deteCted. A comparison of the location of the propane peaks with the disappearance
l(><>€ltion of the larger alkanes in Figure 4-4 indicates that this small alkane is a product of
me decomposition of the larger ones. The contribution of ethylene in the production of
propane is well pronounced in Figure 4-11b. The non-sooting methane flame doped with
j et file] and the Utah/Y ale surrogate had almost 3 times as much propane as the sooting

flalhes (Flames B and C here), which can be attributed to impurities in the methane
Soul‘ce used in those experiments [36]. Also, in the non-sooting flames, small amounts of
iso4butane, pentane and hexane were detected, which can be attributed to the fuel source

ltrlIntrities or their production in the flame.

4~S. C2 Hydrocarbons

Figure 4-12 compares the measured C2 species in Flames A to D and Flame A*.
E-thylene is a major species in the fuel stream and its trend was presented in Figure 4—2. In
file previous experiments on non-sooting methane flames doped with either jet fuel or the
htah/Y ale surrogate, the latter overpredicted the ethylene concentrations, but accurately
I\fiproduced the ethane and acetylene concentrations [36]. Figure 4-12a shows a good

agreement between ethane profiles in all the incipient sooting ethylene flames. Ethane is

46

 

 

 

 

 

 

 

 

 

 

0.001
0.0008 -
80.000633 . + -
3 it ' El I. + ll I
9 ll 0' DD Inu'lhn
"' t
0 I
‘6 0.0004 - ' D
5 re-
_ "fig.
0.0002 - than
‘ I
“tr.
0....I...rlrirriuiit..ttliuirriw,
0 1 2 3 4 5 6 7 8 9 10
Distance from fuel inlet (mm)
(a)
0.012 r
0.01 :-
0.008 :- u
.5 g ' .+
g r- ID-
..I: 0.006?
0
s c ._
0.004 f '_ .[3
Cl
0.002 - I I
- I
' I
2 3 4 5 6 7 8 9
Distance from fuel inlet (mm)

(b)
I§igure 4-12. Profiles of molar fractions of a) ethane and b) acetylene: Flame B (full black

fiymbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+).

47

an irnpun'ty in the fuel source, and it shows trends similar to ethylene, that is, its
maximum concentration occurs at the fuel boundary, its concentration decreases

throughout the flame zone, and the dopants do not make a noticeable contribution to its

concentrations in the baseline flame.

Acetylene, shown in Figure 4-12b, presents a unique behavior that differs in two
main ways fiom the non-sooting methane flame. First, the peak of the acetylene profile
in the sooting flames is nearly 19 times larger than the acetylene peak in the non-sooting
flanles, as expected in these sooting flames, since acetylene is a major surface growth
Species in soot, formation. Second, the profiles show a multimodal behavior as evidenced
by the continuous lines in the figure. At first glance this behavior may seem to be an
eI'KI'Jerimental artifact. However, it is reproducible. For Flames A-D, at least two sets of
experiments were conducted to confirm the reproducibility of data and this behavior was
01’served in all experiments. Also, if it were an artifact of the GC/MS performance or the

experimental technique, we would detect multimodal behavior in other minor species
Such as methane or ethane with much lower absolute concentrations than acetylene, but
We do not (see Figure 4-3 and Figure 4-12a). The results for ethane in Figure 4-12a at
diStances less than 3 mm from the fuel inlet indicate the approximate random uncertainty
in the measurements, which is 4 %; the changes in the acetylene concentration due to the
IIlilltimodal behavior are much larger, 20-50 %. Moreover, this multimodal behavior was
th observed in the non-sooting experiments using the same instrument, leading us to
Qanlude that the acetylene participation in soot growth pathways may cause this
15

ehavior. The first drop in acetylene profiles occurs at z: 3.5-4.5 mm. The peak of the

aQetylene mole fraction is at 2:: 5.5-5.8 mm for all flames and the bulk of the acetylene

48

profile occurs where ethylene is still present (ethylene disappears roughly at Z: 7 mm). It
is difficult to discriminate between the peak concentrations of acetylene in Flames A-D.
This suggests that acetylene, which is present at very high concentrations in these flames,

is mainly produced by ethylene.

Some important reactions that form single ring aromatics involve acetylene [56].
Also, acetylene is a particle surface growth species, as further discussed in the next
section. The first drop in the acetylene profiles (Figure 4-12b) can be attributed to the
early stages of formation of benzene and other large pyrolysis products through reactions
involving acetylene. Pyrolysis reactions of acetylene can form many products other than
benzene. This drop is right before the benzene peak. The second drop may be a result of
participation of acetylene in surface growth. Future studies with the use of Laser-Induced

Incandescence (LII) to measure soot would confirm this hypothesis.

4.6. Aromatics

Flames A-D were deliberately selected to be at the onset of soot formation to assess
the behavior of jet fuel and its surrogates with respect to the formation of critical
precursors in soot formation, such as aromatics. The non-sooting methane flames doped
with jet fuel and the Utah/Y ale surrogate revealed significant discrepancies with respect
to benzene and toluene [3 6], which was cause for concern in the soot context. The present
study considers also the Aachen surrogate (Flame D) that was found to reproduce

successfully soot volume fraction profiles of jet fuel flames [11].

Calibration data is available for the following aromatic species and, therefore, they

can be quantified:

49

1) Tetralin (1,2,3,4-Tetrahydro naphthalene, Clonz)
2) m-xylene (1,3-dimethyl benzene, Cngo)

3) p-xylene (1,4-dimethyl benzene, Cngo)

4) 1,2,4-trimethyl benzene (TMB, C9H12)

5) Benzene (C6H6)

6) Toluene (C7H3)

Some other aromatics and PAH’s (Polycyclic aromatic hydrocarbons) are important
as soot precursors. These may include styrene (ethenyl benzene) and some PAH ’s such as
naphthalene, anthracene, phenanthrene, pyrene, chrysene, tetracene, and coronene. Some
of these species can be identified by the MSD, but no calibration data is available for
their quantification. For instance, calibration of naphthalene (a critical bottleneck to soot)
poses significant difficulties. Identification of some larger PAH’s needs different

analytical approaches.

Figure 4-13 shows a comparison of three aromatics, namely, tetralin, m—xylene and
1,2,4-trimethylbenzene (TMB), among Flames B-D, with different ordinate scales in
Figure 4-13a and in its magnified version, Figure 4-13b. None of these aromatic species
were detected in the baseline ethylene flames (Flames A and A*). P-xylene can also be
detected and quantified in the doped flames, but it has much lower concentrations and
follows a profile similar to that of m-xylene in Flame D. The accuracy of the tetralin
calibration is uncertain and its profile is qualitative. In Flame B, tetralin, m-xylene, and

TMB were identified and quantified. These species are considered tracers of the

50

 

 

 

 

 

 

500
- I m-xylene
. o TMB
400 f. o A tetralin
‘ O
E ti . 0
o. . ° 0 .
3 300 ~
8 P D D C] .
.. ” D D
p 0 (HP
3 m '
*5 200 - ”a
s : a:
100 l a'.
in & AA AA A U .
AA “1b
as
0 .
0 1 2 3 4 5 6 7 8 9 10
Distance from fuel inlet (mm)
(a)
30
; I m-xylene
: A o TMB
25 f A A tetralin
A : ‘ A
g 20 _- A ‘ A
e )-
C E . Q A ‘
9: 15¢ ' 0 '
S _ O .. AA 0
“5 : A °
2 10 - X.
é I I I . I . .I ‘.. [g [9
II I
5 L- . I
P- i ‘
L .r I. §
0 t '
MW.

 

O
_s
N
(a)
k

5 6 7 8 9 10
Distance from fuel inlet (mm)

(b)
Figure 4-13. Comparison of profiles of m-xylene, tetralin, and 1,2,4-trimethylbenzene

with compressed (a) and expanded (b) ordinate scales: Flame B (full black symbols),
Flame C (open symbols) and Flame D (red symbols).

5]

multitude of aromatics in the jet fuel. They decay roughly at 2'2 5.1 mm. In Flame C that
was doped with the Utah/Y ale surrogate, m-xylene and tetralin are the only aromatic
components of the surrogate. Since these components represent a large group of
aromatics in the jet firel, the disagreement of the profiles near the fuel inlet, away from
the reaction zone, is not surprising. The same consideration applies to the aromatic

component of the Aachen surrogate.

In Flame D, which is doped with the Aachen surrogate, TMB is the sole aromatic
compound, which decays at z: 5.8-6 mm, which is close to where the only aliphatic
compound, n-decane, decays. A small peak of m-xylene (5-6 ppm) is observed in Flame
D at Z13 5.1 mm, which is close to the peaks of benzene and toluene. Xylene can be
formed from TMB by chemical pathways that replace a methyl side-chain with an H
atom. Very small quantities of styrene were detected within 2: 3.6-5.1 mm, but no
detectable tetralin and naphthalene were found in Flame D. Also, styrene and naphthalene
(not shown here) were detected in Flame C, within the regions of z: 3.3-5 mm and z:
1.6-4 mm, respectively. Naphthalene could not be measured closer to the reaction zone
because its GC/MS retention time is too close to that of dodecane. No naphthalene and
styrene could be measured anywhere in the jet fuel flame (Flame B) because of the

difficulty in species separation for jet fuel.

4.6.1. Benzene and toluene

The two aromatic compounds that could be measured cleanly in all of the doped
flames, and therefore that could be used as tracers of aromatics and soot formation, are
benzene and toluene. Figure 4-14 illustrates how these compounds are related to the

formation of the two-ring aromatic naphthalene, which is a critical bottleneck to soot

52

formation. Benzene is an intermediate in the growth of aliphatics (e. g.
methylcyloohexane, iso-octane, and dodecane) to naphthalene, and toluene is a byproduct
of the growth of n-alkylbenzenes (e.g., ethylbenzene, propylbenzene) to naphthalene via
benzyl radical. We focus on toluene instead of benzyl radical because we cannot detect

radicals.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Aliphatics n-alkylbenzenes V Isopropyl benzenes
Benzene , Benzyl radical Styrene
' I
I
' """"" ':
: Toluene .
---------a
v
Naphthalene
r
Soot

 

 

 

Figure 4-14. Global pathways of naphthalene formation. Isopropyl benzenes and species
with similar structure are rare in the fuels under study.

A comparison for benzene and toluene among the various flames is shown in
Figure 4-15 and Figure 4-16, respectively. Significantly, even in the case of the jet fuel
(Flame B), whose chromatograrns are the most difficult to analyze, benzene and toluene
can be measured with small uncertainties, since they do not have multiple isomers and
they appear early in the chromatogram, before the time when the “grassy” baseline arises
(see Figure 4-6). In fact, the GC/MS has a better performance in separating the jet fuel

intermediate species as 2 increases and the large fuel compounds and their isomers are

53

pyrolyzed and partially oxidized. Indeed, the profiles of carbon count in Figure 4-7
suggest that beyond 2 =1 5.3 mm, the total carbon counts are consistent between Flames B,
C and D and the comparison is quantitative. This region is in the vicinity of peaks of

benzene and toluene mole fractions.

Figure 4-15 shows good agreement between the surrogates and jet fuel with respect
to the location and magnitude of the benzene mole fraction peak. The doped flames
produce much more benzene than the two baseline ethylene flames (A and A*), which
means that maximum concentrations in the doped flames reflect aromatics formation

from the dopants. The agreement is better for Flame C (Utah/Yale surrogate) than for

 

160,
E .II-
140': -I DC]
I E] El.
. D
120;-
? : DD I'D
3 100E .r- a:
c C . [5
{=3 80: ' _
E : +
q’ 60; a +I"'-+ Cl
6 I I I E
2 h I
40; Iti' i.-
20} I- I
: E" q].
OiW-l-luuruuruu'filIIII l.

 

 

 

012 3 4 5 6 7 8 910
Distance from fuel inlet (mm)

Figure 4-15. Comparison of profiles of benzene mole fractions: Flame B (full black

symbols), Flame C (open symbols), Flame D (red symbols), Flame A (blue symbols) and
Flame A* (+).

54

Flame D (Aachen surrogate). Although the benzene profiles in the Aachen surrogate
flame (Flame D) and the jet fuel flame (Flame B) follow similar trends, the Aachen
surrogate produces nearly 20 % less benzene than jet fuel. The sooting jet fuel doped
Flame B produces nearly 2.5 times more benzene than the non-sooting jet fuel doped

methane flame [36].

Comparison of toluene mole fractions in Figure 4-16 shows reasonably good
agreement between Flame B and Flame C. Small amounts of toluene are present in the
parent jet fuel causing the mole fractions to be nonzero for small values of z. Toluene
was not detectable in Flames A and A*. The much larger toluene/benzene ratios in the
doped flames versus the undoped flames indicate that the toluene in the doped flames is
primarily formed from the aromatic components of the jet fuel and surrogates

independently of benzene formation. The agreement between Flame B and Flame D is

 

 

 

 

 

16_
14}
: I
122- 2.9
E : B]
‘1 10’- D'
3 3— '5‘!-
O
9 ' .
E 6- '- " D
o C]
2 ID.
.ICI- '.D
E? a
PP...L...,.....M........
3 4 5 6 7 8 9 10

 

Distance from fuel inlet (mm)

Figure 4-16. Comparison of profiles of toluene mole fractions: Flame B (full black
symbols), Flame C (open symbols) and Flame D (red symbols).

55

also reasonable, with nearly similar peak location, but the peak magnitude is almost 30 %
smaller in Flame D. These observations suggest that, despite its simplicity, the Aachen
surrogate has an acceptable agreement with jet fuel in incipient sooting conditions with

respect to aromatic formation.

In summary, it would appear that the discrepancies between jet fuel and surrogates
that were observed under non-sooting conditions [3 6] are not present in the current, and

more relevant, situation of incipient sooting.

4.6.2. Effects of component chemical class

Following the kinetic pathways from each surrogate component to benzene or
toluene is difficult. However, it is quite informative to compare the relative contribution
of each chemical class of components to the formation of benzene or other important soot
precursors. This would guide researchers to modify the formulation of the surrogates in
order to match the critical profiles of soot precursors and eventually the sooting behavior.
Table 4-1 presents the concentrations of each chemical class in the jet fuel and two

surrogates used. Each chemical class represents a molecular structure.

Two recent efforts on premixed flames by the Professor Sarofim’s group at
University. of Utah are reported here. As mentioned before, aliphatics are virtually
responsible in benzene formation. Studies of Zhang et al. [65] on premixed flames shed
light to the relative significance of each class of paraffins in benzene formation
suggesting a trend cyclo-paraffins > iso-paraffins > normal-paraffins. Specifically for the
constituents of the Utah/Y ale surrogate in a premixed flame scenario, the study shows

that methylcyclohexane produces much higher benzene concentrations than iso-octane

56

Table 4-1. Concentrations of each chemical class in the jet fuel and two surrogates.
Values are in vol%.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

“Average” jet
fuel from Wright-
Patterson Air Utah/Y ale Aachen
Force Base Surrogate Surrogate
(WPAF B), POSF
No. 4658
(n- and i-) paraffins 55.2 72.98 82.76
monocyclo-paraffins 17.2 13.6
dicyclo-paraffins 7.8
alkyl benzenes 12.7 9.78 17.24
indans and tetralin 4.9 3.64
naphthalenes and trycylco-paraffins 2
alkyl benzenes + indans + tetralin= 17.6 13.42 17.24

while iso-octane forms relatively more benzene as compared to dodecane. As a result, the
adjustment of relative concentrations of these cyclo, iso, and normal paraffins may

increase or decrease the total formation of benzene. Moreover, noteworthy is the recent
study of Zhang et al. [66] on 22 premixed flames of Cl-Clz fuels, which corroborates the

structure-dependency of benzene formation, for example, when a kerosene mixture, n-

decane and iso-octane were compared.

Extension of data available for benzene formation in premixed flames to non-
premixed flames (similar to Flames A-D) is not always accurate. One reason might be
different levels of H and OH radical concentrations. For non-premixed flames, only
general sooting tendencies, which have long been studied, are reported. It should be noted
that overall (global) measures of sooting tendency such as smoke point are not

representatives of the trends in benzene formation or early stages of soot formation.

57

Early studies on the effect of molecular structure of pure hydrocarbons on sooting
tendency in laminar difliwion flames show that the general sooting tendency trend for
hydrocarbons is [67]
n-paraffins < branched paraffins z naphthenes < alkenes < alkynes < alkylbenzenes <

naphthalenes.

This trend is not strictly followed because different hydrocarbon classes may
behave differently as carbon number increases. The sooting trend for individual

components in the Utah/Y ale surrogate is
tetralin > m-xylene > iso-octane > tetradecane > dodecane > methylcyclohexane.

When these pure hydrocarbons are mixed the general sooting tendency should be

assessed for the mixture.

The focus of this study was not on the final stages of soot formation. Instead,
we considered the decomposition and oxidation of fuels leading to the formation of small
aromatics that are important since they grow eventually to soot. Tentative trends for soot
formation using a global scale of sooting tendency are briefly discussed in the next

section.

4.6.3. Threshold sooting index ('1‘ SI)

To study the sooting tendency of surrogate fuels, Yang et al. [67] used the
threshold sooting index (T81) [68]. The TSI, which is an arbitrary scale for the sooting

tendency of hydrocarbon firels, for diffusion flames is defined by

TSI=a (MW/SP) + b, (4-1)

58

where MW is the fuel molecular weight [g/mol], SP is the fuel smoke point [69], and a
and b are constants dependent on the apparatus used for smoke point measurements (and
scale the TSI from 0 to 100). Ethane and methylnaphthalene are assigned the TSI ratings
of 0 and 100, respectively. Table 4-2 lists the TSI, aromatic content, H/C ratio, and cetane
number for jet fuel and surrogate mixtures. The TSI of the JP-8 is somewhat uncertain
since its composition varies, but values in the literature range from 15 to 26 [70]. We
have calculated the TSI of the Utah/Y ale and Aachen surrogates using the mixing rule

[67],
TSImix=inTSIi, (4-2)

where X, is the mole fraction of the individual Components. TSI values for pure

hydrocarbons are obtained from the work of Olson et al. [71]. The TSI of the Utah/Y ale

surrogate, which contains 13.4 (liq. Vol) % aromatics, is 14.6, while that of the Aachen

Table 4-2. TSI, aromatic content, H/C ratio, and cetane number for jet fuel and surrogate

 

 

 

 

 

 

 

 

 

mixtures.
TSImix Ag)??? H/C Ratio CNm,x
JP-8 15.72-25.66 17.86 1.909 31.8-56.8
J et-A 19.71-22.17 1 7.58 Not reported Not reported
Utah/Y ale surrogate 14.63 13.4 1.99 61.57
Aachen surrogate 15.11 17.24 2.017 65.22

 

 

59

 

surrogate which contains 17 .2 (liq. Vol) % aromatics, is 15.1. Aromatics account for
roughly 18 vol % of jet fuel [61]. By consideration of this global measure of sooting
tendency one would expect a tentative trend for soot formation: jet fuel > Aachen

surrogate > Utah/Y ale surrogate.

Some researchers [70] have utilized cetane numbers (CN) as representatives for

autoignition when various surrogate mixtures were compared. The mixture cetane

number (CNmix) is calculated using
CNmix=ZviCNi, (4'3)

where vi is the volumetric fi‘action of the individual components. CN values for pure

hydrocarbons (CNi) are taken from the work of Santana et al. [72]. Our study took

advantage of the detailed probing of the diffusion flame as well as spatial temperature
measurements. Indeed, this detailed flame structure study provided a wealth of
information that would better illustrate the ignition behavior of complex fuels as opposed

to the utilization of a global measure like the cetane number.

4.6.4. Possible role of the antistatic additive

A final issue to consider is the influence of trace amounts of sulfur compounds in
jet fuel (about 490 ppm) on the formation of soot precursors. The sulfur trioxide reaction
H+SO3—->OH+SOZ, in the late pyrolysis process of diffusion flames forms the hydroxyl

radical that attacks soot precursors and suppresses soot formation [57]. The good
agreement in aromatics behavior and the similarity of flame appearance under incipiently

sooting conditions suggest that the presence of sulfirr in the parent liquid plays a

60

negligible role. However, the situation is more complex in the present experiments
relying on the electrospray for fuel dispersion. The electrospray can operate successfully
so long as the fluid has a finite electric conductivity, which is not the case for
hydrocarbons. Consequently, we need to add 0.05% (by volume) of an antistatic additive,
Octel Stadis 450, to elevate the liquid electrical conductivity. This additive is also present
in some jet fuels (such as JP-8) as a static dissipater for in-flight refueling, at
concentrations not exceeding 80 ppm. It contains up to 30% in sulfur compounds [61,
73], which corresponds to a maximum of 0.018 % (by weight) in a surrogate mixture.
These sulfur compounds are present in an oil-soluble sulfonic acid, which is a constituent
of the Stadis. According to Henry [61, 73], a suitable sulfonic acid is mono and di
sulfonates of alky benzenes, and in one aspect is ' dinonylnaphthyl sulfonic acid or
doecylbenzene sulfonic acid. The Octel Stadis 450 that was used in the lab contained the
former. To ascertain if the additive played any role, we performed the following
experiment: we selected a flame with boundary conditions similar to Flame B except for
the dopant that was jet fuel at 3000 ppm, to ensure a more pronounced soot-laden zone.
In a separate experiment, the liquid fire] was spiked with 4-dodecylbenzene sulfonic acid
(C13H3OO3S, CAS Number: 121-65-3), yielding a five-fold increase in the sulfur
contribution as compared to that in jet fuel and from the Stadis. No distinct changes in the
soot zone luminosity (see Figure 3-4) and thickness were observed, which suggests that
the observed trends in terms of soot and soot precursors are not an artifact associated with

the antistatic additive.

61

Part II

Biogas Combustion and Chemical
Kinetics

62

Chapter 5 Background and Introduction to Biogas
Combustion Research

Increasing global energy demands along with limited resources and enviromnental
concerns, has led to the exploration of environmentally-friendly new and diverse energy
resources. Consequently, there is a challenge for combustors to use fuels with variable
compositions from diverse energy resources. There is an anticipated increase in the use of
renewable energy. In recent years, there has been a significant emphasis on the study of
production, commercialization, and combustion characteristics of bio-derivative fuels.
Biogas is a renewable and biodegradable energy source that can be used for the
transportation sector, farm communities, small-scale power generation, and large gas
turbines as a complementary fuel (e.g., to natural gas). Constraints like the cost of

cleaning and upgrading to remove C02, limit the use of biogas. Most newly built

stationary power-generating gas turbines use natural gas incorporating dry low NOx
technologies based on lean-premixed combustion to meet emissions regulations.
Therefore, the combustors operate under conditions near the lean limit of their stability
domains where flame blowotf, noise, and instability can adversely influence engine

performance and reliability [74].

There is an interest in using industrial fuel-flexible gas turbines near the alternative
fuel production sites which can produce their own fuel, for example farm communities.
The capability of new gas turbine technologies should be extended to a wider range of

potentially commercial, low quality fuels such as biogas. An acute issue in the

63

combustion of biogas as well as its integration with electricity generation that must be
investigated is varying gas compositions and heating values. Biogas composition depends
on many factors, particularly the management of the process and the nature of the
feedstock material. Composition variability places considerable challenges on the gas
turbine industry because low emission combustion systems are typically optimized to
operate with a single specific fuel (such as natural gas) with tight specifications. Also,
combustion instability problems become more serious in biogas-driven gas turbines as a

result of fuel composition and fuel-air ratio variability.

In this chapter, composition, production and pre-processing of biogas are
explained. The econorrric and environmental benefits of the utilization of biogas as a
renewable energy resource, as well as its potentials to replace fossil fuels for power
generation are discussed. The status of current gas turbines that use a wide range of fuels,
specifically biogas of different sources, is reviewed. Some challenges of fuel flexible gas

turbines utilizing biogas are addressed.
5.1. Biogas Production and Benefits

5.1.1. Biogas composition and upgrading

Biofuel is any fuel that is derived fi'om biomass. It is a renewable and
biodegradable energy source, unlike other natural resources such as petroleum, coal and
nuclear firels. One definition of biofuel is any fuel with an 80% rrrinimum content by
volume of materials derived from living organisms harvested within the ten years

preceding its manufacture [75].

64

Biogas typically originates from the anaerobic digestion of biomass and diverse
organic wastes by micro-organisms. Sources of organic waste include [76]: agricultural
wastes and manure, animal by-products, industrial wastes, municipal solid wastes,
sewage sludge and other biodegradable wastes. Different articles in the literature report
diverse constituents for biogas because each considers a particular organic source of
interest. The corresponding composition varies because of that specified source.
Wellinger and Lindberg [77] report a very complete description: Biogas produced in

anearobic digestion plants or landfill sites is primarily composed of methane (CH4) and
carbon dioxide (C02) with smaller amounts of hydrogen sulphide Gigs) and ammonia
(NH3). Trace amounts of nitrogen (NZ), hydrogen (H2), carbon monoxide (CO), oxygen
(02), and dust particles are occasionally present in the biogas, which is usually saturated

with water vapor. It may also contain siloxanest. Biogas is sometimes called swamp gas,
landfill gas or digester gas. It can be called renewable natural gas (RNG) when its

composition is upgraded to a higher standard of purity.

Constituents in biogas are sometimes similar to those of a raw (non-conventional)

natural gas. Raw natural gas leaving the well contains a series of other species [78]: (a)
C02 up to 25 vol. % (aged natural gas), (b) N2 up to 70 vol. % (weak natural gas), (c)
sulphur compounds: HZS (sour natural gas), (d) ethane, liquefied petroleum gas, and

condensates (rich natural gas). These constituents are usually removed before

 

*
Siloxanes (CxHySisz) are used in household and commercial products and find their way into
wastewater and landfill solid waste, thus, volatilize into digester gas and landfill gas. During combustion

SiOz is formed and found in the form of a white powder in gas turbine hot section components and as a

light coating on post—combustion catalysts. Microturbine and catalyst failures have focused industry-wide
attention on Siloxanes.

65

liquefaction or compression. Some gas turbines can burn a raw natural gas fuel. Similar

to a raw natural gas, biogas is a viable alternative candidate fuel for gas turbines.

In this work, the interest is in the study of biogas originating from anaerobic

digestion from the following sources:

i- Biogas of methanization: Anaerobic micro-organisms in the absence of oxygen
digest wet organic material (in a digester). Anaerobic digestion is also called

fermentation or methanization. It needs a warm environment to typical temperatures of at

least 30 oC.

ii- Landfill biogas: Landfills are a large source of biogas. Here, anaerobic digestion
of wastes occurs naturally. Gas collection is practical for landfills that are at least 40 feet

deep and contain at least one million tons of waste [79]. Anaerobic digestion of
municipal solid waste in this oxygen fiee environment primarily produces CH4 and C02.
Despite the fact that producer gas or syngas, produced fiom slow pyrolysis or gasification
of organic wastes or wood (with main constituents, H2 and CO), is sometimes named
biogas, our work is solely focusing on biogas from biomass anaerobic digestion or

landfill gas.

Scott et a1. [80] report an average composition for biogas produced in Dairy
Anaerobic Digestion (AD) Plants, which is CH4 = 60.3%, C02 = 38.2%, N2 & 02: 1.5%,
H28 = 1984 ppm. Figure 5-1 summarizes the methane content of landfill gas extracted

from various types of landfills and current available energy recovery technologies.

66

Methane Content (%) in Landfill Gas (LFG)
A

10%< Old Closed Landfills > 35%

25% <ecently Closed Landé9 40%
m m.

1.5% = Fuel Flexible Microturbines : Up to -———> 1M%
38% <—Non Flexible Microturbine: —/
40%4— IC Engines ———v’

 

(

 

 

 

 

 

 

Figure 5-1. Landfill gas characterization by methane content (%) and technologies gas
operating ranges.

Biogas composition strongly affects the thermo-chemical properties of the fuel/air
mixtures (as will be discussed in the following chapters of this thesis). For instance, inert"
gases like N2, C02 and steam reduce the lower heating value (LHV) of the fuel. Biogas in
general has a small to medium LHV. Figure 52 adapted from [81], shows LHVs of some
alternative gas turbine fuels. The LHV of the waste methanization biogas (averaging at
about 20,000 kJ/m3) is less than natural gas and slow pyrolysis biogas (producer gas)
[81].

Corrosive impurities and small LHVs, compared to natural gas, make biogas
unsuitable to be compressed or injected into a pipeline system without prior cleaning and
upgrading. As a result, for many biogas applications, cleaning and upgrading are
required; Cleaning foremost implies the separation of water, hydrogen sulphide, and

particles. Upgrading means removal of C02 to raise the LHV of the fuel [82]. After a

 

t
It will be discussed later that these so called inert gases participate in the reactions to some extent.

67

50000
40000
30000
20000
10000

LHV (lem3)

 

Figure 5-2. LHV’s of alternative gaseous firels compared to natural gas adopted from
81].

considerable amount of cleaninyupgrading, biogas (sometimes called biomethane) can

be used in any application that consumes methane as a fuel [82].

It is of great interest to utilize biogas-driven gas turbines close to biogas production
plants, which produce their own biogas fuel. Constraints on gas turbine efficiency,
operability, and emissions should be met with as minimal biogas cleaning/upgrading as
possible. This objective has been limited by some technical and economic obstacles.
According to Sarmaa [83], the construction of digesters and the transportation of biomass
(especially for large-scale biogas plants) are costly. The upgrading is associated with high
compression powers (in most upgrading methods). According to Wellinger and Lindberg
[77], at present four different methods are used commercially for removal of carbon
dioxide from biogas (upgrading) either to reach vehicle fuel standard or to reach natural

gas quality for injection to the natural gas grid. These methods are:

68

1) Water absorption or scrubbing (with associated compression to about 10 bar and
pmnping),
2) Polyethylene glycol (selexol) absorption or scrubbing (similar to water scrubbing),

3) Carbon molecular sieves (often called pressure swing adsorption or PSA with

compression up to 6 bar),

4) Membrane separation (includes high pressure gas separation with compressions up to
36 bar). Gas-liquid absorption membranes [77] developed recently for biogas

upgrading work at atmospheric pressures.

5.1.2. Biogas benefits

The biogas possesses three general benefits: i) fossil fuel displacement; ii)

economic development; iii) environmental.

Renewable bio-based fuels such as biogas will reduce the society’s dependence on
non-renewable resources such as coal, petroleum, and fossil fuel-derived natural gas.
Biogas has been used for applications including heating, lighting and power generation.
Among biofuels the benefits of biogas are similar to those of natural gas: increased
energy security and improved public health and environment through reduced emissions

[76].

Treatment of organic waste is an “enviromnentally—friendly” way to produce
biofuels. The stabilized solid residue can be used as a soil conditioner material (compost),
a nutrient-rich liquid fertilizer, or as fodder. The growth of the dairy industry has been
limited as a result of difficulties to manage the livestock manure. Bio gas facilities provide

more flexibility in manure disposal, lowering the need for commercial fertilizer, and

69

lowering odor [83]. A study in Denmark [83] has shown that conversion of animal waste
to biogas will provide added value to manure as an energy resource and reduce
environmental problems associated with untreated animal wastes. Biogas production also

creates jobs and benefits the rural economy.
Burning biogas is a means for reducing greenhouse gas (GHG) emissions. It will be
elaborated in Chapter 7 that burning biogas reduces the net C02 release and prevents CH4

release to the atmosphere (CH4 is a much stronger greenhouse gas than C02). Thus, it is

also a means to satisfy various national and international ecological constraints such as

the Kyoto protocol. As an example, Gerin et al. [84] showed that there is a net reduction
in C02 and a net production of renewable energy as compared to fossil fuels, when maize
and grass are used as the energy crops to feed anaerobic digesters. Borjesson and
Berglund [85] analyzed fuel-cycle emissions of C02, CO, nitrogen oxides (NOX), sulphur
dioxide (80;), hydrocarbons (HC), CH4, and particles fi'om a life-cycle perspective for

different biogas systems based on different digestion technologies and six different raw

materials. Borjesson and Berglund [85] suggest that the overall environmental impact of
biogas systems depends largely on the status of uncontrolled losses of CH4, the end-use

technology that is used, the raw material digested, and the energy efficiency in the biogas

production chain. Combustion of bio gas can reduce the flame temperature, which will
reduce NOx emissions because the thermal NO mechanism (at combustion temperatures
roughly above 1800 K) is the main pathway of NOx formation [86]. Conversion of

animal waste to biogas reduces water pollution by using feedstock that would end up in

rivers and lakes and reduces pollution of ground water as well [83]. For landfills, it

70

reduces the cost of complying with US. Environmental Protection Agency landfill gas

combustion requirements and reduces the amount of material that must be landfilled [87].

5.2. Global utilization of biogas

A study group commissioned by the German government shows a potential of up to
35% of total energy consumption by Germany of biofuels by 2030 [88]. At the end of
2006 about 3500 biogas plants with a total capacity of 1100 MW have been operating in
Germany. Energy crops are the main substrate: manure has less than approximately 50 %
share. In 2006, cultivation of non-food crops in Germany was 13.2% of German arable
land [89]. Denmark is also one of the most advanced countries in biogas technology.
Biogas plants in Denmark were well developed until 1998, but their expansion was

slowed down afterward [83].

Sweden is presently a leading nation in biogas research and uses it for both
transportation (trains, garbage trucks, automobiles) and power generation. In the last five
years, Sweden has focused on biogas and bio-ethanol development [88]. Biogas can be
used as a firel in vehicles specially adapted to methane gas. According to Persson, [82]
biogas has been used in Sweden as a vehicle fuel in large scale systems for buses and
other vehicles since 1996. Before using biogas as a vehicle fuel, cleaning and upgrading

of the gas is needed to avoid corrosion and mechanical wear, and to meet quality
requirements of gas applications. The requirement for CH4 content is 97d:2 vol. %. In

2007, there were over 30 upgrading plants in operation or in the construction phase in
Sweden, and during 2006, 54 % of the gas delivered to vehicles was biogas [82]. Biogas

as vehicle fuel is given more and more interest worldwide and last year both Germany

71

and Austria set up national targets of 20 % biogas in the fuel sold to vehicles [82]. Most
of the above technologies need to convert biogas to a quality suitable for either vehicles
or pipeline distribution, which imposes tight standards of upgrading/cleaning at least to a

conventional natural gas quality.

In some European countries, government incentives (subsidies) support the biogas
industry, which may not be economically feasible for other countries. There is a lack of
comprehensive datasets in the literature demonstrating how much cleaning and upgrading
is required for power generation by gas turbines in an economically, environmentally
viable manner. The following sections address the latest status of technology and

research relevant to burning biogas as an alternative gas turbine fuel.
5.3. Biogas potential as an alternative gas turbine fuel

5.3.1. Fuel-flexible gas turbines utilizing biogas

Worldwide demands for low emission and fuel flexible gas turbines are increasing.
Gokalp et al. [81] report the preliminary progress of the major European project AFTUR
(Alternative fuels for industrial gas turbines), begun in February 2003, in the selection
and characterization of potential liquid and gas, alternative fuels for industrial gas
turbines. However, their report is limited to the LHVs and C/H ratios for gaseous
alternative fuels including biogas. Moliere [90] provided comprehensive information as
well as an essential technical description of alternative gas turbine fuels ranging from
volatile firels, weak gas fuels extracted from the coal/iron industry, ash-forming oils,
hydrogen-rich byproducts from refineries or petrochemical plants, liquid biofuel (e.g.,

ethanol), and biogas.

72

Studies are needed to assist governmental and private agencies in determining the
types and characteristics of combustors most suitable for bio gas power generation. For
example, for the California Energy Commission, using a computer code developed for

simple-cycle gas turbines, Yomogida et al. [91, 92] estimated that the thermal efficiency

of a 125 kW Solar gas turbine (Titan© series) for biogas power generation is about 14%.

They conducted a sensitivity analysis of the operating conditions and design parameters
with the greatest influence on thermal efficiency. Some studies have focused on
performance analysis and technological assessment of industrial gas turbines utilizing
small LHV fuels. Palmer and Erbes [93] suggested a modeling approach and simulated a
GE PG6541B gas turbine for both natural gas and low-BTU biogas fuel. Hanagudu, [94]
in a more general study, reviewed the advantages, technical considerations and
applications of biogas-fired industrial gas turbines with focus on the improvement of the
electricity cogeneration for Indian cane sugar distilleries with mandatory treatment units.

The biogas obtained was traditionally fired in boilers to generate steam.

5.3.2. Biogas-driven micro-gas—turbines (MGT)

Some current Micro-CHP (Combined Heat and Power or Cogeneration)
installations use micro-gas-turbines (MGT) technology, which can be powered with
different fuels and will provide on-site power and thermal energy with low emissions and
low maintenance requirements. Some studies have focused on the use of biogas-driven
micro-gas-turbines (MGTs) because biogas fuel applications may represent the best

market for microturbines. At some facilities such as large dairy, sewage or wastewater

73

treatment plants, the use of anaerobic digesters and rrricroturbine cogeneration systems

provides opportunities for reduction of both waste/manure treatment and energy usage.

In the U.S., the first rrricroturbine tested on a biogas was a Capstone 30 kW unit at
the Palmdale wastewater treatment in Los Angles County [95]. MGTs require the biogas
to be compressed to about 6 atrn and require the compressed gas to be dried (i.e., no
moisture) [95]. Krautkremer et al. [96] state that MGTs are able to cope with up to 70
ppm H2S, a value never attained in biogas. Waste heat available in the exhaust can be
used, for example, in absorption chillers. Goldstein [97] mentions that Capstone Turbine
Corporation's latest offering is the 65 kW nricroturbine that uses waste flare gases from
landfills (landfill gas) or wastewater treatment plants to create renewable energy. At

multiple sites, Capstone experienced siloxane induced turbine failures. As a result,
Capstone has established a fuel specification that requires less than ~ 0.03 mgr/m3 of

siloxane [98]. Ingersoll Rand (IR) markets two microturbine units referred to as the
MT70 series with a capacity of 70 kW of continuous on-site electrical power and the
MT250 series with capacity of 250 kW. Ingersoll-Rand commissioned the construction of
a firel mixing facility to test these commercial MGTs while operating on diluted
alternative fuels such as biogas. Dolak and Armstrong [99] provide an overview of this
fuel mixing facility, its capabilities, and a discussion of diluted fuel compositions
typically found in field applications. They demonstrate that pollutant emissions are
closely tied to both the magnitude and the composition of diluents. Biogas production is
high in the hot season, whereas the thermal energy demand is highest in the cold season.
Naing et al. [100] investigated the performance of a MGT cogeneration system in cold

regions and the applicability of such a system to a sewage treatment plant which produces

74

biogas from anaerobic digestion. They examined biogas hydrate formation as a cost-
effective biogas storage solution to adapt to the delay between peak energy supply and
peak energy demand with simulations of integration into an existing cogeneration system.
Bohn and Lepers [101] overviewed low-BTU fuels suitable for utilization in gas turbines
and investigated biogas utilization for a model integrated microturbine plant. Using a
combustion model based on a systematically reduced 6-step reaction mechanism, they
analyzed the change of emissions of NOx and CO. They also investigated the effects of
low-BTU fuels (as compared to natural gas) on gas turbine materials and gas turbine

efficiency.

5.3.3. Premixed turbulent biogas combustion

Experimental and theoretical studies dealing with flame structure of turbulent lean
premixed flames are mostly limited to specific fuel compositions and they cannot be

transferred to gas turbine combustion with variable composition gases such as biogas.

Ultra-low emissions targets have been set for gas turbines with NOx emissions <5
ppm (@ 15% 02). A recent technology that can be used for low-emissions biogas-driven

gas turbines is the Low-Swirl Injector (LS1). Johnson et al. [102] investigated a new
approach for low emissions gas turbines by modifying a typical production high swirl
injector [103] for gas turbine combustors to operate in a novel low-swirl stabilization

mode. This LSI was investigated using PIV at elevated combustion inlet temperatures
(230 < To < 430 °C) and pressures (6 < P0 < 15 atrn). It was reported that the LSI emits

NOx levels about 60% lower than from the HSI with no effect on C0. According to

Johnson et al. [102], the lack of a large dominant strong recirculation zone and the shorter

75

residence time within the LS1 may provide an explanation for the NOx reduction.

Recently, Littlejohn and Cheng [104] continued the earlier work of Johnson et al. [102] to

investigate the fuel effects for seven diluted and undiluted fuels including a 50%CH4-
50%C02 fuel (representing biogas). The LSI was shown to be capable of supporting

stable hydrocarbon flames that emit <5 ppm NOx (@ 15% 02) with CO well below

acceptable limits. The LSI did not need to undergo significant alterations to operate with
the different hydrocarbon fuels used in their study. It was demonstrated that NOx
emissions of biogas are lower, and CO emissions are higher than methane and propane

flames of the LS1 with increasing q).

Few fimdamental studies (experimental and numerical) provide data for lean
premixed turbulent bio gas combustion in gas turbine combustors at typical conditions,
e. g. temperature and equivalence ratio ((p) of inlet flow, and combustor pressure. Lafay et

al. [86] , under the AFTUR [81] project, compared the stability combustion domains,
flame structures and dynamics between CH4/air flames and biogas/air flames in lean
premixed combustion conditions of gas turbines. Their experiments were conducted with
five gas compositions: CH4, a biogas (issued from waste methanization) with CH4 = 61,
C02 = 34 and N2 = 5 (vol. %), three C02 diluted-methane-air flames with 12, 20, and 30

% by volume of C02. For the same (p, the addition of C02 implied strong modification of

the reaction zone location and the reaction intensity. In addition, the C02 showed a

stabilizing effect on pressure fluctuations. It was found that the main parameter to predict

the flame structure is the laminar flame speed which depends on both (p and biogas

76

"U

I")

'II

composition [86]. For this reason, laminar flame speeds of some biogas mixtures are

calculated in this thesis (see Chapter 6).

Plasma Assisted Combustion (PAC) and/or H2 enrichment could allow burning in
ultra lean conditions[105]. Leung et al. [106] showed that the stability of a biogas non-
premixed jet flame is enhanced sigrrificantly by introducing 10 (vol. %) H2 into the fuel
stream. More investigations, especially on premixed flames, are needed to understand the

PAC and H2 enrichment effects.

Further experiments in large-scale gas turbines with real biogas fuels (extracted
from various biogas plants) can provide quantitative suggestions to modify the biogas
composition and the combustor design. Moreover, combustion instabilities should be

investigated for a wider range of compositions for biogas.

5.4. Discussion: Biogas research objectives

Gaseous alternative fuels can be produced by a variety of techniques, but attention
here has been focused on biogas originating from anaerobic digestion. The desirability of
utilizing the biogas near its production site was discussed, including the alignment of

such physically proximate uses with the favorable attributes of gas turbines. It is desirable

to expand the flexibility of gas turbines to handle a fuel with impurities (e.g., HZS,

particles and Siloxanes) and diluents (e.g., C02 and N2). Biogas has been demonstrated as

a viable, practical alternative fuel that is relatively commonly used in Germany and the
Nordic/ Scandinavian countries for transportation and also for stationary power
generation. Cleaning and upgrading of biogas fuel is necessary for using it as a vehicle-

grade fuel: despite this requirement, numerous plants are in operation and a significant

77

fraction of North-European heavy-duty vehicles use biogas for vital transportation needs.
Gas turbines, which are widely (and ideally, or optimally) used for stationary power
generation, would seem to be a good candidate for the use of replacement (biogas) fuel.
However, gas turbines can be fuel-sensitive and variations of relative percentages of
constituent combustibles can alter performance characteristics negatively. For this reason,
basic research is needed. One specific application occurs in what are called micro gas
turbines (MGTs), for which some preliminary research has been conducted. Much of this
research is empirical and non-systematic (i.e. trial-and-error): not many systematic
studies of biogas fuel in MGTs have appeared. Applied research (consisting mainly of
modifications to existing equipment) has been carried out but studies are few, possibly

because modification of existing systems is expensive.

Technical research issues of practical concern include the production of soot, NOx,
and green house gases and the possible influences of biogas C02 content on emissions

and engine operation. These issues warrant the attentions of applied research (i.e., modify
turbines and then test with various fuels). They also warrant the focus of basic research
on fundamental combustion processes occurring in the combustors of such turbines. This
study presently turns toward addressing the fundamental question of biogas combustion
characteristics and chemical kinetics. The objective will be to compare and contrast
various basic metrics of biogas combustion with those of established, commonly used
fuels (e.g., natural gas). Both premixed and non-premixed configurations will be utilized
for this study in the subsequent chapters. Non-premixed scenarios, although not common
in the current stationary gas turbines, are of significance in some other devices burning

blends of diluted hydrocarbons (e. g., biogas). Laminar counterflow flames are commonly

78

used to study chemical kinetics and reaction sets when both chemistry and transport are

important (as used in Part I).

79

Chapter 6 Chemical Kinetics Study of Simulated
Biogas in Premixed Combustion

The goal in this chapter is to study chemical kinetics of diluted hydrocarbon blends

with focus on fuel blends that represent a biogas. Diluted hydrocarbon blends may be
found in biogas combustion diluted mostly by C02, exhaust gas recirculation (EGR)
diluted by combustion products, [107] and SOF Cs (solid oxide firel cells) operating on
biogas or diluted methane [108-110]. Other applications of the research on C02 dilution
include C02 recovery in semi-closed gas turbines with recirculating C02 and topping
combustors [111]. Some studies [112-114] show that C02 diluted oxy-fuel combustion

for pulverized fuels facilitates C02 separation and eventual sequestration. Therefore, oxy-

fuel combustion involves high proportions of C02. Also, the addition of C02 and H20

through flue gas recirculation (FGR) reduces oxides of nitrogen [115].

6.1. Biogas as a diluted hydrocarbon blend

Biogas can be simulated as a diluted hydrocarbon blend assuming that H2S and

particles have been separated and it does not contain NH3, H2, 02, and CO. For the
understanding of the impact of fuel blends on combustor operability, traditional

engineering design parameters such as heating value, the Wobbe Index: and autoignition

are ultimately insufficient [116]. The cumulative thermal, transport and chemical effects

 

I
The Wobbe Index (W1) is the main indicator of the interchangeability of gaseous fuels such as natural gas.
It is frequently defined in the specifications of gas supply and transport utilities. If HHV is the higher

heating value, and S0 is the specific gravity, the Wobbe Index, W7, is defined as WI = HHV/ JSG . For
example, the WI value for methane is 12,735 (kcal/m3).

80

of the constituents combine to produce a completely different behavior in blends
compared to individual constituents. Properties like ignition delay times may behave in a
highly nonlinear manner as fuel mixtures change. It is important to gain more
understanding of the relationship between firel mixture composition and properties such

as flame speed and ignition and chemical delay times.

In this chapter, fuel blends include diluted methane mixtures whose main

constituents may include CH4, C02, N2, and H20 (constituents similar to water saturated

biogas from digestion without drying). The consequences of C02 addition to hydrocarbon
fuels have been studied in both premixed and non-premixed configurations. Several
studies have demonstrated that C02 is not inert and participates in reactions [117-125].

Studies in non-premixed configurations will be reported in the next chapter.

Liu et al. [120] have numerically investigated the chemical effects of C02
replacement of N2 in air on the burning velocity of lean to stoichiometric CH4/
COz/Olez and H2/COz/Oz/N2 mixtures at 1 atrn. They focused mostly on the chemical
effects of the addition of (:02 to CH4 indicating that the addition of (:02 to air or

replacement of N2 in air by C02 may affect the burning velocity through three
mechanisms: 1) The variation of the transport and thermal properties of the mixture, 2)

The possible direct chemical effect of C02, and 3) The enhanced radiation transfer by
C02. The reaction C02+HHCO+OH was shown to be the most important reaction
associated with the direct chemical participation of C02 in these premixed flames.

Hermann et al. [126] investigated the effect of four diluents (H20, C02, N2 or Ar) on the

81

NOx formation rate in premixed laminar natural gas and hydrogen flames. They
addressed that, in certain conditions, the rate of formation of NO is reduced by the

addition of any diluents at constant combustion temperature.

In this chapter, some detailed and reduced mechanisms are explained, compared,
and evaluated (e.g., GRI-Mech 3.0 and San Diego detailed mechanisms as well as GRI-
Mech 1.2 reduced mechanism) in regard to their applicability for a biogas chemical
kinetics study. For the evaluation of reaction mechanisms before their utilization in 2-D
or 3-D CF D modeling or to find laminar flame speeds and ignition delays, a counterflow
diffusion flame configuration, which is a fluid-dynamically simple and tractable l-D
configuration, is chosen. Gas phase equilibrium calculations are then performed to find
adiabatic flame temperatures of biogas flames with Variable compositions. Ignition delay
times and laminar flame speeds are also calculated. Moreover, this study provides

correlations to predict the behavior of individual diluents in biogas.

6.2. Gas phase chemical kinetics mechanisms

6.2.1. Standard-state thermodynamic properties in CHEMKIN

The molar heat capacity at constant pressure for species k is defined using

arbitrary-order polynomial fits,

C° k M _

_Rf_ : Z“kak(m 1) (6-1)
m=l

82

The superscript 0 refers to the standard-state, which is an ideal gas at 1 atmosphere

for gas-phase species. The standard state molar enthalpy and entr0py can be obtained

from following equations:
H; = ITK €12de + H]: (0) S]: = ITK EflidT + S]: (0) 6-2)
0 298 T

The CHEMKIN Gas-Phase Kinetics package is designed by default to work with

thermodynamic data in the form used in the NASA chemical equilibrium code [127]. In
this case, seven coefficients, a1-7, are needed for each of two temperature ranges. These

fits take the form in equations (6-3) to (6-5), where the temperatures are in Kelvin.

O

C

 

k 2 3 4 _
% = alt + ‘12ka + 03ka + a4ka + a5ka (6 3)
Hk =alk+flTk+flTk2+fliTk3+flTk4+flL (6'4)
R7), 2 3 4 5 Tk
%zalklnTk+£12ka+%Tk2+-a:—ka3+a—Z-k-Tk4+a7k (6'5)

Accounting for the appropriate pressure and entropy-of-mixing terms, the entropy is

Sk = S; —Rln Xk —Rln(P/Patm) (6-6)

Thus, mixture-averaged entr0py, Gibb’s and Helmholtz free energies are found. The

mixture-averaged entropy is

83

K
S = 2(5). -RlnXt -R1n(P/Patm))Xk (6-7)
k=1

The therm.dat file provided with CHEMKIN version 4.1 includes seven coefficients for

each of two temperature ranges and has 778 species.

6.2.2. Detailed, reduced and skeletal reaction mechanisms

Some well-examined methane (or natural gas) reaction mechanisms that can be

utilized for biogas flames are introduced in this Section.

1- GRI-Mech 3.0 mechanism

GRI-Mech 3.0:. (which replaced versions 1.1, 1.2 and 2.11) from the Gas Research
Institute (GRI) is a compilation of 325 elementary chemical reactions and associated rate
coefficient expressions and thermochemical parameters for the 53 species involved. It
differs from the previous release in that kinetics and target data have been updated,
improved, and expanded. Propane)r and C2 oxidation products have been added and new
formaldehyde and NO formation and rebum targets included [128]. It has nitrogen

chemistry with 106 reactions belong to the nitrogen chemistry part.

 

*
The Gas Research Institute discontinued its support of the GRI-Mech project in February 2000 after
which no further development continued.

T 4 species have been added to 49 species of GRI-Mech 2.11. Acetaldehyde (CH3CHO) and vinoxy
(CHZCHO) chemistry are included to better describe ethylene oxidation, and this path is included among

the Ox + CzHy reaction products. C3H7 and C3H3 also added as well as a minimal set ofpropane kinetics
to model propane, as a minor constituent only.

84

GRI-Mech 3.0 has been optimized for methane and natural gas as fuel and includes
NO formation and rebum chemistry. Reactions that are involved in the combustion of
other hydrocarbon constituents of natural gas (e.g., ethane and propane) are also included.
However, since the optimization did not include targets relevant to other fuels, GRI-Mech
should not be used to model combustion of pure methanol, propane, ethylene, and

acetylene, even though these compounds are on the GRI-Mech species list [128].

The conditions for which GRI-Mech 3.0 has been optimized, limited primarily by
the availability of reliable optimization targets, are roughly 1000 to 2500 K, 10 Torr to 10
atm, and equivalence ratio fiom 0.1 to 5 for premixed systems [128]. In this work, GRI-
Mech 3.0 is used inside its optimization and validation ranges. The soot formation for
natural gas is not described by GRI-Mech 3.0 as so the chemistry involved in selective
non-catalytic reduction of NO, which may be important in natural gas reburrring at lower
temperatures. Thermodynamic data is provided from GRI-Mech 3.0 Thermodynamics

released 7/30/99 based on NASA Polynomial format for CHEMKIN-II.

2- San Diego (SD) detailed mechanism

The San Diego mechanism [129] (Release 2005/12/01)* provides a detailed
chemistry to focus on conditions relevant to flames, high temperature ignition and in the
case of supersonic combustion, detonations. It is composed of 6 elements including N, H,
O, C, Ar and He, 46 species, and 235 reactions. Helium is an addition to the GRI-Mech

3.0 mechanism, which includes only the first 5 elements. Pressure dependent reactions

 

*
In the latest version of the San Diego Mechanism (released 2005/ 12/01), reactions related to combustion

of ethanol were added to the existing San Diego Mechanism. Most of those reactions were adopted from
the Ph.D. thesis of Juan Li [130]. Thermodynamic and transport data for the additional species associated
with ethanol have also been added.

85

are described by the TROE formulation [131]. All reactions except one

(CH3O+02—->CHZO+H02) are reversible and their backward rates can be calculated from

the forward rate and the thermodynamic data.

The San Diego mechanism does not include nitrogen chemistry and NOx
formation. However, the group offers a NOx mechanism (Release 2004/12/09) to enable
researchers to model NOx in flames. This nitrogen chemistry is based on the previous
work by Hewson and Bollig [132]; however, this mechanism has not yet been evaluated
thoroughly. One reaction (HCCO+NOHHNCO+CO) has been added based on the GRI
2.11 rates [133]. In the revision of March 2005, the pre-exponential in this reaction has

been revised based on the evaluation by the San Diego Group. This value has been

modified from 2.35E+13 to 2.00E+13 [(mol/cm3)-1/s].

Uncertainties in rate parameters of many steps can increase the uncertainties in the
predictions when large numbers of steps are included. Also, it has been understood from
experience that the rate parameters of a relatively small number of elementary steps are
of crucial importance to the predictions. Figure 6-1 shows a distribution of scaled
reaction rates for the 235 reactions of the San Diego mechanism at stoichiometric

methane/air adiabatic flame temperature (Tad=2224 K). To scale values at T, each

reaction rate is divided by the maximum reaction rate at T (RRmax) and then multiplied

by e. An investigation of this figure renders that only 7 reactions are in the range of (5%
- RRmax - RRmax) and only 23 reactions are in the range of (2% RRmax - RRmax). In a

sense, these reactions represent the global thermo-chemical behavior of the flame.

Cumulative effects of small contributions from a large number of steps are seldom of

86

much significance (also seen in Figure 6-1). Therefore, if the numbers of species and
reactions are kept as much as possible to the minimum needed to describe the systems,
uncertainties in the rate parameters employed are minimized. This statement
characterizes the philosophy underlying the San Diego mechanism. It differs from a
number of other data bases, many of which seek completeness, attempting to include all

potentially relevant elementary steps [129].

 

Ln( R.R.IR.R.max*e 1

 

I
‘
(a,
--—-——--—-

 

 

 

Reactions

Figure 6-1. Distribution of the scaled reaction rates for San Diego mechanism reactions at
Tad of stoichiometric methane/air mixture (Tad: 2224 K).

3- Reduced GRI-Mech methane mechanism
A reduced mechanism [134] which includes 19 species (+N2 and Ar) and 84

reactions (released 1994) is also used*. This mechanism has been reduced from GRI-

 

I
This mechanism can also be considered a skeletal mechanism with respect to very smaller reduced
mechanisms.

87

Mech 1.2 by the group who developed GRI-Mech mechanism. This mechanism,
however, does not include nitrogen chemistry (the detailed GRI-Mech 1.2 mechanism did

not include a nitrogen chemistry).
4- San Diego skeletal and reduced mechanisms including N Ox Chemistry

San Diego group offers a skeletal mechanism describing hydrocarbon and nitrogen
chemistry, which includes 65 steps for the Cl and C2 hydrocarbon chemistry plus 55
steps for the nitrogen chemistry. This mechanism has been derived for non-premixed
combustion. Some reactions relevant to premixed burning have been left out (i.e.
CH4+O=CH3+OH). Also, some reactions related to ignition chemistry have been
neglected. Hewson and Bollig [132] derived a reduced 5 step hydrocarbon mechanism
from the above mentioned skeletal mechanism to model non-premixed methane
combustion. A 6 step reduced mechanism for nitrogen chemistry in diffusion flames has
also been offered. Reduction attempts continued to a one-step reduced NOx mechanism

which is N2 + 02 —>2 NO.

6.3. Evaluation of reaction mechanisms

In this section, a counterflow diffusion flame with methane as fuel is modeled and
compared to a benchmark to evaluate and compare the reaction sets. The flame structure
study of Sung et al. [135] at Princeton is the benchmark for comparison. In the next step,
after validation of both the computational model and the reaction set, this flame system

will be employed to study simulated bio gas mixtures.

The CHEMKIN based numerical code OPPDIF [136], originally developed by Kee

et al. [137] for premixed opposed-flow flames, is employed. The governing equations are

88

found in Kee et al. [137]. Radiative heat loss is not considered. The transport and

thermochemical properties are obtained from the CHEMKIN database [125].

Fuel enters from the origin of the coordinate system. Calculations are performed
with one atmosphere pressure and 300 K upstream temperatures. The burner separation,
L, is 13 mm. A plug flow boundary condition is selected for this study (Sung et al. [135]
demonstrate that despite the differences in the plug flow or potential flow boundary

conditions, velocity and temperature fields are identical within the thermal mixing layer).

Two computational cases of interest have velocities of VF=V0=45 crrr/s and VF=VO=25.5
cm/s, which correspond to strain rates of a=56 s'1 and a=42 s", respectively. The

stoichiometric mixture fraction, zf, for these flames is 0.3. The fuel and oxidizer streams

 

 

 

 

 

 

0.3 1800
t :
3 J 1600
0.25 - .
E— _______ \ .3 1400
t: 0.2 _- - g
g _ J 1200 a,
E 015 P 1 1%
L; ' - J 1000 6
—- - Q.
o - l E
2 r - a:
0.1 :- - 800 *-
; :
. - 600
0.05 r j
i 3 400
- I 4
0 I l I l J L I l _ . L l I I
0 0.2 0.4 0.6 0.3 1 1.2

Distance from Fuel Nozzle (cm)

Figure 6-2. Temperature and major species profiles for (23%CH4-7 7%N2)/ Air flame at 1
atrrr, a =56 s'l (GRI-Mech 3.0 mechanism).

89

consisted of 23% methane in nitrogen and 23% oxygen in nitrogen, respectively, both by

volume. Figure 6-2 plots the temperature and major species profiles when the GRI-Mech

3.0 mechanism is used.

Figure 63 presents a comparison of two computed temperature profiles corresponding to
two strain rates case a) a=56 s'I and case b) a=42 5'1 (two inlet velocities) and the
experimental data for a=56 s'' [135]. The full width at half maximum (F WHM) criterion

is used for the flame thickness, 5. The flame thickness is approximately 0.21 cm for case

a (higher strain rate) and 0.29 cm for case b. Temperature profiles become narrower when

the strain rate is increased.

1 800

 

 

1600- ————— - eeeeeeeeeeeeeeeee .-, v=45cmrs --

9 Exp. Data 1-

}[ —V=25cmls l
1400s..-.m--. .................. l

 

1200 .

   
 

1000 —

800 —

Tam peramre [K]

000 -

400 —

 

 

 

 

 

200 - I l
0 0.2 0.4 0.6 0.8 1 1 .2

Distance from Fuel Nozzle (cm)

Figure 6-3. Computed temperature profiles for (23%CH4-77%N2)/Air flame at 1 atm,
with GRI-Mech 3.0 mechanism: a) VF=V0=45 cm/s (line, computed), b) VF=VO=25.5
cm/s (line, computed) and c) VF=VO=45 crrr/s (solid squares, experiments of Sung et al.
[135])

90

Comparison of the temperature and major species profiles with those of
experimental and computational study of Sung et al. [135] shows a good agreement in the
oxidizer side 'and very good agreement in peak points. A slight offset from the
experimental data points is observed in the fuel side. According to Sung et al. [135], the
characteristic flame thickness should scale withJa— , that is 6 x J— : Const. An order of

15% deviation is found for this correlation based on the calculated values fiom Figure

6-3. These slight deviations may have two reasons:

1) Uncertainty of the velocities in the experiments leading to slight mismatching of

strain rates between the benchmark and computations.

2) The effect of the underlying chemistry and reaction mechanism.

The Princeton group used a C2 scheme consisting of 29 species and 152 elementary

reactions [135]. The effect of reaction mechanism selection on the offset in the fuel side
is investigated by using the detailed San Diego as well as the reduced GRI-Mech 1.2
mechanisms. The San Diego mechanism produces similar profiles as the GRI-Mech 3.0

mechanism with negligible difference in major species, a slight shift in the temperature
profile, and a small difference in temperature peak (ATmax~7 K, 0.4 %). As seen in
Figure 6-4, the reduced mechanism shows the same behavior for the temperature (close
values for Tm“). A similar behavior is also observed for the major species (not shown
here). However, Figure 6-5 (profiles of H radicals) shows that the reduced mechanism
overpredicts the mole fraction of some important radicals. H radicals are chain branching

radicals which affect flame speeds, which we intend to predict. Our goal in this chapter is

lean premixed flame modeling. Thus, the San Diego skeletal mechanism, which includes

91

1725

1700 -. ~

1650

Temperature [K]

1625 1

1 600

1675 1

 

—Sen Diego
— GRI-mach 3.0

— Reduced Gil-
mech 1 .2

 

 

 

 

  
  

 

 

 

0.6 0.62

0.64

 

r

0.66 0.7

0.68

Distance from Fuel Nozzle (cm)

Figure 6-4. Comparison of computed profiles of temperature near the peak point for

(23%CH4-77%N2)/Air flame at 1 atm (a =56 s") with GRI-Mech 3.0, reduced GRI-Mech
1.2, and San Diego mechanisms.

Mole Fraction

 

 

 

  
   
 

 

 

 

 

 

0.0018 .
7 —San Diego ll
0.0016 ~
—G!l-Mech 3.0 3
0.0014 « ]
—GRl-Mech 1.2 l
4
0.0012 Reduced ;
0.0010 .
0.0008 -
0.0006 ~
0.0004 .
0.0002 4
0.0000 . . .
0.55 0.6 0.65 0.7 0.75 0.8

Distance from Fuel Nozzle (cm)

Figure 6-5. Comparison of computed profiles of H radicals near the peak point for
(23%CH4-77%N2)/Air flame at 1 atm (a =56 s") with GRI-Mech 3.0, reduced GRI-Mech

1.2, and San Diego mechanisms.

92

65 steps for the C1 and C2 chemistry plus 55 steps for the nitrogen chemistry, is not

employed here because it has been derived for non-premixed combustion.

6.4. Carbone dioxide diluted methane/air counterflow flames

A configuration similar to the configuration in Section 6.3 is selected to study
methane diluted counterflow flames. Only a case similar to case (a) above with
VF=V0=45 cm/s at 1 atm pressure is considered, but, unlike the previous problem, fire]
stream is not diluted with nitrogen and air containing 21% 02 enters from the oxidizer
side. Computations are performed for four fuel mixture compositions, mixtures 1 to 4
summarized in Table 6-1. GRI-Mech 3.0 kinetic mechanism is employed. Figure 6-6
illustrates profiles of temperature and major species for four compositions of simulated
biogas flames (mixtures 1-4).

Figure 6-7 and Figure 6-8 show, respectively, profiles of H radical and CH4 mole
fi'actions for four compositions (mixtures 1-4) in the thermal mixing layer. Since the
breakdown of CH4 is mainly due to its attack by OH and H radicals [138], the peak
values of OH and H occur on the oxidizer side of the flame. When the C02 proportion in
the fuel stream increases, the peak of H reduces and the peak locations move toward the
oxidizer side. This suggests that the mechanism with which C02 influences the flame
structure is as follows: When C02 concentration rises respect to CH4 concentration (from

20% to 50%, respectively), the competition of reactions like C02+H—>CO+OH with the

chain branching reactions like H+Oz—+H+OH reduces the concentrations of important

radicals i.e. H, O, and OH present in the thermal mixing layer. When the total amount of

93

these radicals is less, CH4 is attacked less and can penetrate more towards the oxidizer

side while consuming radicals and moving their peak locations. This shift in radical peaks

is not significant as compared to the shift due to flame stoichiometry changes resulting

from C02 dilution of the fuel stream.

Table 6-1. Selected compositions for simulated biogas used in the computational study.

 

 

 

 

 

 

 

Biogas mixture No. Simulated biogas blend composition (Vol. %)
1 50% CH4-50% 002
2' 60% CH4-40% co;
3* 70% CH4-30% C02
4 80% CH4-20% €02
5 50% CH4 + 16.6% €02 + 16.6% N2 + 16.6%H20
6it 61% CH4+34% C02 +5 %N2

 

 

 

 

It is of interest to conduct a scaling analysis for the diluted hydrocarbons, here

methane. Due to the highly nonlinear behavior of species, this is a hurdle in the current

study. Attempts are focused on C02 profiles based on the results of counterflow flame

study. [C02] is scaled by the [COz]1=, which is the molar concentration of the C02 in the

 

*

This composition will be used in Chapter 6 as a typical for biogas from anaerobic digestion (AD) plants
used in dairy industry without upgrading.

Compositions in mixtures 3 and 4 represent biogas from AD plants or landfills with “partial” upgrading.

I A typical biogas issued from waste methanization reported by Gokalp and Iebas

[81] used also in the experiments of Lafay et al. on lean premixed turbulent biogas combustion at CNRS-
LCSR, France [86]. This composition will be used in this chapter to examine equilibrium state, ignition
delays, and laminar flame speeds.

94

2000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-l
.. __ T 3
0.8 . _ ...... 02 1 1800
i ....... H20 3 1600
_ — — — - CH4 :
0.6 - CO -j 1400 Q
g - CO2 : v
-— ~ 9
g -_ 1200 3
LT. ' 3 ‘°
. 6
2 0.4 '- -_' 1000 Q,
o . - E
2 q o
- 1 800 *-
0.2 - g '''''''''''''' '5 600
- -] 400
o r m r r I r r r r I r .L— ((1 .r r r l r r r r l r r: 200
0 0.2 0.4 0.6 0.8 1 1.2
Distance from Fuel Nozzle (cm)
a) Fuel mixture #1: 50% CH4-50% C02
, 2000
.. __ T E
0.8 - _ ______ 02 1 1800
h """" H20 3 1600
_ _ _ — .. CH4 :
0.6 — ——————— \ -— CO -j 1400 g
g - \ C02 3 v
._ \ - e
g . \ -1 1200 3
LT. \ 3 ‘3
2 0.4 \ -j 1000 8
é’ i ‘ E
g m
l. \ -_ 800 *-
i \ 3
0.2 - , ... '''''''''''''''' ‘-j 600
r- "‘ '/
.t i
i \ /'\ 1 400
I, \ / \‘\ :
0‘ \ .1 l I l l l 1 l I l 1‘ 200
0 0.2 0.4 0.6 0.8 1 1.2

Distance from Fuel Nozzle (cm)

b) Fuel mixture #2: 60% CH4-40% C02

Figure 6-6. Temperature and major species profiles for four compositions of C02 diluted
methane flames (P=l atm, VF=VO=45 crrr/s, GRI-Mech 3.0 mechanism).

95

Figure 6-6 Cont’d

Mole Fraction

Mole Fraction

0.8

0.6

0.4

0.2

0.8

0.6

0.4

0.2

 

 

 

   
      

 

 

 

 

 

 

  

 

 

 

 

 

l 1.1
0.2 0.4 0.6 0.8 1 1.2
Distance from Fuel Nozzle (cm)

(1) Fuel mixture #4: 80% CH4-20% C02

96

, 2000
T E
_ ______ 02 -_ 1800
"""" H20 : 1600
‘ ___- CH4 :
CO {1 1400
C02 1
Q 1200
-j 1000
—j 800
................ _: 600
-j 400
rrrrlrrrrla” rlr_. rrrlrlr :200
0.2 0.4 0.6 0.8 1 1.2
Distance from Fuel Nozzle (cm)
c) Fuel mixture #3: 70% CH4-30% C02
2000
T a
" —————— \\ _ ______ 02 1 1800
\ """" H20 3 1600
\\ ____ CH4 :
CO -j 1400
C02 3
—j 1200
a: 1000
-j 800
-—'-‘-'-'-"'-'-€ 600
-j 400
' ' ' 200

Temperature (K)

Temperature (K)

 

0.0040

 

— 50% 002

0.0035 4
0.0030 -

0.0025 ~

 

 

 

0.0020 1.

0.0015 1

[H] Mole Fraction

0.0010 -»

 

 

 

 

0.0000 " A ’ r r
0.6 0.65 0.7 0.75 0.8

Distance from Fuel Nozzle (cm)

Figure 67 Profiles of H radical mole fractions for four compositions of C02 diluted

methane flames in thermal mixing layer (P=l atm, VF=V0=45 cm/s, GRI-Mech 3.0
mechanism).

 

 

 

 

 

 

 

 

 

 

—50%c02 ‘ - _
—40%coz -— ~
:9: —30% 002 - -
0
a _ _
a: _- _ -._ 20% 602 J
a
'5'
I
2
0.6 0.65 0.7 0.75 0.8

Distance from Fuel Nozzle (cm)

Figure 6-8. Profiles of CH4 mole fraction for four compositions of C02 diluted methane
flames in thermal mixing layer (P=l atm, VF=V0=45 cm/s, GRI-Mech 3.0 mechanism).

97

fuel mixture (e.g. for the mixture of 60% CH4-40% C02 , [COZ]F=0.4). Figure 6-9 shows

the profiles of scaled CO2 for four compositions selected. Variations have a similar trend

upstream of the thermal mixing layer, but trends are different within the mixing layer

(x=0.6-0.75 cm).

 

 

 

 

 

Scaled C02

 

 

 

0.0 r Y i
0.4 0.5 0.6 0.7 0.8 0.9

Distance from Fuel Nozzle (cm)

 

fl

Figure 6-9. Profiles of scaled C02 for four compositions of C02 diluted methane flames
in a counterflow configuration (P=l atm, VF=VO=45 cnr/s, GRI-Mech 3.0 mechanism).

6.5. Water vapor and nitrogen influences in biogas combustion

It is informative to investigate the relative contribution of diluents that can be
present in a biogas i.e. C02, H20 and N2, on the flame temperature, NO emissions and
important radicals. This investigation can guide to understand the alteration of biogas

flame characteristics due to the presence of water vapor (or occasional N2 content). For

this study, the configuration in Section 6.4 is utilized with VF=V0=45 cm/s at 1 atm. The

98

011C

370‘.

,.
1.x“

b—Y"

oxidizer is air containing 21% 02. Two biogas mixtures are compared: 1) A
CH4/COz/N2/HZO mixture consisting of 50% CH4 and equal proportions (16.6%) of C02,
N2, and H20 (Vol. %) designated as biogas mixture 5 in Table 6-1; 2) biogas mixture 1

(50% CH4-50% C02) that was modeled before. Both mixtures have the same CH4

proportions. GRI-Mech 3.0 kinetic mechanism is employed. Figure 6-10 presents the

computed temperature, major and minor species profiles for fuel mixture 5.

 

 

 

 

 

 

 

 

 

0.9 _ 2000
0.8 E- ..................... .1 1800
0.7 {- T 1 1600
E 02 3
c 0.6 : H20 1 1400 g
o _ , _
.5 _ , - o
8 0.5 _— ——————— . CH4 1 1200 5
t: : CO : ‘6
g 0.4 E 002 1 1000 8.
2 E --------- N2 3 as:
0.3 :- 1 800 '-
: i
0.2 :- . ______ . ......... '5 600
0.1 [— 1 400
0 n r L r r l r r r r I ‘L r i r I r r r r I r L: 200
0 0.2 0.4 0.6 0.8 1 1.2

Distance from Fuel Nozzle (cm)

(a)

Figure 6—10. Temperature, major species profiles (a) and minor species profiles (b) for
biogas mixture 5 (P=l atm, VF=VO=45 cnr/s, GRI-Mech 3.0 mechanism).

99

Figure 6-10 Cont’d

 

      

 

       
 

 

 

   

 

0.005 1 2000
— 1
_ 1 1800
0.004 - :
I H E
C l. -_ 1400 Q
.9 0.003 r ; I;
E Z l 1200 3
LL _ I S
0) 1 O)
— ‘ — o.
0 0.002 - 1 1000 E
2 _ _ a:
- : '-
1 1 800
0.001 _: 600
,- 1 400
0 r _] L-v'l” I . a I 1 “‘31. I r
0.5 0.6 0.7 0.8 0.9
Distance from Fuel Nozzle (cm)
(13)
6.0E-05 ——————1 1~—— .___--,_-.A__ 2.5E-03
505'” i 2.05-03
:
o G
3 405-05 1 , g
E 1.5E-03 g
2 3.0E-05 ~ 1;
g 1 105-03 3
0 205-05 ~ , 5
z :
1.05-05 » 5‘05“
0.0E+00

 

 

 

 

0.0E+00
‘ 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Distance from Fuel Nozzle (cm)

Figure 6-11. Comparison of NO and H radical mole fractions: biogas mixture 1 (black
lines); biogas mixture 5 (blue lines).

100

Figure 6-11 shows a comparison of NO and H mole fractions between two biogas

rrrixtures. For fuel mixture 1, Tmax=1825 K and for fuel mixture 5, Tmax=l844 K. The
comparison in Figure 6-11 indicates that for a fixed hydrocarbon content (CH4=50%),
replacement of C02 by N2 and H20 increases the NO emissions and H radical

concentrations. This suggests that, among these diluents, C02 is relatively stronger in the

reduction of temperature, NO and H. The reduction of H leads to lower flame speeds in
premixed scenarios. Thus, it is anticipated that the burning velocity of fuel mixture 5 is

bigger.
6.6. Biogas premixed flames: Results and discussion

6.6.1. Equilibrium calculations for biogas/air combustion

Gas phase equilibrium calculations are conducted to find the adiabatic flame
temperature and equilibrium species. GRI-Mech 3.0 reaction mechanism is employed.
Initial conditions for the mixtures are T=298 K and P=1 atm similar to those usually

reported in combustion literature. Two gaseous mixtures are compared: a) pure methane:
CH4=100%, and b) A typical biogas issued fi'om waste metharrization [81] with a
composition CH4=61%, C02=34%, N2=5 % (Vol. %) which has been used in the work of
Lafay et al. [86]. The later mixture was designated as biogas mixture #6 in Table 6-1. As
illustrated in Figure 6-12, the adiabatic flame temperature of this typical biogas mixture is

lower than methane in the whole lean to rich range of (p. This corroborates the statement

made in the introduction of Part II.

101

Figure 6-13 presents profiles of H and OH radicals as well as emissions of NO and
CO. For the lean conditions in methane and biogas combustion, CO is negligible while it
rises significantly at rich conditions with more CO emissions from biogas. Methane
produces more NO emissions than biogas (mixture 6) because the flame temperature that
affects thermal-NO formation is higher in the methane flame. H and OH radicals, which
are important in chain branching reactions, have also been compared. The abundance of
both radicals is greater for pure methane. However, the difference is more pronounced for
H radicals; i.e. the effects of diluents addition to methane can be relatively better

understood via the tracking of H radicals as opposed to OH radicals.

 

2300

2100 .

if?”

 

Adiabatic Flame
Temperature [K]

170° ‘ —o—cr-wAir

- - -I- - - Biogas (61 %CH4 34%002 5%N2) [Air

iii

 

 

g
o
- «I
u .Q.

 

0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

Equivalence Ratio

Figure 6-12. Adiabatic flame temperature of CH4 and a typical biogas (mixture 6) in lean
to rich equivalence ratios.

102

 

0.14 d

0.12 ~

0.10 -

Xi, Mole Fraction

0.02 1
A

 

--0— CH4: [CO]

-----o BIOGAS:[CO]
—-e— CH4: 20*[NO]
m... BIOGAS:20*[NO]
+ CH4: 200*[H]
-----I BIOGAS:200*[H]
+ CH4: 20*[Ol-l]
-----I BIOGAS:20*[OH]

 
 
 
 
 
 
 

 

 

0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Equivalence Ratio

Figure 6-13. Equilibrium mole fractions of H, OH, NO and CO for CH4 and a typical

biogas (mixture 6) in lean to rich equivalence ratios.

6.6.2. Chemical kinetic modeling of biogas ignition

Gas turbine industry has been widely utilizing premixed combustion technology.

Understanding the ignition process for gas turbines is vital because autoignition of firel

and oxidizer mixtures prior to reaching the main burner may enhance pollutant emissions

and damage combustor components. Experimental techniques including constant-volume

bombs, continuous flow devices, and shock tubes have been utilized to measure ignition

delay times (fig). Comparison of computational predictions of ignition delay times to

shock tube experiments (usually by using reflected shocks) is often used in the validation

and testing procedure of detailed chemical kinetics mechanisms and provides an

understanding of the underlying chemistry. For instance, Petrova and Williams [139]

103

systematically applied the technique for propane, propene, allene, and propyne to test a

small detailed chemical kinetics mechanism for hydrocarbon combustion.

Ignition delay characteristics of methane fuels have been comprehensively
reviewed by Spadaccini and Colket [140]. In a fuel blend, variation of combustible
constituents and diluents will change the ignition process. This is elucidated in an

empirical correlation proposed for the ignition delay times of natural gas [140, 141]
rig = 1.77 x10"14 exp(l8693 / T)x [021‘1-°5[CH4]°-66[HC]'°-39, (6-8)

in which concentrations are expressed in mole/cm3. [HC] represents the total molar

concentration of all non-methane hydrocarbons (contaminants or additives including
ethane, propane and n-butane) present in small levels in a natural gas blend. This

correlation is valid for temperatures between 1300 and 2000 K, pressures from 3 atm to

15 atm and <p=0.43-1.25. The overall pressure dependence is P'0'78.

To our knowledge, there is no empirical correlation for the ignition delay times of

typical biogas blends extracted from landfills or digestion plants, which include high
concentrations of diluents (e.g., 40-60% C02). Here, we examine a chemical kinetic
model for ignition delay times of a simulated biogas blend: A typical biogas with a
composition CH4=61%, C02=34%, N2=5 % (Vol. %) which was designated as biogas
mixture #6 in Table 6—1. GRI-Mech 3.0 kinetics mechanism that has performed well in
the prediction of the counterflow diffusion flame structure in this chapter and in the CH4

oxidation modeling is employed.

104

The prediction of ignition delay times is performed using CHEMKIN based on one
of these criteria: 1) A time based on the maximum of certain species concentrations (e.g.,
the peak of the OH or CH profile); 2) A specified rate of increase of temperature (e. g., the
maximum of dT/dt). Ignition times are calculated for a closed homogeneous chamber (a
0-D problem free from transport effects) with uniform pressure. The energy equation is
solved for specified initial pressures and temperatures. Here, two criteria are used for the

ignition delay times
1) The inflection point in the temperature profile (max. dT/dt),
2) The maximum mole fraction of the OH radicals.

Therefore, for each initial condition two values are obtained. The temperature and
OH profiles must be carefully investigated to realize the correct ignition times in case

more than two values are obtained.

Although it is known that tig decreases for both increasing temperature and

increasing pressure [140], the ignition behavior of biogas mixtures for varying pressures
and temperatures should be carefully examined. The ignition delay times of biogas
mixture 6 using the maximum OH radical criterion are delineated in Figure 6-14 when
the mixture temperature is increased from 1000 K to 2000 K. Pressure is maintained at 1
atm and the fuel-to-air equivalence ratio is constant and equal to l. Biogas ignition delay

times are higher than methane (not shown here) i.e. biogas ignition is slower.

Table 6-2 (plotted also in Figure 6-15) summarizes peak gas temperatures and the
ignition delay times (using both criteria introduced before) of biogas mixture 6 at

different fuel-to-air equivalence ratios, (p, fixed atmospheric pressure and fixed initial

105

1 .0307

 

1.0E+06 1
1.0305 1
1.0304 -
1.0303 1
1051-02 —

Ignition delay (us)

 

1.0301 1

 

 

1 .OE-I-OO . t t
4 10

 

6 8
(1044111 [1IK]

Figure 6-14. Ignition delay times of biogas mixture 6 at 1 atm and <p=1 (The peak of the
OH profile criterion was employed).

Table 6-2. Ignition delay times of biogas mixture 6 for different equivalence ratios at l

 

 

 

 

 

 

 

 

 

atm and 1000 K.
Tin::ll= 11%;) K Thermal ignition time [sec]
.. ”Farrier“ W .53:
criterion crrterron [K]

0.6 0.8863 0.8864 2153
0.7 0.9646 0.9647 2271
0.8 1.040 1.040 2362
0.9 1.114 1.114 2425

1 1.185 1.185 2462
1.1 1.256 1.256 2472
1.2 1.325 1.325 2453

 

 

 

 

 

106

 

 

   
 

A

I

N
1
l
I
l
l
l
l
l
l
l
l
I
l
l
l
l

..-_-,._

Ignition Delay Time (sec)

 

 

 

1.1 ...... _______
1~ ————————————— ;. ——————— E- ------------- ir -------
0.0 ----- . ——————— ....... ...... | ....... _______
0.3 i i i i i
0.6 0.7 0.8 0.9 1 1.1 1.2

Equivalence Ratio
Figure 6-15. Ignition delay times of biogas mixture 6 at various equivalence ratios with

P=1 atm, and Tinitial=1000 K.

Table 6-3. Ignition delay times of biogas rrrixture 6 at various pressures and fixed
equivalence ratio and initial temperature.

 

 

 

 

 

 

 

 

 

 

 

 

 

—1
T (11:1 (300 K Thermal ignition time [sec]
initial -
Pressure . Temperature Peak of the OH Peak gas
[atm] rnflectron pornt (max. ro file criterion temperature
dT/dt) criterion P [K]

1 1 .097 1 .097 2462
2 0.5494 0.5494 2493
3 0.3437 0.3437 2510.
4 0.2457 0.2457 2522
5 0.1896 0.1896 2531
6 0.1538 0.1538 2538
7 0.1291 0.1291 2544
8 0.1111 0.1111 2549

 

 

107

 

 

1.2

 

 

 

«1 2550
1 d g
—> «e 2530 3’
48 h
3 0.8 1 g
a, - 2510 3
>4
'2 06 5
0 , -1
'0 .- 2490 *5
c :1
g 3
g, 0'4 ‘ + 2470 a
8
G
O
0.2 i 1— 2450 ‘L
0 . . l . 2430
0 2 4 6 8 10

Pressure (atm)

Figure 6—16. Ignition delay times and peak gas temperatures of biogas mixture 6 at
various pressures, (p=1, T=1000 K.

temperature of 1000 K. A nearly linear increase is seen for the ignition delay times of

biogas with respect to the increase of (p.

The effects of raising chamber pressure on ignition delays of biogas are seen in

Table 6-3. The initial temperature is 1000 K and the mixture is stoichiometric. Values are
also plotted in Figure 6-16. An overall pressure dependence P4“ is found from this

figure for biogas mixture 6. Future work will model ignition of biogas at high
temperatures. The same modeling procedure is applied for various biogas mixtures, but

results are not reported in this thesis.

108

6.6.3. Laminar flame speed of biogas/air mixtures

In this section, laminar flame speeds, SL, of simulated biogas mixtures are

calculated. A comparison of our computed laminar flame speed for methane using
CHEMKIN and GRI-Mech 3.0 mechanism (for adiabatic conditions) with the most
referred experimental and computational data in the literature is presented in Table 6-4.
In this computation, the initial temperature and pressure are 298 K and 1 atm with (0:1.
Our computed value lies well in the range of data fi'om experiments and computations

using other kinetic schemes.

Table 6-4. Comparison of calculated laminar flame~ speed for methane with the data in the

 

 

 

 

 

 

 

 

 

 

literature.
No. <p=l, P=1 atm, adiabatic conditions (q”=0) SL [cm/sec]

1 CH4/Air: our computed value = VCH4 39.61

2 CH4/Air: Egolfopoulos et al. 1990, 1991 40 2
Experimental data [142] '

3 CH4/Air: computed by Egolfopoulos et al. 1992 [142] 39.3

4 CH4/Air: Egolfopoulos et al. 1990 computed with a 39 2
complete C3 scheme [143] °

5 CH4/Air: Egolfopoulos et al. 1990 computed with a 39 5
complete C2 scheme [143] '

5 Biogas mixture 5 [61% CH4 +34% co; + 5% 26.26=
N2]/Air, our computed value z 0-66 x VCH4

 

 

109

A comparison of the computed results for biogas mixture 6 with calculations of
Lafay et al. [86] using a different code (CANTERA) and the GRI-Mech 3.0 mechanism

shows negligible difference (not reported here). The calculated downstream (products)
temperatures from the solution of the governing equations for the CH4/Air and biogas
mixture 6/Air blends are 2230 K and 2113 K, respectively, which are very close to the
adiabatic flame temperatures (Tad=2224 K and Tad=2114.4 K, respectively), confirming
the assumption of adiabaticity in calculations. The discrepancies of SL values are
negligible suggesting that the computational technique as well as the reaction mechanism
can be employed for an extensive investigation of biogas laminar flame speeds at various
compositions.

Figure 6-17 depicts the calculated laminar flame speeds for CH4/C02 mixtures with

C02 proportions (in Vol. %) varying fiom 0 to 50. SL decreases as the C02 percentage is

increased; a quadratic fit is found (although it seems at a first look that the graph is

linear)

SL=-24.0233x [cog/2-2 7.201 Ix [C02] +39. 5432. (6-9)

As illustrated in Figure 6-18, the concentration of H radicals also decreases as well as the

downstream (product) temperature. Liu et al. [120] found that the competition of C02 for
H radicals through the reaction CO+OHHC02+H with the most important chain
branching reaction H+02HO+OH plays a significant role that reduces the overall rate of
combustion. As a result, laminar flame speeds or flame thicknesses are affected by C02

dilution.

110

Laninar Flame Speed [cuisec]

 

 

 

 

1 8 20 22 24 26 28 30 32 34 36 38 40
1 *lr: i I fil 1 i 17+ i i I I; fir I i 17 " o
0.95 . : : : : : : : : )I’ ~ 0.05
r: 0.9 4 I‘ I I i I I 'I‘ I ~ 0.1
o I I I I I I I l I
'3‘ 0.85 4 I 'i I I l L I 5 i ._ 0.15
o I . I I I I I/ I I
E 0.8 1 f : l ; ,0} . ; ; ~~ 0.2
l I I ‘ l l I I I I
2 0-75 . l : t : : ,g I f : r 0-25
o I l l I I I | l l |
2 0-7 * :* t : : ,x.‘ 1 : : : r 0-3
3*: 0.05 « i I 4 Jr 1 1 ‘ i i — 0.35
i i ' z’.’ i I i i I
o 0.6 ‘ I I ”b | I I I I I F 0.4
0.55 I t ’1’ : : : : : : : r 0.45
0.5 P ‘ ' ‘ ‘ ' 0.5

C02 Mole Fraction

Figure 6-17. Effect of C02 concentration on the laminar flame speed of CH4/C02
mixtures simulating biogas.

[H] Mole Fraction

 

0.0004 ~

0.0003 -«

0.0002 --

 

0.0001

2200

 

 

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
[002]

2250

I 2000

T [K]

Figure 6-18. Effect of C02 concentration on the concentration of H radicals and
downstream temperature for CH4/C02 mixtures simulating biogas.

lll

It is also of interest to investigate the effect of nitrogen content when the biogas

methane content remains constant. Figure 6-19 demonstrates that with a fixed CH4 mole
fraction, when the N2 mole fraction increases with respect to C02, the laminar flame

speed increases; a quadratic curve fits well, which has the equation

SL=9.2424x [N212 +16.3597x [N2] +24. 8265. (6-10)

The replacement of C02 by N2 will increase the downstream temperature (not shown

here).

 

0.4
0.35 ~~ ;
0.3 .
0.25 .
0.2 -
0.15 I
0.1 ~
0.05 -
0

C02 Mole Fraction
N2 Mole Fraction

 

 

 

 

24 25 26 27 28 29 30 31 32 33
Laminar Flame Speed [cm/sec]

Figure 6-19. Relative effectosz and C02 concentration on the laminar flame speeds of
CH4/C07] N2 mixtures with constant CH4 concentration, [CH4]=O.6.

112

In the calculations presented so far, the initial temperature was 298 K and the
pressure was 1 atm. Some correlations have been reported in the literature, which relate
the laminar flame speed with the temperature and pressure. For methane/air mixtures,

Kobayashi and co-workers [144] report one such correlation fi'om [145],
SL = SL0(T/TO)’"(P/P0)" (6-11)

where P and T are the pressure and temperature of the mixture, and To, Po, and Sm are,

respectively, 300 K, 0.1 MPa, and laminar flame speed at these conditions as reference
values (very close to our calculated values at 298 K, 1 atm). Kobayashi and co-workers
report that the temperature exponent, m, and pressure exponent, n, used are 1.9 and -0.5,
respectively, based on the experimental data [144] and numerical analysis. The accuracy

of such correlations to predict the SL for blends of diluted hydrocarbons is not certain and

needs more investigation.

113

Chapter 7 Thermal and Chemical Structure of
Biogas Counterflow Diffusion Flames

In this chapter, the thermal and chemical structure of biogas counterflow diffusion

flames is studied [146]. C02 diluted methane blends, which are typical biogas blends, are

of interest to this study. The chemical and thermal influences of the addition of C02 as a

diluent are investigated in a counterflow diffusion flame configuration. The
computational investigation uses a CHEMKIN [125] based numerical code. Two detailed
reaction mechanisms and a reduced mechanism are evaluated and compared to
experiments [36]. Four non-sooting biogas diffusion flames are compared to

quantitatively examine thermal and chemical effects of C02 content of biogas on flame

temperature, major species, minor species, and emissions of NOx, soot and green house

gases.

Most studies on the effects of diluents in non-premixed flames include either

counterflow or coflow configurations with diluents added to either the fuel or oxidizer

streams. Rortveit et al. [147] studied a H2 counterflow diffusion flame diluted in the fuel

stream with N2, C02, or He. All three diluents reduced the flame temperature. C02, the

most efficient diluent, reduced the temperature largely through dissociation. The heat

capacity of a diluent also plays a significant role in altering the temperature. Recently,

Berry Yelverton and Roberts [148] measured the soot surface temperature in He, Ar, N2,

and C02 diluted ethylene jet diffusion flames. Among these diluents, the effects of heat

114

capacity as well as small contributions fiom radiative heat loss made the C02 diluted
flame the coolest. It is known that C02 directly participates in chemical reactions
primarily through the elementary reaction C02+H—>CO+OH [117, 118, 149] and to a

lesser extent C02+CH—>CO+HCO [119]. Earlier studies have shown that C02 dilution of
fire] in a diffusion flame can suppress soot and NOx [150, 151]. The reaction

H+NO—+HNO was important in the destruction and formation of prompt NO [123].

In diffusion flames, addition of a diluent to a fuel may produce four categories of
effects: [120, 151] (1) transport (or dilution) effects by reduction in concentration of
reactive species and thus their collision fiequencies; (2) thermal effects by changing the
flame temperature; (3) radiative effects by changing the radiation transfer fi'om the hot

products; and (4) direct chemical effects by participation of diluents in chemical
reactions. For C02, attempts have made to isolate transport, thermal, radiative, and

chemical influences. Two methodologies have been proposed in the literature: (1) The
technique used in the experiments of Du et al. [151] to isolate dilution, thermal and

chemical effects by comparing results of a pure ethylene flame and three diluted ethylene

flames: a C02 diluted flame, a C02 diluted flame with the same temperature as pure fuel,
and an N; diluted flame with the same temperature as pure fuel. Controlling the flame

temperature was achieved by the adjustment of Ar and N2 concentrations in the oxidizer
stream; (2) The methodology proposed by Liu et al. [119] to numerically identify

chemical effects of C02 addition to ethylene. They identify a normal C02 that
participates in reactions and a fictitious C02 that has exactly the same thermochemical

and transport prOperties as the normal C02, but is chemically inert. The radiative

115

properties of the normal and fictitious C02 were added for similar studies by Park et al.

[121] who implemented this method.

Biogas may have variable compositions and low qualities. A drawback with biogas
combustion, in general, is its small LHVs and diminished flame temperatures. Upgrading
costs and heat release reduction are sometimes masked since the environmental

advantages are well highlighted; the reduction of NOx emissions, for instance, has been
related to lower flame temperatures. Biogas upgrading to remove C02 should be
optimized in such a manner that ensures: (1) Minimum possible upgrading to reduce cost
and (at the same time) yielding the benefit from the influences of C02 on NOx emissions

reduction and soot suppression; (2) Operability and stability of the combustor in as wide
a range as possible, obtaining target heat release regions, and (for premixed flames)

proper ignition delay times and flame speeds for efficient operation.

The focus of this study is on exarrrining the thermo-chemical structure of biogas

diffusion flames by considering carefully selected boundary conditions. Criteria used to
select these conditions are explained in Section 7.1 .1. Any change in C02 concentrations

and boundary conditions in the inlet fuel stream (reactant gases) of a biogas diffusion
flame results in a change in biogas composition and mass flow rate. In gas turbines, the

mass flow rate is related directly to the compression work. Although various effects of
C02 addition to fuels such as CH4, C2H4 and C3H3 have been studied for various

aforementioned applications, there remains a lack of detailed understanding of the

mechanisms through which biogas utilization can reduce emissions of soot, NOx, and

116

green house gases (GHG). In the case of the green house gases (C02, CH4 and N20) it is

vital to indicate whether or not their net release is reduced.

This study examines the chemical kinetics of blends of diluted hydrocarbons that
simulate biogas. A biogas fuel can be simulated as a blend of CH4, C02, N2, and H20
(constituents similar to biogas from digestion saturated with water vapor). Here we
examine C02 diluted CH4 blends, which describe a typical dry biogas. In order to
investigate the chemical and thermal influences of the addition of C02 as a diluent, a
counterflow diffusion flame is selected for analysis. Four non-sooting biogas diffusion
flames are compared to quantitatively examine the thermal and chemical effects of C02
content of biogas on flame temperature, major species, minor species, and emissions of
NOx, soot and green house gases. Major and certain C2 species for one of the flames are

compared with available experimental data [36].

7 .1. Numerical Approach

In our previous study [152], we employed a laminar counterflow diffusion flame

with C02 proportions varying fiom 20 to 50 (vol. %) in a fuel stream comprising only

CH4 and C02. Here, we use a fuel stream mixture of CH4 and C02 highly diluted by N2

in order to study the detailed chemical kinetics of biogas. A schematic of our laminar
counterflow diffusion flame configuration is shown in Figure 1. A planar, steady, non-
sooting flame is generated by impinging a uniform oxidizer jet against a uniform
biogas/N2 jet. The stoichiometric mixture fraction, Z], and the density corrected strain

rate, a, for our counterflow flame are employed as defined by Eq. ((3-4) and Eq. ((3-3),

117

respectively. The CHEMKIN based numerical code OPPDIF [136], is utilized. Radiative

heat loss is not considered.

The method of isolating chemical and thermal effects of biogas C02 content
follows that of Du et al. [151] However, when we systematically add C02, unlike Du et

al. [151] and our previous study [152], the reactive species mole fraction (here CH4) is

kept constant or is changed very slightly to minimize the alteration of the peak flame

 

 

 

N2+02
I ' r ’
L ! Z SP
!
v , I \l’ >-
N2+Biogas (CH4+C02)

Figure 7-1. Schematic of the counterflow burner setup for the chemical kinetics study
(L=14.1 mm; SP: stagnation plain). A nitrogen shroud is usually used in experiments.

temperature, the flame shape, and its location. This allows us to capture the chemical and

thermal influences of C02 addition to the fuel stream while ensuring minimal changes to

118

the flame stoichiometry and the entire temperature field. When C02 is added the

adjustment of the flame temperature is achieved not by changing the oxidizer
composition [151] (which may change the transport properties) or by preheating the fire],

but by a slight increase in the CH4 mole fiaction.

7.1.1. Criterion for biogas flame comparison

Four flames are numerically investigated and compared, called Flames 1 to 4. For
all four flames, L=14.1 mm and the fuel steam enters at the origin of the coordinate
system. Boundary conditions are listed in Table 1. Boundary conditions for Flame 1 are
selected similar to those of a baseline methane counterflow diffusion flame in the

experiments of Bufferand et al. [36] (See FlameA in Table 3-1). This flame has been
experimentally established (by adjusting zf, a, and feed stream compositions) in such a

way as to generate non-sooting, laminar, steady, stable conditions. The chemical structure
of this methane counterflow diffusion flame has been analyzed by gas sampling via
quartz microprobes (CD. =3 340 um) and subsequent GC/MS analysis. A gas
chromatograph equipped with thermal conductivity (TCD), flame ionization (F ID) and
mass spectrometry detectors (MSD) have been utilized for the chemical analysis. Details

of the experimental methodology as well as a schematic of the setup can be found in
Chapter 3 and in [36, 37]. For Flame 1, 2f: 0.76 and a = 134.4 3']. The oxidizer
composition is unaltered between Flames 1-4. These flames are very lean with a global
equivalence ratio, (0, of about 0.27. As seen in Table 1, strain rates of the four flames are
very close with a maximum 2% deviation. This suggests that we should expect very close

flame thicknesses since the flame thickness, 6, varies inversely with a [135]. The values

119

Table 7-1. Boundary conditions of Flames 1—4

 

 

 

 

 

 

 

Flame 1 Flame 2 Flame 3* Flame 4
Biogas composition (40% C02+ (39% C02+ (60% C02+
(% CH4+ % €02). Vol. % (100% CH.) 60% end 61% (mt 40% c1165
Molar Composition
CH4 0.1029 0.1029 0.1080 0.1029
coz 0 0.0686 0.0686 0.1543
N 0.8971 0.8285 0.8234 0.7427
:3 2
2 B' CH4+CO ) Mass
3 ”gas ( 2 2, 0.17 0.48 0.49 0.87
"“ Flux (g/(cm .min))
ma] MESS ,Flux 2.77 2.89 2.88 3.03
(g/(cm .mrn))
Temperature (K) 380 380 380 380
Molar Composition
é N2 0.2271 0.2271 0.2271 0.2271
15 02 0.7729 0.7729 0.7729 0.7729
_N
g Mass Flux (g/(cm2.min)) 3.19 3.19 3.25 3.19
Temperature (K) 340 340 340 340
Strain Rate (s") 134.4 135.8 137.0 137.4
2, 0.76 0.77 0.76 0.78

 

* Note that the temperatures of Flame 1 and Flame 3 are identical.

1' Typical compositions for biogas from anaerobic digestion (AD) plants.

§ Typical compositions for biogas from active or recently closed landfills.

 

of zf increase very slightly for Flame 2 and Flame 4, yielding nearly fixed flame

locations.

Flame 1, Flame 2, and Flame 4 have the same oxidizer and fuel flow velocities. In

Flames 2 and 4 the C02 mole fraction is 6.86 % and 15.43 %, respectively, which

corresponds to a 60%CH4-40%C02 (vol. %) biogas and a 40%CH4-6O%C02 (vol. %)

120

biogas in the fuel stream. On the one hand the 6O%CH4-40%C02 composition is typical
for biogas extracted from digesters in anaerobic digestion (AD) plants. On the other hand,

the 4O%CH4-60%C02 composition is typical for biogas from active or recently closed
landfills. In Flames 1, 2 and 4, the CH4 mole fraction is deliberately kept constant (to

ensure as small stoichiomtry changes as possible), however, substitution of N2 by C02

reduces the peak flame temperature by 41 K and 85 K in Flames 2 and 4, respectively. To
properly compare Arrhenius kinetics and in order to fix the flame location, temperature
profiles and mixture fractions should be unaltered [36]. Flame 3 is obtained by an

adjustment of the Flame 2 boundary conditions that ensures nearly identical temperature
profiles (temperature fields) and 2f values with ~Flame 1. We increased the CH4 mole

fi'action by 5% to raise the temperature. To move the flame, we increased the oxidizer

velocity by 2%. Similar adjustment strategies to fix the flame temperature and location
have been employed by previous investigators [36, 37]. The C02 mole fraction is

identical for Flames 2 and 3.

7.2. Comparison of reaction mechanisms for Flame 1

For Flame 1, three reaction mechanisms are evaluated and compared: (1) the GRI-
Mech 3.0 detailed mechanism [79], which includes 325 chemical reactions and associated
rate coefficients and thermochemical parameters for the 53 included species. It consists
of 106 reactions including nitrogen chemistry; (2) the San Diego (SD) detailed
mechanism [129], which is composed of 46 species and 235 reactions; (3) the GRI-Mech

1.2 reduced mechanism (released 1994) [134] which includes 22 species (+N2 and Ar)

and 104 reactions. Mechanisms (2) and (3) do not include NOx chemistry.

121

A comparison of the calculated major species profiles (using three reaction
mechanisms) for CH4, C02, CO, and 02 with the experimental data of Bufferand et al.
[36] in Figure 7-2 shows good agreement. All mechanisms show negligible differences
for these major species (see Figure 7-2). The calculated CO mole fraction peak is nearly
1.7x greater than that of the experiments. It is an established fact that the CO oxidation

can be catalyzed by moisture [49]. It is very sensitive to small deviations in relative

 

  
    

 

 

 

 

 

  

 

humidity and
0.12
“ Peak temperature
: I: CH4 region
0.1 " U a D C] O 02
_ A CO
008 L . 002
C ' _ 0 CH4
.9 -
g 006 _ Fuel Ox1_dIzer
3 ' - snde sude
'5 .
2 -
0.04 F
~ 002
0.02 -
O h 1 I r 1 r I L L 1 Ii

0 2 4 6 8 10 12 14
Distance from fuel inlet (mm)

Figure 7-2. Profiles of major species (02, CH4, C02, and CO) for Flame 1. Comparison
between a) solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism; c) dotted
lines: reduced GRI 1.2 mechanism; d) symbols: experiments of Bufferand et al. [36].

other factors that are not yet incorporated in the standard reaction codes. Temperature
peak locations are almost identical between both detailed mechanisms (GRI-Mech 3.0

and SD). The flame sheet (the peak flame temperature region) locates at z~6.6 mm.

122

Small differences in the peak magnitude (ATmax ~ 8 K) are revealed (the temperature

profile for the GRI-Mech 3.0 mechanism is shown in Figure 7-5). However, the peak
temperature for the reduced mechanism is almost 15 K and 23 K lower than the GRI-

Mech 3.0 and SD mechanisms, respectively. The flame sheet is slightly shifted toward

 

 

 

 

 

the fuel side.
0.0016
: Peak temperature
- a C H r ion
0.0014 - 2 6 eg
I 0 C2H4
0.0012 - A 02112
c 0.001 1 1’]
.9 ' .,
.2 0.0008 - Fuel ‘. = Oxidizer
2 : s1de _ I s|d9
o _
2 0.0006 _-
0.0004 :-
0.0002 :-
Z 6
0
3 4 7 8 .

 

Distance from fuel Inlet (mm)

Figure 7-3. Profiles of C2 species (blue: C2H2, red: C2H4, black: C2H6) for Flame 1.
Comparison between a) solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism;
c) dotted lines: reduced GRI 1.2 mechanism; d) symbols: experiments of Bufferand et al.

[36] The firel supply in experiments contained about 120 ppm of ethane.

Figure 7-3 illustrates profiles of three major “C2 species” formed in methane
flames, Csz, C2H4 and C2H6. The fire] source in the experiments [36] contained about

120 ppm ethane impurities. Profiles of C2 species in experiments and computational

results with the GRI-Mech 3.0 mechanism agree satisfactorily. The acetylene mole

123

fraction is overpredicted by the reduced mechanism and underpredicted by the SD
mechanism; however, the latter performs relatively better. Both the reduced GRI-Mech
and SD mechanisms overpredict the ethylene mole fraction. All three mechanisms
perform fairly in predicting the ethane distribution, possibly due to the ethane impurities

of the fuel supply (in the experiments).

Profiles of H, OH and CH3 radicals are presented in Figure 7-4. Slight
discrepancies between both detailed mechanisms (GRI—Mech 3.0 and SD) are observed in
the profiles of OH, H, and CH3. With respect to the detailed GRI-Mech mechanism, the

reduced GRI-Mech mechanism always overpredicts these radicals. Thus, the reduced

 

0.007

0.006

   
  

Peak temperature

0-005 region

0.004

Oxidizer
side

Fuel
side

 
 

 

0.003

Mole fraction

0.002

0.001

 

 

TI[IIIlIU'IITTIITII'IIIT'IIIIIIIII

 

01
O)
\l
(D
(0

Distance from fuel inlet (mm)

Figure 7-4. Profiles of H, OH, and CH3 radicals for Flame 1. Comparison between a)
solid lines: GRI 3.0 mechanism; b) dashed lines: SD mechanism; c) dotted lines: reduced
GRI 1.2 mechanism.

124

mechanism fails to perform adequately for the purpose of study, which aims at tracking

the behavior of every major and minor species, and all important radicals.

Contrasting computational results with experiments in the preceding profiles
suggests that the GRI-Mech 3.0 mechanism performs slightly better than the SD
mechanism for this study. Also, GRI-Mech 3.0 has an advantage over the SD mechanism:
it includes NOx chemistry. GRI-Mech 3.0 is therefore employed for the remainder of our

computations.
7.3. Influences of CO; content of biogas; Comparison of Flames 1-4

7.3.1. Temperature and velocity

Figure 7-5a presents profiles of temperature and axial velocity for Flames 1-4. The
stagnation plane for Flame 1 is at 2 ~ 7.3 mm. All flames lie on the fuel side of the
stagnation plane. Except for the oxidizer velocity in Flame 3, which was adjusted to be
2% more than that of other three flames, all other oxidizer boundary conditions are
identical for Flames 1-4. As a result, no distinction is seen between profiles of
temperature on the oxidizer side of the flame. The volume flow rate of the fuel stream is
constant for all flames. The full width at half maximum (FWHM) criterion for
temperature profile is used to define the flame thickness, 5. Here, 8 ranges between 2.5-
2.6 mm for Flames 1-4. As seen in this figure and in its magnified version, Figure 7-Sb,
the entire temperature fields of Flames 1 and 3 are indistinguishable, which ensures that
the strategy we used to match these profiles is successful. Comparison of temperature

profiles in Figure 7-5b shows a maximum offset of 0.14 mm in peak locations between

125

 

 

 

 

 

 

 

 

 

 

 
   

 

 

 

 

1800 L, SP for Flame 1 - 50
- 5'!" i 40
I :! Flame 1 E 30
1500 f ,3,” ......... Flame 2 g A
- j I: - O
A - " I . ~---~----- Flame 3 -. 20 3
>5 : 3' ‘ Flame 4 E E
g 1200 - ',-' ‘ '5 10 v
5 ~ .i -: 0 133‘
a . ‘j . : 2
e r Fuel ,1 . Oxidizer -I -10 2
,2 900 " side I side : a
I- : - .-
- _I 1 -20 2
- .I a
- f1 1 -30
600 - i 5
. ,I' \5
- ' -50
am 1_LJ_L 1 1 l 1 J 1 l 1 1 1 l r L r l r 1 1 l 1 r 1
0 2 4 6 8 10 12 14
Distance from fuel inlet (mm)
(a)
1860
Z Flame 1
1830 -
: Flame 3
as 1800 l
9 ' . .
a ; F... Ogggger
(u .
8 I sude
o. 1770
g T
l- : ,.
1740 l
;
1710 -
6 6.5 7 7.5

Distance from fuel inlet (mm)

(b)

Figure 7-5. a) Comparison of temperature and axial velocity profiles for Flames 14; b)
Magnified version of temperature profiles near the peaks. The stagnation plane for Flame
1 is at 2 ~ 7.3 mm. All flames are on the fuel side of the SP.

126

 

8.00E+09

 

 

 

700 ‘—Flame1I
.E+097-~~~~» "I »
§ —Flame2‘
g 6.00E+09~~ , , ‘—Flame3‘
I

<5 I—Flame4‘
2 5.00E+09 .- 2--
E’

8.

4, 4.00909

“

E

3 3.00E+09

N

2

2 2.001309

1'6

0

I 1.005+09~

0.001900 . , f , p
5 5.5 6 6.5 7 7.5 a

Distance from fuel inlet (mm)

Figure 7-6. Comparison of heat release profiles (signatures) for Flames 1-4 in the reaction
zone.

 

  
 
 
 
   

 

0.2
— Peak temperature
: /.. region for Flame 1
0.99.2 ............... 5 Flame 1
0.15 -
7 --------- Flame 2
.1. -------------------------- Flame 3

_ -. _ -------- — Flame 4
0.1 i

Mole fraction

0.05

 

 

 

Distance from fuel inlet (mm)

Figure 7-7. Comparison of the mole fractions of C02, H20 and CO for Flames 1-4. Solid
lines for CO and C02 profiles in this Figure and Figure 7-2 are the same.

127

these four flames. This is the direct consequence of fixing the CH4 mole fraction for
Flames 1, 2, and 4. The offset in the temperature peaks as well as the reduction in peak
magnitude is attributable to the increase in the C02 content. Figure 7-6 delineates heat

release signatures (profiles) for Flames 1-4 in the reaction zone. In addition to the

temperature fields, heat release signatures of Flames 1 and 3 are nearly identical.

7.3.2. Major species, radicals and intermediate species

Figure 7—7 illustrates profiles of C02, CO and H20 for Flames 1-4. As expected,

the C02 mole fiaction has the maximum value in Flame 4. Flame 3 has a nearly 21 %
greater CO mole fraction than Flame 1. This difference is mainly due to the conversion of
C02 in the fuel stream of Flame 3 and its chemical participation through reactions such

as

R1: C02+X —>XO+CO
R2: COz+CH -—>HCO+CO
R3: C02+CH2* —>CO+CHzO

where X may be one of these radicals: H, OH or 0. Figure 7-8a presents a comparison
between mole fractions of the important chain branching radicals; H, O and OH for
Flames 1-4. Comparison between Flames 1, 2, and 4 with different temperatures reveals
that C02 dilution reduces the mole fraction of these radicals. The maximum reduction
between Flames 1-4 is 20 %, 21%, and 34% for O, OH, and H radicals, respectively.

Comparison of H profiles in Figure 7-8b, which is a magnified version of Figure 7a,

128

shows that, in general, C02 dilution has a greater effect on H radicals than on OH and O

radicals.

As demonstrated in Figure 7-8b, the difference between the H profiles for Flame 1

and Flame 3 (with identical temperature profiles as in Figure 7-5b) indicates the chemical

efi’ects of the C02 addition (dilution). The difference between the H profiles for F lame 2
and Flame 3 (with identical inlet C02 mole fractions as in Figure 7-7) indicates the
thermal eflects of the C02 addition. This interpretation is used in the discussion of this
and all subsequent figures. With respect to O and OH, the chemical influences of C02

addition are the most effective for H. This is exhibited in the comparison of Flames 1

and 3. When Flame 1 is diluted by C02 to form Flame 3, the peak of the H profile is

reduced by approximately 7.6% whereas the reduction in the peak of O radicals (Figure

7-8a) is much smaller, and the OH profiles can hardly be discriminated. The thermal

effects of C02 dilution on H, O, and OH radicals are nearly of the same order.

Figure 7-9a presents profiles of methyl radicals (CH3) and formaldehyde (CHZO)

for Flames 1-4. The peak CH3 for Flame 1 is at z~6.15 mm. C02 addition shifts the CH3

profile toward the oxidizer side. The peak magnitude is also reduced; the reduction is

modest for Flame 3 (there was a 5% difference between Flames 1 and 3 in the mole
fractions of CH4, the primary source of CH3). This amplifies the notion that the direct
chemical effect of C02 on the concentration of CH3 radicals is modest by contrast with
its thermal effects (compare Flames 2, 3). This is consistent with Figure 7-8 and

highlights the influence of H, O and OH radicals on CH3 concentration through important

129

chain branching reactions such as CH4+H—->CH3+H2, CH4+O—>CH3+OH, and

CH4+OH—*CH3+H20.

In an interesting trend, the addition of C02 increases the CHZO peaks as seen in
Figure 7-9a. CHZO is an important intermediate in the CH4 reaction pathways. Both
Flames 2 and 3 with the same C02 content in the fuel stream produce very close CHzO
peaks. This result undermines the thermal effects of C02 dilution on the formation of
CHzO (i.e., C02 produces a weakly temperature-dependent effect on CH20). Chemical
influences of C02 on CHzO formation are more pronounced and favor its formation e.g.,
through the reaction (R3).

Profiles of CH radicals are shown in Figure 7-9b. Both Flames 1 and 3 peak at
almost the same location (2 ~ 6.3 mm). With respect to Flame 1, the peak magnitude for
Flames 2, 3, and 4 reduces by 38%, 21% and 66%, respectively. The CH radical

(methyne) is the dominant immediate precursor to prompt NOx [115]: its reduction is
expected to reduce NOx. This will be elaborated in Section 3.2.4. 61%CH4-39%C02
biogas (typical for AD plants) in Flame 3 significantly reduces the CH radical
concentration. This reduction can be attributed to the direct chemical influence of C02 on
CH through reactions such as COz+CH —-»HCO+CO or to its indirect influence to a lesser

extent through the reduction of H and OH radicals.

130

 

 

    
     
 
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.006 L
: Flame 1
0005 _'_ --------- Flame 2
~ “”“""“"' Flame 3 Peak temperature
: - ........ _ Flame4 region for Flame1
0.004 f
C
.9 1
1'5 -
g 0.003 -
2 Z
c
2 I
0.002 _- /’
0.001 _- w
o - 1 1 1 1 I J 1 1 j ’ 1 -
5 5.5 6 6.5 7 7.5 8
Distance from fuel inlet (mm)
(a)
0.002
' ure CH t
- p 4 /\ [Chemical Effect
. 6.86% 002:
, . [\I Thermal Effect
0.0015 - 63649027 ~
I. o
c i 15.4% co; _;._,.. Ham" 1
g : :7 --------- Flame2
g 0.001 - 5 ,-°' -~------Flame3
Q L o' :
3 _ :. i! - -------- -FIame4
2 : i
" Peak temperature 4 '/,/l
i region for Flamet ,- 1’
0.0005 - . i
_ : i
. .' j
- l 0"
- .’ /'
01111111 tof'O111I1111I114_1 ,
5 5.5 6 6.5 7 7.5 8
Distance from fuel inlet (mm)
(b)

Figure 7-8. a) Comparison of the mole fractions of H, O and OH radicals for Flames 1-4;
b) Magnified version of Figure (a) showing the mole fractions of H radicals. The
percentages of C02 in the fuel stream have been written. Note the quantitative definitions
of “chemical effect” and “thermal effect”, which are employed (without explicit
demarcation) in all subsequent figures.

131

0.0014

0.0012

Mole fraction

2.4E-06

2E-06

1 .6E-06

1 .2E-06

Mole fraction

8E-07

4E-07

 

III1'1III'IIrlllllvlllllllllllrllr

 

 

55. 6 6.5 7 7.5 8

 

Flame 1
--------- Flame 2
~~~~~~~ — Flame 3
— -------- — Flame 4

   
  

. Peaktemperature
‘. = region for FIameI

 

Distance from fuel inlet (mm)

 

lfilrrllrrrllIlllrllljv‘rI—r'

rlr

 

 

    

(a)
N0 002*” Flame1
--------- Flame 2
m~~~ Flame 3
6.86% cof— _ ........ _ Flame 4
Peak temperature

6.86% COZ‘; region for Flame 1

 

 

15.4% coz<

 

 

'0
\1-_L11l1111l1111

 

01

5.5 6 6.5 7 7.5 8

Distance from fuel inlet (mm)

(b)

Figure 7-9. Comparison of the mole fractions of a) CH3 radicals and CHZO; b) CH
radicals for Flames 1-4. The percentages of C02 in the fuel stream have been written.

132

7.3.3. Soot precursors

C2 species, mainly C2H2, C2H4 and C2H6, are soot precursors and the main source
of the CH and CH2 radicals that are responsible for “prompt NO” production [115].

Figure 7-10 shows the mole fractions of acetylene as well as the decay of CH4 for Flames

1-4. The reduction in the peak of acetylene, which is an important soot precursor and
surface growth species, suggests that biogas may be beneficial for soot suppression. This

statement is consistent with the recent study of Guo et al. [122] that showed the chemical
effect of C02 addition to an ethylene/air diffusion flame suppresses soot inception and

surface growth rate whereas it has only a negligible influence on soot oxidation. More

studies on sooting or incipiently sooting biogas flames are needed to ascertain whether

the C02 addition can also alter rebum of either acetylene (through reactions like

C2H2+O—>OH+C2H) or soot particulates.

The observed shift in the peak locations of radicals (see Figure 7-8 and Figure 7-9)

is consistent with the shift in the decay of CH4. Flames 1 and 3 show close peak locations
for radicals (see Figure 7-8 and Figure 7-9) and for Csz (see Figure 7-10). This is
consistent with the CH4 trend in Flames 1 and 3. Flame 1 produces nearly 2.4>< as much

acetylene as Flame 4. Since the primary source of acetylene is methane, and the CH4

mole fraction is fixed for Flames 1 and 4, the reduction must be caused by the presence of

C02 in the fire] stream of Flame 4.

133

0.0008

 

 

 
   
   
  
  
  
   

 

 

  

I I
I Flame1 ZEDCH4
* --------- Flame2 ; ;
' .................. F|am93 I: I_
_ :i Peak temperature
0'0006 — -------- — Flame4 I 1 region for Flame1
' : i
c I i
4% — a I
2 0.0004 - i i
0 I :
'5 ‘ E l
2 : !
'. i
_ : I
0.0002 - 0sz - 5 I
'.,'. |_
' £1.11
I i." i.
- . it
0 J_l I 111111111 \.\¢m11l1111l1111

 

 

4 4.5 5 5.5 6 6.5 7 7.5 8
Distance from fuel inlet (mm)
Figure 7-10. Comparison of the mole fractions of acetylene for Flames 1-4. The decay of
CH4 profiles is also demonstrated in this Figure.

The flame temperature has a profound influence on soot formation. Flames 1 and 3
have indistinguishable temperature-time histories. The acetylene peak for Flame 3,
however, is 11% smaller than that of Flame 1. This reduction is therefore attributed to the
chemical effects of C02 in biogas. The comparison of Flames 2 and 3 shows that the

thermal effects, which reduce the flame temperature (Figure 7-5b), are comparatively

greater.

Removal of C2 precursors of prompt NOx from the diffusion flames may also be

beneficial for NOx reduction [115]. It was suggested in the discussion of Figure 7-9 that
NOx could be reduced when using biogas even when the temperature does not change.

NOx reduction is discussed in the next section.

134

7.3.4. N Ox emissions

Figure 7-11 shows a comparison of the N radical and NO mole fractions for Flames
1-4. The peak NO for Flame 1 locates at z~6.55 mm. The peak NO for Flames 1-4 is in
close proximity to the temperature peak. We expect a decrease in the NO concentrations
for Flame 2 and especially for Flame 4. The peak for the N radical in Flame 4 is

approximately 15% of the N peak in Flame 1. The N radical plays an important role in

“thermal NO” formation through the reaction N+Oz—2NO+O. The chemical influences of
C02 on the reduction of N radicals predominate: compare Flames 1, 3 whose thermal
profiles are essentially identical. Flame 4 shows an NO peak approximately 1/4 of Flame
1. This substantial reduction does not appear to be a result of N2 concentration changes in
the fuel stream. Flames 2 and 3 with very close N2 mole fractions in the fuel stream
produce very different NO peaks. This reduction is due to both thermal and chemical
effects of C02 dilution. Flame 4 uses a low-quality biogas (landfill biogas) with a mass
flux almost 5X greater than Flame 1. Comparing the pure CH4 flame (Flame 1) and the

landfill biogas flame (Flame 4) shows that there exists a peak temperature reduction of
AT=85 K and a large difference in the mass flux. The reduction in the peak of NO (25%
ratio) is feasible only if the associated loss in heat release and rise in mass flux are

acceptable for a specific application. It is more economical to use: 1) lower mass fluxes;
2) less biogas upgrading, which means a higher C02 content and a lower quality.

Contrasting Flames 2 and 4 shows that there is a tradeoff between these two factors; for
the lower quality biogas the reduction of NO is larger, but the mass flux (and thus the

required compression work) is higher.

135

 

 

 

 

  

 

 

 

 

 

 

 

60 2507

- Peak temperature _

C region for Flame 1 _

C Flame 1 ‘
......... Flame 2 _ 1 .5E'O7 :
i ‘ .9.
4° .' 6.86% 002: —-—-——-—- Flame 3 g
’5‘ C / ' , - -------- - Flame 4 e:
a. - - 2
°- - _ o
5’ 3° - 6.86% cor f, 3 _ "5'07 E
Z Z -‘ gi' ‘ ' g
- '4 .0
20 - ' '-.,__ g
' I : ‘ 2

15.4% 002: : , 3, - 5E-O8

E $.- ’1 :1" '. \._\.. 1

.. ‘; ,&1 ‘~.\ NO ,

" :' i .

1' .1 l I 1 1‘1”“: 0
7 8 9 10

 

Distance from fuel inlet (mm)

Figure 7-11. Comparison of the mole fractions of N radicals and NO (in ppm) for Flames
1-4. The percentages of C02 in the fuel stream have been written.

Comparison of Flames 2, 3 shows, as expected, that the flame temperature, which
is directly altered by dilution, plays a significant role in NO reduction. Flames 2 and 3
use a biogas with compositions more similar to a typical biogas from the anaerobic
digestion process. For these flames, which have very similar biogas compositions and
only a 1.7% difference in the biogas mass flux in their fuel stream, AT is 40 K while the

NO peak shows a 26% difference.

Flame 3 has a similar temperature profile to Flame 1 with a biogas mass flux nearly
2.9x larger than Flame 1, but its NO peak is less than Flame 1 by about 26%.
Comparison of Flames 1 and 3, however, lends additional credence to the beneficial

influence of biogas in NO reduction. In this case a temperature decrease cannot explain

the NO reduction. It must be explained by the chemical influence of C02 on NO.

136

Reduced concentrations of CH radicals and C2 species (Figure 7-9b, Figure 7-10) as well

as N radicals suggests a possible reduction in both prompt and thermal NO formation.

Figure 7-12 presents profiles of nitrogen dioxide (N 02) and nitrous oxide (N 20) for
Flames 1-4. In Flame 1, the peaks of N02 and N20 are located at z~ 8.2 mm and z~ 7.2
mm, respectively. Both peaks are on the oxidizer side of the flame. N02 shows a small

second peak at z~ 5.5 mm on the fuel side. N02 may be attacked by H and O radicals
between these two peaks (where radicals are abundant) and converted to NO or other

species. N02 shows a trend similar to NO with a decrease in peak values when C02

 

 

 

 

  

 

2E-06
- Flame1 Only CH 4
--------- Flame 2
I ~—---———--- Flame 3
1_ E. b _ ........ _
5 06 _ Flame4 6.86% 002
5 Peak temperature / \ o
‘6 - region for Flame 1 f x I ' 36 4’ CO2
.33 1E-06 - i; 1,1,
2 ~ 1’: :\
o I
2 j- ,1: " |\‘\
. ,1 I, ‘-. 2 15.4% 002
5E'07 " I: \._ ‘;-.,
t j: ‘. \, ‘
' "if i &‘ NO2
_ N02 -' \.
I" . /' 5 \~
0 ..- ....-._ .. —::i—-'- ". L . 1 . . I ' .
4 5 6 7 8 9 10

Distance from fuel inlet (mm)

Figure 7-12. Comparison of the mole fractions of N02 and N20 for Flames 1-4. The
percentages of C02 in the fuel stream have been written.

137

increases (The N02 peak magnitude for Flame 4 is 33% of Flame 1). Chemical effects of
C02 on N02 reduction are readily seen in the comparison of Flames 1 and 3. C02
dilution chemically influences the formation of nitrous oxide (N20); the peak of N20 in

Flame 1 is reduced by almost 5% when compared to Flame 3. As observed in Figure

7-12, the thermal effects of C02 addition on N20 are less pronounced than they are for

N02.

7.3.5. Greenhouse gas emissions
Three greenhouse gases are relevant to this study: carbon dioxide, methane, and
nitrous oxide. According to EDGAR [153], C02, CH4 and N20 respectively account for

72%, 18% and 9% of the total Kyoto-Protocol greenhouse gas emissions. Landfill biogas

contains methane, which has the potential to be produced naturally from biodegradable
wastes (e. g. municipal solid wastes). The collection of biogas in landfills reduces the CH4

release to the atmosphere.

Nitrous oxide (N20) is an important greenhouse gas. As seen in Figure 7-12, biogas
use reduces N20 emissions. Nitrous oxide is reduced by C02 dilution mainly due to its
chemical effects (with only a weakly temperature-dependent influence).

Combustion-produced C02 is a source of environmental C02. We require a rational

scale to identify whether burning a candidate fuel can potentially reduce the net C02

release to the atmosphere. Thus, we define a carbon dioxide emission index:

138

 

mCOZ ,emitted

mF,burned

EICOZ = (7-1)

EIC02 indicates the ratio of the net emitted mass of C02 leaving a combustor to the
mass of fuel (here CH4) burned. The net emitted mass of C02 can be found as

mC02,emitted = (mg-02 )exhaust —(mCO2 1",“. For the counterflow configuration, the EIC02

will be

L , 0
2J0 WCOZ ardz " mC02,inlet

mE'H 4 ,burned

(7-2)

 

151cc2 =

where p is the gas mixture density, WCO; is the mass fiaction of C02, a, is the radial

velocity gradient, and 1518025121.»: and mammmed are the mass fluxes of C02 and

CH4 in the inlet fuel stream, respectively. For Flames 1-4, Table 2 lists EICOz and the

calculated values of the inlet C02 mass flux (6123023,,80 as well as the exhaust C02

mass flux (7712302,,3xhaust). As seen in Table 2, the exhaust C02 mass flux from Flame 1,

which has no C02 addition, is 0.08 g/(cmZ-min). This amount is produced by the
oxidation of CH4 in Flame 1. With respect to Flame 3, adjustment of the boundary
conditions of Flame 2 to increase its flame temperature changes the EICOz only very
slightly from 0.415 to 0.411 (a 1% change). Flame 1, which contains pure CH4, has the

highest EICOz (0.469) whereas Flame 4, which utilizes landfill biogas (a low-quality

139

biogas), shows the lowest value (0.352). Flames 2 and 3 (using a typical biogas from AD

plants) lie in between.

Table 7-2. C02 exhaust and inlet mass fluxes, and EIC02 for Flames 1-4*

 

Flame 1 Flame 21 Flame 3t Flame 4§

 

Exhaust C02 mass flux

2 , 0.0801 0.3836 0.3862 0.7633
(g/(cm .min))

Inlet cozszs flux 0 0.3127 0.3125 0.7032
(g/(cm .min))

EICOZ 0.469 0.415 0.411 0.352

 

* Note that the temperatures of Flame 1 and Flame 3 are identical. Flames 2, 3, and 4 contain
6.9%, 6.9%, and 15.4% C02 in their firel stream, respectively.

1' Typical compositions for biogas fiom anaerobic digestion (AD) plants.

§ Typical compositions for biogas from active or recently closed landfills.

EICOz can be used as an assessment tool to compare the net C02 release fiom the
renewable fuel-based flames (Flames 2-4 bunting biogas) versus the fossil fuel-based
flame (Flame 1). Flame 1 utilizes pure CH4 (similar to natural gas, considered as a fossil-
based fuel) whereas all other flames burn biogas and produce a lower EICOZ. This
suggests that burning the renewable fuel (biogas) by the combustion process reduces the

net release of C02 as opposed to burning a fossil fuel (methane).

A final issue to mention is the tradeofl that exists between (1) biogas consumption rate

and work associated with its compression; (2) flame temperature (heat release); (3) the

140

reduction of greenhouse gases and NOx emissions. For example, a comparison of Flame
2 biogas from AD plants with Flame 4 landfill biogas reveals that a 2.5 % reduction in
the flame temperature reduces EIC02 by 15 %. However, Flame 4 has a 1.8x greater
biogas mass flux than Flame 2. Figure 7-13 illustrates the tradeoff between three scaled
quantities: peak flame temperatures, peak NO productions, and biogas mass flow rates. In
a real-world application of biogas combustion, the reduction of GHG and NOx emissions,
a possible increase of the compression work, and also a possible loss of heat release
should be all taken into account. It should be noted that typical compositions for biogas
from landfills and AD plants in this work are based on zero upgrading (i.e. a significant

reduction of cost).

Similar kinetics studies on the flame structure of C02 diluted methane blends can

be employed for biogas-fired fuel cells and C02 diluted oxy-fuel combustion as well as

the EGR modeling for IC engines.

141

 

 

 

 

 

 

 

Scaled peak flame
temperature=
T /T
max max (Flame 1) 0.94
0.92
1 2 3 4
1 1
0.8 -
Scaled peak NO= 0.6
NOmax/Nomax (Flame 110 4 ,
0.2
o _
1 2 3 4
6
5 s
Scaled mass flow 4 ‘
rate= 3 _
m/ m (Flame 1) 2 -
1 _
o _
1 2 3 4
Flame No.

Figure 7-13. Tradeoff between scaled flame temperatures, scaled NO, and scaled mass
flow rates for Flames 1-4.

142

Part III

Flame Spread over Solid Fuels in
Microgravity

143

Chapter 8 Background and Introduction to Flame
Spread in Microgravity Conditions

Flames that interact strongly with surfaces are by far the most prevalent both in
nature and technology. They include attached burner flames in combustors, torches and
heaters (industry), forest fires (nature), flames in rockets or missile combustion champers
(space), and attached flames in fires and flame spread (safety) [154]. For several decades,
the science of flame spread over solid materials has been of interest to researchers and
professionals who have responsibility for fire safety including accident investigators,
insurance companies, vehicle designers and building code authors. Flame spread is by
nature complicated due to the coupled physical processes involved. In the flame spread

literature, simplified theories and apparatuses for certain limiting cases and certain

classifications (e. g., thermally thick vs. thint materials) have been developed [155]. In a
spreading flame, the flame and flow direction can assume any angle depending on the
particular setup, but two flows are of dominant interest: (1) opposed flow, in which the
direction of the oxygen flow is against the direction of flame spread; (2) concurrent flow,
in which the flame advances in the direction of oxygen flow [156]. Concurrent flow is
usually thought to be more hazardous since its gas phase forward heat transfer
mechanism is convection (excluding radiation) with a much faster spread rate [157].
Much attention has been devoted to the study of both concurrent and opposed flow flame

spread in the past decades [157]. The focus in this thesis is on opposedflow flame spread.

 

*
In a thermally-thin surface the front and rear surface temperatures are nearly identical during heating.

144

A recent trend in this research is emphasis on the three dimensional structure of the
spreading flames, although this limits many analyses to purely numerical work. Although

models of flame spread have been developed for “generic” problems such as the “ideal

vaporizing solid” (e. g., PMMA") [158], it is well known that few materials burn in this
way. Real materials regress, liquefy, and form spotty regions with various degrees of
char-ring and liquefaction. Real materials are also heterogeneous mixtures (e.g.,
propellants) and their burning is extremely complicated as recent research demonstrates

[159]. Real fires also spread over heterogeneous dissimilar materials.

For these reasons it is believed that before developing detailed 3-D solution codes,
it is imperative to understand three things: (1) The “flame microstructure” which includes
attachment characteristics, local reactivity, local temperature, enthalpy, species, reaction
rate contours, and characteristic (time/length) scales. (2) The detailed heat and mass
transfer responses of a surface during combustion. The detailed response includes the gas
phase and it should produce an explanation of the basic structure of the flame attachment
process and its requirements. (3) The reason why real flame spread at nearly all length
scales is actually a 3-D (not 2-D) process. This requires performing actual 3-D analyses
(which can be presently accomplished only numerically) and also examining the causes
for the formation of 3-D flames (fires) fi'om initial 2-D flames. The mathematical and

numerical models examined in this Part attempt to address questions (1)-(3).

 

t
Poly methyl methacrylate.

145

In the study of diffusion flames attached to surfaces, recent work by our research

group at MSU, under the NASA project ATHINA* [160] has shown that complicated
burning patterns can occur. Analyses of the phenomenon by our research group at MSU
and NASA can be found in Refs. [155, 160]. A complicated “flamelet” form of flame
spread has been reported, in which stable two-dimensional flame fronts break apart and
form flamelet fionts, in which isolated individual and small 3-D “flame caps” spread over
the surface, consuming the fuel as they generate a complicated burning pattern (See
Figure 10-3). The reasons why the 2-D flame front eventually become unstable and

fi'agment into irregular 3-D fronts will be addressed.

Many of these complications arise when the inflow rate of oxidizer is reduced
[160]. However, it is clear that other influences, such as fuel thickness; fuel homogeneity
and heterogeneity; fuel regression pattern; melt layer (or char layer) formation; volatile
transport/reactivity in the gas; fluid-dynamical flow pattern near the surface; will all
affect the flame structure. The reason is that flame spread over surfaces involves a
complicated feedback process, in which the flame heats the surface, which decomposes to
volatiles. These volatiles are transported by diffusion and convection into the gas leading
to the formation of mixing layers and combustible gas mixtures, which facilitate reaction
and transport of heat (thermal energy) to the solid fuel [158]. Flames over non-
homogeneous fuels such as propellants can also show complicated surface burning

behaviors, and these behaviors are only beginning to be studied [161].

 

*
ATHINA stands for the NASA Microgravity Combustion project: "Analysis of Thermo-diffusive and
Hydrodynamic Instabilities in Near-extinction Limit Atmospheres”, started in January, 2003.

146

As elaborated before, the combustion of a solid fuel is an extremely complicated
process. The goal of this research is to begin to understand this process in terms of simple
descriptions that isolate key features and focus on their rigorous description, instead of
trying to solve the entire problem at once. The approach taken here is to simplify the

problem and to attempt to rigorously describe it.

In the flame spread model, the entire downstream surface is considered to gasify, as
in all “classical” flame spread models [162, 163]. Long and co-workers at MSU [156,
164] examined a simplified 2-D mathematical model of classical flame spread over thin
cellulosic fuels including both gas phase flow and solid phase degradation. The
theoretical calculations were shown to be in excellent qualitative agreement with the
numerical computations, suggesting that the physical features of such models are
adequately characterized. All quantities important for the macroscopic and microscopic
structure were found and analyzed. The diagnostic parameters can be used to examine

flame structure for more complicated flame geometries.

We perform a numerical and theoretical study, in Chapter 9, for another kind of
flame model in which the surface pyrolyzes over only a finite segment (or several
segments) of filel. This model describes (characterizes) flarnelet formation and spread,
flame burnout lengths [165] and heterogeneous materials [166], none of which support
large and uniform flaming. Our theoretical and numerical study of flame spread over
single and multiple fuel segments include an open environment. Simulations are
conducted for zero gravity (g = O) to mimic microgravity conditions although the flow in
open environments on earth is strongly driven by buoyancy. Convective flows

transporting volatiles and oxidizer are assumed to be weak in the theoretical models.

147

This research finally turns toward the study of flames forced to spread in a confined
space, which has important implications for fire safety in normal gravity. In Chapter 10,
we will examine the flame spread in narrow gaps, or rectangular channels. The fuel and
flame are close to other surfaces resulting in large heat losses to the nearby walls. As a
result, in narrow gaps, the flame becomes “near-limit”. The designation “near-limit” in
Part HI indicates limiting conditions that weaken the flame including large heat loss or
poor oxygen delivery (produced by a reduction of the oxygen inflow to levels low enough
to starve the flame of oxygen). A motivation for our research is that “near-limit”
conditions are achieved in the most deadly fires. One of these fires caused the crash of
the 1998 Swissair Flight 111 into the Atlantic Ocean while attempting an emergency
landing at Halifax International Airport. The crash killed all 229 passengers aboard [167].
A small, weak flame originated due to a wire arcing event in the cockpit slowly spread,
driven by ventilation flows between the walls. This near-limit flame received a much
greater airflow in the narrow gap after breaching a silicon vent cap, leading to a rapidly
growing and spreading flame [155]. Other examples include flames in small gaps (in the
presence of wire bundles or electronic equipments) or in regions where materials are in

close proximity, such as inside a wall.

A new experimental apparatus, a Narrow Channel Apparatus, or NCA, has been
built and tested at MSU and at NASA. The NCA allows well-controlled experiments to
be conducted [155, 168]. The research at MSU using the NCA along with parallel efforts
of our colleagues, Drs. Olson and Miller, at NASA [155] using drop tower experiments
(actual microgravity conditions) and a NCA they constructed after MSU’s original

version [168] was designed to investigate whether the NCA is well suited to study flame

148

spread in simulated microgravity. A scaling analysis is performed to determine the
characteristic magnitudes of the various velocities that occur in the NCA. It will be
shown that a NCA operating in normal gravity on earth can be used to test microgravity

flame spread, which is an acknowledged space-fire hazard [169].

149

Chapter 9 Numerical and Analytical Study of
Diffusion Flames Spreading near Solid Fuel Surfaces in
Microgravity

Flame attachment to surfaces in problems such as spreading flames, fires, and
surface-burning propellants is a complicated process. The flame heats the surface, which
decomposes into volatiles, which leave the surface as the gaseous fuel that mixes with
incoming oxidizer, feeding further combustion. Although classical flame spread models
usually consider gasification over the entire downstream surface, in certain applications
the burning of finite segments (one or more) of solid fuel is of practical significance.
Much attention has been devoted to examining near surface diffusion flames analytically

using simplified non-dimensional governing equations [170-172].

Heterogeneous combustion processes will produce interacting distributed flames,
although little work has been done on this problem. Flame-flame interaction between
single flames on parallel solid surfaces has been studied, but these works examine
vertical surfaces in l g conditions [173]. Rouvreau et al. [174] numerically simulated a
microgravity diffusion flame of ethylene over a flat plate with the flow of oxidizer
parallel to its surface using the Fire Dynamics Simulator (FDS) code. Their work, which
has been compared to a detailed experimental study [175], examined in detail the
influence of flow perturbations and the validity of the boundary-layer assumption based

on the fuel injection velocity.

In this chapter, diffusion flames near solid fuel surfaces are investigated

numerically and analytically in a 2-D planer domain. An attempt is made to describe 2-D

150

flames, in which oxidizer is swept past a finite-size fuel-delivery segment (or segments)
embedded in a non-combustible “binder” material. The goal is to determine whether such
flames can resemble separate flamelets, one behind the other, or whether the combination
of downstream combustion product gases and air vitiation produce conditions unsuitable
for successive flamelets to survive. The “flamelet”, as mentioned before, is an interesting
flame shape appearing in narrow-channel combustion, microgravity combustion, and

combustion over heterogeneous fuels.

The sensitivity of the 2-D model to the downstream conditions suggests that the
actual localized flames and flamelets are likely to be 3-D. Nevertheless, the 2-D model
may be used to achieve insight to the flame structure near the attachment region for

single and multiple fuel segments.

To construct an analytical model of surface-attached solid fuel flames, a segment
of fuel adjacent to an insulated, non-reactive binder material, and located in an infinite
gas-phase environment, is examined. An exact analytical solution is derived for the
mixture fi'action equation when convective terms are discarded. In the next step, and to
take into account the role of weak convective terms, the mathematical model is solved
analytically and compared to numerical solutions. Numerical solutions of the Z (mixture
fraction) equation are obtained for a finite-height channel with an Oseen flow (uniform
streamwise velocity), with or without a constant transverse velocity. In the more complex
case of multiple fuel segments, exact solutions for the Z equation are unattainable,

whereas the numerical modeling is still applicable.

Comparisons are also made with the results of numerical modeling of a

qualitatively similar problem using FLUENT, which includes a surface segment (or

151

segments) of a gaseous hydrocarbon (not generated volatiles of a solid fuel at the
pyrolyzation temperature) in a cross flow of air. The equations for conservation of mass,
momentum, energy and species are solved. Since the gas phase is of principal interest,
various inlet fuel and air velocities are considered. Results of analytical and numerical
models are compared near the flame upstream leading edge (attachment region) and near
the trailing edge, where the flame can reattach to the surface. The ultimate purpose of this
study is to establish an analytical model to examine heat transfer, flame structure and
reaction rate in flames attached to solid fuel surfaces, particularly surfaces with complex

structure and flames that consist of many flame fi'agments, or flamelets.

9.1. Model Problem

In the model problem considered here the solid and gas interact through boundary
conditions that render the gas phase non-isenthalpic. In other words, the gas loses
enthalpy to the solid. Viewed sirnplistically, the energy lost from the bunting gas phase
flame is used to produce volatiles from the condensed fuel. Within this class of
approximations there are simple models that have not yet been examined for flamelet

spread. One of these models will be formulated here.

When the cold fuel is heated to the pyrolysis temperature as a result of heat transfer
from the flame, the heated surface generates fuel vapor. The focus will be on the gas
phase. Consider, therefore, a uniform source of fuel vapor that feeds the flame from an
isolated surface segment. To produce a combustible mixture, the fuel vapors mix with the
oxidizer, which exists either as: (1) a stationary-infinite source, or (2) enters the domain
as a forced flow in a finite region. Our model problem examines the former case. In this

model, the convective inflow velocities are negligible, thus the species equations can be

152

simplified and analytical solutions can be derived. However, this is rare. When exact
solutions cannot be derived, either simplified versions of the models can be solved
approximately or numerical methods can be employed to develop solutions for such

attached flames.

Two regions in particular are investigated: (1) The upstream flame leading edge
near the fi'ont of the fuel segment; (2) The flame trailing edge where it can reattach to the
surface downstream of the fuel inflow (blowing) segment. Theoretical and numerical

investigations are conducted to examine various quantities in these and other regions.

9.2. 2-D theoretical model

To construct a theoretical model, a one-step, irreversible chemical reaction in the
form F + v0 -> (I +v)P (mass basis) is considered for the gas phase. Here, P denotes the
combustion products. The governing equations for the continuity of species are written

with the assumption pD = const. and Fick’s law for mass diffusion [176]. The equations

for the conservation of species are:

ZYl—.+u§£+v-a—YL=DV2
at 6x 6y

Y,- _Vir , i=F, 0 (9'1)

Here, v1: = 1 and v0 =v, and r is a source term pertaining to the reaction

stoichiometry. Equations (9-1) can be combined to produce a single homogeneous

equation for the mixture fraction, Z, which is defined as

VYFF +Y000

 

153

In this equation, Ypp is the fuel mass fiaction at the fuel inlet (the “pyrolyzing”

surface segment) and Yoga is the oxygen mass fraction in the far field, which is the

oxidizer source. The equation for Z is simply Eq. (9-1) without r, viz.

-a—Z—+u§§+va—Z=DV22. (9‘3)
at 6x 6y

At the flame the Z-value is given by (Y F=YO=O)

zf = Y0°° = 1 , (9-4)
VYFF+YOoo l-IrllYplr/Y0,Jo
where the subscript “f” denotes the values of Z at the flame sheet. The ratio

(15 =VYFF/Y0m is defined as the stoichiometric index. The condition Z f = (1+ J )_1

locates the flame sheet, since Z=Z(x,y) and Z(x, )=Zf defines the mathematical

relationship between the coordinates x,y when Z=Zf.

The numerical method is a second order central difference scheme for the x and y
directions. For Neuman boundary conditions near the wall (62/6y = 0), a second order
forward difference scheme is applied. The Alternate Direction Implicit (ADI) method
with successive over-relaxation (SOR) is employed here as the iterative method. The

numerical domain in non-dimensional space is 10x10 including 100x100 uniform grids.

9.2.1. Analytical solution of V22 2 0

In the steady state, with zero bulk flow streamwise and transverse velocities

(negligible convective terms), Eq. (9-3) reduces to the Laplace equation,VzZ = O.

154

Figure 9-1 shows the boundary conditions for this problem. In the analytical solution, the

length and width of the domain boundaries extend to infinity. The analytical solution is

 

 

 

 

y i
Y0=Y0002>Z=0
YF=O
6x
Y0=O:>Z=1
YF=YFF \ %zo 2513203954)
5y ’ 5y 5y
\ 1 I

 

Figure 9-1. Boundary conditions for a single filel segment to solve mixture fraction

equation V 2Z = 0 . The “fuel inflow segment” is shown in red. The rightmost boundary
is for downstream (x—mo) and the topmost boundary is for transverse (y—+oo).

2 _. _
20 = ;tan 1(#0600), (9 5)

where the subscript ‘0’ denotes zero convection. ,u is defined implicitly in the next

equation. Lines with constant 20 values have ,u=c0nstant. At the flame sheet 11 =

tan(7rZOp’2). The functional relation between coordinates x and y at the flame sheet is

obtained by an algebraic calculation in which ,u is an input parameter:

155

 

2 2
y2_\/(x2+1+2# )2+x2(1+:’ _x2)
1“ (9-6)

The equationVZZ = O can also be solved numerically subject to the boundary
conditions shown in Figure 9-1. In the numerical solution, finite values must be assigned
to the domain length and width, i.e., the topmost and the rightmost boundaries are now

located a finite distance from the origin x = y = 0.

9.2.2. Numerical and analytical solution of 6(62 / 6x) + d (62 / 6y) = V 2Z

Here, we assume that the convective terms in both streamwise (x) and transverse
(y) directions are weak. Assuming steady state conditions and constant velocities in the x

and y directions, the mixture fi'action governing equation is

e(6z /6x) + d(6z / 6y) = V22 , (94)

where e denotes a uniform streamwise velocity (in the x—direction) and d indicates a
constant transverse velocity (in the y direction). In the physical problem corresponding to
actual flamelets, the fuel inflow (blowing) only occurs in the region 0 < x < 1 (identical
to the red region in Figure 9-1) and the transverse velocity V is identically zero in the
region x > 1. Thus, the assumption of uniform V for all values of x in this simplified
model is not in accordance with reality for flamelets. However, this model is employed in

order to provide some physical insight to the study of flame shape. Let

156

Z = er+Jy¢(x, y) (9'8)

For the case e —) 0, d —) 0 and ¢ is an arbitrary function. It can be shown that the

lowest order approximate analytical solution for Z is given by

(3x+iy) (99)
Z = e 2 2 20

The function Z0 is the solution for the zero convection case (Eq. (9-5)). At the

flame sheet, where Z = Zf, the flame location is found by solving the equation

11 = tan [(TtZfllZ) exp (-ex/2-dy/2)] (9-10)

To find the flame sheet curves (lines of constant Zf) Eq. (9-6) is used while the

values of p are deduced from Eq. (9-10).

9.3. Discussion: 2-D Theoretical model
A comparison of the numerical solution with the exact analytical solution
ofVZZ =0 in Figure 9-2 validates the numerical scheme used to solve the governing

equationVZZ = O , especially for points close to the flame leading edge. This figure is the

leading edge part of the 10x10 domain. Flame sheets for nine values of Zf are shown in

this figure. This figure demonstrates, however, that applying the far field boundary
conditions at a finite distance from the origin produces the largest deviations between

analytical and numerical results in the far field. This is to be expected.

157

 

 

 

 

4 l r l l I I l
3.5"

3- """"" J1 """"""
2.5- ,x’/ \

>' 2‘ I Zfincreases

1.5- -----------

1~ ‘

‘ \ 0.2\

05L \

o A 1 1 ‘i l 1 L 1

0 05 1 15 2 25 3 35 4

Figure 9-2. Comparison of numerical and exact analytical solutions of the equation

V2 Z = 0 for Zf = 0.1-0.9. (Solid line: numerical solution, Dashed line: exact analytical
solution).

Figure 9-3 shows a comparison of the approximate analytical solution of Eq. (9-7)

and its numerical solution for a zero transverse velocity (d = 0). This figure implies that

for values of Zf near unity (which characterize flame sheets near the surface); the

agreement of results is better. Moreover, near the upstream leading edge where x is small
(the flame attachment region), the numerical and (approximate) analytical solutions show
good agreement. Near the trailing edge, where the flame can, in principle, reattach to the

surface, the (approximate) analytical solution does not predict a nearby reattachment

point for low Zf values, see Zf = 0.1 in Figure 9-3. For other small values of Zf a direct

comparison shows that near the trailing edge the agreement of the (approximate)

158

analytical solution with the numerical solution is often poor. Reattachment is not
observed for larger values of e. The exponential term in Eq. (9-10) increases dramatically

as x increases in the downstream direction.

 

 

 

 

6 I I I I I
5- .
4~ ,-------~------~——c
z’I”
’1’,
/’
> 31- // -n
/
/
/
/
/
/
/
l/
2— / -
/ ,,,,,,,,,,
/ fl,”d \‘\\
/’ \\\
\\
F- -------- ‘ \ 'l
1 \\\ \
—“~‘\ \ \\
a \ \ x
\ \ \
\ \ \ \ K
1 \ LL 1 1 -1 1
0 1 2 3 4 5

Figure 9-3. Comparison of approximate analytical solution and numerical solution of

e(62 /6x) + d (62 / 6y) = V2 Z for e = 0.1 and d = 0. (Solid line: numerical solution;
Dashed line: lowest order analytical solution given by Eq. (9-9)).

Figure 9—4 and Figure 9-5 give contours of Zf for different streamwise and

transverse velocities e, d. When e = d = 0, the curves are exactly similar to those of

Figure 9-2. The red dashed line in Figure 9-4 represents the approximate analytical

159

solution when Zf= 0.5 (i.e. the stoichiometric index is unity). As illustrated in Figure 9-4,
for small velocities, the approximate analytical solution and the numerical solution show
good agreement, and both solutions can be employed to study the problem. The Zf

contours sweep to the right and shrink when e increases for fixed d (relevant figures not
shown here). For large e-values, the approximate analytical solution is not applicable to

solve Eq. (9-7) especially to predict the behavior in the trailing edge (larger x values).

 

 

 

5 l I I l l
4.5“ J
4” -1
3.5”. Zf = 0.5 "

A
3* I
/\ 0.1

Zf increases 0
. \ 4

 

 

 

Figure 9-4. Contours of constant Zf from the numerical solution of

6(62 / 6x) + d(6Z / 6y) = V2 Z for e = 0. 05, d = 0.05. Red dashed lines represent the

approximate analytical solution for Zf= 0.5.

160

 

 

 

 

5 f I I I I
4.5.. /
0.1
4» _
3.5- _
3” ZZfincreases ‘
>' 2.5‘ 0 2 \‘
2' i
1.5 i
0.3
1r -
0.5~ -
o , , 1 1 1 ‘
0 1 2 4 5 6

X0:-

Figure 9-5. Contours of constant Zf from the numerical solution of

e(6Z/ 6x) + d (6Z / 6y) = V2 Z for e = 0. 8, d = 0.2. Red dashed lines represent the
approximate analytical solution for Zf= 0.5.

9.4. Flame-Surface Interaction over Single and Multiple Solid Fuel Segments in a

Channel Cross Flow

In solid fuel combustion, the surface heated by the flame decomposes to small
volatile molecules. These molecules leave the surface and serve as gaseous fuel that

mixes with incoming oxidizer. In the simplified model constructed to analyze the

diffusion flame in the gas phase, a uniform inlet flow of a gaseous fuel (Y f: l) injected

through the surface through one, two or three injection ports along the solid-gas interface
represents the released volatiles. This numerical model is essentially a study of mixing

and combustion of a fuel jet (or a series of jets) in a weak convective oxidizer crossflow.

161

In the model examined here, the incoming oxidizer is air (Y0=0. 233), which enters

the channel at 5 cm/s. The inflow Reynolds number is 630, which ensures that the flow is
laminar. Simulations are conducted using FLUENT version 6.2.16 [177], the mesh being
developed in GAMBlT. The “cold” solution gives mixing and velocity contours without
any reaction taking place. The flow is steady and the geometrical configuration of the
computational domain is intended to mimic the domain of the theoretical solution except
upstream of fuel inlet slot, where a 1 cm inert wall is placed. This 2-D computational

domain with rectangular grids is illustrated in Figure 9-6.

0.1 ,,

 

—> E 0.06
_. ;
—>
0.04 '

Oxidizer flow 5
Fuel Inlet

Slots 0'02

00 0.02 0.04 0.06 0.08 0.1 0.12
Inert upstream wall X (m)

Figure 9-6. Schematic of the upstream section (12 cm x 10 cm) of computational domain
with three 1 cm the] inlet slots and 240 x 150 grids. Grid compression is applied in the y
direction.

162

The gas phase flow field is computed using the finite volume SIMPLE-based
scheme with under-relaxation techniques. The density is found by the incompressible
ideal gas assumption (i.e. p=f(T)). The mixing law is applied to find the specific heat
capacity. The thermal conductivity and viscosity are assumed constant. The constant
dilution approximation is employed for the mass diffusivity. In this study, the fuel and
oxidizer are ethane and air, respectively. A one step, second order, global gas-phase

reaction is assumed that obeys Arrhenius kinetics. Zero gravity conditions are applied
(gr-0).

In the study of cross flow jets, a significant parameter that affects the flow field is

the ratio of the jet to cross flow momentum, r, defined as
2
PIV} )1/2 (9-11)

where subscripts j and of denote the jet and cross flow, respectively. In the case of equal

densities, Eq. (9-11) reduces to r = Vj/ch.

Boundary conditions: the isothermal, uniform inflow boundary condition is
applied for the fuel and oxidizer inlets. The lower walls are insulated with zero fluxes and
a no-slip condition. The upper wall is no-slip and isothermal. The outlet boundary uses a

pressure outlet boundary condition.

The numerical model represents forced convective flow generated in a slow wind
tunnel with finite dimensions in a realistic experiment (the 5 cm/s flow speed is a typical
flow speed in the MSU and NASA NCAs under “flamelet” conditions). The goal, as

Stated earlier, is to examine combustion for single and multiple streamwise fuel inlet

163

segments. Using this model, the flame shape, mixing, and flow fields for three cases with
lean, stoichiometic and rich conditions are numerically investigated. These cases use ¢=
(0.5, I, I. 7). In the fuel rich condition, the momentum ratio is r = 1.025, which decreases
to 0.6 and 0.3 for stoichiometric and lean conditions, respectively. In simulations for two
or three the] segments, the fire] inflow temperature is taken to be 600 K, which is close to

a solid fuel decomposition temperature for a typical solid phase hydrocarbon fuel.

9.5. Discussion: Simulation for one fuel segment

Figure 9-7 to Figure 9-9 present temperature and oxygen mass fraction contours.

For one fuel segment simulations, the fuel inflow temperature is taken to be 300 K. As
illustrated in these figures, Yo z 0 underneath the flame as expected. Also, Y1: z 0 in the

air above the flame. Comparison of Figure 9-7 to Figure 9-9 indicates that the flame
moves downward toward the fuel surface for fuel lean conditions and it lies closest to the

fuel surface in this condition. However, the global flame structure remains similar for all
cases and does not vary significantly. The maximum flame temperature (T f) occurs at a

location where the flame sheet has a large curvature. The fuel inlet segment is located

underneath this maximum Tf area.

At the flame leading edge, where the flame is attached to the surface, a comparison
of theoretical and numerical results shows good agreement and either approach can be
employed to predict the flame shape as well as the local concentration field. There are
important differences, however, in the numerical model, the flame attachment point is in
close proximity to the fuel inlet slot and attached to the upstream inert wall whereas it

was located right at the fuel segment corner in the theoretical model. The numerical

164

0.08

 

 

 

 

0.06
E
>. 0.04
0.02
0
0 0.02 0.04 0.06 0.08 0.1
X(m)
(a)
0.1 —
0.23
0.08 + 8.30
0.19
84%
A 0-06 ’ 0.14
E 81:
>' 0210
0.09
0.07
0.06
0.04
0.03
0.01
0.00

 

 

0 0.02 0.04 0.06 0.08 0.1 0.12
Km
(b)

Figure 9-7. Contours of a) temperature and b) oxygen mass fraction in fuel rich
conditions ((0 = 1.7).

165

0.1

0. 08

0.06

Y(m)

0. O4

0. 02

   

0 0.02 0.04 0.06 0.08 0.1 0.12
MW
(at)

Y (m)

0.04 *-

0.02 -

 

QQQQQQQQQQQQQQQQP
89828388363836888

 

0 0.02 0.04 0.06 0.08 01 0.12
X(m)
(b)

Figure 9-8 Contours of a) temperature and b) oxygen mass fraction in stoichiometric
conditions ((p = 1).

166

0.1

 

 

 

0.08
0.06
E
>.
0.04
0.02
0 _
0 0.02 0.04 0.06 0.08 0.1 0.12
X(m)
(a)
0'17
0.23
0.08~ 8230
0.19
84%
..0'06” 014
s 81:
>_ .
0.10
0'04i 0.09
.- 0.07
, 0.06
_ . 0.04
0.02 A 003
0.01
0.00
0o 0.02 0.04 0.06 0.08 0.1 0.12
X0“)
(1))

Figure 9-9 Contours of a) temperature and b) oxygen mass fraction in fuel lean conditions
( a) = 0.5).

167

graphs show that near the trailing edge, the flame does not reattach to the surface. The

agreement ranges widely and is often poor in this area.

This study demonstrates that theoretical predictions disagree in important features
with numerical simulations. The suggestion is that the simulations are “sensitive” to

certain quantities. Such a level of sensitivity suggests (but does not prove) instability.

The two—dimensional model, with its sensitivity to downstream conditions, suggests
that a 2-D flamelet structure would be difficult to achieve in practice. The model
examined here hints that the actual localized flames and flamelets attached to surfaces are
more likely to be 3-D entities. This is, in fact, the case in laboratory experiments on
flames in “near-limit” conditions, and it is also the case for flames observed over most
propellants (see the extensive experimental analyses using advanced diagnostic

techniques in [166]).

9.6. Discussion: Simulation for multiple fuel segments

Figure 9—10 shows the contours of oxidizer and fuel mass fractions near the flame

leading edge. The simulations, just as for flame spread, show that Y0 z 0 underneath the

flame and Y}: z 0 above the flame. The flame shifts downward and it lies closest to the

fuel surface for the fuel lean condition. The flame structure remains the same for all three

stoichiometries: its shape does not vary significantly. Similar to the previous case with

one fuel inlet, the maximum Tf occurs above the fuel inlet slot, where the flame has a

large curvature, in agreement with flame-spread results presented in our ASME

publication [1 64] .

168

 

0.01

O 008 - / /
O / /
_. / / /
/ / / /
/ /
/
/ / / / //i
— / / / / /// ///
/ / / / // ///////A
00006 - / / / // //// ///
/ / / / ' ’/
h / / / / //.
E __ / /////// 0 02
> ._ / / ////

/ / // // /
r //// /
/ / //// /// 0
0.004- / / xx Wm 05
.. / /////

/

   
   

' / / / /////// /

0.002 - / / / /////// /

/ / /////////
/ / /////////
’ / / /////////
L— / / /////////
1 1 il l ildlhlll' 11

0
0.004 0.006 0.008 0.01 0.012 0.014 0.016
X(m)

 

 

 

Figure 9-10. Y0 and Y F contours near the leading edge (dotted line: oxidizer; solid line:
fuel)

At the flame leading edge attachment point, both theoretical and numerical
approaches can be employed to predict the flame shape and structure. However, near the
flame trailing edge, the numerically-established flame does not reattach to the surface,
being instead almost parallel to the wall. As illustrated in Figure 9-4 and Figure 9-5 for
the approximate analytical solution with constant streamwise and transverse velocities,
the flame trailing edge plunges towards the surface when the convective flow amplifies.
A reattachment point in fact existed in those graphs, in direct contradiction with the

present numerical simulations.

169

0.06

 

IIII'IIIIIITIY'IIII

p

O

—A
llll'lllll

 

 

 

/
/I-/._ 1

0.01002 0.03 0.04 0.05 0.06 0.07
X(m)

 

C

Figure 9-11. Temperature contours upstream of the flame for the case with two fuel inlet
segments ((p = 1).

Figure 9-11 and Figure 9-12 illustrate contours of temperature for two and three
fuel inlets, respectively. All cases here describe the stoichiometric condition (go = 1).

Figure 9-13 presents reaction rate contours for the case with three fuel inlets. This

figure shows that the maximum reaction rate is near the upstream fuel inlet. Product
dispersion, observed in Figure 9-14, shows that product (here, C02) concentration field

broadens toward the downstream.

170

 

 
   
 
 

 

 

 

0.06 _
0.05 _—
0.04 :-
E 0.03 i
0 02 L AAA 589
. ' /
0.01 _- 00/:////7
' ‘6 'fl/rozz/m “
~ 3
o_ .I»-__. 1/1111 llllK7§Jhll
O 0.01 0.02 0.03 0.04 0.05 0.06 0.07

X (m)

Figure 9-12. Temperature contours upstream of the flame for the case with three fuel
inlet segments ((p = I ).

The study carried out here suggests: (1) The 2-D flame structure is unstable to
minor changes in surface and far-field boundary conditions; (2) The flame-surface system
with multiple fuel inlet segments in Section 9.6 can not resemble separate flamelets, one
behind the other, displaying instead a single, attached flame. Separate flames cannot, it
seems, exist directly downstream of one another unless the material contains its own
oxidizer (e.g., propellants). Even for propellants the downstream flame(let)s must
contend with passing product flow; (3) Multiple flames likely exist in the form of arrays

in which individual flame fronts (flamelets) propagate toward the incoming oxidizer.

171

 

Y (m)

 

0.2 0.3 0.4
x (m)

00 0.1

Figure 9-14. Product dispersion plot showing broadening of the C02 mass fraction ((p =
1). Three fuel inlet segments are employed (Figure is not to scale).

172

Unless the heterogeneous material contains oxidizer, the array is simply an advancing
fiont [160]. In order to describe the burning of heterogeneous fuels and fuel/oxidizer

mixtures, it is necessary to examine fully three-dimensional models.

173

Chapter 10 Flame and Flamelet Spread over Thin
Fuels in Simulated Microgravity Conditions

Most previous research on flame spread over solid surfaces (e. g., Chapter 9) has
examined flames in open areas. In this study, by contrast, the flame spreads in a narrow
gap (e.g., the NCA), as occurs in fires behind walls or inside electronic equipment. This
geometry leads to interesting flame behaviors not typically seen in open flame spread,

and also reproduces some of the conditions experienced by microgravity flames.

Research groups at MSU and NASA have conducted two sets of experiments
involving flame spread in a Narrow Charmel Apparatus (N CA, available at MSU and
NASA) in normal gravity, and the others taking. place in actual microgravity in the
NASA drop facilities. Three important variables are varied: oxygen concentration, flow
velocity, and heat loss. Complete details of experiments in both facilities can be found in

[155].

The research to be described in this chapter addresses flame and “flamelet” spread
over thermally-thin fuels [160, 168, 178-182] in simulatedreduced buoyancy conditions.
When the oxidizer flow is reduced in the NCA, the initially uniform flame front becomes
corrugated and breaks into separate flamelets. “Flamelets”, as introduced before, are
small, 3-D flame fragments that are formed when a wide, 2-D flame front weakens to the
point where it can no longer be sustained, and breaks up into individual flames separated
by distinct gaps of non-burning material [160]. This breakup behavior allows the flame to
keep propagating below standard extinction limits by increasing the oxidizer transport to

the flame, but has not been observed in other microgravity experiments due to the narrow

174

samples employed. Breakup cannot be studied in typical (i.e., “0pen”) normal gravity test
facilities due to buoyancy-induced opposed flow velocities that are larger than the forced

velocities in the flamelet regime.

In the experiments, three contributors to flamelet formation are varied. These
contributors, which are controlling variables in real fires in confined spaces, are: (l)
ambient oxygen concentration. Reducing the oxygen concentration weakens the flame,
making it more susceptible to quenching; (2) the flow velocity of the incoming oxidizer.
If the velocity is sufficiently high, the flame front is uniform. If it is too low, the flame
may not receive sufficient oxygen for combustion [183]: at low enough flow rates it may
break into flamelets or extinguish. Thus, flow velocity and oxygen concentration are
related, in that each controls the amount of oxidizer reaching the flame, but in addition
the flow contributes to convective heat loss from the flame; (3) heat loss to a nearby
object. A metal or other substrate placed behind the thermally-thin test sample draws
thermal energy, weakening the flame and making it more likely to fi'agment. A thick

sample serves as its own heat sink by conducting heat into the interior of the solid.

Flamelets exist on the margins of flammability. The flamelet regime extends the
range of material flammability to lower opposed flow velocities, lower oxygen
concentrations, or higher heat loss by increasing the dimensionality of the flame (surface
to volume ratio = S/V) and thus enhancing oxygen transport to the flame zone. Our
colleagues, Drs. Olson and Miller, have evaluated the surface to volume ratio (SN) for
an ideal cylindrical flame (i.e. uniform flame) compared to the nearly spherical flamelets
[155]. The increase in SN ratio enhances the oxygen transport to the flamelets through

the flame surface while focusing the heat release in a small volume, which then can

175

 

provide sufficient heat flux to the unburned fiiel
beneath. This allows the flarnelets to survive.
They occur only under low oxygen flux
conditions. Thus, the flamelet’s multidimensional
adaptation extends the material’s flammability

beyond the uniform flame extinction limit.

From a fire safety viewpoint, flamelets are
by nature small and hard to detect, since they
occur at the limits of extinction. They have been
found for opposed flow and concurrent flow under
similar low oxygen flow conditions [160]. They
can persist indefinitely under the right conditions,
and can flare up rapidly into a large fire when the
environmental conditions become favorable. See
Figure 10-1. It is standard practice on US
spacecraft for the astronaut crew to turn off the
ventilation to help with the extinguishment of a

fire, both to eliminate the fresh oxygen supply and

 

Figure 10-1. A single flamelet (~6 mm wide) propagates steadily in an air flow of 0.5
cm/s in the NASA NCA. When the air flow is abruptly increased to 50 cm/s (100x), the
flarnelet grows 200% in less than 10 seconds. (Each photograph is 2 sec afier the
preceding one) [155].

176

to reduce the distribution of the smoke. If some tiny flamelets go undetected until the
ventilation system is reactivated, the sudden increase in flow might allow the enduring

flamelets to grow into a large fire very rapidly, posing a significant fire hazard.

In this chapter, the MSU NCA is briefly explained. A scaling analysis is carried out
to determine the characteristic magnitudes of the various velocity sources in this NCA.
The goal is to examine the suppression of buoyancy. It also aims at understanding the
conditions under which flamelet phenomenon appears in experiments. Appendix B
discusses the flammability maps generated from experimental data in the NCA and actual

microgravity tests conducted in the NASA 5.18 second drop facility.

10.1. The Narrow Channel Apparatus (N CA) for normal gravity experiments

A Narrow Channel Apparatus (N CA) has been developed to study the
phenomenon of flamelet spread in narrow gaps that have suppressed-buoyancy, high heat
loss conditions [160, 168]. The NASA NCA, shown schematically in Figure lO-2a,
reduces buoyant convection by limiting the channel vertical gap height (Figure 10-2b) to
a value where vertical cellular flow caused by a hot source is small compared to the
horizontal (streamwise) forced flow. This streamwise convective flow is forced through
the duct at a controlled low speed between the copper bottom plate and the quartz top
plate, with the sample held midway between the plates (Figure 10-2b). Tests can be done
in either an upward facing orientation where the sample is viewed from above (quartz
window on top, copper on the bottom), or the downward facing orientation, where the

sample is viewed from below (copper on the top, quartz window on the bottom). Ignition

occurs at a high flow (~ 20 cm/s), followed by a linear flow ramp down (~ 0.5 cm/sz) to

177

   
 
  

“Ty/l Outlet (wl honeycomb

Window
Inlet (wl honeycomb flow flow straightening section)
’. .

stralghtening section) _

 

 

 

(a)
Duct Top (Quartz Window) Gravity
_’ Flame _ kGap
Forced ”CM—a Thin Fuel 6 ‘LHeiaht
Duct Bottom (Copper)
0))

Figure 10-2. Schematic of the Narrow Channel Apparatus (NCA). a) Flow enters the left
side through holes in the tube, and is distributed in the plenum section, straightened
through screens and honeycomb, and then enters the test section. The flow exhausts from
the right plenum. b) Side view of test section, showing the thin firel suspended in the
center of the narrow channel.

either a fixed flow for steady spreading at that flow, or a continuous very slow ramp
down rate (~0.03 cm/sz) until extinction occurs.

A color digital video image is recorded for each test, imaging through the quartz
window onto the surface of the fuel sample. All video records from the normal gravity

(upward facing or downward facing) and microgravity tests are analyzed. The analysis

includes flame/flamelet spread rate, flarnelet size, and burned fraction.

178

Microgravity experiments, conducted by our colleagues, Drs. Olson and Miller,
in a low-speed wind tunnel drop rig [183] in the NASA Glenn Research Center’s 5.18
second Zero Gravity Research Facility (ZGRF), are not reported in detail in this
dissertation. The test procedure is briefly explained here. Whatman 44 ashless filter
papers and PMMA were used as fuel samples by NASA. During each test, the flow was
first allowed to stabilize. The sample was then ignited in normal gravity so that a stable,
uniform, propagating 2-D flame front was established prior to the drop. Once stable
flame propagation was observed and the flame spread away fi'om the igniter (~5-10
seconds), the experiment package was released. Two on-board video cameras captured
the flame before and during the dr0p, one fiom the side and one from the top. At the end

of the drop, the chamber was evacuated to extinguish the flame.

10.2. Flame and flamelet spread in the NCA

Flames spreading across samples held in the center of the NCA (as shown in Figure
10-2b) break up into flamelets just as the microgravity flames described in [155].
However, we have time to observe the spread across the entire sample that took almost
250 seconds (4+ minutes) for the case shown in Figure 10-3, instead of only a few

seconds as in the microgravity tests. For this test, the sample was ignited at 8 cm/s. The

flow was ramped down to 4 cm/s at a rate of 0.5 cm/s2 to reach the flamelet regime,

followed by a slower ramp to 2 cm/s at the rate 0.03 cm/sz. The flamelets continued to

spread in the 2 cm/s air flow. As they approached the end of the sample, the air flow was
quickly ramped back up (not shown) and the flamelets merged back into a single flame

front. A gap spacing of 5 mm (i.e. h=l cm in Figure lO-2b) on both sides of the sample

179

was used to mimic the heat loss from the microgravity test in which a substrate at 2.5 mm

existed on only one side of the sample.

Mr 111])“

Ilzmiclct spread

l8.5 cm —>

Bifurcation

Extinguishment

 

Figure 10-3. Flamelet tracks on the filter paper in the normal gravity NCA, illuminated
with green LEDS. The uniform flame consumes the entire sample (black solid area at the
left side of the image) and then breaks up into ~10 flamelets as the flow is reduced. The
unburned paper between flamelets appears green, while the flamelet tracks are black. The
flamelets are the luminous tips at the ends of the branching pattern [155]. The falmelet
burned fraction is approximately 0.6.

As shown in Figure 10-3, some flamelets may bifurcate and some may extinguish.
A detailed analysis has shown that there is extensive order in the seemingly random
patterns [160, 179-182]. Olson and Miller’s recent analysis [155] has also shown that the
flarnelet phenomenon is not an ‘instability’ consisting of a transitional state heading to
extinction. Rather, the flamelet state is steady in terms of persistence, spread rate, fraction
of sample burned, etc. Thus, the flamelet regimes in the flammability maps are a stable,

near-limit, multidimensional flame adaptation to the limiting oxygen environment.

The flame and flamelet spread rates for the NCA (with 5 mm gap) agree well with

Zero Gravity Research Facility (ZGRF) data for a comparable 2.5 mm substrate gap

180

[155]. Experiments with PMMA in the ZGRF showed that flamelets are not unique to

cellulosic or charting fuels.

10.3. Velocity scaling analysis in Narrow Channel Apparatus
In the NCA, there are at least five velocities that can be used in a scaling analysis.
As depicted in Figure 10-4, these are Vflow, Vbuoya Vfiwrm, Vdifir, and Vblow- Our scaling

analysis evaluates the magnitudes of these characteristic velocities for typical NCA

conditions. The scaling analysis is canied out by performing numerical

Quartz
z/ ‘ Buoyant Flow Vbuoy

Thermal Expansion Flow
Vtherm

WDiffusion Flow Vd/‘ff
Blowing Outflow Vblow
- - "'T-“I:;--.;.-.:-:..3.:.-.r.:.-_.:.:.. C h a rred S u rfa Ge

 

Inflow Flame

  

Vflow

 

 

‘ .. . . I . - . . .
_ .‘ . , . . , _~
. v . - . . . - n V
I . . 1‘ . x
.’n." il I I". I'-

Line of Symmetry

 

Figure 10-4. Definition of the various velocities that occur in the NCA. The sample is

placed between top (quartz) and bottom walls. Inflow is left to right, while the flamelet

moves right to left. In this figure, gravity vector points downward; buoyant velocity is
upward.

simulations of narrow channel duct flows over short heated sections of the duct lower
surface. These heated duct wall sections (we use L= 0.5 cm length of surface heated to
1500K) simulate the flamelets, see Figure 10-5. The strong local lower-wall heating

produces thermal expansion and, in lg, vertically directed buoyant flow. Our model does

181

not consider lower surface blowing (i.e., mass addition fiom thermal decomposition of

the sample) although we estimate Vblow and VM using alternative methods. Three inflow
speeds are used: Vflow = (5, 10, 20) cm/s. The simulations are 2-D, hence the duct is

assumed infinitely wide. Simulations are conducted for zero gravity and normal earth
gravity.

The simulation employs the FLUENT code, version 6.2.16 [177]. The inflow air
enters the narrow channel 5 cm upstream of the hot spot leading edge. The slug inflow

velocity profile evolves into a parabolic profile before it reaches the hot spot. No-slip

conditions are applied along the channel surfaces. The constant pressure outflow

 

 

 

 

 

 

 

 

 

 

}'f
0 Not to scale
A
y=2h/3
h
y=h/3
x=0.5 cm
‘ : X
f 5.0 cm v3.5 ch; 6.0 cm '
hotspot

Figure 10-5. Schematic of geometry of numerical simulation used for velocity scale
analysis. The channel height used was h=5 mm, and the hot spot was also 5 mm centered
at x = 0 (the vertical (y) axis is located at x = 0). Flow velocity (5, 10, 20 cm/s) and
gravity level (0g, lg) were varied.

condition is applied at the channel exit 6 cm downstream of the hot spot trailing edge. For

the energy equation: the inflow has constant (ambient) temperature; the outflow has zero

182

temperature gradient; the hot spot temperature is 1500K; the lower surfaces upstream and
downstream of the hot spot are insulated; the top surface was insulated (zero normal
gradient), except for a comparison test with constant temperature (300K). Under most

flow conditions the differences caused by changing the top-surface boundary condition

from T = 300K to zero heat flux (q”= 0) are negligible. As illustrated in Figure 10-6, for

the lower-speed inflows the condition T = 300K changes the buoyant flow slightly. When

 

  
 
   

2.1 i . —o—T=Const., 09
o ° 0 . .__°‘__qfl=o’og

1.7 . +T=Const.,1g
q"=0. 1g

 

 

 

 

1.3 -
a
E
8 0.9 —
>~i
>

0.5 ~

0.1 ~

0.0 0.1 0.2 0.3 0.4 0,5

 

-0.3 -
Y (cm)

Figure 10-6. Numerical results for vertical velocity at x=0 for isothermal and adiabatic
top wall boundary conditions under two inflow (forced flow) velocities: Vfl0w= 5 cm/s
and Vfl0w= 20 cm/s.

183

the inflow speed is higher (between 10-20 cm/s) the difference in the flow fields between

the two sets of boundary conditions is small (see Figure 10-6). Thus, we consider only

q ”= 0 at the top channel wall.

The characteristic magnitudes of Vbuoy and VMem, are determined by comparing the

velocity distributions for zero and normal gravity for identical values of Vflow and h. The

velocity distributions are evaluated along imaginary lines in the duct: two horizontal lines

at (1/3)h and (2/3)h and two vertical lines located at the middle of the heated section
simulating the flamelet (x = 0), and at x = 2.5 mm behind the downstream edge of the

heated section, respectively. These locations are illustrated in Figure 10-5.

10.3.1. Results and discussion: Velocity scaling analysis in NCA

All surfaces except the heated section simulating the flamelet (T = 1500K) are

insulated (q”= 0). Profiles of the heat flux (—k6T/6Y ) along the heated surface (not

shown here) delineate that the difference between the two gravity levels, 0g and l g, for
inflow velocities 210 cm/s is nearly negligible. The heat flux profiles (also called
signatures) are valuable to determining how a solid material will heat up and can be

eventually used to determine the ignition delay times.

Shown in Figure 10-7 is Vy as a fimction of x at y = (I/3)h; results at y = (2/3)h are

qualitatively similar. This figure shows that at zero gravity there is no downflow
upstream of the flame, whereas at normal gravity there is a small downflow for the lowest
value of the inflow velocity (5 cm/s); otherwise, there is not a substantial difference

between normal and zero gravity profiles. The results also show that the largest and

184

smallest values of Vy along these lines are always of greater magnitude for the higher

flow velocity values. Although these extrema increase with Vflow the increase is not
linear. The relationship is demonstrated for the zero-gravity case in Figure 10-8. The 2nd
order fit through the data points yields a negative coefficient for the quadratic term. This
indicates that as the forced flow is increased, it gradually overcomes the flow due to
thermal expansion (since there is no buoyant flow in zero gravity). Note that the reason
the thermal expansion velocity increases at all as the forced flow velocity is increased is
because the heated section maintains a constant temperature. Hence, more thermal energy

enters the flow as the velocity is increased.

 

Vy, [cm/S]

 

 

 

 

 

x [cm]

Figure 10-7. Numerical predictions of the vertical velocity component along the line y =
(1/3)h for three values of the inflow velocity at both normal and zero gravity.

185

 

 

 

 

 

3 _.__
= - 0.0013x2 + 0.16x + 0.1333
2.5 ~
2 a
E"
3
x 1.5 -
1
,-
>
1 a
0.5 « ’5 <> (1/3)h
A (2/3)h
-— Poly. ((1/3)h)
- . . . Poly. ((2/3)h)
0 I l I T Hi
0 5 10 15 20 25
In-Flow Velocity (cm/s)

Figure 10-8. Maximum velocity for three values of the inflow velocity at y = (1/3)h and y
= (2/3)h for zero gravity conditions.

Figure 10-9 shows the vertical version of Figure 10-7 at x = 0. The velocities are of

the order of approximately I cm/s. At Vflow = 10 and 20 cm/s the normal and zero gravity

vertical velocities are essentially identical, the maximum difference being ~ 0.2 cm/s. At
an inflow of 5 cm/s, however, buoyancy plays a significant role and shifts the peak of the

velocity to a lower height in the channel and raises the maximum value.
In Figure 10-10, the maximum and minimum Vy differences [Vy(1 g) — Vy(0g)] are

examined along the line y = (1/3)h. The quantity [Vy(1 g) -— Vy(0g)] in Figure 10-10 is

186

2.0 .

Vy 1.0:
[cm/s]
0.5
0.0‘

 

 

 

y 1ch

Figure 10-9. Numerical results of vertical velocity component for the six cases along a
vertical slice centered above the hot spot at x = 0.

interpreted as the buoyant velocity. When g = 0 there is only thermal expansion: at
normal gravity there is also buoyant flow. The magnitude of the buoyant velocity is
larger at the lowest forced flow rate, and drops as the forced flow is increased. Negative

values indicate the down-flow that is seen in the 1g case ahead of the heated section.

The inset to Figure 10-10 shows the maximum and the minimum buoyant velocity
as a function of forced flow, along with their difference. Here we note that the difference
in the extrema approaches zero as the forced flow is increased. Contrasting Figure 10-8
and Figure 10-10, the thermal expansion velocity effect drops slowly as the forced flow
velocity is increased, whereas the buoyant velocity drops more rapidly. At 10 cm/s forced

velocity, for instance, the maximum buoyant velocity is less than I cm/s, while the

187

maximum thermal expansion velocity is 1.5 cm/s. Both are about 10% of the forced flow

 

velocity.
1.5
--—- 5 cmls
—-"10 cmls
1 "" 20 cmls

0.5

 

 

1 -o— Vbuoy.max
—a— Vbuoy. min
~ —-— Delta Vbuoy ‘

Vbuoyant " Vy(1g)-Vy(09) [cm/sec]
o

 

 

 

-0. 5
y
'1 Buoyant Flow, [cmls]/ _1 //
0 0

lLLLLIlIllllllLllllllIIllll|l

10
Vflow: [cmls]

-1.5 ""l""l""l|"'11"I'l'lll'lllll1111111111"
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3

x [cm]

 

 

 

 

Figure 10-10. Numerical results of the buoyant velocity, calculated as the quantity

[Vy(1 g) — Vy(0g)] along horizontal slice y=(1/3)h. Inset shows the trend in maximum
buoyant flow as a function of forced flow.

For Vblow a different approach is used. Here, 3 Stefan flow is used to scale Vblow.
. 2 .
First, we assume that Df ~ 0.3 cm /s for a fuel vapor of moderate molecular weight at ~

400 0C, a reasonable near-surface fuel temperature. Then, with Pa = 1 atm the fuel

188

partial pressure gradient may be estimated as (1-0) atm / 0.2 cm = 5 atm/cm using pure

fuel at 1 atm at the fuel surface that is totally consumed in the flame 2 mm above the fuel.

We find the result VbIow z (Df /Pa)de /dy ~ (0.3/U5 = 1.5 cm/s. Thus, Vblow is

substantially smaller than, Vflow- For a Vflow value of approximately 10 cm/s, Vblow is

between 10-20% of this value.

It is also possible to evaluate Vdifl in which constituents (e.g., oxygen) diffuse
toward the flame. Here we use Fick’s law: Vdijf = - Dar/ax. Putting 6Y/6x ~ (0.233 —
0)/(0.5-0) ~ 0.5 cm-1 and multiplying by D~ 3 cmz/s (oxygen at flame temperature) gives
defi” 1.5 cm/s, which is of the same order of magnitude as Vbuoy and Vthem.

These calculations suggest that when buoyancy is suppressed the other velocities
are all of the same order of magnitude (e. g., 10-20% of the forced inflow velocity). When

buoyancy is not suppressed, and the channel is not narrow, even though the forced inflow

is reduced, Vbuoy can become of the order of 30 cm/s [184]. This value is large enough to

dominate these other velocities and to prevent the appearance of flamelets. This is the
case in ordinary flame spread under standard 1 g conditions. If, on the other hand,
buoyancy is suppressed with a NCA, all of the velocities are of the same order of
magnitude as the slow inflow rates. As a direct consequence the flamelet phenomenon

OCCllI'S.

These results are in agreement with experiments of Olson and Miller [155] using a

NCA in normal gravity and the NASA drop tower microgravity facility. The experiments

189

 

confirmed that the NCA quantitatively captures the essential features of the microgravity
tests for thin fuels in opposed flow. The NCA essentially suppresses buoyancy and serves
to produce test conditions that can, to a satisfactory degree of accuracy, simulate the
conditions achieved in actual microgravity. Under suppressed buoyancy conditions, a
uniform flame fi'ont becomes corrugated and breaks into separate flamelets as the
opposed flow is reduced. Flamelets were found under the same conditions in both the
NASA drop facilities and in a NCA which suppresses normal gravity buoyancy. Good
agreement was found between flame and flamelet spread rate and flarnelet size between
the two facilities. Experimentally-generated flammability maps (see Appendix B)
constructed by Olson and Miller [155] delineate the uniform regime, the flamelet regime,
and extinction limits for thin cellulose samples in microgravity and in the normal gavity
NCA as a function of opposed flow velocity, heat loss, and oxygen concentration. The
three regions representing (uniform) flame spread, flamelet spread and extinction

generally agree over all classes of tests.

NASA’s low gravity testing of spacecraft materials faces many limitations,
including short test time and high cost. The short available test times in ground-based
microgravity facilities make it difficult to study the flammability of thick materials or
those of low flammability. The NCA, shown here to simulate microgravity conditions,
provides an extremely long test time (of the order of minutes to tens of minutes)
compared with the few-second time intervals available in ground-based drop facilities.
Thus, the NCA provides a new method to study non-buoyant flame spread without the

costs and limitations of actual low- gravity testing.

190

The applicability of the “buoyancy free” condition is not restricted to microgravity
conditions. In fact, flame spread in tight gaps and narrow spaces is an important fire
safety challenge in normal gravity conditions. In normal gravity, the flamelets are a fire
hazard since they can persist in small gaps where they are hard to detect. When the air

flow is increased, the flamelets can grow abruptly and generate a deadly fire.

191

Chapter 11 Conclusions

11.1. Conclusions: Jet fuel combustion

In Part I, combustion of jet fuel and jet fuel surrogates, which are blends of large

molecular weight fuels with complex kinetics, was experimentally studied.

A well-defined baseline ethylene flame under incipiently sooting conditions was
perturbed with the addition of 2000 ppm on a molar basis of either jet fuel or two jet fuel
surrogates: 1) the six-component Utah/Y ale surrogate that has previously shown good
agreement for temperature profiles and extinction limits and has captured oxidation and
pyrolysis trends of jet fuel in non-sooting diffusion flames; 2) the two-component Aachen
surrogate that has been reported to mimic not only conditions of extinction and
autoignition, but also to match the soot volume fraction behavior of j et fuel. In addition to
attempting to validate the surrogate formulations with respect to the flame structure, a
first experimental dataset for the pyrolysis and oxidation behavior of jet fuel in a
diffusion flame at the onset of soot formation (incipient sooting conditions) is provided
and can be used in parallel research efforts. The similarity in profiles of temperature,
major species, and one minor species in all flames lends credence to the use of this
perturbation approach to treat the flame as a flow reactor, with virtually identical
conditions, except, of course, for the imposed perturbation. Principal conclusions follow

for this part of the research.

The data for C7-C12 alkanes in Chapter 4 are consistent with the typical

decomposition pattern for large alkanes with both surrogates showing reasonably good

192

qualitative agreement with jet fuel in their pyrolysis trends. More quantitative agreement
is difficult to achieve, because of the complex chromatograms of jet fuel with

overlapping peaks due to isomerism.

Olefins are formed as the firel alkanes decompose. Both jet firel and surrogate
doped flames show good agreement in their l-pentene profiles. The agreement with
respect to l-hexene, on the other hand, is relatively poor with the Utah/Y ale surrogate-
doped flame and the Aachen surrogate-doped flame having concentrations a factor of 2.7
and almost 1.5 larger than jet fuel-doped flame, respectively. The agreement improves for
smaller alkenes, such as propene, probably because the specific details of the fuel alkane

structure are less important since many pathways can produce the smaller intermediates.

The concern in terms of significant aromatics discrepancy between jet fuel and the
Utah/Y ale surrogate that was reported for non-sooting flames in a previous study at Yale
are not applicable in the present, and more relevant, situation of incipient sooting. In fact,
good agreement was found between the jet fuel-doped and the surrogate-doped flames
with respect to the location and magnitude of the benzene mole fraction peak, with the
Aachen surrogate performing slightly worse and producing about 20 % less benzene than
the others. Comparison of toluene mole fractions shows also reasonably good agreement
between flames, with the Aachen surrogate still underperforrning and underpredicting the
peak magnitude by almost 30 %. Nevertheless, the somewhat worse, but still adequate,
performance of the Aachen surrogate is compensated by its simplicity, since it consists of

only two components as compared to the six components of the Utah/Y ale surrogate.

Acetylene presents a unique behavior with peak concentrations nearly 19 times

larger than in the previously studied non-sooting methane flames, as expected, since

193

acetylene is a major contributor to soot formation. Furthermore, the profiles show a
multimodal behavior, with a first local minimum that may be attributable to acetylene
participation in the formation of soot precursors such as benzene and other large
pyrolysis products, and a subsequent local minimum, which may be the result of its

participation in surface growth of soot particles.

The influence of trace amounts of sulfur compounds in jet fuel or in the antistatic
additives on the formation of soot precursors was examined. No distinct changes in the
soot zone luminosity and thickness were observed, which suggests that the observed

trends in terms of soot and soot precursors are not an artifact associated with the antistatic

additive.
l 1.2. Recommendations and future work: Jet fuel combustion

Recommendations for the future work on jet fuel and jet fuel surrogate research are

presented in a point form below:

1- Other surrogate candidates (e. g., MURI teams, SERDP) and their individual

components can be exarrrined systematically.

2- To the knowledge of the author, there exist no studies on the detailed chemical
structure of F-T jet fuel flames. Moreover, surrogate formulations for F-T fuels
have not been studied as comprehensively as JP-8 or Jet A surrogates. When these
alternative fuels are commercialized, a daunting challenge will emerge: The need
for proper (more complete) surrogate formulations with subsequent compilation of
detailed reaction mechanisms from available (if available for iso-paraffinic

components) mechanisms for surrogate components. F-T jet fuels and their

194

surrogates can be systematically studied without a need to significant changes in the

experimental procedure used in this work.

Preliminary computational predictions of the surrogate-doped ethylene flames by
Professor Smooke at Yale did not show an acceptable agreement with experiments.
The chemical kinetic mechanism in the computational model should be fine-tuned

to improve agreement with experiments.

For guidance in the chemical analysis, sensitivity tests in the chemical kinetic
model can be applied. Once more important species are realized, calibration curves

are found for quantification of these species.

Specifically for the evaluation of acetylene profiles, Laser-induced Incandescence

(LII) to measure soot volume fractions can be utilized.

Similar work can be performed at high pressures (up to 40 atm). A high pressure
counterflow setup has been recently constructed at Yale. The GC/MS setup should
be modified in such a way to make the sample loops portable for collection of

gaseous samples from the hi gh-pressure counterflow burner.

A reduced kinetics strategy can be implemented for both the Utah/Y ale and the

Aachen surrogates.

11.3. Conclusions: Biogas combustion and kinetics

The main conclusions of the technical research carried out in Part II on biogas

combustion and chemical kinetic modeling are summarized below.

1) Biogas, produced from anaerobic digestion of biodegradable materials such as

agricultural and municipal wastes, is presently a viable alternative fuel (e.g., for gas

195

turbines). It has principal chemical constituents CH4, C02, N2, and can be saturated with
water vapor.

2) A laminar counterflow diffusion flame, as a fluid-dynamically simple
configuration that intrinsically couples both chemistry and transport, was used in Chapter
6 to evaluate three reaction mechanisms and to understand thermo—chemical effects of
each component. Both detailed mechanisms (GRI-Mech 3.0 and San Diego) give very
close quantitative results. The GRI-Mech 1.2 reduced mechanism overpredicts the
concentration of some minor species such as H radicals (important for chain branching

reactions) although it satisfactorily predicts flame temperatures.

3) In Chapter 6, combustion characteristics such as equilibrium state properties,
ignition delay times, and laminar flame speeds were investigated and characterized for
blends of air premixed with diluted methane simulating biogas using a systematic

chemical kinetics study and GRI-Mech 3.0 detailed kinetics mechanism. Correlations

were found to predict the behavior of individual diluents (e.g., C02 or N2) in a simulated
biogas.

4) Detailed kinetic modeling yields to a comprehensive knowledge about chemical
effects of C02, N2, and H20 addition to a methane gas. Besides chemical effects, each

component in a fuel blend possesses thermal and transport effects as well as radiative

effects.

5) The thermal and chemical structure of simulated biogas in diffusion flames was

studied in Chapter 7. The numerical procedure allows quantitative comparison of the

196

chemical and thermal effects of the C02 content of biogas. Quantification of the chemical

influences of biogas C02 content is vital because C02 dilution was shown to reduce soot
precursors, NOx emissions, and greenhouse gas emissions even when flame temperature
reduction is not feasible. The biogas, simulated as a CH4-C02 mixture, is highly diluted

with N2 in the fuel stream. Four flames with nearly identical strain rates and

stoichiometric mixture fi'actions are considered. These flames (Flames 1-4) have C02

mole fiactions 0, 0.069, and 0.15, respectively. The boundary conditions for one of the

biogas flames (Flame 3) are adjusted in order to ensure indistinguishable temperature

profiles between that flame and the pure CH4 flame (F lame 1). The principal conclusions
of the research presented in Chapter 7 are:

5-1) As discussed in Section 7.2 the reduced mechanism with 22 species and 104
reactions fails to perform adequately for the chemical kinetics study of the biogas
diffusion flame because it poorly tracks the minor species and chain branching radicals

important in the detailed study of a simulated biogas.

5-2) The C02 dilution reduces the concentrations of H, OH, and O radicals. The

thermal effects of C02 dilution on H, O, and OH radicals are nearly the same, whereas

for H radicals, chemical influences predominate. The direct chemical effect of C02 on

the concentration of CH3 radicals is modest compared to its thermal effect. However,

chemical effects of C02 are more pronounced for some intermediate species such as

197

CHzO. The Flame 3 biogas (61%CH4-39%C02) reduces the CH radical concentration
suggesting a possible reduction of prompt NO.

5-3) The reduction in the peak of acetylene, an important soot precursor and smface

growth species, suggests that biogas may be beneficial for soot suppression.

5-4) The NO profile shows a trend similar to the N02 profile with a significant

decrease in peak values when C02 increases, even with the same temperature

distributions. A typical AD plant biogas (Flame 3) having a flame temperature nearly

identical to a pure CH4 gas (in Flame 1) can reduce NO and N02 peaks by about 25%

although it needs a mass flow rate almost 3X greater than the pure CH4 flame. The NO

and N02 mole fractions are reduced through both thermal and chemical effects. The

thermal and chemical influences of C02 dilution on NO are nearly of the same order.

However, it is demonstrated in the comparison of Flames 1 and 3 that the chemical

influences of C02 slightly dominate the reduction of N02.

5-5) Biogas utilization reduces the net release of three green house gases (GHG):

C02, CH4, and N20. Burning CH4, which is a very strong green house gas, prevents its

release to atmosphere in landfills. A carbon dioxide emission index, EICOZ, is defined,

which quantifies the ratio of the net emitted mass of C02 that leaves a counterflow

domain to the mass of CH4 burned. Flame 1, which utilizes pure CH4, has the highest

EICOZ whereas the low quality landfill biogas Flame 4 shows the lowest value.

198

Contrasting four flames suggests that burning the renewable biogas fuel by the

combustion process reduces the net release of C02 as opposed to burning a fossil fuel
(methane or natural gas). Also, two different biogas compositions are compared; with a
2.5 % reduction in the flame temperature, the EIC02 reduces by 15 % indicating that the
relative release of C02 to atmosphere can be reduced with a slight diminishment of the

heat release. The use of biogas reduces emissions of nitrous oxide (NZO) mainly due to

the chemical effects of C02 with only a weakly temperature-dependent influence.

5-6) The biogas mass flux is correlated to its consumption rate and the work
associated with compression of biogas. Comparison of these flames suggests that in a
practical design of biogas-driven combustors, a [tradeoff between three factors emerges:
the flame temperature, the biogas mass flow rate (and thus the required compression
work), and the possible reduction of greenhouse gas emissions, and of NOx emissions

and soot precursors.
6) Similar kinetics studies on the flame structure of C02 diluted methane blends

can be employed for biogas-fired fuel cells and C02 diluted oxy—fuel combustion as well

as the EGR modeling for IC engines.

11.4. Recommendations and future work: Biogas combustion and kinetics

Biogas chemistry is highly non-linear in the oxidation zone; however, scaling

analysis, although rare in the context of chemical kinetics, can provide qualitative

199

correlations to predict the biogas combustion behavior. More work is planned in the

future study on scaling analysis along the lines of Section 6.4.

A complete sensitivity analysis to the heats of formation and reaction rates can be

conducted for all reactions involving H20, C02 and N2. This provides information to

indicate the most responsible reactions for the chemical effects of each diluent in a

simulated biogas.

Preliminary kinetic modeling of simulated bio gas in a Well-Stirred Reactor (W SR)
was performed, but not reported in this thesis, to study lean premixed biogas combustion.
A WSR setup is being developed at MSU for subsequent validation of this kinetic

modeling.

Large-scale gas turbine tests with real biogas fuels give qualitative suggestions
especially to reduce emissions by modifications in turbine configuration and biogas
compositions. Combustion instabilities in large-scale gas turbines utilizing a wide range

of biogas compositions can also be investigated.

11.5. Conclusions: Flame Spread in Microgravity

In Part III, microgravity flame spread over solid fuels was investigated in
collaboration with NASA Glenn Research Center. Diffusion flames near solid fuel
surfaces were investigated numerically and analytically in two environments, namely, an

“open” environment and a “confined near-limit” environment.

In Chapter 9, Surface-attached solid fuel flames in open environments were studied

theoretically and numerically considering weak convection. An exact analytical solution

200

was written for the mixture fraction equation (V2Z = 0) when convective terms were
considered negligible. This equation was also solved numerically and results were
compared to the exact analytical solution to establish an applicable numerical scheme for
weak convective flows. To take into account the role of the weak convective terms, a

simple mathematical model was then solved both numerically and analytically.

Comparisons were made with the results of numerical modeling for a qualitatively
similar problem in which a uniform inflow (or inflows) of a gaseous hydrocarbon fi'om a
surface segment (or segments) is issuing in a cross flow of air. Since the gas phase is of
interest, various equivalence ratios were considered corresponding to various fuel inlet
velocities. Near the flame upstream leading edge (attachment region), the results of the
analytical and numerical models demonstrate Strong agreement. Near the trailing edge,
where the flame can reattach to the surface, the agreement is often poor; the flame
simulated numerically stands nearly parallel to the surface and does not reattach to it. For
the 2-D case, multiple firel segments (slots) were subjected to a preliminary examination
in order to mimic flames that consist of many consecutive fragments (flamelets). An
interesting aspect of the multiple segments with fuel injection is the absence of

significant reactant mixing prior to the reaction-rate maximum.

Perhaps the most important long-term impact of the research on flame and flamelet
spread in a confined space (as conducted in Chapter 10) addresses the fire safety
implications of flames and flamelets in narrow spaces in which combustible materials are
placed in both close proximity and high density. Examples of fires originating in
confined, narrow spaces abound, whether rare cases like the Swissair flight 111 accident

of 1998 or simple domestic and industrial electrical or appliance fires originating

201

undetected in tight spaces. In fact, since these flames/flamelets operate or “survive” in
near-limit environments, detection is an important practical matter. A miniature flamelet
is essentially an active ignition source or pilot flame, which will ignite into full flaming
with inflow of fresh oxidizer. Once ignition occrns the flamelet undergoes transition into
full, active flaming as a nascent fire. This scenario represents the most deadly feature of

such near-limit flames.

Two-dimensional simulations in a narrow channel demonstrate, for small gap
heights, that most of the gravitationally-induced buoyant flow caused by the flame is
suppressed and has a negligible effect on the flow pattern for inlet velocities 2 5 cm/s.
These calculations suggest that when buoyancy is suppressed or eliminated the other
velocities in the system (blowing, thermal expansion, and diffusion) are all of the same
order of magnitude, approximately 10-20% of the inflow velocity, and that when the
inflow is correspondingly reduced to this level the flamelet phenomenon appears in the
experiments. The results indicate that the buoyancy-suppressing NCA quantitatively
captures the essential features of microgravity tests for thin fuels, and thus provides a
viable method to study non-buoyant flame spread in a normal gravity laboratory without
the limitations of actual low-gravity testing in ground-based drop facilities including the

test time and cost.

11.6. Recommendations and future work: Flame Spread in Microgravity

Although the applicability of 2-D FLUENT simulations to the case of 3-D
flamelets is not certain, the current 2-D model represented at present a viable simulation.
A numerical effort at full-scale simulation can be attempted by upgrading the simulation

strategy used in this work. Preliminary 3-D simulations using the NIST Fire Dynamics

202

Simulator (FDS) code were attempted. F larnelets were successfully produced over thin
cellulosic fuels, however, at least four drawbacks were realized: 1) the solid thermal
decomposition and the oxidizing kinetics of volatiles are simple and empirical even in the
MSU modified code using the two-step Broido-Shafizadeh reaction scheme for cellulosic
solid phase pyrolysis; 2) flamelet phenomenon was reproduced only with a non-realistic
oxygen concentration of 47%; 3) the radiation mechanism was simple; 4) The

computational time was excessively long. More work is needed in the 3-D modeling.

For the 2-D scaling analysis, it would be interesting to investigate whether the
Narrow Channel in [g is the same as (or similar to) not having a channel top wall at all in
0g. This can be shown by simulating the narrow channel with a large gap spacing (e. g., h

= 50 mm) in zero-gravity.

203

Appendix A

. H, ' I "t ' i ‘ A 3:37.
Figure A-l. Photograph of the experimental setup at Yale University. Right: the
counterflow burner setup; center: the GC/MS setup; left: the data acquisition system.

    

   

. “ 7 T'T/I F. A .
Figure A2 The counterflow burner setup at Yale University.

204

Appendix B

Flammability maps: The Flammability regime

Flamelets have been observed in NASA over a range of substrate spacings, flow
velocities, and oxygen concentrations in both the ZGRF tests and the NCA in both the
upward facing and downward facing configurations [155]. The test results are used to

make maps of the flamelet regime, which invariably occurs at the edge of flammability.

In Figure B-l, the flamelet regime is delineated by flow velocity and heat loss. The
latter is inversely proportional to substrate spacing, as shown on the right axis. As the
velocity decreases for a given heat loss (moving left on the figure), the flame becomes
corrugated, breaks into flamelets, and eventually extinguishes at sufficiently low
velocities. For a given velocity (5 cm/s), as the heat loss is increased (substrate spacing
reduced, moving up on the figure), the same sequence of events occurs, with eventual

extinguishment at sufficiently high heat loss values.

In addition to the baseline upward facing tests in the NCA, downward facing tests
have been conducted where the apparatus was turned upside down so that the positions of
the copper substrate and quartz viewing window were reversed. The goal of the
downward facing tests was to determine whether the gravitational influence in the 1g
tests had been eliminated altogether, or whether the reversal of g would manifest itself in
a variation in flame behavior. The results shown in Figure B-l indicate that the flamelet
and quenching boundaries are similar, though slightly narrower in the downward facing

geometry. These differences indicate a limited role of gravity in the flame behavior.

205

 

 

 

 

 

§ E
a
O .C
8, d)
C
8 '6
3 8
45 (I)
o a
:1: a
N 9
E c
2 —
E
0.001 . ........ . .......-0.01
1 10 100

Opposed Flow Velocity, cm/s

 

Dg ZGRF extinguished
[lg ZGRF flamelets

0g ZGRF uniform

0g [11] extinguished
0g [11] flarnelets

0g [11] uniform

1g NCAU extinguished
1g NCAd extinguished
lg NCAu flamelets
1g NCAd flamelets

1g NCAU uniform

lg NCAd uniform

<>D<l DOOHQQEKJO

 

 

 

Figure B-l. F lammability map with opposed flow velocity and heat loss as axes, showing
good agreement between Zero Gravity Research Facility tests and normal gravity tests in
the NCA, both in the ‘u’=upward facing orientation and ‘d’=downward facing orientation
[180, 181]. The flamelet regime is a near-limit zone. 0g data with 0.02 mm- spacing is
from [185].

206

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

References

G. Coppola, B. Coriton, and A. Gomez, "Highly turbulent counterflow flames: A
laboratory scale benchmark for practical systems," Combustion and Flame, vol.
156, pp. 1834, 2009.

T. Edwards and L. Q. Maurice, J. Propuls. Power, vol. 17(2), pp. 461-466, 2001.

M. Colket, T. Edwards, S. Williams, N. P. Cernansky, D. L. Miller, F.
Egolfopoulos, P. Lindstedt, K. Seshadri, F. L. Dryer, C. K. Law, D. Friend, D. B.
Lenhert, H. Pitsch, A. Sarofim, M. Smooke, and W. Tsang, presented at 45th
AIAA Aerospace Sci. Meet. Exhibit, 2007.

P. Dagaut and M. Cathonnet, Prog. Energy Combust. Sci., vol. 32, pp. 48-92,
2006.

C. P. Wood, V. G. McDonell, R. A. Smith, and G. S. Samuelson, "Development
and Application of a Surrogate Distillate Fuel," J. Propuls. Power, vol. 5, pp.
399-405, 1989.

W. D. Schulz, ”Oxidation Products of a Surrogate J P-8 Fuel," ACS Petroleum
Chemistry Division Preprints, vol. 37, pp. 383-392, 1991.

A. Violi, E. G. Eddings, A. F. Sarofim, S. Granata, T. Faravelli, and E. Ranzi,
Comb. Sci. and T ech., vol. 174, pp. 399-417, 2002.

A. Agosta, N. P. Cemansky, D. L. Miller, T. F aravelli, and E. Ranzi, Exp. Therm.
Fluid Sci, vol. 28, pp. 701-708, 2004.

C. J. Montgomery, S. M. Cannon, M. A. Mawid, and B. Sekar, "Reduced
chemical kinetic mechanisms for JP-8 combustion," presented at 40th AIAA
Aerospace Sci. Meet. Exhibit, Reno, Nevada, 2002.

S. Humer, A. Frassoldati, S. Granata, T. Faravelli, E. Ranzi, R. Seiser, and K.
Seshadri, Proc. Combust. Inst, vol. 31, pp. 393-400, 2007.

207

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

S. Honnet, K. Seshadri, U. Niemann, and N. Peters, "A Surrogate Fuel for
Kerosene," Proc. Combust. Inst, vol. 32, pp. 485-492, 2009.

R. H. Natelson, M. Kurrnan, N. P. Cemansky, and D. L. Miller, Fuel, vol. 87, pp.
2339-2342, 2008.

A. T. Holley, Y. Dong, M. G. Andac, F. N. Egolfopoulos, and T. Edwards, Proc.
Combust. Inst, vol. 31, pp. 1205—1213, 2007.

S. S. Vasu, D. F. Davidson, and R. K. Hanson, Combust. Flame, vol. 152, pp.
125-143, 2008.

E. Ranzi, available at: ht_tp://www.chem.mlimi.it/CRECKModelingllginetic.htrnl,
2006.

H. R. Zhang, B. G. Eddings, and A. F. Sarofim, Proc. Combust. Inst, vol. 31, pp.
401—409, 2007.

S. S. Vasu, D. F. Davidson, Z. Hong, V. Vasudevan, and R. K. Hanson, Proc.
Combust. Inst, pp. in press, 2009.

S. S. Vasu, D. F. Davidson, and R. K. Hanson, Combust. Flame, pp. in press,
2009.

T. Edwards, D. Minus, W. Harrison, E. Corporan, M. DeWitt, S. Zabarnick, and
L. Balster, "F ischer-Tropsch Jet Fuels — Characterization for Advanced Aerospace
Applications," presented at 40th AIAA/ASME/SAE/ASEE Joint Propuls. Conf.
Exhibit, 2004.

A Review of United States Air Force and Department of Defense Aerospace
Propulsion Needs: Air Force Studies Board, The National Academic Press,
Washington, DC, 2006.

L. Shafer, R. Striebich, J. Gomach, and T. Edwards, "Chemical Class
Composition of Commercial Jet Fuels and Other Specialty Kerosene Fuels," AIAA
Paper 2006- 79 72, 2006.

208

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

T. J. Bruno and B. L. Smith, "Improvements in the Measurement of Distillation
Curves - Part 2: Application to Aerospace/Aviation Fuels RP-l and S-8," Ind.
Eng. Chem. Res, vol. 45, pp. 4381-4388, 2006.

R. H. Natelson, M. S. Kurman, M. D. L., and N. P. Cemansky, "Oxidation of
Alternative Jet Fuels and their Surrogate Components," presented at 46th AIAA
Aerospace Sciences Meeting and Exhibit, 2008.

M. S. P. Kahandawala, M. J. DeWitt, E. Corporan, and S. S. Sidhu, "Ignition and
emission characteristics of surrogate and practical jet fuels," Energy and Fuels,
vol. 22, pp. 3673, 2008.

M. A. Mawid, "Development of a detailed chemical kinetic mechanism for JP-8
and Fischer-Tropsch-derived synthetic jet firels," presented at 43rd
AIAA/ASME/SAE/ASEE Joint Propulsion Conference, Reston, VA 20191-4344,
United States, 2007.

V. R. Katta, M. Mawid, B. Sekar, E. Corporan, J. Zelina, W. M. Roquemore, and
C. J. Montgomery, "Comparison of chemical-kinetics models for JP-8 fuel in
predicting prerrrixed and nonpremixed flames," presented at
AIAA/ASME/SAE/ASEE 42nd Joint Propulsion Conference, Reston, VA 20191-
4344, United States, 2006.

H. Pitsch, "FlameMaster, A C-H- Computer Program for 0D Combustion and 1D
Laminar Flame Calculations, Tech. rep., Institut f 'ur Technische Verbrennung,
RWTH Aachen, may be downloaded from
hth)://www.stanford.edu/groun/pitsch/CES.htrn."

CHEMKIN 4.1 by Reaction Design Inc.

P. Dagaut, A. El Bakali, and A. Ristori, "The combustion of kerosene:
Experimental results and kinetic modelling using 1- to 3-component surrogate
model fuels," Fuel, vol. 85, pp. 944, 2006.

R. J. Kee, J. F. Grcar, M. D. Smooke, and J. A. Miller, "PREMIX: a Fortran
program for modelling steady laminar one-dirnensional premixed flame," Sandia
Report SAND85-8240. Livennore, CA: Sandia National Laboratories, 1985.

209

[31]

[32]

[33]

[34]

[351

[36]

[37]

[38]

[39]

[40]

P. Glarborg, R. J. Kee, J. F. Grcar, and J. A. Miller, "PSR: a Fortran program for
modeling well-stirred reactors," Sandia Report No.8AND86-8209. Liverrnore,
CA: Sandia National Laboratories, 1986.

A. E. Lutz, R. J. Kee, J. F. Grcar, and F. M. Rupley, "OPPDIF: A Fortran
Program for Computing Opposed-Flow Diffusion Flames," SAND96-8243,
Unlimited Release, 1996.

J. A. Cooke, M. Bellucci, M. D. Smooke, A. Gomez, A. Violi, T. Faravelli, and E.
Ranzi, Proc. Combust. Inst, vol. 30, pp. 439-446, 2005.

H. Bufferand, L. Tosatto, B. L. Mantia, M. D. Smooke, and A. Gomez, Revised
paper submitted to Combust. Flame, 2008.

D. E. Rosner, Transport Processes in Chemically Reacting Flow Systems.
Mineola, New York: Dover Publications Inc., 2000.

H. Bufferand, L. T osatto, B. La Mantia, M. D. Smooke, and A. Gomez,
"Experimental and computational study of methane counterflow diffusion flames
perturbed by trace amounts of either jet fuel or a 6-component surrogate under
non-sooting conditions," Combustion and Flame, vol. 156, pp. 1594-1603, 2009.

S. Jahangirian, C. S. McEnally, and A. Gomez, "Experimental study of ethylene
counterflow diffusion flames perturbed by trace amounts of jet fuel and jet fuel
surrogates under incipiently sooting conditions," Combustion and Flame, vol.
156, pp. 1799—1809, 2009.

T. Lu and C. K. Law, "Toward accommodating realistic fuel chemistry in large-
scale computations," Progress in Energy and Combustion Science, vol. 35, pp.
192, 2009.

T. Lu and C. K. Law, "Strategies for mechanism reduction for large
hydrocarbons: n-heptane," Combustion and Flame, vol. 154, pp. 153, 2008.

T. Lu and C. K. Law, "Linear time reduction of large kinetic mechanisms with
directed relation graph: n-Heptane and iso-octane," Combustion and Flame, vol.
144, pp. 24, 2006.

210

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[481

[49]

[50]

T. Lu and C. K. Law, "On the applicability of directed relation graphs to the

reduction of reaction mechanisms," Combustion and Flame, vol. 146, pp. 472,
2006.

P. Pepiot-Desjardins and H. Pitsch, "An efficient error-propagation-based
reduction method for large chemical kinetic mechanisms," Combustion and
Flame, vol. 154, pp. 67, 2008.

P. Pepiot-Desjardins and H. Pitsch, "An automatic chemical lumping method for
the reduction of large chemical kinetic mechanisms," Combustion Theory and
Modelling, vol. 12, pp. 1089, 2008.

A. Gomez, B. L. Manita, L. Tosatto, and H. Bufferand, "Manual of the GC/MS
System for Probing the Cherncical Structure of Flames," Yale University 2008.

L. Tosatto, B. L. Mantia, H. Bufferand, P. Duchaine, and A. Gomez, Proc.
Combust. Inst, in press, 2009.

L. Tosatto, B. L. Mantia, H. Bufferand, P. Duchaine, and A. Gomez, "Chemical
structure of a methane counterflow diffusion flame perturbed with the addition of

either JP-8 or a jet fuel surrogate," Proceedings of the Combustion Institute, vol.
32,pp.1319,2009.

G. Bikas, Kinetic Mechanisms for Hydrocarbon Ignition, Doktor der
Ingenieurwissenschaften thesis, http://depositddb.de/cgi-

bin/d0l3erv?idn=96493285 Z RWYHAachen University, Fakult "at f 'ur
Maschinenwesen, 2001.

E. Ranzi, M. Dente, G. Bozzano, A. Goldaniga, and T. Faravelli, Prog. Energ.
Combust. Sci, vol. 27, pp. 99—139, 2001.

C. K. Law, Combustion Physics. New York: Cambridge University Press, 2006.

K. Seshadri and F. A. Williams, "Laminar flow between parallel plates with
injection of a reactant at high Reynolds number," Intl. J. Heat Mass Transfer, vol.
21, pp. 251-253, 1978.

211

[51]

[52]

[531

[54]

[55]

[56]

[57]

[58]

[59]

[60]

[61]

[62]

[63]

M. H. Duby, W. Deng, K. Kim, T. Gomez, and A. Gomez, "Monodisperse
Electrospray Multiplexing in the Multi-Jet Mode," Journal Aerosol Sci, vol. 37,
pp. 306-322, 2006.

C. Shaddix, K. Brezinsky, and I. Glassman, Proc. Combust. Inst, vol. 24, pp.
683—690, 1992.

R. G. Butler and I. Glassman, Proc. Combust. Inst, pp. in press, 2009.

A. Hamins, D. T. Anderson, and J. H. Miller, "Mechanistic studies of toluene
destruction in diffusion flames," Combust. Sci. Technol, vol. 71, pp. 175—95,
1990.

C. S. McEnally and L. D. Pfefferle, Combust. Sci. T echnol., vol. 131, pp. 323—44,
1998.

C. S. McEnally, L. D. Pfefferle, B. Atakan, and K. Kohse-Héiinghaus, Prog.
Energy Combust. Sci, vol. 32, pp. 247-294, 2006.

I. Glassman, Combustion, 3rd ed: Academic Press, 1996.

A. Gomez, G. Sidebotharn, and I. Glassman, Combust. Flame, vol. 58, pp. 45-57,
1984. l

R. L. Axelbaum, W. L. Flower, and C. K. Law, "Dilution And Temperature
Effects Of Inert Addition On Soot Formation In Counterflow Diffusion Flames,"
Combust. Sci. T echnol., vol. 61, pp. 51-73, 1988.

B. M. Kumfer, S. A. Skeen, and R. L. Axelbaum, Combust. Flame, vol. 154, pp.
546—556, 2008.

T. Edwards, J. Propuls. Power, vol. 19, pp. 1089—1107, 2003.
D. C. Rapp: Pennsylvania State University, PA, 1996.

R. M. Fristrom, Flame Structure and Processes. New York: Oxford University
Press, 1995.

212

[64]

[65]

[66]

[67]

[68]

[69]

[70]

[71]

[72]

[73]

[74]

B. A. V. Bennett, C. S. McEnally, L. D. Pfefferle, M. D. Smooke, and M. B.
Colket, "Computational and experimental study of axisymmetric coflow partially
premixed ethylene/air flames," Combust. Flame, vol. 127, p. 2004-2022, 2001 .

H. R. Zhang, B. G. Eddings, and A. F. Sarofim, "Criteria for selection of

components for surrogates of natural gas and transportation fuels," Proceedings of
the Combustion Institute, vol. 31, pp. 401, 2007.

H. R. Zhang, B. G. Eddings, A. F. Sarofim, and C. K. Westbrook, "Fuel
dependence of benzene pathways," Proceedings of the Combustion Institute, vol.
32,pp.377,2009.

Y. Yang, A. L. Boehman, and R. J. Santoro, Combust. Flame, vol. 149, pp. 191—
205, 2007.

H. F. Calcote and D. M. Manos, Combust. Flame, vol. 49, pp. 289-304, 1983.

S. R. Turns, An Introduction to Combustion, 2nd ed., McGraw—Hill, 2000.

Z. Yang, M. Chaos, and F. L. Dryer, “Surrogate Composition Effects: A
Comparison of Surrogate Mixture Candidates with Combustion Targets (H/C,

T SI, CM ", presented at the Surrogate Fuels Working Group meeting, January 6‘"
2008, Reno, Nevada. Available at: MURI website, 2006.

D. B. Olson, J. C. Pickens, and R. J. Gill, Combust. Flame, vol. 62, pp. 43-60,
1985.

R. C. Santana, P. T. Do, M. Santikunapom, W. E. Alvarez, J. D. Taylor, E. L.
Sughrue, and D. E. Resasco, "Evaluation of different reaction strategies for the
improvement of cetane number in diesel fuels," Fuel, vol. 85, pp. 643, 2006.

C. P. Henry, Enhanced Antistatic Additives for Hydrocarbon Fuels and Solvents,
US Patent No. US200 7/0220803 A], 2007.

R. K. Cheng, D. Littlejohn, W. A. Nazeer, and K. O. Smith, "Laboratory Studies
of the Flow Field Characteristics of Low-swirl Injectors for Adaption to Fuel-
Flexible Turbines," presented at Proceedings of GT2006 ASME Turbo Expo
2006, GT2006-90878.

213

[75]

[76]

[77]

[73]

[79]

[80]

[81]

[32]

[83]

[84]

[35]

<httD://bio-firelsources.conr/>.

P. Griebel, R. Bombach, A. Inauen, Scharen R., S. Schenker, and P. Siewert,
"F lame characteristics and turbulent flame speeds of turbulent high-pressure lean
premixed methane/air flames," ASME paper GT 2005-685 65, 2005.

A. Wellinger and A. Lindberg, "Biogas Upgrading and Utilization," [EA
Bioenergy Task 24: Energy from biological conversion of organic waste.

A. Rojey, Natural Gas Production, Processing and Transport. Technip, Paris,
1996.

G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gardiner, V. V. L. Jr.,
and Z. Qin, "GRI-Mech 3.0," http://www.me.berkeley.edu/grifmech/. 2000.

N. R. Scott, S. Zicari, K. Saikkonen, and K. Bothi, "Characterization of Dairy-
Derived Biogas and Biogas Processing," presented at 2006 ASABE Annual
International Meeting, Portlan, Oregan, 2006.

I. Gokalp and E. Lebas, "Alternative Fuels for Industrial Gas Turbines
(AFTUR)," Applied Thermal Engineering, vol. 24, pp. 1655, 2004.

M. Persson, "Biogas Upgrading and Utilization as Vehicle Fuel," Proceeding of
The Future of Biogas in Europe III: Visions and Targets 2020, European Biogas
Workshop, University of Southern Denmark. Esbjerg, Denmark, 2007.

M. N. Sannaa, "The Development of Biogas Technology in Denmark:
Achievements & Obstacles," in Department of Environment, Technology and
Social Studies, Roskilde University RUC, Denmark, 2004.

P. A. Gerin, F. Vliegen, and J.-M. Jossart, "Energy and C02 Balance of Maize
and Grass as Energy Crops for Anaerobic Digestion," Bioresource Technology,
vol. 99, pp. 2620, 2008.

P. Bdrjesson and M. Berglund, "Environmental systems analysis of biogas
systems-Part I: Fuel-cycle emissions," Biomass and Bioenergy, vol. 30, pp. 469,
2006.

214

[86]

[87]

[88]

[891

[90]

[91]

[92]

[93]

[94]

[95]

Y. Lafay, B. Taupin, G. Martins, G. Cabot, B. Renou, and A. Boukhalfa,
"Experimental study of biogas combustion using a gas turbine configuration,"
Experiments in Fluids, vol. 43, pp. 395, 2007.

United States Enviromnental Protection Agency (EPA) web page. Available
online at www.gpa. gov.

R. Biitikofer, "Biotech to Biofuels: How New Technologies are Changing
Transatlantic Relations," Energy Security and Global Warming Conference, San
Diego, 2007.

J. B. Holrn-Nielsen, "The future of biogas in Europe III: Visions and Targets
2020," European Biogas Worla'hop and Study Trip, University of Southern
Denmark, Esbjerg, Denmark, 2007.

M. Moliere, "Expanding fuel flexibility of gas turbines," Proc. IMechE Vol. 219
Part A: J. Power and Energy, pp. 109, 2005.

D. Yomogida and N. D. Thinh, "Thermal analysis of a simple-cycle gas turbine in
biogas power generation," presented at Proceedings of the Heat Transfer and
Fluid Mechanics Institute, 1995.

D. E. Yomogida, N. D. Thinh, V. M. Tiangco, and Y. Lee, "Thermal efficiency
projections of a simple-cycle gas turbine in biogas power generation," presented
at American Society of Mechanical Engineers, Bunningharn, UK, 1996.

C. A. Pahner and M. R. Erbes, "Simulation methods used to analyze the
performance of the GE PG6541B gas turbine utilizing low heating value fuels,"

presented at American Society of Mechanical Engineers, International Gas
Turbine Institute (Publication) IGTI, Portland, OR, USA.

H. Hanagudu, "Biogas fired industrial gas turbines a technological and potential
assessmen ," presented at American Society of Mechanical Engineers (Paper),
Cincinnati, OH, USA.

G. Wiltsee and H. Emerson, "Clean power from microturbines using biogas,"
BioCycle, vol. 45, pp. 53, 2004.

215

[96]

[97]

[98]

[99]

[100]

[101]

[102]

[103]

[104]

[105]

B. Krautkremer and J. Muller, "Mikrogasturbine - Eine Alternative zur Nutzung
von Biogas (Micro gas turbine - An alternative use of biogas)," VDI Berichte, pp.
181, 2005.

N. Goldstein, "Microturbines, gas engines link biogas to the grid," BioCycle, vol.
47, pp. 59, 2006.

E. Wheless and J. Pierce, " Siloxanes in Landfill and Digester Gas Update," Los
Angeles County Sanitation Districts and SCS Energ, California, 2004.

E. J. Dolak and J. P. Armstrong, "The correlation of IR microturbine combustion
performance while burning diluted gaseous fuel supplied by a fuel mixing
facility," presented at Proceedings of the ASME Turbo Expo, Reno-Tahoe, NV,
United States.

S. Naing, T. Yarnada, and K. Nakanishi, "Analysis of micro gas turbine
co generation system with biogas storage," presented at Collection of Technical

Papers - 4th International Energy Conversion Engineering Conference, San
Diego, CA, USA.

D. Bohn and J. Lepers, "Effects of biogas combustion on the operation
characteristics and pollutant emissions of a micro gas turbine," presented at
American Society of Mechanical Engineers, International Gas Turbine Institute,
Turbo Expo (Publication) IGTI, Atlanta, GA, United States.

M. R. Johnson, D. Littlejohn, W. A. Nazeer, K. O. Smith, and R. K. Cheng, "A
comparison of the flowfields and emissions of high-swirl injectors and low-swirl
injectors for lean premixed gas turbines," Proceedings of the Combustion
Institute, vol. 30, pp. 2867, 2005.

H. C. Mongia, Held, T.J., Hsiao, GO and Pandalai, R.P., "Challenges and
progress in controlling dynamics in gas turbine combustors," Journal of
Propulsion and Power, vol. 19(5), pp. 822-829, 2003.

D. Littlejohn and R. K. Cheng, "Fuel effects on a low—swirl injector for lean
premixed gas turbines," Proceedings of the Combustion Institute, vol. 31, pp.
3155, 2007.

Y. Lafay, "Biogas Combustion in Gas Turbines," P. Communication, Ed.,
October 2007.

216

[106]

[107]

[108]

[109]

[110]

[111]

[112]

[113]

[114]

[115]

[116]

T. Leung and I. Wierzba, "The effect of hydrogen addition on biogas non-
premixed jet flame stability in a co-flowing air stream," International Journal Of
Hydrogen Energy, vol. 33, pp. 3856-3862, 2008.

M. Zheng, G. T. Reader, and J. G. Hawley, "Diesel engine exhaust gas
recirculation--a review on advanced and novel concepts," Energy Conversion and
Management, vol. 45, pp. 883, 2004.

J. Staniforth and K. Kendall, "Biogas powering a small tubular solid oxide fuel
cell," Journal of Power Sources, vol. 71, pp. 275, 1998.

J. Staniforth and K. Kendall, "Cannock landfill gas powering a small tubular solid
oxide fuel cell -- a case study," Journal of Power Sources, vol. 86, pp. 401, 2000.

K. Kendall, C. M. Finnerty, G. Saunders, and J. T. Chung, "Effects of dilution on
methane entering an SOFC anode," Journal of Power Sources, vol. 106, pp. 323,
2002. .

H. Kobayashi, H. Hagiwara, H. Kaneko, and Y. Ogami, "Effects of C02 dilution
on turbulent premixed flames at high pressure and high temperature,"
Proceedings of the Combustion Institute, vol. 31, pp. 1451, 2007.

B. J. P. Buhre, L. K. Elliott, C. D. Sheng, R. P. Gupta, and T. F. Wall, "Oxy-fuel
combustion technology for coal-fired power generation," Progress in Energy and
Combustion Science, vol. 31, pp. 283, 2005.

F. H. V. Coppens and A. A. Konnov, "The effects of enrichment by H2 on
propagation speeds in adiabatic flat and cellular premixed flames of
CH4+02+C02," Fuel, vol. 87, pp. 2866-2870, 2008.

P. Glarborg and L. L. B. Bentzen, "Chemical effects of a high C02 concentration
in oxy-fuel combustion of methane," Energy & Fuels, vol. 22, pp. 291-296, 2008.

S. C. Li and F. A. Williams, "NOx formation in two-stage methane-air flames,"
Combustion and Flame, vol. 118, pp. 399, 1999.

T. Lieuwen, V. McDonell, E. Petersen, and D. Santavicca, "Fuel flexibility
influences on premixed combustor blowout, flashback, autoignition, and
stability," Proceedings of the ASME Turbo Expo, vol. 1, pp. 601-615, 2006.

217

 

[117]

[118]

[119]

[120]

[121]

[122]

[123]

[124]

[125]

[126]

C. Zhang, A. Atreya, and K. Lee, "Sooting structure of methane counterflow
diffusion flames with preheated reactants and dilution by products of
combustion," Twenty-Fourth Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, pp. 1049, 1992.

A. R. Masri, Dibble, R.W., Barlow, R.S., Combustion and Flame, vol. 91, pp.
285-309, 1992.

F. S. Liu, H. S. Guo, G. J. Smallwood, and O. L. Gulder, "The chemical effects of
carbon dioxide as an additive in an ethylene diffusion flame: Implications for soot
and NOx formation," Combustion And Flame, vol. 125, pp. 778-787, 2001.

F. Liu, H. Guo, and G. J. Smallwood, "The chemical effect of C02 replacement
of N2 in air on the burning velocity of CH4 and H2 premixed flames,"
Combustion and Flame, vol. 133, pp. 495, 2003.

J. Park, D. J. Hwang, K.-T. Kim, S.-B. Lee, and S.-I. Keel, "Evaluation of
chemical effects of added C02 according to flame location," International Journal
of Energy Research, vol. 28, pp. 551, 2004.

H. Guo and G. J. Smallwood, "A numerical study on the influence of C02
addition on soot formation in an ethylene/air diffusion flame," Combustion
Science and T echnolog, vol. 180, pp. 1695-1708, 2008.

DJ. Hwang, J. Park, C.-B. Oh, K.-H. Lee, and S.-I. Keel, "Numerical study on
NO formation in CH4-Oz -N2 diffusion flame diluted with C02," International
Journal of Energy Research, vol. 29, pp. 107, 2005.

J. Park, S.-G. Kim, K.-M. Lee, and K. Kim, "Chemical effect of diluents on flame
structure and NO emrssron characteristic in methane-air counterflow diffusion
flame," International Journal of Energy Research, vol. 26, pp. 1141, 2002.

J. Park, J. S. Park, H. P. Kim, J. S. Kim, S. C. Kim, J. G. Choi, H. C. Cho K W
Cho, and H. S. Park, "NO emission behavior in oxy-fuel combustion recir’culated
with carbon dioxide," Energy and Fuels, vol. 21 , pp. 121, 2007.

F. Hermann, T. Zeuch, and J. Klingrnann, "The effect of diluents on the formation

rate of nitrogen oxide in a premixed laminar flame," resented t A
Expo 2004, Vienna, Austria, 2004, p a SME Turbo

218

[127]

[128]

[129]

[130]

[131]

[132]

[133]

[134]

[135]

[136]

[137]

S. Gordon and B. J. McBride, "Computer Program for Calculation of Complex
Chemical Equilibrium Compositions, Rocket Performance, Incident and Reflected
Shocks and Chapman-Jouguet Detonations," NASA Report SP-273 1971.

G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M.
Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gardiner, V. V. L. Jr.,
and Z. Qin, "GRI-Mech 3.0," http://www.me.berkeley.edu/gri_mech/.

San Diego reaction mechanism, Center for Energy Research, UC San Diego,
Available online at: http://maeweb.ucsd.edg/~combr_istion/.

J. Li, "Experimental and Numerical Studies of Ethanol Chemical Kinetics," PhD.
Thesis, Princeton University, 2004.

R. G. Gilbert, K. Luther, and J. Troe, "Theory of thermal unimolecular reactions
in the fall-off range. ii. weak collision rate constants," Ber. Bunsenges. Phys.
Chem, vol. 87, pp. 169-177, 1983.

J. C. Hewson and M. Bollig, "Reduced mechanisms for NOx emissions from
hydrocarbon diffusion flames," Twenty-Sixth Symposium (International) on
Combustion, The Combustion Institute, pp. 2171— 2179, 1996.

C. T. Bowman, R. K. Hanson, D. F. Davidson, J. W. C. Gardiner, V. Lissianski,
G. P. Smith, D. M. Golden, M. Frenklach, and M. Goldenberg, "Gri-mech 2.11,"
in http://www.me.berkeley.edu/gri mech/.

A. Kazakov and M. Frenklach, "Reduced Reaction Sets based on GRI -Mech
l . 2 ," http://www. me. berkelev. edu/drm/.

C. J. Sung, J. B. Liu, and C. K. Law, "Structural response of counterflow
diffusion flames to strain rate variations," Combustion and Flame, vol. 102, pp.
481, 1995.

A. E. Lutz, R. J. Kee, J. F. Grcar, and F. M. Rupley, "OPPDIF: A Fortran
Program for Computing Opposed-Flow Diffusion Flames," SAND96-8243,
Unlimited Release, 1996.

R. J. Kee, J. A. Miller, and G. H. Evans, "A computational medel of the structure
and extinction of strained, opposed flow, premixed methane—air flame,"

219

[138]

[139]

[140]

[141]

[142]

[143]

[144]

[145]

[146]

[147]

Proceedings of the 22nd Symposium (International) on Combustion. The
Combustion Institute, pp. 1479, 1988.

N. Peters and R. J. Kee, Combustion and Flame, vol. 68, pp. 17-29, 1987.

M. V. Petrova and F. A. Williams, "A small detailed chemical-kinetic mechanism
for hydrocarbon combustion," Combustion and Flame, vol. 144, pp. 526, 2006.

L. J. Spadaccini and M. B. Colket Iii, "Ignition delay characteristics of methane
fuels," Progress in Energy and Combustion Science, vol. 20, pp. 431, 1994.

E. L. Petersen, J. M. Hall, S. D. Smith, J. de Vries, A. R. Amadio, and M. W.
Crofion, "Ignition of lean methane-based fuel blends at gas turbine pressures,"
Journal Of Engineering For Gas Turbines And Power-Transactions Of The Asme,
vol. 129, pp. 937-944, 2007.

F. N. Egolfopoulos, D. X. DU, and C. K. Law, Combustion Science and
Technology, vol. 83, pp. 33-75, 1992. '

F. 'N. Egolfopoulos, D. L. Zhu, and C. K. Law, Twenty-Third Symposium
(International) on Combustion, The Combustion Institute, pp. 471-478, 1990.

H. Kobayashi, K. Seyarna, H. Hagiwara, and Y. Ogami, "Burning velocity
correlation of methane/air turbulent premixed flames at high pressure and high
temperature," Proceedings of the Combustion Institute, vol. 30, pp. 827, 2005.

M. Metghalchi and J. C. Keck, Combustion and Flame, vol. 38, pp. 143—157,
1980.

S. J ahangirian, A. Engeda, and I. S. Wichman, "Thermal and chemical structure
of biogas counterflow diffusion flames," Energy and Fuels, Published online
October 9, DOI:10.1021/ef9002044, 2009.

G. J. Rortveit, J. E. Hustad, S.-C. Li, and F. A. Williams, "Effects of diluents on
NOx formation in hydrogen counterflow flames," Combustion and Flame, vol.
130,pp.48,2002.

220

[148]

[149]

[150]

[151]

[152]

[153]

[154]

[155]

[156]

[157]

[158]

T. L. Berry Yelverton and W. L. Roberts, "Soot surface temperature

measurements in pure and diluted flames at atmospheric and elevated pressures,"
Experimental Thermal and Fluid Science, vol. 33, pp. 17, 2008.

F. Liu, H. Guo, G. J. Smallwood, and O. L. Gulder, "The chemical effects of
carbon dioxide as an additive in an ethylene diffusion flame: implications for soot
and NOx formation," Combustion and Flame, vol. 125, pp. 778, 2001.

K. P. Schug, Y. Manheirner-Timnat, P. Yaccarino, and l. Glassman, "Sooting
Behavior of Gaseous Hydrocarbon Diffusion Flames and the Influence of
Additives," Combustion Science and Technology, vol. 22, pp. 235, 1980.

D. X. Du, R. L. Axelbaum, and C. K. Law, "The influence of carbon dioxide and
oxygen as additives on soot formation in diffusion flames," Symposium
(International) on Combustion, vol. 23, pp. 1501-1507, 1990.

S. Jahangirian and A. Engeda, "Biogas Combustion and Chemical Kinetics for
Gas Turbine Applications," Proceedings of 2008 ASME International Mechanical
Engineering Congress and Exposition, Boston, USA, IME CE2008-6666 7.

J. A. van-Aardenne, F. D. Dentener, J. G. J. Olivier, J. A. H. W. Peters, and L. N.
Ganzeveld, "Emission Database for Global Atmospheric Research," The EDGAR
3.2 Fast Track 2000 dataset (32F T2000), 2005.

Y. Long and I. S. Wichman, "Structure of Attached Flames Spreading over
Condensed Fuels," presented at Central States Sections Meeting, The Combustion
Institute, 2005.

S. L. Olson, F. J. Miller, S. Jahangirian, and l. S. Wichman, "Flame spread over
thin fuels in actual and simulated microgravity conditions," Combustion and
Flame, vol. 156, pp. 1214, 2009.

Y. Long, "A Numerical Study of Flame Spread over Thin Cellulosic Fuels in
Microgravity," in Michigan State University, Ph.D. Thesis, 2006.

H. D. Ross, Microgravity Combustion. San Diego: Academic Press, 2001.

I. S. Wichman, "Theory of Opposed-Flow Flame Spread," Progress in Energy
and Combustion Science, vol. 18, pp. 553-593, 1992.

221

[159]

[160]

[161]

[162]

[163]

[164]

[165]

[166]

[167]

[168]

[169]

L. Massa, T. L. Jackson, J. Buckrnaster, and M. Campbell, "Three Dimensional
Heterogeneous Propellant Combustion," Proc. Comb. Inst, vol. 29, pp. 2975-
2983, 2002.

S. L. Olson, F. J. Miller, and I. S. Wichman, "Characterizing Fingering Flamelets
Using the Logistic Model," Combustion Theory and Modelling, vol. 10, pp. 323-
347, 2006.

J. Buckrnaster, "The Modeling of Heterogeneous Propellant Combustion,"
presented at Central States Sections Meeting, The Combustion Institute, 2006.

J. N. DeRis, "Spread of a Laminar Diffusion Flame," Proc. Comb. Inst, vol. 12,
pp. 241-252, 1969.

I. S. Wichman and F. A. Williams, "A Simplified Model of Flame Spread in an
Opposed Flow along a Flat Surface of a Semi-Infinite Solid," Combustion Science
and Technology, vol. 32, pp. 91-123, 1983.

Y. Long, S. Jahangirian, and I. S. Wichman, "Flarne-Surface Interaction and
Flamelet Microstructure over Single and Multiple Solid Fuel Segments in a
Channel Cross Flow," Proceedings of the 2006 ASME International Mechanical
Engineering Congress and Exposition, Chicago, USA, IME CE2006-I 5 068, 2006.

S. Bhattachajee and R. A. Altenkirch, "The Effect of Surface Radiation on Flame
Spread in a Quiescent Microgravity Environment," Combustion and Flame, vol.
84, pp. 160-169, 1991.

V. Yang, T. B. Brill, and W. Z. Ren, "Solid Propellant Chemistry, Combustion,
and Motor Interior Ballistics," Progress in Astronautics and Aeronautics, vol.
185, 2000.

Transportation Safety Board of Canada Report Number A98H0003, 1998.

L. M. Oravecz, I. S. Wichman, and S. L. Olson, Proceedings of the 1999 ASME
International Mechanical Engineering Congress and Exposition, vol. 364-4, pp.
183-187, 1999.

R. Friedman, B. Jackson, and S. Olson, "Testing and Selection of Fire-Resistant
Materials for Spacecrafi Use," NASA T M-2000-2097 73, 2000.

222

[170]

[171]

[172]

[173]

[174]

[175]

[176]

[177]

[178]

[179]

[180]

[181]

R. Vance and I. S. Wichman, "Heat Transfer Analysis of a Diffusion Flame
Leading Edge Near a Cold, Chemically Inert Surface," International Journal of
Heat And Mass Transfer, vol. 43, pp. 921-933, 2000.

I. S. Wichman, Z. Pavlova, B. Ramadan, and G. Qin, "Heat Flux from a Diffusion
Flame Leading Edge to an Adjacent Surface," Combustion and Flame, vol. 118,
pp. 652-668, 1999.

I. S. Wichman, "On Diffusion Flame Attachment Near Cold Surfaces,"
Combustion and Flame, vol. 117, pp. 384-393, 1999.

D. L. Urban, J. S. Goldrneer, and Z. G. Yuan, "Interactions between Flames on
Parallel Solid Surfaces," presented at 4th Inter. Microg. Comb. Workshop. NASA
CP——10194, 1997.

S. Rouvreau, P. Joulain, H. Y. Wang, P. Cordeiro, and J. L. Torero, Proc. Comb.
Inst, vol. 29, pp. 2527-2534, 2002.

P. Cordeiro, G. Legros, S. Rouvreau, P. Joulain, and J. L. Torero, " Detailed
Description of the Structure of a Low Velocity Laminar Diffusion Flame in
Microgravity," presented at 7th International Workshop on Micro gravity Comb.
and Chemically Reacting Systems, Cleveland, 2003.

K. K. Kuo, Principles of Combustion: John Wiley & Sons, 1986.
FLUENT Software, by ANSYS, Inc., WWW.FLUENT.COM.

I. S. Wichman, L. M. Oravecz-Simpkins, and S. Tanaya, "Experimental Study of
Flamelet Formation in a Hele-Shaw Flow," presented at 3rd Joint Mtg. U.S.
Sections Combustion Institute, 2003.

S. L. Olson, F. J. Miller, and I. S. Wichman, "Describing Near-Limit Flamelet
Fingering Behavior Using Bio-Mathematical Population Models," presented at
Fourth International Symposium on Scale Modeling, 2003.

S. A. Tanaya, M.S. Thesis, in Michigan State University, 2004.

K. L. Aditjandra, M.S. Thesis, in Michigan State University, 2005.

223

[182]

[183]

[184]

[185]

I. S. Wichman, S. Tanaya, K. Aditjandra, L. Yang, S. L. Olson, and F. J. Miller,
"A Simulated Zero-Gravity Flame Spread Apparatus," presented at Central States
Section of the Combustion Institute Annual Meeting, Cleveland, OH, 2006.

S. L. Olson, Combust. Sci. T ech., vol. 76, pp. 233-249, 1991.

T. Hirano, S. E. Noreikis, and T. E. Waterman, "Measured Velocity and
Temperature Profiles Near Flames Spreading Over a Thin Combustible Solid,"
Combustion and Flame, vol. 23, pp. 83-96, 1974.

S. L. Olson, T. Kashiwagi, O. Fujita, M. Kikuchi, and K. Ito, Combustion and
Flame, vol. 125, pp. 852-864, 2001.

224