Punthesis PDF

Measurements of
Thermo-Acoustic Coupling
Thesis by
Winston Pun
In Partial Fulfillment of the Requirements

for the Degree of
Doctor of Philosophy
California Institute of Technology

Pasadena, California
2001
(Defended May 24, 2001)
ii
2001
Winston Pun
All Rights Reserved
iii
ACKNOWLEDGEMENTS
This work could not have been completed without the help and encouragement of many
of those around me. I would like to express my gratitude to my advisor, Prof. Fred
Culick, for his guidance during my graduate career and in the preparation of this thesis. I
would also like to thank Professors Mory Gharib, Melany Hunt, and Richard Murray for
serving on my thesis committee.
I would like to acknowledge the financial support of AFOSR, DURIP, the Department of
Energys AGTSR program, ENEL, L.A. County Sanitation District, and the California
Institute of Technology.
Those in the JPC research group have also contributed greatly to this work, as well as
contributing to an enjoyable working environment.
In particular, I would like to
acknowledge the collaboration of Steve Palm with the OH PLIF measurements. I greatly
value both his technical expertise and friendship. Thanks also are due to Konstantin
Matveev and Guido Poncia for their significant contributions to the work with the Rijke
tube. I would like to thank Giorgio Isella and Claude Seywert for occasionally hauling
me out to the Red Door and contributing to my social (SC) as well as professional
development. I would also like to thank Al Ratner, Grant Swenson, Sanjeev Malhotra,
and Olivier Duchemin for their numerous discussions and for always being available
whenever heavy lifting was required in the lab.
A special thanks to my family and to Suzie, for their support and encouragement.
Thank you all for helping me bring this thesis into reality!
iv
ABSTRACT
The problem of combustion instabilities has existed since the early 1940s, when they
were observed during the development of solid and liquid rocket engines. While various
engineering solutions have served well in these fields, the problem is revisited in modern
gas-turbine engines. The purpose of this work is to provide experimental measurements
of laboratory devices that exhibit thermo-acoustic coupling, similar to the interaction
observed during combustion instabilities, which will aid in the design and development
of stable systems.
Possibly the simplest device which exhibits these characteristics is a Rijke tube. An
electrical, horizontally mounted, 1 m long version of the original Rijke tube is presented,
with measurements taken during unstable and stable operation. An accurate stability
boundary with uncertainty is determined for a heater position of x/L = , as a function of
mass flow rate and heater power. Hysteresis, not previously reported, is observed at flow
rates above 3 g/s.
A one-dimensional model of the stability boundary with linear
acoustics is shown to have qualitative agreement with experimental data.

A novel technique has also been devised which can provide insight into the local dynamic
response of a flame to an acoustic field.
In the experiments, a test chamber is
acoustically excited by a pair of low-frequency drivers. The response of the flame is

visualized by two techniques; chemiluminescence and planar laser-induced fluorescence
(PLIF) of the hydroxyl (OH) radical, both of which are well-known indicators for heat
release in flames. The resulting images are phase-resolved and averaged to yield a
qualitative picture of the fluctuation of the heat release. The images are correlated with a
pressure transducer near the flame, which allows stability to be evaluated using
Rayleighs criterion and a combustion response function.
This is the first known
measurement of the combustion dynamics of a flame over a range of frequencies.

Results indicate that the drive frequency and burner configuration have a pronounced
effect on the response of the flame. Drive frequencies ranging from 22 Hz to 55 Hz are
applied to the jet mixed burner, supplied with a premixed 50/50 mixture of methane and
carbon dioxide at a Reynolds number of 20,000.
The burner is operated in two
configurations; with an aerodynamically stabilized flame and with a flame stabilized by

two protruding bluff-bodies. Results indicate that in general the bluff-body stabilized
flame is less sensitive to chamber acoustic excitation.
vi
TABLE OF CONTENTS
Acknowledgements ...................................................................................... iii

Abstract......................................................................................................... iv
Table of Contents......................................................................................... vi
List of Figures .............................................................................................. xi
List of Tables ............................................................................................ xviii
CHAPTER 1 INTRODUCTION.................................................................1
1.1 MOTIVATION ............................................................................................................. 1
1.2 OBJECTIVES .............................................................................................................. 4
1.3 RAYLEIGHS CRITERION........................................................................................... 5
1.4 PREVIOUS WORK ...................................................................................................... 7
1.4.1 The Electrical Rijke Tube ................................................................................... 7
1.4.2 Flares ................................................................................................................... 8
1.4.3 Combustion Dynamics of Unsteady Flames ....................................................... 9
CHAPTER 2 THE RIJKE TUBE .............................................................13

2.1 EXPERIMENTAL SETUP ........................................................................................... 13
2.1.1 Heat Source ....................................................................................................... 14
2.1.2 Air Flow ............................................................................................................ 15
vii
2.1.3 Pressure Transducers......................................................................................... 16

2.1.4 Thermocouples.................................................................................................. 17
2.1.5 Data Acquisition System................................................................................... 18
2.2 EXPERIMENTAL PROCEDURE ................................................................................. 19
2.3 RESULTS .................................................................................................................. 20
2.3.1 Hysteretic Behavior........................................................................................... 25
2.3.2 Stability Boundary ............................................................................................ 33
2.4 PREDICTION OF THE STABILITY BOUNDARY .......................................................... 35
2.5 COMPARISON OF PREDICTION WITH EXPERIMENTAL RESULTS ........................... 39
2.6 SUMMARY ................................................................................................................ 42
CHAPTER 3 FULL-SCALE FLARE AND SUB-SCALE MODEL ....44

3.1 FULL-SCALE FLARE ................................................................................................ 45
3.1.1 Introduction ....................................................................................................... 45
3.1.2 Description of Full-Scale Flare ......................................................................... 46
3.1.3 Diagnostics........................................................................................................ 47
3.1.4 Results ............................................................................................................... 48
3.1.5 Possible Driving Mechanisms........................................................................... 55
3.2 SUB-SCALE FLARE MODEL..................................................................................... 55
3.2.1 Apparatus and Diagnostics................................................................................ 55
3.2.2 Scaling of the Model ......................................................................................... 58
3.3 CONCLUSIONS ......................................................................................................... 60
CHAPTER 4 CHEMILUMINESCENCE MEASUREMENTS ...........62
viii
4.1 EXPERIMENTAL SETUP ........................................................................................... 63

4.1.1 Acoustic Driving System .................................................................................. 63
4.1.2 Acoustic Cavity................................................................................................. 64
4.1.3 Test Burners ...................................................................................................... 65
4.2 ACOUSTIC PROPERTIES .......................................................................................... 68
4.3 DIAGNOSTICS .......................................................................................................... 71
4.3.1 Pressure Transducers......................................................................................... 71
4.3.2 Data Acquisition System................................................................................... 71
4.3.3 High-Speed Video Camera ............................................................................... 72
4.3.4 Additional Electronics....................................................................................... 72
4.3.5 Shadowgraph Setup........................................................................................... 73
4.3.6 Chemiluminescence Measurements .................................................................. 74
4.4 SHADOWGRAPH RESULTS ....................................................................................... 75
4.5 CHEMILUMINESCENCE RESULTS ............................................................................ 76
4.5.1 Two-Dimensional Flame Structure ................................................................... 76
4.5.2 Axial Flame Structure ....................................................................................... 83
4.5.3 Modified Rayleigh Index .................................................................................. 91
4.5.4 Global Forced Rayleigh Results........................................................................ 92
4.5.5 Spatially Resolved Forced Rayleigh Results .................................................... 93
4.6 SUMMARY ................................................................................................................ 95
CHAPTER 5 OH PLIF MEASUREMENTS ...........................................97

5.1 PLANAR LASER-INDUCED FLUORESCENCE OF OH................................................ 97
5.1.1 PLIF Theory ...................................................................................................... 97
ix
5.1.2 Laser System ................................................................................................... 103

5.1.3 Optics .............................................................................................................. 104
5.1.4 ICCD Camera.................................................................................................. 107
5.2 EXPERIMENTAL PROCEDURE ............................................................................... 108
5.3 DATA REDUCTION ................................................................................................. 110
5.3.1 Phase Characterization .................................................................................... 110
5.3.2 Image Processing ............................................................................................ 111
5.4 OH PLIF RESULTS ............................................................................................... 112
5.4.1 Pressure and Heat Release Measurements ...................................................... 112
5.4.2 Forced Rayleigh Index .................................................................................... 127
5.4.3 Global Rayleigh Results.................................................................................. 127
5.4.4 Spatially Resolved Rayleigh Results .............................................................. 129
5.4.5 Combustion Response..................................................................................... 139
5.5 SUMMARY .............................................................................................................. 145
CHAPTER 6 CONCLUDING REMARKS .......................................... 147
Appendix A Mechanical Drawings ........................................................ 152
Appendix B Gas Mixture Viscosity Calculation.................................. 156
Appendix C Chemiluminescence Rayleigh Indices ............................. 158
Appendix D Locally Normalized Combustion Response.................... 167
REFERENCES ......................................................................................... 170
xi
LIST OF FIGURES
Figure 1-1: Effect of equivalence ratio on NOx production (from Rosfjord 1995). ........... 2
Figure 1-2: Combustor system............................................................................................ 3
Figure 2-1: Electrical Rijke tube experimental setup. ...................................................... 14
Figure 2-2: Expected stability boundary........................................................................... 19
Figure 2-3: Steady state stable Rijke tube data recordings. (Mean power = 995 W, mean
mass flow = 3.3 g/s). .................................................................................................. 22
Figure 2-4: Steady state unstable Rijke tube data recordings. (Mean power = 995 W,
mean mass flow = 3.1 g/s).......................................................................................... 23
Figure 2-5: FFTs of an unstable case in the Rijke tube. ................................................... 24
Figure 2-6: Bulk RMS pressures at a mean flow rate of 3.15 g/s..................................... 26
Figure 2-7: Bulk RMS pressures at a mean flow rate of 2.44 g/s..................................... 27
Figure 2-8: Frequencies of oscillations for first two dominant modes at 3.15 g/s............ 29
Figure 2-9: RMS pressures of oscillations for first two dominant modes at 3.15 g/s....... 30
Figure 2-10: Frequencies of oscillations for first two dominant modes at 2.44 g/s.......... 31
Figure 2-11: RMS pressures of oscillations for first two dominant modes at 2.44 g/s..... 32
Figure 2-12: Stability boundary for Rijke tube at x/L = . .............................................. 34
Figure 2-13: Efficiency factor, E, for r* = 1.0. ................................................................. 40
Figure 2-14: Stability boundary prediction. Solid black line is for T = 600K, and the
dashed black line for T = 300K.................................................................................. 41
Figure 3-1: Landfill flare station....................................................................................... 45
Figure 3-2: Instrumentation layout on full-scale flare. ..................................................... 47
xii
Figure 3-3: Flare bulk temperature profile........................................................................ 49

Figure 3-4: Flare pressure traces (low pass filtered at a cutoff of 200 Hz). ..................... 50
Figure 3-5: FFT of pressure traces in full-scale flare. ...................................................... 52
Figure 3-6: Flare pressure traces, filtered using a 6th order butterworth phase-preserving
bandpass filter between 7-10 Hz (1st mode).............................................................. 53
Figure 3-7: Flare pressure traces, filtered using a 6th order butterworth phase-preserving
bandpass filter between 21-24 Hz (2nd mode). ......................................................... 54
Figure 3-8: Schematic of sub-scale flare model assembly................................................ 57
Figure 4-1: Schematic of overall test section.................................................................... 64
Figure 4-2: Bluff-body stabilized burner flameholder detail, viewed from the upstream
side. ............................................................................................................................ 66
Figure 4-3: (a) Aerodynamically stabilized burner (b) bluff-body stabilized burner. ...... 67
Figure 4-4: Gas feed system. ............................................................................................ 68
Figure 4-5: Peak-to-peak pressure amplitudes in the chamber, driven at various
frequencies (no flame)................................................................................................ 69
Figure 4-6: RMS pressure in the chamber, when driven at various frequencies (no flame).
.................................................................................................................................... 69
Figure 4-7: Shadowgraph imaging setup. ......................................................................... 73
Figure 4-8: Shadowgraph results above the burner tube for (a) aerodynamically and (b)
bluff-body stabilized cases......................................................................................... 75
Figure 4-9: Chemiluminescence contour plots at 22 Hz for (a) aerodynamically stabilized
and (b) bluff-body stabilized cases. ........................................................................... 78
xiii
Figure 4-10: Chemiluminescence contour plots at 27 Hz for (a) aerodynamically

stabilized and (b) bluff-body stabilized cases. ........................................................... 79
Figure 4-14: Mean axial intensities at 22 Hz (a) aerodynamic (b) bluff-body. ................ 86
Figure 4-15: Normalized axial intensities at 22 Hz (a) aerodynamic (b) bluff-body. ...... 86
Figure 4-24: Chemiluminescence global forced Rayleigh indices. .................................. 92
Figure 4-25: Chemiluminescence 2D forced Rayleigh indices at 32 Hz for (a)
aerodynamic and (b) bluff-body stabilized burners. .................................................. 94
Figure 4-26: Chemiluminescence axial forced Rayleigh indices. .................................... 94
Figure 5-1: Simplified energy level transfer diagram for LIF. ........................................ 99
xiv
Figure 5-2: PLIF system. ................................................................................................ 105

Figure 5-3: Mixer/doubler system (Stages available but not used are indicated by dashed
lines)......................................................................................................................... 106
Figure 5-4: OH PLIF optics arrangement. ...................................................................... 106
Figure 5-5: Phase-binning procedure for OH PLIF images............................................ 111
Figure 5-6: OH PLIF images over a period of a sinusoidal pressure oscillation for the
aerodynamically stabilized burner at 32 Hz. The intensity scale is in number of
counts, and the x and y coordinates are in pixels. .................................................... 114
Figure 5-7: Pressure and heat release traces and power spectrums for the aerodynamically
stabilized burner driven at 22 Hz. ............................................................................ 115
Figure 5-8: Pressure and heat release traces and power spectrums for the bluff-body
Figure 5-9: Pressure and heat release traces and power spectrums for the aerodynamically
Figure 5-11: Pressure and heat release traces and power spectrums for the
aerodynamically stabilized burner driven at 32 Hz.................................................. 119
xv
Figure 5-17: Phase relationship between the 1st mode of pressure and heat release for the
(a) aerodynamically stabilized and (b) bluff-body stabilized cases at 22 Hz. Heat
release traces have been scaled for ease of comparison........................................... 126
Figure 5-18: Frequency driven global Rayleigh index. .................................................. 128
Figure 5-19: Contour plot of Rf for (a) aerodynamically and (b) bluff-body stabilized
burners at a driving frequency of 22 Hz................................................................... 131
Figure 5-24: Aerodynamically stabilized 2D Rayleigh plots. ........................................ 138
Figure 5-25: Bluff-body stabilized 2D Rayleigh plots. .................................................. 138
xvi
Figure 5-26: Axial Rayleigh index plot: Solid lines are the aerodynamically stabilized
burner, and dashed lines are the bluff-body stabilized case..................................... 139
Figure 5-27: Global combustion response function........................................................ 140
Figure 5-28: Combustion response magnitude (a) aerodynamically stabilized (b) bluffbody stabilized. ........................................................................................................ 142
Figure 5-29: Combustion response phase (a) aerodynamically stabilized (b) bluff-body
stabilized. ................................................................................................................. 143
Figure A-1: 1/12th scale flare model: chamber section. ................................................. 153
Figure A-2: 1/12th scale flare model: burner section. .................................................... 154
Figure A-3: Instrumentation boss. .................................................................................. 155
Figure C-1: Chemiluminescence axial forced Rayleigh indices at 22 Hz. .....159
Figure C-5: Chemiluminescence axial forced Rayleigh indices at 55 Hz. .........161
Figure C-6: Chemiluminescence 2D forced Rayleigh indices at 22 Hz for (a)
aerodynamically and (b) bluff-body stabilized burners. .....162
xvii

Figure D-1: Locally normalized combustion response magnitude (intensity
is on a log scale) (a) aerodynamically stabilized (b) bluff-body stabilized. ......169
xviii
LIST OF TABLES
Table 1-1: Previous work in oscillating flames. ............................................................... 11
Table 2-1: PCB 112A04 pressure transducer properties with 422D11 charge amp. ........ 17
Table 2-2: Data acquisition analog measurements. .......................................................... 18
Table 4-1: Flame base position and oscillation................................................................. 84
Table 5-1: Smoothing filter weighting matrix. ............................................................... 112
Table B-1: Coefficients for viscosity quadratic fit, valid for P = 1 atm, 5 C < T < 45 C.
.................................................................................................................................. 157
Chapter 1
Introduction
1.1 Motivation
Of the six principle air pollutants tracked by the EPA (carbon monoxide, lead, nitrogen
oxides, particulate matter, sulfur dioxide, and volatile organic compounds), all have made
significant reductions in emissions since the passage of the Clean Air Act in 1970, except
for nitrogen oxides1. NOx, consisting of NO and NO2 are contributors to stratospheric
ozone depletion, acid rain, and smog. Stricter government regulations regarding pollutant
emissions and in particular oxides of nitrogen have come into effect. As a result, the gasturbine industry is seeking ways to reduce NOx emissions.
From document EPA-456/F-98-006 (September 1998)
Several techniques have been explored to lower pollutant production, including catalytic
combustion, rich-burn/quick-quench schemes, fuel staging, water or steam injection, and
lean premixed combustion. Of these, the most promising technique is believed to be the
operation of combustors in the regime of lean, premixed flames (Correa 1992). The
general strategy is to reduce flame temperatures, and thereby minimize NOx production
due to the thermal (or Zeldovich) mechanism. The other pathways for production of
nitrogen oxides, the prompt or Fenimore mechanism, and the nitrous oxide mechanism
are typically not significant contributors to NOx formation under gas-turbine combustion
conditions. Another important consideration in lean combustion is the uniformity of the
fuel-air mixture ratio, since both spatial and temporal fluctuations in mixture ratio will
result in higher NOx production (Swenson et al. 1996; Fric 1993).
Figure 1-1: Effect of equivalence ratio on NOx production (from Rosfjord 1995).
Operation near the lean blowout limit is desirable to minimize high flame temperatures.
This circumstance gives rise to conditions under which combustion instabilities are more
likely to occur in lean premixed systems, and have in fact been observed in new
combustors operating under these conditions, as shown in Figure 1-1.
The term
combustion instability refers to the presence of a large amplitude pressure oscillation

inside the combustion chamber and does not indicate unstable burning of the flame. The
pressure excursions associated with these instabilities can cause unacceptable levels of
vibration and increased rates of heat transfer. Vibration levels may induce mechanical
failure, and enhanced heat transfer can cause components already operating near high
temperature limits, such as the liner or turbine blades, to fail catastrophically.
Lean premixed combustion systems are especially susceptible to combustion instabilities,

since small pressure oscillations in the chamber can also affect air feed lines, causing
variations in the fuel-oxidizer ratio, which in turn will produce fluctuations in the heat
release rate, which can further drive the pressure oscillations. Incomplete premixing of
the fuel and oxidizer can have a similar effect, creating pockets of burning mixtures at
different equivalence ratios, which can drive acoustic modes in the chamber.
This
interaction suggests a coupling or feedback loop that exists between combustor dynamics
and combustion dynamics, as shown in Figure 1-2.
External
Inputs
Combustor
Dynamics
q, Energy
Addition
Combustion
Dynamics
Figure 1-2: Combustor system.
1.2 Objectives
The mechanisms causing combustion instabilities in gas turbine combustors are not well
understood. Although similar in principle (Raun et al. 1983) to a Rijke tube (a heatdriven acoustic oscillation), the added geometric complexities and injector configurations
of a practical combustor make their dynamical behavior unpredictable. Current industry
design techniques are largely empirical and not clearly defined with respect to
combustion instabilities. Ultimately, industrial combustor designs are finalized without a
clear measure of the stability margins of the system. A method for predicting and
evaluating the stability characteristics of a given combustor configuration is needed for
more robust designs. The central objectives of this work are
To provide accurate measurements of nonlinear instability phenomena of

laboratory devices for use in model validation.
To develop tools for evaluation of the combustion response of a burner in a

variety of combustor configurations.
To evaluate the use of chemiluminescence and OH PLIF as indicators for the

response of a combustion system in a forced acoustic environment.
It is anticipated that this work and future work derived from it will lead to design rules of
thumb that will improve current design techniques and the understanding of acousticflame interactions.
1.3 Rayleighs Criterion

When considering the phenomena of heat-driven pressure oscillations, Rayleigh stated
his now well-known criterion in 1878:
If heat be given to the air at the moment of greatest rarefaction, or be
taken from it at the moment of greatest rarefaction, the vibration is
encouraged. On the other hand, if heat be given at the moment of greatest
rarefaction, or abstracted at the moment of greatest condensation, the
vibration is discouraged.
The mathematical development of Rayleighs criterion follows from the conservation
equations. Following the framework of Culick (1976), the mass, momentum, and energy
equations with sources can be written as
+ u = W
t
(1-1)
r
u
+ u u = p + F
t
(1-2)
(1-3)
p
+ p u = u p + P
t
where
(1-4)
W = ws u
(1-5)
r
F = Fs uws
(1-6)
P=
R
CV
u2
u
F
Q
w
+
s
s
s
2
where , p , and u, are the density, pressure, and velocity in the gas phase, ws, Fs, and Qs
are sources of mass, momentum, and energy. The analysis can be carried out in one-
dimension for convenience, following the development of Sterling (1987) and Culick
(1987). Expanding the momentum and energy equations (1-2) and (1-3) in terms of mean
and fluctuating quantities, multiplying them by u and
p
respectively, and adding them
po
together without mass sources, we arrive at

R
up po
u 2 p 2
=
Qp (up)
+ uo o +
x
x 2
2po CV po
po x
t
(1-7)
p 2
uuo + ou2 +
po
where
( )o
are mean quantities and
( )
uo
,
x
represent fluctuating quantities. Assuming small
changes in a correspondingly low mean flow, orthogonality between u and p , and

integrating over the combustor volume we are left with
(1-8)
o u 2 p 2
R
=
Qp .
+
2po CV po
2
Integrating equation (1-8) over the volume of the combustor, V, and a cycle of the
oscillation, , we arrive at
(1-9)
E =
( 1)
t +
dV t pQdt ,
po
where E is the energy added to the system during a cycle. Equation (1-9) is an explicit
expression of Rayleighs criterion.
1.4 Previous Work

1.4.1 The Electrical Rijke Tube
A Rijke tube is comprised of a tube with a mean airflow and a heat source, often a heated
wire gauze. At a particular heater position, temperature, and airflow, the tube can be
made to sing as acoustic modes within the tube are excited. Through consideration of
Rayleigh's criterion, Collyer and Ayres (1972) were able to excite the 1st through 3rd
modes of a Rijke tube consisting of 3.66 cm diameter Pyrex glass tubes, of length 0.79 m
and 1.54 m. The fundamental mode can be produced by placing the heater at L/4, the 2nd
mode by a heater at L/8 or 5L/8, and the 3rd mode with a heater at L/12, 5L/12, or 3L/4.
It was noted that the frequency of the modes increased slightly due to higher air
temperatures, so a compensating heater to increase the air temperature at lower power
levels was introduced just below L/2. Though not ideal, the addition of the second heater
had the general desired effect.
Katto and Sajiki (1977) performed a wide range of experiments in an electrically driven
Rijke tube. Their experimental arrangement consisted of a compressor supplying airflow
rates up to 50 L/min through a large surge tank and into a steel tube with an inner
diameter of 30 mm, with lengths of 310 mm to 2810 mm. The kanthar wire heaters were
in a spiral or coil configuration. Stability boundaries are given, but without temperature
profiles, a clear criterion for the onset of instability, or limit cycle amplitudes. Work by
Madarame (1981a, 1981b), involved a similar apparatus, but added measurements of the
linear growth rate and excited frequencies. A summary of work done on Rijke tubes and
similar devices is contained in the review paper of Raun et al. (1993). Chapter 2 contains
the results and development of a model used to predict the stability boundary of a Rijke
tube.
1.4.2 Flares
A device similar to a Rijke tube is the biogas flare, which is used to dispose of gases
produced in a landfill site. Typically landfill flares use an enclosed design, with all
combustion taking place inside a refractory-lined chamber, such that no flame is visible.
This configuration allows control of product gas residence time and temperatures in order
to minimize pollutant emissions (John Zink 1988). A consequence of this configuration
is that the flare burner may interact with the feed system and chamber acoustics to
produce an unstable system. While other types of unstable systems may be corrected by
trial and error, unstable flares are often left to run below full capacity due to government
restrictions on noise levels near residential areas. For testing purposes, landfill gas
composition is taken to contain 50% methane and 50% carbon dioxide (Christo et al.
1998). At these compositions, the flame has substantially different properties than typical
methane-air flames. Work by Qin et al. (2001) has shown that the presence of high CO2
quantities decreases laminar flame speeds and extinction strain rates, while increasing
NOx production. Further work regarding unstable flares has been notably absent in the
literature. Data taken from an unstable flare site, and sub-scale modeling attempts are
presented in chapter 3.
1.4.3 Combustion Dynamics of Unsteady Flames

In order to study the unsteady dynamics of a combustion chamber, a reliable technique to
visualize the combustion process and its response to an oscillating pressure field is
required.
Two techniques that can be used to perform these measurements are
chemiluminescence and planar laser-induced fluorescence (PLIF).
McManus et al.
(1995) give a review of these techniques as they are applied in modern combustion
research.
Chemiluminescence of various radicals of combustion, in particular CH, is an excellent

marker of the reaction zone, and has been used by a number of researchers to study heat
release in unsteady flames. One major drawback of chemiluminescence is the inability of
the measurements to provide high spatial resolution, since line-of-sight integration occurs
at the detector. These experiments can be categorized into two groups; measurements
using a photo-multiplier tube (PMT) with a slit obscuring a portion of the flame to obtain
some spatial (typically axial) resolution (Poinsot et al. 1987; Sterling 1991; Chen et al.
1993; and Kappei et al. 2000), and full two-dimensional imaging using a charge-coupled
device (CCD) based camera (Broda et al. 1990; Kendrick et al. 1999; and Venkataraman
et al. 1999). Of these works, only Chen et al. (1993) involved an acoustically forced
flame, but used a PMT with a slit configuration that obtained only integrated onedimensional information. Poinsot et al. (1987) had the capability of injecting acoustic
waves into their system, but did not do so. They later made use of this capability to
perform combustion instability active suppression control experiments (Poinsot et al.,
1989). Venkataraman et al. (1999) used phase-resolved CH chemiluminescence to study
10
the instability characteristics of a dump combustor. They determined that swirl tends to
induce combustion instabilities near the lean blowout limit.
Dyer and Crosley (1982) performed the first demonstration of 2D (or planar) laserinduced fluorescence (LIF) of the hydroxyl radical in a flame. Spatial resolution is
defined by the resolution of the detector, and the laser sheet, typically several hundred
microns in width. The PLIF technique has been used since then to measure a variety of
chemical species in unsteady reacting flows, including OH as a measure of the heat
release (Cadou et al. 1991; and Shih et al. 1996), and NO seeded fuel to measure the
temperature field (Cadou et al. 1998). PLIF measurements can discriminate strongly
between different chemical species, while chemiluminescence measurements often
contain overlapping source contributions to the signal. A summary of these various
works involving both chemiluminescence and PLIF is provided in Table 1-1, including
the acoustic frequencies examined in the studies.
While chemiluminescence measurements are more readily obtained, since they do not
require a costly laser pump source, they have several disadvantages. Chemiluminesence
measurements cannot capture fine structures in flames, since the signal is integrated
through the depth of the flame. PLIF images are obtained of only a very specific plane
where the laser sheet illuminates the flame. Another disadvantage of chemiluminescence
is that the signal is several orders of magnitude weaker than PLIF. This decreases the
temporal resolution of measurements, since longer integration times are required to
obtain a sufficiently strong signal. A typical integration time for a single shot using
11
chemiluminescence is on the order of approximately 100 s, versus 100 ns when

performing PLIF.
Chemiluminescence
Naturally
Unsteady
Acoustic
Forcing
PLIF
Poinsot et al. (1987) (440-590 Hz)
Cadou et al. (1991) (43 Hz)
Sterling and Zukoski (1991) (188 Hz)
Shih et al. (1996) (400 Hz)
Broda et al. (1998) (1750 Hz)
Cadou et al. (1998) (328 Hz)
Kendrick et al. (1999) (235 Hz, 355 Hz)
Venkataraman et al. (1999) (490 Hz)
Kappei et al. (2000) (370-460 Hz)
Chen et al. (1993) (300 Hz, 400 Hz)
Cadou et al. (1998) (360 Hz, 420 Hz)
Table 1-1: Previous work in oscillating flames.
Most experimental work characterizing various combustor configurations has been

performed on naturally unstable systems (see Table 1-1). However, these results are
specific to the combustors tested and provide little insight to how a particular injector or
burner design will behave in a different combustor. A study of the acoustic coupling
between fuel injectors and an applied acoustic field has been carried out by Anderson et
al. (1998), but only for cold flow experiments.
Work by Chen et al. (1993) with
premixed flames was specifically designed to simulate solid rocket propellants and used
the same apparatus as Sankar et al. (1990). It produced one-dimensional spatial results
and used only two forcing frequencies. The study by Cadou et al. (1998) was based on a
specific 2D dump combustor configuration and showed little response to nonresonant
forcing.
A more generalized body of work is required to provide a scientific
understanding of guidelines that will be useful in designing stable combustion systems.
12
Although OH radicals have been used by other researchers (Yip et al. 1994) as a marker
of the reaction zone, there is some question as to its validity, since OH is known to persist
in high-temperature product gas regions (Allen et al. 1993; Barlow et al. 1990).
However, in non-premixed flames, the OH radical quickly vanishes on both sides of the
reaction zone (Cessou 1996). Since the burner configuration is only partially premixed in
this study, we assume OH to be sufficient as an indicator for the heat release, as is
commonly the case.
The purpose of this study is to demonstrate a novel technique that can be used as part of a
method to assess stability margins over a range of frequencies for various burner designs.
It is anticipated that this technique will provide sufficient temporal and spatial resolution
that can be used to improve predictive capabilities and correlate experimental results with
numerical simulations. A burner using a mixture of methane and CO2 is operated in two
configurations: aerodynamically stabilized and stabilized with a bluff-body. The burner
is subjected to a forced acoustic field with frequencies ranging from 22 Hz to 55 Hz. The
configuration discussed here has been chosen to simulate a practical application. It
serves as a relatively simple device for which the new diagnostics can be tested with
minimal difficulties arising with the test apparatus. Results from chemiluminescence
imaging are presented in chapter 4, and laser diagnostics involving OH PLIF are
presented in chapter 5.
13
Chapter 2
The Rijke Tube
This chapter concerns perhaps the simplest device that exhibits a thermo-acoustic
instability, namely the Rijke tube. The experimental campaign, results, and model used
to predict the stability boundary are discussed.
2.1 Experimental Setup

The classic Rijke tube (Rijke 1859) consists of a vertically mounted glass tube with a
wire gauze suspended inside. The gauze is heated using a flame, which causes the tube to
sing in certain cases. The essential characteristics of the original device are retained
with several modifications to better quantify the phenomenon. One of the major changes
is orienting the tube horizontally, which removes the mean flow induced by convection.
This enables a quantitative investigation of the effect the mean flow has on the system.
14
The other major change is use of an electrically heated nichrome grid as a heat source
instead of a flame since the power input into the system is better characterized.
The electrical horizontal Rijke tube (pictured in Figure 2-1) consists of a 9.5 x 9.5 cm
square aluminum tube, 1.0 m in length. Air is sucked in from one end of the tube (taken
as the origin, x = 0) into a large plenum, which acts as a damping chamber to decouple
the blower dynamics from the tube.
Damping chamber
Thermocouple array
Rijke tube
Heater power rods
P2
Pressure
transducers
P1
Air flow
Blower
Figure 2-1: Electrical Rijke tube experimental setup.
2.1.1 Heat Source

As previously mentioned, the heat source consists of a nichrome grid (40 mesh) with a
wire diameter of 0.01". It is silver brazed to two strips of copper, which form the positive
and negative terminals of the heater. The heater is suspended in a frame machined from
15
macor in order to withstand high temperatures. Two long copper rods, welded directly to
the copper strips on the heater, form a solid physical and electrical connection from the
heater to the power source. The power source consists of two TCR-20T250 high current
power supplies, each capable of producing 500 amps of current. The power supplies are
load balanced and operate in parallel, enabling the system to draw up to 1000A. The
actual power supplied is dependent on the resistance of the nichrome grid, which changes
with temperature.
The power supplies are computer controlled using a software-
implemented controller to stabilize the output power, although fluctuations on the order
of 1% can occur.
The heater is located at a position of x/L = . This is the ideal location for driving the
fundamental mode of an open-open Rijke tube, according to Rayleighs criterion. Input
power is determined by directly measuring the voltage between the copper rods and
measuring the current through one of the rods using a current sensor (Amploc, CL500).
The final input power measurement is corrected to account for power dissipation along
the copper rods.
2.1.2 Air Flow
The mean air flow through the Rijke tube is provided by a GAST R1102 blower,
operating at 3450 rpm with a maximum throughput of 0.0127 m3/s at standard
atmospheric conditions. The blower is operated at full capacity with a 2" by-pass ball
valve controlling the amount of air drawn through the damping chamber, or from the
atmosphere. A large plastic shroud (not pictured) is placed above the entrance to the
Rijke tube to minimize air current effects on the system.
16
The flow rate is measured using a laminar flow element (Meriam 50MW20) and a
differential pressure transducer (Honeywell Microswitch). This measurement takes place
between the damping chamber and the blower. A thermocouple, located upstream of the
laminar flow element, is used to correct for air density and viscosity to produce the total
air mass flow rate.
2.1.3 Pressure Transducers
Selection of the proper pressure transducers used in this experiment was critical since
they must provide accurate measurements in a hot environment. The transducers used
were PCB model 112A04, coupled with a 422D11 charge amplifier and a 482A20 signal
conditioner.
Table 2-1 lists some of the important characteristics of the pressure
transducers. Charge-mode piezoelectric transducers were used, since the majority of the
electronics is located in a separate charge amplifier, increasing the operating temperature
range while retaining relatively high sensitivities. The two pressure transducers, flush
mounted in the tube at positions x/L = 0.15 and x/L = 0.80, are labeled P1 and P2
respectively (see Figure 2-1). In most cases, P1 is on the cold side of the tube and P2 is
on the hot side.
17
Sensitivity
100 mV/psi
Maximum Pressure
5000 psi
Linearity
< 1% FS
Temperature Range
-400 to +600 F
Flash Temperature
3000 F
Resonant Frequency
> 250 kHz
Rise Time
< 2 s
Table 2-1: PCB 112A04 pressure transducer properties with 422D11 charge amp.
2.1.4 Thermocouples
An array of 15 type K thermocouples is suspended from the top of the tube to the
centerline, at positions of x = 5, 10, 15, 22, 27, 30, 35, 40, 45, 50, 56.7, 63.3, 70, 76.7,
and 90 cm. An additional thermocouple is located just before the laminar flow element
that measures the mean flow through the tube. The odd spacing results from a desire to
place more thermocouples nearer to the heat source, as well as to allow the heater to be
located at key locations without interfering with the thermocouples.
Since the
thermocouples have a relatively large time constant, they are multiplexed and sampled at
2 Hz (i.e., all 16 thermocouples are read in 0.5 s). It is not possible for thermocouples to
respond quickly enough at the acoustic time scales required in the experiment. They are
used solely for bulk temperature measurements.
18
2.1.5 Data Acquisition System

In order to provide accurate measurements of the acoustic pressures and other relevant
phenomenon in the Rijke tube, a fast sampling system is required. The data acquisition
system is based on a Pentium III 700 MHz computer.
A Computer Boards CIO-
DAS1602/12 (12 bit) data acquisition board is installed in the machine, using Sparrow
(Murray, 1995) as the software interface. An EXP-16 expansion board accommodates
the 16 thermocouples in a multiplexed array and also provides cold junction
compensation. The channels acquired are listed in Table 2-2.
Channel
Measurement
System thermocouples
Cold junction compensation
Pressure transducer P1
Pressure transducer P2
Heater voltage
Flow rate (LFE pressure drop)
Heater current
Table 2-2: Data acquisition analog measurements.

The DAS1602/12 is operated in single-ended mode, giving a total of up to 16 analog
input channels. It also contains two analog output channels, one of which is used to
control the power supplies. In this configuration, data could be acquired in short bursts at
over 8000 Hz, and for extended periods of time streaming to the hard drive at over 4000
Hz. For this Rijke tube, the primary frequencies are the first two modes at approximately
19
190 Hz and 380 Hz. These frequencies were easily captured at the data rates capable by
the data acquisition system.
2.2 Experimental Procedure

In order to carefully map the stability boundary, a methodical system of acquiring data
was employed. Variables in the Rijke tube experiment are the heater power (P), air flow
rate or velocity (V), and heater position (x). Based on previous works (Katto and Sajiki
1977; Madarame 1981a), the stability region should be similar to the one schematically
shown in Figure 2-2.
P
Unstable
region
V
x
Figure 2-2: Expected stability boundary.
For this work, only one heater position (x/L = ) is investigated, although the experiment
is easily modified to include other heater positions, which can preferentially drive
alternate harmonics in the tube.
Before commencing an experimental run, the tube is subjected to a warm-up procedure,

in order to minimize temperature variations as the power input is increased. The warmup depends on the actual conditions under which the experimental run will take place,
based on anticipated stability boundaries. Typically, the power input will be set to
20
approximately 200 W below the unstable point for a particular flow condition, and the
tube run at that rate for 20 minutes. If the stability boundary is not known or incorrectly
selected, a more conservative estimate of the stability boundary is used at the expense of
an increased duration for the experiment.
Since the flow rate is manually controlled, the flow condition is set, and the power
increased via computer control.
Initial power increments of 50 W are used while
relatively far from the stability boundary. As the boundary is approached, the power
increments are reduced to the limits of resolution of the controller and power supplies.
The tube is held steady at each condition for approximately 120 seconds until the system
temperature field has settled and is quasi-static, before data is acquired. Due to the
presence of hysteresis at certain conditions, there exist two possible states for the system.
The small steps in power followed by a long waiting period prevent transient thermal
effects from triggering (Burnley and Culick 2000) the transition to the unstable regime.
Once the tube has become unstable, a few more data points are taken within the
instability boundary, and then the power increments are reversed to determine the return
to stability boundary. In a similar fashion, power decrements are initially large and then
subsequently refined as the boundary is approached. At each point, a full set of data is
acquired for post-processing.
2.3 Results
The raw data obtained by the data acquisition system is post-processed, with the variables
converted into appropriate units. The pressure signals are filtered using a 5th order
21
Butterworth highpass filter with a cutoff frequency of 20 Hz, to eliminate low frequency
noise and environmental effects.
Examples of data traces taken at a stable and unstable condition are shown in Figure 2-3
and Figure 2-4 respectively. Notice that while both conditions are very similar, their
behavior is quite different. This is due to hysteresis effects, which must take into account
the time history or evolution of a particular condition. This will be explored in greater
detail in the next section.
Notice in the stable case (Figure 2-3), there is no coherent pressure oscillation. The low
frequency drift in the signal is due to thermal drift and noise induced in the pressure
transducers. The unstable case (Figure 2-4) shows a well-defined pressure oscillation in
both transducers. Differences in amplitude are due to the position of the transducers with
respect to the modeshape of the acoustic modes, as well as transducer P2 being located in
the hot section. The temperature profiles are characterized by a large jump at x = 25 cm
(the position of the heater gauze) of approximately 300 K. There is significant cooling of
the air as it progresses downstream in the tube, which is often neglected in modeling
efforts. The mass flow rate fluctuates with a period characteristic of the blower (RPM
and number of fan blades). In the unstable case (Figure 2-4), the flow rate oscillates
approximately out of phase with the pressure oscillations, as would be expected. In both
cases, the power fluctuates with a small amplitude high frequency superimposed over a
lower frequency higher amplitude, which is due to the action of the controller on the
system.
22
P ress ure #1 - f34p1040
P ress ure #2
0.01
0.01
0.005
0.005
ps i
0.015
ps i
0.015
-0.005
-0.005
-0.01
-0.01
-0.015
10
20
30
tim e, m s
40
50
-0.015
140
300
120
200
100
100
20
30
tim e, m s
40
50
S ound P ressure Level (P 1)
400
db
cels ius degs.
Tem perature P rofile
10
80
20
40
60
80
60
100
10
cm
M ass Flowrate
20
30
tim e, m s
40
50
40
50
P ower
3.45
1060
3.4
1050
3.35
W
g/s
1040
3.3
1030
3.25
1020
3.2
3.15
10
20
30
tim e, m s
40
50
1010
10
20
30
tim e, m s
Figure 2-3: Steady state stable Rijke tube data recordings. (Mean power = 995 W,
mean mass flow = 3.3 g/s).
23
P ressure #2
0.2
0.05
0.1
ps i
ps i
P ressure #1 - b34p1040
0.1
-0.05
10
20
30
40
tim e, m s
Tem perature P rofile
-0.2
50
400
150
300
140
200
130
dB
-0.1
-0.1
100
10
20
30
40
tim e, m s
S ound P res sure Level (P 1)
120
20
40
60
80
110
100
10
20
30
tim e, m s
P ower
40
50
10
20
30
tim e, m s
40
50
cm
M ass Flowrate
3.6
1030
1020
3.4
1010
W
g/s
3.2
3
1000
990
2.8
2.6
50
980
10
20
30
tim e, m s
40
50
970
Figure 2-4: Steady state unstable Rijke tube data recordings. (Mean power = 995
W, mean mass flow = 3.1 g/s).
24
Power Spectrum of Pressure #2

100
50
50
dB
dB
Power Spectrum of Pressure #1 - b34p1040

100
-50
-100
-50
100
200
300
Frequency [Hz]
400
-100
500
250
100
200
50
150
-50
200
300
Frequency [Hz]
400
500
Power Spectrum of Power
150
dB
dB
Power Spectrum of Flow Rate
100
100
100
200
300
Frequency [Hz]
400
500
50
100
200
300
Frequency [Hz]
400
500
Figure 2-5: FFTs of an unstable case in the Rijke tube.
For the unstable case shown in Figure 2-4, the corresponding FFTs are plotted in Figure
2-5. The first and second modes are most prominent (higher modes are not significant,
and are not shown) at approximately 190 Hz and 380 Hz respectively. The spectrum of
the flow rate shows a response at the fundamental excitation frequency, but significantly
lower excitation of the second mode when compared with the relative magnitudes of the
pressure responses, possibly attributable to the damping chamber.
25
2.3.1 Hysteretic Behavior

As the Rijke tube transitions from a stable to an unstable operating condition, it is
important to identify as precisely as possible when this transition actually takes place.
One possible measure of the instability is the amplitude of the pressure oscillation. A
problem with this definition arises since noise and flow-induced vortices can give rise to
increasing pressure amplitudes measured by the transducers.
Thus, the appropriate
threshold that defines the transition to instability is not well defined.
Instead, a
combination of approaches is taken, using observation of both the amplitude of the

oscillation as well as the frequency of the strongest mode excited. Once the Rijke tube
goes unstable, the frequency locks in to an unstable mode of the tube. In this case,
since the Rijke tube is in an open-open configuration with the heat source located at the
point, the main unstable frequency will be approximately a half-wave (with appropriate
end corrections) as dictated by Rayleighs criterion and chamber acoustics.
Hysteresis implies that the history of the system is important in determining the current
state of the system. Hysteresis has been viewed previously in a dump combustor (Isella
et al. 1997) and in a Rijke burner (Seywert 2001). An example of the hysteresis in the
electrical Rijke tube is shown in Figure 2-6. The cold section represents data taken by
transducer P1 at x = 0.15 m, and the hot section by transducer P2 located at x = 0.8 m.
The triangles pointing up indicate data points taken as the power is being increased.
Since the transitional points are of primary interest, many of the preliminary settings are
not recorded and often by-passed as described in the warm-up procedure. Once the
26
(a) C old S ection
RMS Pressure [psi]
0.05
0.04
0.03
Increasing Power
Decreasing Power
0.02
0.01
0
600
650
700
750
800
850
900
950
1000
1050
1100
Power [W ]
(b) Hot Section
0.06
RMS Pressure [psi]
0.05
0.04
Increasing Power
Decreasing Power
0.03
0.02
0.01
0
600
650
700
750
800
850
900
950
1000
1050
1100
Power [W ]
Figure 2-6: Bulk RMS pressures at a mean flow rate of 3.15 g/s.
27
(a) C old S ection
RMS Pressure [psi]
0.025
0.02
0.015
Inc reasing P ower

Dec reasing P ower
0.01
0.005
0
435
440
445
450
455
460
465
470
475
P ower [W ]
(b) Hot Section
0.03
RMS Pressure [psi]
0.025
0.02
Inc reasing P ower

Dec reasing P ower
0.015
0.01
0.005
0
435
440
445
450
455
460
465
470
475
P ower [W ]
Figure 2-7: Bulk RMS pressures at a mean flow rate of 2.44 g/s.
28
unstable point has been reached and slightly exceeded, characterized by a sharp increase
in pressure amplitudes at approximately 1075 W, the input power is decreased. As the
power is gradually decreased, indicated by the downward-pointing triangles, the
amplitude of the limit cycle slowly decreases. A large hysteresis loop exists, from
approximately 1050 W down to 650 W, when the Rijke tube returns to stable operation.
Note, that the amplitude of the pressure oscillations decreases with decreasing power,
which could produce uncertainty as to when the system is again stable. Though in this
case the transition is well defined, in other cases, such as the one in Figure 2-7, the
transition may not be as obvious. At a flow rate of 2.44 g/s, it is not entirely clear
whether the data point at approximately 443 W should be classified as stable or unstable.
If an additional data point existed at a power level of 440 W, the ambiguity would be
even greater.
As stated above, the definition chosen requires analysis of the frequencies produced
corresponding to the first mode of the pressure oscillation. Returning to the first example
at a mass flow rate of 3.15 g/s, Figure 2-8 and Figure 2-9 show the breakdown of
frequencies and pressure amplitudes for the two most dominant modes in the system. It
is clear from examination of the first mode, that the system has become unstable at a
power level of 1070 W as the power is increased, and has returned to stable operation as
the power is decreased at 650 W. In a similar fashion, Figure 2-10 and Figure 2-11
reproduce the same plots for the 2.44 g/s condition. It is now possible to classify the
somewhat ambiguous point at a power of 443 W, observed in Figure 2-7. The frequency
of the first mode in Figure 2-10 shows that at 443 W the system is still locked to the
29
(a) First Harmonic

220
Inc reasing P ower
Dec reasing P ower
Frequency [Hz]
210
200
190
180
170
600
650
700
750
800
850
900
950
1000
1050
1100
950
1000
1050
1100
P ower [W ]
(b) Second Harmonic
420
Inc reasing P ower
Dec reasing P ower
Frequency [Hz]
410
400
390
380
370
360
350
600
650
700
750
800
850
900
P ower [W ]
Figure 2-8: Frequencies of oscillations for first two dominant modes at 3.15 g/s.
30
(a) First Harmonic
RMS Pressure [psi]
0.05
Inc reasing P ower

Dec reasing P ower
0.04
0.03
0.02
0.01
0
600
650
700
750
800
850
900
950
1000
1050
1100
1000
1050
1100
P ower [W ]
(a) Second Harmonic
RMS Pressure [psi]
0.025
Inc reasing P ower
Dec reasing P ower
0.02
0.015
0.01
0.005
0
600
650
700
750
800
850
900
950
P ower [W ]
Figure 2-9: RMS pressures of oscillations for first two dominant modes at 3.15 g/s.
31
(a) First Harmonic

205
Inc reasing P ower
Dec reasing P ower
Frequency [Hz]
200
195
190
185
180
175
170
435
440
445
450
455
460
465
470
475
P ower [W ]
(b) Second Harmonic
410
Frequency [Hz]
400
Inc reasing P ower

Dec reasing P ower
390
380
370
360
350
435
440
445
450
455
460
465
470
475
P ower [W ]
Figure 2-10: Frequencies of oscillations for first two dominant modes at 2.44 g/s.
32
(a) First Harmonic
RMS Pressure [psi]
0.02
0.015
Inc reasing P ower
Dec reasing P ower
0.01
0.005
0
435
440
445
450
455
460
465
470
475
470
475
P ower [W ]
(a) Second Harmonic
RMS Pressure [psi]
0.02
0.015
0.01
Inc reasing P ower

Dec reasing P ower
0.005
0
435
440
445
450
455
460
465
P ower [W ]
Figure 2-11: RMS pressures of oscillations for first two dominant modes at 2.44 g/s
33
unstable mode of the Rijke tube. This observation is corroborated by the plot of pressure
amplitude (Figure 2-11), which shows that a limit cycle still exists significantly over
noise levels at 443 W. It can be reasonably concluded that the system has returned to
stable operation at 437 W.
2.3.2 Stability Boundary

The records presented in the previous section show the presence of hysteresis in the Rijke
tube system. This however is not always the case. At lower mass flow rates, there is no
hysteresis behavior. The results of mapping the stability boundary of the Rijke tube in
two dimensions, power and mass flow rate (heater position is left for future work), are
summarized in Figure 2-12. Red lines indicate increasing heater power and the point
when the system goes unstable. Blue lines indicate the point when the system first
returns to stable operation, as the heater power is being decreased.
Errors bars are included, and are taken from the rms values generated during the data
collection process. Points below mass flow rates of 0.5 g/s were not taken due to high
power requirements coupled with the low flow rates, which produced extremely high grid
temperatures and risked overheating various elements of the apparatus. At high flow
rates, the experiments were limited by the maximum throughput capacity of the blower.
34
1400
P ower inc reasing
P ower dec reasing
1200
P ower [W ]
1000
800
unstable
600
400
stable
200
0.5
1.5
2.5
3.5
Mass Flow Rate [gm/sec]
Figure 2-12: Stability boundary for Rijke tube at x/L = .
Of particular interest is the large hysteresis region that appears at mass flow rates greater
than approximately 3 g/s.
This is most likely due to nonlinearities arising in the
acoustics, but may also be partially attributable to transitions to turbulence, postulated by

Poncia (1999). Mass flow rates from 3.0 to 3.5 g/s correspond to velocities of 40 to 50
cm/s, and Re DH (Reynolds number based on hydraulic diameter) in the range of 1200 to
1500. Transitional Reynolds numbers are typically in the range of 2000 to 3000 for
normal pipe flows. However, the presence of the grid and frame may trip the system to
turbulence earlier. The flow field is further complicated since the flow will not become
fully developed in the relatively short (1 m long) tube length. At moderate flow rates,
35
there is no significant hysteresis beyond the uncertainty present in the experiment. At

very low mass flow rates, the extreme slope of the curve also does not appear to contain
hysteresis.
2.4 Prediction of the Stability Boundary

A condition for the stability boundary of the Rijke tube can be found by consideration of
the energy added and dissipated by the system. Starting with the conservation equations
of mass and momentum (equations 1-1 and 1-2) and expanding in terms of small
fluctuations, a linearized wave equation with a heat source can be derived assuming a
one-dimensional model, with a nearly uniform temperature profile. The linearized wave
equation containing a heat source (Maling 1963) and generic linear damping (Howe,
1998) is given by
(2-1)
p 'tt (t , x) + p 't (t , x) a 2 p' xx (t , x) = ( 1) Q& t ' (t , x) ,
where p is the sound pressure, is the gas (air) density, is the ratio of specific heats, Q&
is the heat release rate per unit mass, represents generic linear damping, and
( ) indicates fluctuating quantities.
The solution of equation (2-1) is sought as a Fourier
expansion of the eigenmodes of the chamber, following the method of Culick (1976).
(2-2)
p
= n (t ) n (x )
po n=0
&n (t ) d
n (x )
2
n=1 k n dx
(2-3)
u =
(2-4)
n ( x ) =sin (k n ( x + lc )) ,
36
where the acoustic field has been represented by the modeshape , with a time-varying
amplitude, , and a length correction lc, has been introduced to compensate for the nonideality of the node locations of the acoustic mode outside the chamber with respect to
the actual length, L. With the assumption of a uniform temperature profile along the
tube, the speed of sound is constant and the modeshapes are orthogonal. Considering the
simplest case of the existence of only one unstable mode, namely the first mode of the
system (n=1), equation (2-1) can be converted from a partial differential equation into an
ordinary differential equation.
Substitution of equation (2-2) into equation (2-1),
multiplying both sides of the resulting equation by 1 ( x) , and integrating over the
effective length of the tube, a dynamic equation for the amplitude of the first mode is
obtained (indices dropped for convenience)
(2-5)
&& + & + 2 =
1 2 L+lc &
Qt ' (t , x) sin(k ( x + lc ))dx ,
po L + 2lc lc
where k=/a represents the wave number.
A common model for the steady heat release in the Rijke tube is that the heat release is
coupled to the flow velocity in the system. In a similar way, the unsteady heat release
couples with the instantaneous velocity with the addition of a time delay (Putnam and
Dennis 1954). Due to symmetry considerations, it is evident that the heat release should
be independent of the direction of flow, but rather proportional to the magnitude of the
flow. It is also assumed that the heat release takes place in an infinitesimally thin region
characterized by the location of the heater grid, lg. With these considerations, the heat
release rate per unit mass can be expressed by
37
(2-6)
P
u ' (t , x )
Q& (t , x) =
( x lg ) ,
1+
S
uo
where P is the electric power supplied to the grid and S is the cross-sectional area of the
Rijke tube. The acoustic velocity of the first mode can be found from equation (2-3), and
is given by
(2-7)
u ' (t , x) =
&
cos(k ( x + lc )) .
k
Since we are interested in determination of the stability boundary, the acoustic velocity
can be considered to be much smaller than the mean flow velocity. Linearizing equation
(2-6) to form Q& and making use of equation (2-7) for the acoustic velocity, the integral
on the right-hand side of equation (2-5) can be solved, resulting in
(2-8)
&&(t ) + & (t ) + 2 (t ) =
1 1 P
sin (2k (l g + lc ))&&(t )
2 pouo S
= c&&(t ) ,
where the forcing and damping parameters are represented by c and respectively.
They are assumed to be of low enough magnitude, that they can be considered to be small
corrections to the undamped linear oscillator equation. The mode amplitude can be
approximately considered to vary harmonically in time, with a growth parameter A
yielding
(2-9)
(t ) = A sin ( nt ) .
Making use of equation (2-9), the energy dissipated per cycle is
38
2 /
Wd =
(2-10)
& (t )& (t )dt = A .

2
Similarly, the energy added to the acoustic mode per cycle is given by
2 /
Wa =
(2-11)
c&&(t )& (t )dt = c
sin( ) A 2 .
The condition for the onset of instability occurs when Wa > Wd. Comparing equations
(2-10) and (2-11) and using equality of Wa and Wd to indicate the stability boundary, the
system first goes unstable when
2 (l g + lc )
1 1 P
sin(
) sin( ) .
2 pouo S
L + 2lc
(2-12)
Taking into account damping in wall boundary layers (other losses typically have
negligible contributions), the losses are modeled classically by (Howe, 1998)
L 2
=
R
(2-13)
( + ( 1) ) ,
aL
where is the thermal diffusivity, is the kinematic viscosity, and R is the tube radius.
Thus we arrive at
(2-14)
2 (l g + lc )
1 1 P
L 2
sin(
) sin( )
2 pouo S
L + 2lc
R
( + ( 1) ) ,
aL
which is an explicit criterion for instability involving all relevant parameters: supplied
power, mean flow velocity, grid location, natural frequency, time delay, system geometry
and fluid properties.
39
Note that the time delay is generally found from previous experiments and simulations,
to be less than a quarter of a period, so the phase can be considered to lie between 0
and / 2 . Several observations can be drawn from equation (2-14).
A necessary condition for instability is the location of the grid in the first half of
the tube (in accordance with Rayleighs criterion).
No limit cycle is possible using a linear stability analysis.
In the limits u 0 or P 0 with the other parameters fixed, the system is

stable.
In the limit as P the system is unstable.
The limit u 0 0 cannot be analyzed with this approach, since it violates the initial
assumption that the acoustic velocity is a small fluctuation imposed over the mean flow.
It is evident that a stability diagram with flow velocity and power input as the variable
parameters will result in a linear relationship. However in reality, the time delay and
amount of energy transported to the acoustical mode are not constant, so that the exact
stability boundary may not be predicted by equation (2-14).
2.5 Comparison of Prediction with Experimental Results

In order to reconcile the model of the previous section with experimental results, a better
model must be constructed for the heat release, which takes into account the time delay
and change in heat transfer characteristics at low flow rates. Kwon and Lee (1985)
performed numerical simulations to compute the heat transfer from an isothermal wire to
an acoustic wave. They define an efficiency factor, E, resulting from their simulations,
which relates the amplitude of the heat release and phase delay as a function of non-
40
dimensionalized mean flow velocity, uo* = uo / , and heater wire radius, r * = r / .

Transforming equation (2-14) for use with the model of Kwon and Lee, the efficiency
factor can be defined as
E=
(2-15)
R
a
sin ( )
2uo L
and is plotted for r* = 1.0 in Figure 2-13.
0.1
0.08
0.06
0.04
0.02
0.5
1.5
2.5
*
u
o
Figure 2-13: Efficiency factor, E, for r* = 1.0.
Recall the assumption that the temperature is approximately uniform in the tube.
Temperature selection affects the physical properties of the air, which in turn changes the
scaling of the non-dimensional parameters.
Two mean temperatures were selected,
41
which are representative of the low and high range of temperature in the tube. For the
experimental conditions used, r* varies from approximately 0.5 at 600K up to 1.0 at
300K. The difference in the efficiency curve for r* between 0.5 and 1.0 is very small, so
only the efficiency factor for r* = 1.0 is considered.
The resulting stability boundary is plotted in Figure 2-14 for both temperatures.
Qualitatively, the predicted curves show the correct shape, but fail to accurately predict
1400
P ower inc reasing
P ower dec reasing
1200
P ower [W ]
1000
800
600
400
200
0.5
1.5
2.5
3.5
Mass Flow Rate [gm/sec]
Figure 2-14: Stability boundary prediction. Solid black line is for T = 600K, and the
dashed black line for T = 300K.
42
the stability boundary. It can be argued that the lower temperature curve is more suitable
for low power inputs, but it still over-predicts the power required to cause instability. In
general, the model over-predicts power requirements at low flow rates, and underpredicts power requirements at higher flow rates.
2.6 Summary
The experimental apparatus and approach for collecting a set of accurate measurements
that characterize the Rijke tube have been presented. The experiment is capable of
capturing both steady-state and transient behavior, spanning the relevant parameters
(mass flow rate, heater power input, and heater position) over a wide range of values. In
addition, bulk temperature profiles are collected along the centerline of the tube, which
have not been measured previously. Experimental results show the presence of hysteresis
at high mass flow rates (above 3 g/s). A stability curve is presented summarizing the
stability characteristics of the Rijke tube, with the heater at a position of x/L = .
A one-dimensional model using linear acoustics is used in conjunction with numerical

heat transfer results from Kwon and Lee (1985) to attempt to predict the stability
boundary. It is based on the physical parameters of the experiment, and is not limited to
this configuration only. The results agree qualitatively, although accurate prediction of
the stability curve was not achieved. It is evident that a more detailed model, involving
an accurate heat transfer model including radiation
fluid mechanical considerations
at least two-dimensional, possibly three-dimensional effects
nonlinear acoustics and multiple acoustic modes
43
will be necessary to predict accurately the limit cycle amplitudes and stability
characteristics of the Rijke tube.
44
Chapter 3
Full-Scale Flare and
Sub-Scale Model
This chapter describes a practical, industrial application of a device exhibiting
characteristics similar to a Rijke tube, namely a large flare. A description of the device
and the data collected on site is contained in this chapter. This chapter also describes the
sub-scale design and modeling efforts, as well as the results that were obtained.
Ultimately, this motivates the more advanced diagnostic techniques that were employed
in Chapter 4.
45
3.1 Full-Scale Flare

3.1.1 Introduction
One of the nations largest landfill sites, local to Caltech, uses a pair of large flares to burn
off its production of landfill gas. As solid waste material decomposes, landfill gas is
collected by a system of gas wells, trenches, and collection pipes. A portion of the
landfill gas (composed primarily of methane and carbon dioxide) is used to power a 50
MW power plant and on-site vehicles. The remainder is sent to the flare station, shown
in Figure 3-1, to be burnt under controlled conditions, minimizing emission of unburnt
hydrocarbons (UHC), NOx, and other undesirable pollutants.
This particular landfill site receives in excess of 12,000 tons of solid waste per day. As
the decomposition rates are anticipated to continue to increase, so will the production of
Figure 3-1: Landfill flare station.
46
landfill gas. In anticipation of increased capacity requirements, each flare is designed to

process landfill gas at a maximum rate of 5500 SCFM. However, at approximately 50%
of maximum capacity, the flares emit a low frequency rumbling, which disturbs local
homeowners. This acoustic, low frequency noise is a result of the flares being unstable.
3.1.2 Description of Full-Scale Flare
The flare can be considered to be a type of Rijke tube, with closed-open boundary
conditions, versus the Rijke tubes open-open boundary conditions. Each flare stands at a
height of 45 feet, with an inner diameter of 12 feet. The burner inside the flare consists
of an array of 180 fuel spuds (see Figure A-2 for more details) distributed evenly on ten
welded straight lengths of 12-inch steel pipe. Each spud is 4.5 inches in height with an
inner diameter of 0.688 inches. The tops of the spuds extend to a height of 5.7 feet
relative to the ground. Landfill gas exits the spuds into a burner block section, which
extends to a height of 8.8 feet relative to the ground. The burner block contains an
eductor section, which accelerates the flow and allows it to jet-mix with entrained air.
Air enters the flare through two sets of horizontal louvers, located on opposite sides at the
bottom of the flare. The smaller louver is 3 feet high by 3 feet wide, while the larger
louver is 4 feet high by 7 feet wide. During the tests undertaken on-site, analysis of the
landfill gas mixture showed 40% methane, 33% carbon dioxide, and the balance made up
of nitrogen and trace gases (Bjerkin, 1999).
47
3.1.3 Diagnostics
In order to characterize the flare system, sets of instrumentation arrays were designed to
slide into access ports in the side of the flare and on the gas feed line.
Each
instrumentation pair consisted of a pressure transducer (RE Technologies, Model PTX1)

recessed to prevent damage from high temperatures and a type-K thermocouple. The
Scaffold
Instrument Pair 1
(at 37 feet)
Instrument Pair
2 (at 9.2 feet)
Burner Block
Pressure Transducer
4 (on gas inlet)
Inlet air
Instrument Pair
3 (at 1 foot)
Figure 3-2: Instrumentation layout on full-scale flare.
48
thermocouples extended approximately 2 feet into the chamber. The pairs were located
at heights of 37 feet, 9.2 feet, and 1 foot, labeled 1, 2, and 3 respectively. An additional
pressure transducer was located on the gas inlet line and is referred to as transducer 4.
Data was acquired on site with a Computer Boards CIO-DAS1602/12 data acquisition
board and a CIO-EXP16 expansion board for the thermocouples, both installed in a
Pentium II 400 MHz computer. Pressure data was acquired at 1 kHz, while temperature
data was multiplexed and acquired at 2 Hz. A diagram showing the instrumentation on
the flare and a few of the flares physical characteristics is shown in Figure 3-2.
3.1.4 Results
The flare was run at mass flow rates of 2500-3100 SCFM. The data presented here is for
the 3100 SCFM flow condition, since it is the highest mass flow condition tested and is
representative of the system instability.
Fuel composition was approximately 40%
methane and 33% carbon dioxide with a balance of nitrogen and trace gases. The
temperature distribution at the three data ports is shown in Figure 3-3. This represents
the bulk temperature in the flare, since the time constant of the thermocouples is too large
to respond to acoustically driven temperature fluctuations. The large jump in temperature
at 9.2 feet is due to the presence of the burner block section just below this data port
(Figure 3-2).
Pressure data taken from the site is shown in Figure 3-4, for each of the four
measurement locations. This data has been low-pass filtered with a cutoff frequency of
200 Hz to remove any high frequency noise.
Note the lower amplitudes of the
oscillations for Pressures 1 and 3, since Pressure 1 is situated near the top of the tube
49
(atmospheric boundary condition), and Pressure 3 is located at the bottom of the tube.
Even though there is an acoustically solid boundary at the bottom of the flare, the
presence of the open louvers introduces a mixed boundary condition, at least with respect
to the pressure transducer, which only penetrates 2 feet into the chamber through the
bottom louver. Pressure 4, located on the gas inlet line, is similar to Pressure 2, even
though there is the burner section in between the two transducers. The pressure traces
show these positions to be of similar amplitude and qualitatively, to follow the same
pattern.
Figure 3-3: Flare bulk temperature profile.
50
Figure 3-4: Flare pressure traces (low pass filtered at a cutoff of 200 Hz).
51
FFTs for the pressure traces are shown in Figure 3-5.
They indicate peaks at
approximately 8.6 Hz and 21.5 Hz, corresponding to the first and second modes of a
closed-open acoustic system.
Making use of the FFT results, the pressure traces can be selectively bandpass filtered
about their most prominent frequencies.
A 6th order butterworth phase-preserving
bandpass filter is applied to the pressure data about the first and second acoustic modes of
the system, and displayed in Figure 3-6 and Figure 3-7 respectively. Figure 3-6 clearly
shows all pressures to be in-phase, as would be expected for a quarter-wave (the 1st mode
of an open-closed system). Corresponding to this result is the bandpass filtered data for
the 2nd mode. Figure 3-7 indicates that Pressure 1 and Pressure 2 are generally speaking
in-phase, which suggests that the pressure node of the 2nd mode occurs somewhere
below Pressure 2. Pressure 3 is out-of-phase with the pressures in higher sections of the
flame, as expected, with the inlet Pressure 4 again mirroring what is seen at Pressure 2.
52
Figure 3-5: FFT of pressure traces in full-scale flare.
53
Figure 3-6: Flare pressure traces, filtered using a 6th order butterworth phasepreserving bandpass filter between 7-10 Hz (1st mode).
54
Figure 3-7: Flare pressure traces, filtered using a 6th order butterworth phasepreserving bandpass filter between 21-24 Hz (2nd mode).
55
3.1.5 Possible Driving Mechanisms

There exist a number of potential mechanisms that could be driving the flare towards
instability. These may be working independently or acting in concert to exceed the
natural losses associated with the flare system. They can be summarized as follows:
Indirect Energy Transfer: Combustion induced buoyancy creates the draft of air; air
flowing past physical edges causes vortex shedding that then excite acoustic waves.
This is analogous to the operation of wind instruments or sirens.
Feed System Coupling: Fluctuations of pressure and/or velocity in the fuel supply
system and in the air flow through the louvers cause fluctuations of the fuel/oxidizer
mixture ratio, which then cause fluctuations in the energy release that further pump
the acoustic resonance.
Chemical and/or Heat Transfer Sensitivity: Combustion processes in the burners are
sensitive to pressure and velocity fluctuations, producing an internal feedback path
which causes the combined system (burner and flow dynamics) to be linearly
unstable.
Attempts to clarify which mechanism is responsible for the instability through sub-scale
modeling of the system are detailed in the next section.
3.2 Sub-Scale Flare Model

3.2.1 Apparatus and Diagnostics
A 1/12th sub-scale model of the flare was designed and constructed, in an attempt to
characterize the acoustic properties of the chamber and its interaction with the burner,
and is shown fully assembled in Figure 3-8. The model consisted of a 12.5-inch diameter
56
quartz tube, 42 inches in length, with six -inch laser drilled holes along one side for
instrumentation access. The quartz tube rests inside an aluminum base assembly that has
front and back cutouts to simulate the louver action of the full-scale flare. The aluminum
base and quartz tube together form the chamber section, shown in Appendix A,
Figure A-1.
The corresponding burner section is shown in detail in Figure A-2. It is comprised of

eight lengths of 1-inch diameter aluminum tubing, welded together to form the inlet feed
system of the burner. There are five spuds on each length of tubing, which serve to
provide fuel. The burning region, which occurs above the two rings, is shown in the
Section A drawing of Figure A-2. These two rings form the eductor section of the
burner, which was only approximately modeled in this apparatus and will be explored
further in the next chapter.
The data acquisition system used on the sub-scale flare was essentially the same as that
on the full-scale flare, though the specific transducers differed.
The type-K
thermocouples (Omega KMQSS-062U-6) and pressure transducers (PCB 112A04 with

422D11 charge amplifiers) were mounted in water-cooled instrumentation bosses (Figure
A-3) that attached to the quartz tube through the laser drilled instrumentation ports.
57
Exhaust
Model Ring
Burner
Quartz Tube
(12.9inch I.D.,
42 inch length)
Aluminum Base
(5 inch length)
Air Inlet
Louvers (2 PL)
Figure 3-8: Schematic of sub-scale flare model assembly.
58
3.2.2 Scaling of the Model

The first step in characterizing the sub-scale flare model was an attempt to operate it in an
unstable regime. A wide variety of conditions were experimented with to elicit
oscillations from the model including:
Diffusion flames
Premixed flames
Single and double meshes above the burner spuds
Raising the burner from its natural position to x/L = 1/3 up to x/L = 1/2
Running with and without burner shields
Placing a mesh over the shields
Doubling the chamber length by adding a stainless steel extension tube
Secondary fuel pulsed injection
The strategy behind most of these techniques was to increase the intensity of heat release
in a region that would be favorable to drive an acoustic instability according to
Rayleighs criterion. Switching from a diffusion to a premixed flame and using meshes
both serve to concentrate the heat release in a more localized region. If the heat release is
distributed too evenly, regions of strong damping will be created that will cancel out
driving regions.
Raising the position of the burner placed the heat release in a
geometrically more favorable position for the excitation of the 2nd mode. This did not
induce oscillations, hence, changes were next applied to the chamber geometry rather
than just the burner.
59
Doubling the chamber length served two purposes to lower the fundamental frequency
of the system, as well as increasing the aspect ratio, L/D. This is significant due to the
scaling of the acoustic radiative losses out the open end of the chamber. According to
Clanet et al. (1999), the characteristic time for radiative damping can be given by
(3.1)
1
rad
(D )2 ,
8aL
where is the acoustic frequency, D is the chamber diameter, a is the speed of sound,
and L is the chamber length. The acoustic radiative losses are proportional to 2/L.
Increasing the chamber length by a factor of 2 correspondingly decreases the fundamental
frequency by a factor of 2. Therefore, a doubling of the chamber length results in an
eight-fold decrease in acoustic radiative losses. Similarly, Clanet et al. (1999) show that
diffusive losses due to the presence of viscous and thermal boundary layers at the lateral
walls are proportional to . As a result, lowering the fundamental frequency will also
lower acoustic diffusive losses.
Another technique that was tested was to inject secondary fuel into the system, using an
automotive fuel injector. Secondary fuel injection was superimposed over the main fuel
feed, so the system was subjected to a mean fuel flow rate with a fluctuating component.
The objective was to enhance natural resonant frequencies, by providing small-scale
disturbances that would hopefully be further amplified as the instability developed into a
limit cycle. This concept had physical relevance, since the landfill gas entering the fuelscale flare would have a varying fuel composition and thus a fluctuation in fuel flow rate.
60
None of the mentioned techniques were successful in generating a naturally selfsustained limit cycle behavior in the sub-scale flare model.
The most promising
technique was secondary fuel injection coupled with the extended chamber length,
producing measurable pressure oscillation in the chamber at 130 Hz. This case, however,
was deemed to be too artificial, since the amplitude of the fluctuating fuel flow rate was
of the same order as the mean flow rate. As a consequence, any attempts to stabilize the
sub-scale model would not be relevant to the full-scale flare.
3.3 Conclusions
Measurements taken of the full-scale flare operating in its unstable regime indicate the
presence of modes at 8.6 Hz and 21.5 Hz. This corresponds well to the natural first (1/4
wave) and second (3/4 wave) modes that would arise in a closed-open system. The
frequencies do not match precisely due to changes in the speed of sound due to
temperature gradients in the chamber.
The pressure amplitudes do not correspond
completely to a closed-open system, since at the closed boundary the presence of louvers
produces a mixed boundary condition.
Although the bottom of the flare is closed
acoustically to a longitudinal mode, the louvers open the bottom of the flare to the
atmosphere, greatly dampening the amplitude of the pressure oscillations.
The sub-scale model of the flare was unsuccessful at developing self-sustained pressure
oscillations. This was due in large part to the distributed nature of the heat release
(approximately over an 18 inch high region) in the model, compared to the relatively
concentrated heat release that occurs in the flare. Another factor was the acoustic losses
due to changes in the frequencies being studied. Artificial pumping of the acoustic
61
modes in the sub-scale model was achieved using a fuel injector; however, this would not
be useful in a study to suppress oscillations in the full-scale flare. It therefore becomes
necessary to take a closer look at the dynamics of the flame in order to improve its
stability characteristics.
62
Chapter 4
Chemiluminescence
Measurements
This chapter describes the methods used to visualize a flame under an acoustically forced
pressure field. It relates to Chapter 3 in that the burner configuration is modeled after an
individual spud from the flare and is motivated by a need for a more fundamental
understanding of combustion dynamics in flames in unstable systems. The flowfield is
imaged using two different techniques: shadowgraph and chemiluminescence.
IN
addition, chemiluminescence gives a measure of the heat release rate in the flame.
Details regarding the experimental setup and diagnostics are also provided in this chapter.
63
4.1 Experimental Setup

The test section, shown in Figure 4-1, consists of three major components: the acoustic
driving system; the acoustic cavity; and the burner section. It is based in part on the subscale model of the flare burner, discussed in Chapter 3.
4.1.1 Acoustic Driving System
The acoustic driving system is mounted above the acoustic cavity on the outer quartz
tube.
It consists of a large tubular stainless steel section in the shape of a cross,
approximately 12 inches in diameter, which extends the overall length of the acoustic
chamber an additional 24 inches.
The exhaust section is open to the atmosphere,
providing an acoustically open exit condition. A pair of acoustic drivers are sealed to a
pair of air jet film cooling rings (to prevent failure of the drivers), which are in turn
sealed to opposite sides of the steel structure.
The acoustic drivers are 12 inch
subwoofers (Cerwin-Vega model Vega 124), with a sensitivity of 1 W @ 1 m of 94 dB,

and a continuous power handling capacity of 400 W.
A 1000 W power amplifier
(Mackie M1400i) and a function generator (Wavetek 171) provide the power and signal
to the acoustic drivers. Significant power is required to provide reasonable amplitude
pressure oscillations.
The amplitude of the fundamental driving mode is actively
controlled by custom-designed electronics (see section 4.3.4), which measure the

pressure in the acoustic chamber at the burner with a pressure transducer (PCB 106B50),
and appropriately scale the power output of the speakers. The signal from the transducer
is notch-filtered to ensure the intended driving mode is correctly amplified or attenuated.
64
Exhaust
Film
Cooling
Ring
Power
Amplifier
Acoustic
Driver
System
Function
Generator
Controller
12 in, 800
Watt
Acoustic
Driver (2 PL)
Pressure Transducer
Flame
Burner
Inner Quartz Tube

(2.2in I.D., 4.5in L)
Eductor
Block
Fuel Spud
Air Inlet
Louvers
(2 PL)
Stainless
Steel Tubes
(24 in L)
Outer Quartz Tube
(12.9in I.D.,
42in L)
Acoustic
Cavity
Aluminum
Base (5in L)
Figure 4-1: Schematic of overall test section.
4.1.2 Acoustic Cavity

The acoustic cavity consists of an aluminum ring, closed at the bottom end. It has two
sets of inlet louvers cut on opposing sides to allow air to flow into the tube, while
providing an acoustically closed end condition. A large diameter-matched quartz tube
rests in a thin register on the aluminum ring, and extends for an additional 42 inches.
65
Quartz was used in order to withstand high flame temperatures, as well as to allow
transmission of the ultraviolet laser sheet and fluorescence signal. The tube also has
several laser-drilled holes at various locations to provide instrumentation entry ports. See
Appendix A for more precise details on the dimensions of the acoustic chamber.
4.1.3 Test Burners
The burner sections are shown in Figure 4-3, in the aerodynamically stabilized and bluffbody stabilized configurations used.
The design allows for a variety of different
flameholder configurations to be easily tested. Elements common to both arrangements

include a fuel spud that ejects a premixed jet of fuel and an additional gas into the eductor
block made of machinable ceramic. The jet entrains air as it enters the eductor, where it
is jet-mixed, resulting in a partially premixed flame.
For the aerodynamically stabilized flame, the flame is stabilized above the recirculation
zone created as the flow exits the eductor and expands into the 4.5 inch tall burner tube.
In the bluff-body stabilized burner, two additional tabs (constructed of machinable
ceramic) are provided in the stabilization zone, which can provide a stronger recirculation
zone for the flame to attach itself. In this case, the burner tube is 3.75 inches tall, with
the remaining height taken up by a small quartz piece and the ceramic flameholder to
bring the total height to 4.5 inches. The tabs are approximately 0.5 x 0.5 x 0.5 inches in
size yielding a blockage of approximately 7.4%, and are tapered on the upstream side.
More details of the bluff-body tabs can be seen in Figure 4-2. Each burner has two sets
of burner tubes in which the flame is stabilized, one made of pyrex (I.D. 2.17 inches,
O.D. 2.35 inches) and the other one of quartz (I.D. 2.17 inches, O.D. 2.33 inches). The
66
pyrex burner tubes are used for the shadowgraph and chemiluminescence measurements.
Quartz tubes are required for the PLIF experiments, since they need to be able to transmit
UV light. They also have two 1/8-inch slits cut on opposite sides in order to allow the
laser sheet to pass through and illuminate the flame. The slits eliminate luminescence of
the quartz tube caused by the laser sheet, which interferes with the fluorescence signal.
Figure 4-4 provides details of the gas feed system. Fuel for the burner is 50% methane
premixed with 50% CO2 gas to increase the mass flow. The premixer inlets for each gas
are choked in order to prevent disturbances from propagating upstream and affecting flow
rates. The mixture is subsequently passed through a laminar flow element (Meriam
Model 50MJ10 Type 9). The temperature of the mixture is measured by a type-K
thermocouple and a digital thermometer (Analog Devices, Model AD2050-K), while the
pressure drop is measured by a barocel pressure sensor (Datametrics model 590D-10W3P1-H5X-4D, 1400 electronic manometer). From these measurements, the flow rate is
determined.
The flow then exits the fuel spud and entrains atmospheric air.
The
volumetric flow rate through the spud is 2.14 SCFM, yielding a jet velocity of 30 m/s (Re
= 20,000).
Figure 4-2: Bluff-body stabilized burner flameholder detail, viewed from the
upstream side.
67
(a)
(b)
Figure 4-3: (a) Aerodynamically stabilized burner (b) bluff-body stabilized burner.
68
To Burner
Thermocouple
Meriam Laminar
Flow Element
Impinging Sonic
Jet Premixer
Flow Interrupter
Valve
Flow Control
Valve
Regulated Pressure
Indicator
P
Pressure
Regulator
Bottle Pressure
Indicator
P
CO 2
CH 4
Figure 4-4: Gas feed system.
4.2 Acoustic Properties

In order to determine the acoustic properties of the chamber, a second pressure transducer
was used to traverse the height of the chamber, while using the acoustic drivers to excite
the system during a cold test.
Figure 4-5 shows the peak-to-peak pressure amplitudes under identical power conditions
to those in the hot flame conditions. The driving frequencies used in this section were 22,
69
x 10
-3
M odes hape of A cous tic Cham ber
P ress ure A m plitude (peak -to-peak ) [psi]
3.8
22
27
32
37
55
3.6
3.4
Hz
Hz
Hz
Hz
Hz
3.2
3
2.8
2.6
2.4
2.2
10
20
30
Height [inc hes ]
40
50
60
Figure 4-5: Peak-to-peak pressure amplitudes in the chamber, driven at various

frequencies (no flame).
RM S P ress ure [psi]
12
x 10
-4
RM S P ress ures in A cous tic Cham ber
10
22
27
32
37
55
Hz
Hz
Hz
Hz
Hz
10
20
30
Height [inc hes ]
40
50
60
Figure 4-6: RMS pressure in the chamber, when driven at various frequencies (no
flame).
70
27, 32, 37, and 55 Hz2. Due to difficulties inherent in producing low frequencies by the
acoustic drivers (a very large excursion range is required), less power is given to the
system at 22 Hz and 27 Hz to avoid driver failure. This is evident in the lower peak
amplitudes at 27 Hz, however it is not as evident at 22 Hz. In the 22 Hz case, nonlinearities in the drivers cause additional modes to be excited and cause the increased
peak amplitudes. Figure 4-6 displays the rms pressures for the same test. Due to the
driver nonlinearity, the rms pressure for 22 Hz matches those for the higher three
frequencies, which are driven at the same (higher) power level. The drivers have a
smoother response at 27 Hz that results in lower rms pressures at this frequency, since it
is driven at a lower power level.
An important result from this test is the modeshape of the acoustic wave that the driving
system establishes in the chamber. In Figure 4-5, the variations in peak amplitude at
different heights and frequencies are shown. The maximum variation in peak amplitude
is less than 3% from the mean amplitude for 32-55 Hz, increases to 6.4% at 27 Hz, and
increases further to 8.1% at 22 Hz. If any of the driving frequencies excited a natural
mode of the system, a distinct modeshape would be apparent from the amplitudes of the
pressure traces at different heights (i.e., nodes and antinodes would be identifiable).
Since the variation in amplitude is relatively low and the curve is flat, it can be
reasonably concluded that the acoustic drivers produce a bulk mode in the system and not
a standing wave.
Actual frequencies were 22.02, 27.02, 32.02, 37.02, and 55.02 Hz, to prevent the system from modelocking with the 10 Hz laser repetition rate.
71
4.3 Diagnostics
This section describes the diagnostics used in the flow visualization and combustion
dynamics experiments, excluding equipment specific to PLIF, which is found in the next
section.
4.3.1 Pressure Transducers
A piezoelectric pressure transducer (PCB Piezotronics, model 106B50) is located at a
height of 3 inches above the fuel spud, where the flame is stabilized in the burner. This
transducer was selected for its high sensitivity (493.3 mV/psi) and thermal
characteristics. An additional piezoelectric transducer (PCB, 112A04) with a sensitivity
of 2258 mV/psi was used to traverse the length of the test section when determining the
modeshape of the system under various forcing frequencies. A PCB model 482A16
power supply and amplifier, using a gain stage setting of 100, powered both transducers.
4.3.2 Data Acquisition System
The computer used in the data acquisition system consisted of an AMD Athlon 650 MHz
processor with 512 MB of RAM and approximately 100 GB of total hard drive space
(image files are large). The system contains a CD-RW drive used to archive the data.
Installed in the computer are two DAS1602/16 (Computer Boards) 16-bit data acquisition
boards. One operates in differential mode measuring quantities including the oscillator
frequency, pressure, laser shot energy, camera triggers, and control effort, while the other
board operates in single-ended mode and measures three type-K thermocouple
temperatures in the acoustic chamber (at heights of 18, 25, and 37 inches). Also in the
same computer is the PCI controller card for the Princeton Instruments ICCD camera and
72
the IEEE 1394 FireWire card used to interface with the Vision Research Phantom V4.0
camera. The software package used to acquire data is National Instruments LabView 5.1.
4.3.3 High-Speed Video Camera
A Vision Research Phantom V4.0 high-speed video camera is used to capture images for
the shadowgraph and chemiluminescence experiments. It is based on a proprietary 512 x
512 pixel monochrome SR-CMOS (Synchronous Recording, Complementary Oxide
Metal Semiconductor) sensor, capable of exposure times as low as 10 s. The camera
contains 256 MB of memory on board, which allow it to acquire data at 1000 frames per
second, for just over 1 second. Higher frame rates are possible by lowering the pixel
resolution. The camera is equipped with a C-mount, and a Nikon 50 mm F/1.4 lens is
used with a C-mount-to-F-mount adapter. Use of a lens with a low f-stop number
increases the light gathering capacity of the lens and decreases its depth of view. A
decreased depth of view is advantageous, since it minimizes line-of-sight integration,
though small contributions out of the focal plane that smear the image will be inevitable.
4.3.4 Additional Electronics
Two custom electronics units used in the experiments were designed and built by coexperimenter Steven L. Palm. The first is an active control unit that regulates the power
amplifier output to the acoustic drivers. Output from the pressure transducer (PCB
106B50) is notch filtered by the controller to determine the amplitude of the fundamental
frequency the acoustic drivers are producing in the chamber. This is compared with an
adjustable preset value in the controller, which uses PI control to regulate the system to
ensure a constant pressure amplitude at the transducer location (a height of 12 inches,
73
approximately level with the flame stabilization zone).
The second unit is a peak
capture/amplifier box, which reads the output from the pyroelectric joulemeter
(Molectron J9LP) measuring laser shot energy and amplifies the signal so it can be read
by the data acquisition system. More details on this equipment can be found in the work
of Palm (in progress).
4.3.5 Shadowgraph Setup
The reacting flowfield was visualized using the high-speed Phantom V4.0 video camera
described above, in a standard Z-configuration shadowgraph arrangement. Two 30 cm
diameter spherical mirrors collimate the continuous light generated by an Ealing 250W
universal arc-lamp supply driving a Mercury arc-lamp (Ealing, Model 27-1031). More
details on the shadowgraph imaging arrangement are shown in Figure 4-7. Due to the
curved surfaces of the burner and acoustic chamber, imaging was performed on an
unforced flame above the burner tube.
Hg lamp
Lens
Pinhole
Spherical
Mirror
Spherical
Mirror
Burner
High-speed
Video camera
Figure 4-7: Shadowgraph imaging setup.
74
4.3.6 Chemiluminescence Measurements

In combustion experiments, a now common method of determining the reaction zone is to
image the light emitted in the combustion zone (McManus et al. 1995). Of particular
importance are C2, CH, and OH radicals, since they are produced as intermediaries of the
combustion process. Hurle et al. (1968) established the linearity between the radiation
emitted and the volumetric heat release.
Numerous studies involving combustion
instabilities have taken place using this relationship (Samiengo et al. 1993; Shih et al.
1996; Broda et al. 1998). The major limitations with this technique are the integration of
the intensities along the line of sight, and relatively long integration times O(~100 s)
due to low signal strength.
Consistent with previous JPC researchers, no filters are used, and the total radiation
emitted from the combustion process is taken to be proportional to the heat release rate.
The measurements taken by Sterling (1987) used a masked photomultiplier to achieve
spatial resolution, and were subsequently improved upon by Zsak (1993) and Kendrick
(1995), with the introduction of the Hycam (high-speed film camera) providing twodimensional spatial resolution. In this work, advances in imaging technology further
simplify measurements by use of the Phantom digital high-speed video camera described
earlier. This provides temporally and spatially resolved measurements of the heat release
subjected to a forced acoustic field, without the jitter introduced by a high-speed
mechanical film camera.
75
4.4 Shadowgraph Results

In this particular configuration, the shadowgraph imaging technique provided limited
results. Results could not be obtained under forced conditions and in the interior of the
burner tube. While the flow fields are similar, there exist some differences between the
aerodynamically and bluff-body stabilized cases, as shown in Figure 4-8. The bluff-body
burner shows wider spreading of the flame outside of the burner and also displays finer
structures in the downstream section. These can be attributed to the stronger recirculation
zones and vorticity generated by the bluff-bodies.
(a)
(b)
Figure 4-8: Shadowgraph results above the burner tube for (a) aerodynamically and (b)
bluff-body stabilized cases.
76
4.5 Chemiluminescence Results

4.5.1 Two-Dimensional Flame Structure
Single shot chemiluminescence images from the Phantom V4.0 camera were smoothed using
a 3 x 3 median filter in Matlab. Images were then averaged by phase-locking to the pressure
signal. Images were selected by locating their temporal locations, and selecting the images to
be averaged based on their proximity to the sixteen phase divisions used. Approximately 15
images were used in the average at each phase position. The maximum phase resolution
jitter was found to be less than 2 degrees in all cases. Contours were computed and plotted
for each case (five forcing frequencies x two burners) showing eight phases in a cycle, and
are displayed in Figure 4-9 through Figure 4-13. Contour levels are plotted at 5, 20, 40, 60,
80, and 95 percent levels of the maximum intensity of the 0 degree phase contour plot for
each case. In all cases, the phases are taken as a sine wave, with 0 degrees corresponding to
a zero crossing with a rising edge.
For forcing at 22 Hz (Figure 4-9), the bluff-body burner shows a larger stabilization zone
than the aerodynamic burner (40% contour), centered at approximately a height of 5 cm.
Note that the center hole at 8 cm at a phase of 0 degrees is actually at a contour of 5%, and
not 40%. Characteristics to note in both cases are the traveling of a wave in the upstream
direction on the outer edge of the flame, from 0 to 180 degrees. At 180 degrees, the wave
reverses itself, and travels back downstream. There is also a distinct change in intensity as
the flame oscillates. The intensity contour at 40% can be seen to grow into the burner tube as
the flame evolves from 0 to 180 degrees, and in the bluff-body case even connects with the
40% contour levels in the stabilization zone. Again, as the pressure changes from 180 to 360
77
degrees, this intensity zone separates from each other and travels back downstream. These
two oscillating characteristics are generally observed for each forcing frequency, except for
the 55 Hz case.
At a frequency of 27 Hz (Figure 4-10), much the same phenomenon is observed, with a

decrease in the amplitude of the outer propagating wave. In addition, there appears to be a
superimposed higher frequency, lower amplitude outer wave, continuously traveling
downstream. As the forcing frequency is increased to 32 Hz (Figure 4-11), the intensity of
the flame in the burner tube has diminished note the decrease in size of the 40% contour in
the flame stabilization zone inside the tube. Once the acoustic oscillations reach 37 Hz
(Figure 4-12), an interesting reversal occurs. The contours in the stabilization zone show a
significantly larger 40% contour for the aerodynamically stabilized burner, than the bluffbody stabilized case. Recall in all previous cases, the bluff-body stabilized burner yielded
stronger stabilization zones or at least zones comparable to that of the aerodynamically
stabilized burner. Finally, for the 55 Hz case (Figure 4-13) there is essentially no change
between contours at different phases. The low amplitude traveling waves on the outer rim of
the flame are present, but none of the bulk oscillations of intensity or flame shape that
occurred in previous situations.
78
(a) 45
(a) 135
14
14
12
12
12
10
10
10
10
8
6
8
6
Height [cm]
14
12
Height [cm]
14
8
6
-5
-5
-5
Radius [cm]
(b) 45
(b) 90
(b) 135
16
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
8
6
4
-5
(a) 180
-5
Radius [cm]
(a) 225
(a) 270
(a) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
(b) 180
-5
Radius [cm]
(b) 225
(b) 270
(b) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Height [cm]
Height [cm]
16
(b) 0
Height [cm]
(a) 90
16
Radius [cm]
Height [cm]
16
Height [cm]
Height [cm]
(a) 0
16
-5
Radius [cm]

stabilized and (b) bluff-body stabilized cases.
79
(a) 45
(a) 135
14
14
12
12
12
10
10
10
10
8
6
8
6
Height [cm]
14
12
Height [cm]
14
8
6
-5
-5
-5
Radius [cm]
(b) 45
(b) 90
(b) 135
16
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
8
6
4
-5
(a) 180
-5
Radius [cm]
(a) 225
(a) 270
(a) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
(b) 180
-5
Radius [cm]
(b) 225
(b) 270
(b) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Height [cm]
Height [cm]
16
(b) 0
Height [cm]
(a) 90
16
Radius [cm]
Height [cm]
16
Height [cm]
Height [cm]
(a) 0
16
-5
Radius [cm]

80
(a) 45
(a) 135
14
14
12
12
12
10
10
10
10
8
6
8
6
Height [cm]
14
12
Height [cm]
14
8
6
-5
-5
-5
Radius [cm]
(b) 45
(b) 90
(b) 135
16
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
8
6
4
-5
(a) 180
-5
Radius [cm]
(a) 225
(a) 270
(a) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
(b) 180
-5
Radius [cm]
(b) 225
(b) 270
(b) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Height [cm]
Height [cm]
16
(b) 0
Height [cm]
(a) 90
16
Radius [cm]
Height [cm]
16
Height [cm]
Height [cm]
(a) 0
16
-5
Radius [cm]

81
(a) 45
(a) 135
14
14
12
12
12
10
10
10
10
8
6
8
6
Height [cm]
14
12
Height [cm]
14
8
6
-5
-5
-5
Radius [cm]
(b) 45
(b) 90
(b) 135
16
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
8
6
4
-5
(a) 180
-5
Radius [cm]
(a) 225
(a) 270
(a) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
(b) 180
-5
Radius [cm]
(b) 225
(b) 270
(b) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
Radius [cm]
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Height [cm]
Height [cm]
16
(b) 0
Height [cm]
(a) 90
16
Radius [cm]
Height [cm]
16
Height [cm]
Height [cm]
(a) 0
16
-5
Radius [cm]

82
(a) 0
(a) 45
14
14
12
12
12
10
10
10
10
8
6
8
6
Height [cm]
14
Height [cm]
14
Height [cm]
8
6
-5
-5
-5
(b) 45
(b) 90
(b) 135
16
16
16
14
12
12
10
10
10
10
Height [cm]
14
12
Height [cm]
14
Height [cm]
14
12
8
6
4
-5
(a) 180
-5
Radius [cm]
(a) 225
(a) 270
(a) 315
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
Height [cm]
16
8
6
4
-5
(b) 180
-5
Radius [cm]
(b) 225
(b) 270
(b) 315
16
14
14
14
14
12
12
12
12
10
10
10
10
Height [cm]
16
Height [cm]
16
8
6
8
6
4
-5
Radius [cm]
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
-5
Radius [cm]
16
-5
Radius [cm]
Radius [cm]
Height [cm]
Height [cm]
(a) 135
16
(b) 0
Height [cm]
16
Radius [cm]
Height [cm]
(a) 90
16
12
Height [cm]
16
-5
Radius [cm]

83
4.5.2 Axial Flame Structure

In order to emphasize the periodic motion contained in the flames, axial plots showing the
mean intensity as a function of height and phase were constructed and plotted on a 2D
contour plot.
Mean intensities were calculated using a threshold intensity of 15,
corresponding to approximately 5% of the maximum intensity level in the flame. Values

below the threshold were considered to be outside the flame zone and not incorporated into
the mean. In addition, the axial contours were averaged over a period, and the averages used
to normalize the plots, which further enhances the periodic motion of the flame (Figure 4-14
through Figure 4-23).
The mean plots of flame intensity are absolute they are not
normalized in any way to enable comparison between different forcing conditions. The
flame base is defined at an intensity level of 15, and can be easily seen on the mean axial
intensity plots. Data at conditions below approximately 2 cm (below the flame base) on the
normalized intensity plots should be disregarded, since they are outside the flame zone
(denoted by dark blue structures). Both sets of plots are repeated over an additional period
for illustrative purposes.
Immediately observed is the fluctuation of the flame base (note: the lowest intensity plotted
is at 15, i.e., the flame base). In each case, the flame base oscillates in a sinusoidal manner,
corresponding to the driving frequency imposed by the acoustic drivers. Table 4-1 compiles
the mean flame height, the amplitude of the oscillation, and the percent changes of these
parameters when transitioning from the aerodynamically stabilized burner to the bluff-body
burner.
84
Mean Height (cm)
Amplitude
of
Oscillation
(peak-to-peak) (cm)
Frequency (Hz)
Aero
BB
Relative
Aero
BB
Change
Relative
Change
22
1.65
1.17
-29%
0.89
1.02
+15%
27
1.88
1.25
-33%
0.56
0.40
-28%
32
2.07
1.90
-8%
0.55
0.24
-56%
37
1.53
2.14
+40%
0.50
0.28
-44%
55
0.37
1.06
+186%
0.14
0.16
+14%
Table 4-1: Flame base position and oscillation.
At low frequencies, the bluff-body stabilizer has the effect of lowering the mean flame base
position. At approximately 32 Hz, a change in the characteristics of the bluff-body flame
seems to occur. At frequencies greater than 32 Hz, the bluff-body case shows increased
flame base positions relative to the aerodynamic case. For frequencies between 22 and 37
Hz, the flame base increases continuously, except for the aerodynamically stabilized burner,
which shows a sharp lowering of the mean flame base at 37 Hz. At 55 Hz, both burners
appear to enter into a different regime from the lower frequencies. In addition, the bluffbody burner generally has a lower amplitude of oscillation than the aerodynamically
stabilized burner.
Figure 4-14 and Figure 4-15 show the mean and normalized axial intensities at 22 Hz. It is
evident from the flame structure that both burners are responding strongly to the acoustic
forcing. The ranges of motion taken from the normalized axial intensity are comparable in
85
size, with the aerodynamic burner tending to oscillate at a higher overall position. Note the
higher angle of oscillation for the aerodynamic case in Figure 4-15. This implied a larger
velocity of the flame, possibly due to flow retardation caused by the enhanced recirculation
off the bluff-body burner.
At 27 Hz, the mean axial intensities (Figure 4-16) show a stronger coupling in the
aerodynamic case. The normalized intensities (Figure 4-17) show comparable angles (i.e.,
velocities) and a similar range of motion. As the driving frequency is increased to 32 Hz,
there is a much stronger motion observed in the aerodynamic case (Figure 4-19). Again, the
velocities of the motion are comparable between the two cases. At 37 Hz, there is much
stronger anchoring of the flame in the aerodynamic case at the flame base, as shown in
Figure 4-20. Figure 4-21 shows significantly more motion in the bluff-body case, which is
observed for the first time. At the highest frequency tested of 55 Hz, Figure 4-22 and Figure
4-23 show very little motion in both cases.
86
(a)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
Figure 4-14: Mean axial intensities at 22 Hz (a) aerodynamic (b) bluff-body.

(a)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
Figure 4-15: Normalized axial intensities at 22 Hz (a) aerodynamic (b) bluff-body.
87
(a)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
16
140
14
120
Height [cm]
12
100
10
8
80
60
4
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]

(a)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
88
(a)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
16
140
14
120
Height [cm]
12
100
10
8
80
60
4
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]

(a)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
89
(a)
16
140
14
120
Height [cm]
12
100
10
8
80
60
4
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
16
140
14
120
Height [cm]
12
100
10
8
80
60
4
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]

(a)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
90
(a)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
16
140
14
120
Height [cm]
12
100
10
8
80
60
40
2
20
0
90
180
270
360
450
540
630
720
Phase [deg]

(a)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
(b)
2
16
14
1.5
Height [cm]
12
10
8
6
0.5
4
2
0
90
180
270
360
450
540
630
720
Phase [deg]
91
4.5.3 Modified Rayleigh Index

Since the chemiluminescence measurements can be used as a measure of the heat release, it
is possible to use the measurements in order to calculate Rayleigh indices. As mentioned
previously, following the development by Culick (1987), Rayleighs criterion can be stated
mathematically as
(4-1)
E =
t +
1
dV pqdt ,
p
t
where E is the incremental energy added to the acoustic field over a period due to the
coupling between the fluctuating pressure, p, and the fluctuating heat release, q. For the
purposes of this work, equation (4-1) can be modified to yield a frequency-driven or forced
Rayleigh index that has been nondimensionalized and normalized to account for the driving
pressure amplitude and period. The dependence on gas composition is also removed to give
1
(4-2)
Rf =
0
p q
d ,
prms q
where prms is the root-mean-square of the amplitude of the driving pressure wave, and q is
the mean intensity of the heat release. p is redefined as the driving pressure amplitude, and
q becomes the fluctuation in heat release. The time dependence has been normalized by the
period of the driving acoustic wave, T, to give a nondimensional time . Rf can be applied
globally to a system to yield a global frequency Rayleigh index, or over a series of small
control volumes to produce a 2D map of the frequency Rayleigh index. This will affect the
definition of q in equation (4-2), but will be valid provided it is defined in a consistent
manner.
92
4.5.4 Global Forced Rayleigh Results

The phase information brings with it the ability to calculate Rf directly for the system. Since
the heat release varies both spatially and temporally, it is important to define how the heat
release is calculated in the modified Rayleigh index, equation (4-2). For the global results,
the heat release is first calculated according to
Lx Ly
q( ) = q2 D ( , x, y ) dy dx ,
(4-3)
0 0
where q is the spatially integrated heat release of the spatially resolved quantity, q2 D . In
order to evaluate contributions only from the driving frequency, Rf is calculated for a
pressure signal that has been bandpass filtered about the fundamental driving frequency.
1
B luff-body
A erodyn
0.5
Rf
-0.5
-1
-1.5
-2
-2.5
-3
20
25
30
35
40
45
50
55
Frequency [Hz]
Figure 4-24: Chemiluminescence global forced Rayleigh indices.
93
The global forced Rayleigh indices are shown in Figure 4-24.
They indicate that the
aerodynamically stabilized burner in general is more damped than the bluff-body case,
except at 37 Hz.
4.5.5 Spatially Resolved Forced Rayleigh Results

A complete set of spatially resolved Rayleigh results can be found in Appendix C. Spatially
resolved 2D contour plots and axial forced Rayleigh indices at 32 Hz are shown in Figure
4-25 and Figure 4-26 respectively. The 2D contour plots are shown with contour lines drawn
at levels of 20 up to +20, in increments of 2. The positive contours as solid lines, and the
negative contours are shown as dashed lines. The 2D indices show a strongly damped
anchoring zone at the base of the flame, followed by a driving region in the upper portion of
the burner tubes. As the flame exits the burner tube, it is again damped. Some pockets of
driving zones also appear, with larger zones appearing in the bluff-body stabilized case.
The axial Rayleigh plots yield similar information, but enable better comparison of the two
burner types. The bluff-body axial Rayleigh index peaks approximately 1 cm earlier than the
aerodynamic burner, and again shows a stronger driving region at the top end of the flame.
94
(a)
(b)
16
16
-6
-2
14
-2
14
-4
-4
12
-4
-1
10
Height [cm]
-2
-2
Height [cm]
-8
8
-8
-8
-2
-1 0
10
-1 2
-4
12
4
6
4
4
-6
-4
-2
-6
-6
-2
-4
2
0
0
-5
0
-5
Radius [cm]
Radius [cm]
Figure 4-25: Chemiluminescence 2D forced Rayleigh indices at 32 Hz for (a)

aerodynamic and (b) bluff-body stabilized burners.
4
aerody n
bluff-body
2
Axial Rf
-2
-4
-6
-8
10
12
14
16
18
Height [cm]
Figure 4-26: Chemiluminescence axial forced Rayleigh indices at 32 Hz.
95
4.6 Summary
This chapter describes the experimental apparatus and diagnostics used to visualize the
two burner types under examination with acoustic excitation. Shadowgraph imagery and
chemiluminescence visualization techniques were employed to provide details of the flow
field and flame locations. In particular, chemiluminescence provided measures of the
flame base location at different forcing frequencies, and the response between the
aerodynamically stabilized and bluff-body stabilized burners. In general, the bluff-body
burner lowered the position of the flame base, except at 37 Hz, where the opposite effect
was observed. Results at 55 Hz showed very little response from the flame to the
acoustic excitation. It appears that the burners enter a different regime when they are
excited at 55 Hz.
The chemiluminescence measurements also provide relative heat release measurements,

which can be used to calculate Rayleigh indices. Global forced Rayleigh indices indicate
that the aerodynamically stabilized burner provide more damping at all frequencies,
except 37 Hz. Spatially resolved Rayleigh indices were also computed, with examples of
2D contours and axially integrated plots presented.
Chemiluminescence measurements raise questions regarding its validity, particularly in

flames that are not two-dimensional in nature, due to the line-of-sight integration that
occurs.
A possible solution in axisymmetric systems is use of an Abel inversion
technique (Smith et al. 2001), which is able to extract planar information from the
chemiluminescent signal. However, in highly turbulent environments such as the flames
96
under study in this work, such a technique is not possible.
Adding to this is the
complexity that the bluff-body stabilized burner is not axisymmetric. These concerns
give rise to more accurate spatially resolved measurements, which are explored in detail
in the next chapter.
97
Chapter 5
OH PLIF Measurements
This chapter describes the experimental results and measurements of combustion
dynamics of a flame under a forced oscillatory pressure field. The test apparatus and the
majority of the diagnostics are the same as in the previous chapter. The focus of this
chapter is on the enhanced spatial and temporal resolution brought about with the
introduction of laser diagnostics and OH PLIF measurements, a more advanced technique
than chemiluminescence.
5.1 Planar Laser-Induced Fluorescence of OH

5.1.1 PLIF Theory
Laser diagnostics have proven to be ideally suited to acquiring chemical species data. In
particular, laser-induced fluorescence (LIF) is capable of resolving minor species
concentrations at low parts-per-million (ppm) levels.
According to Crosley (1993),
98
detection of the hydroxyl radical (OH) in an atmospheric flame is possible at sub partsper-billion (ppb) concentration levels, with a spatial resolution of 1 mm3 and a temporal
resolution of 10 ns, with a signal level on the order of 100 photoelectrons. Other laser
techniques such as Raman scattering (Masri et al. 1996) are sensitive to concentrations on
the order of 1000 ppm, which are adequate for major species detection, but several orders
of magnitude too high for minor species. Raman scattering is also limited by only
providing point-wise measurements. Perhaps the major advantage of laser techniques is
the ability to provide nonintrusive, in-situ measurements. The introduction of a physical
probe will inevitably disturb the flowfield, distorting the physics of the experiment.
Another issue to note is the difficulty for a physical probe to survive in a high
temperature, high pressure combustion environment. The flexibility of laser diagnostics
includes the capability of spreading the laser beam into a sheet. Planar laser-induced
fluorescence (PLIF) can then be performed, yielding species information as a 2-D planar
image, as opposed to LIF that resolves only a single point with each pulse.
PLIF and LIF operate on essentially the same principles, with the primary difference
being the way in which data is collected. In an LIF system, a photomultiplier can be used
as a detector, whereas PLIF requires an intensified CCD camera or some other detector
that provides spatial resolution. Another obvious consequence is the need for additional
optics to produce the laser sheet.
Laser-induced fluorescence involves three essential features. First, the species of interest
must be brought to an excited electronic state, usually via a tunable-dye laser, pumped by
99
an Nd:YAG or Excimer laser. The excited molecules then fluoresce, by emitting photons
and decaying to lower energy states. The emitted photon can be at the same wavelength
as the excitation source, though this is not necessarily the case. In fact, it is more
convenient if it is not, since detection can take place without interference from the laser
source. The last step involves detection of the fluorescence signal. Figure 5-1 outlines
schematically this process.
Excited state
2
Laser source
Absorption
Ground state
1
Q21
Detector
Spontaneous
emission
Figure 5-1: Simplified energy level transfer diagram for LIF.

The quenching (or collisional quenching) rate, Q, is of primary importance in
determining the accuracy of LIF measurements. Quenching represents energy loss of the
molecule by some pathway other than fluorescence. Possibilities include collision with
other molecules, dissociation, ionization, chemical reaction, or even transitions to
unmonitored molecular energy states.
A technique to avoid errors introduced by unknown quenching rates is to perform

saturated LIF. This involves using a high intensity laser such that the quenching rate is
small compared to absorption and stimulated emission rates. This has the additional
advantage of maximizing the strength of the fluorescence signal that is detected. A
problem of saturated LIF is the need for a high-powered, tunable laser with good beam
quality. Further difficulties include the finite time required to achieve saturation and
100
subsequent relaxation of the probed species, especially when temporal accuracy is

necessary (Eckbreth 1988).
Depletion of the laser-pumped level also leads to
fluorescence being a nonlinear function of population fraction (Seitzman and Hanson

1993).
An alternative to saturated LIF is to operate in the linear fluorescence regime. This

allows for the use of comparatively low powered lasers, but does not eliminate the
quenching dependence. Quenching rates can generally be modeled, but the modeling
requires knowledge of the precise state of the system, such as temperature and
concentrations of all other species. For single point measurements, Raman scattering can
be used to determine major species concentrations, however this is not practical in a twodimensional flowfield.
Temperatures in the flowfield can be determined through a
variety of techniques. Seeding the flow with a temperature sensitive tracer molecule such
as NO, and performing PLIF on the tracer gas can yield temperature fields (Cadou 1996).
Another popular method is use of a two-line technique, which measures the rotational
temperature by ratio of the two fluorescence signals (Cattolica 1981; Lucht et al. 1982;
Palmer and Hanson 1996).
Rayleigh scattering can also yield a two-dimensional
temperature field by measuring the density and inferring temperature.
Combining
Rayleigh scattering with a point two-line LIF measurement has been used to improve
precision (Heberle et al. 2000).
A model for the collisional quenching of NO (Paul et al. 1994) and OH (Paul 1994) has
been developed for flame environments. Comparisons with experimental results show
101
relatively good agreement with the empirical correlations proposed (Tamura et al. 1998).
Making use of Pauls models, a successful technique for LIF of NO in high pressure (up
to 10 atm) environments was developed by Battles and Hanson (1995).
According to Allen et al. (1995a), the fluorescence signal can be modeled by a two-level
steady state model, and is given by
(5-1)
Aeff

S f= V c I
j P t i f
4
A + Q( P , P , T )
B, i
(T) Bi g i ( , P, T ) ,
where:
= quantum efficiency of ICCD photocathode
= collection optics solid angle
Vc = collection volume of one detector pixel
I = laser spectral fluence
Aeff = effective Einstein coefficient for spontaneous emission
A = Einstein coefficient for spontaneous emission
Q(P, P, T) = electronic quenching rate
j = mole fractions of measured species j, in measurement volume
Pt = total gas pressure
fB, i(T) = Boltzmann fraction of absorbing species in state i
Bi = Einstein coefficient for stimulated emission for transition i
gi(, P, T) = overlap integral (convolution of absorption and laser lineshape
profiles).
102
It is important to note in equation (5-1), that the fluorescence signal is directly

proportional to the species mole fraction within the probed volume, and thus can be
related to species concentration. Quenching can be accounted for through the use of the
models of Paul previously mentioned. The quenching rate is defined as
(5-2)
Q = <vj> <<>> P/kBT,
where <vj> = (8kBT/mj)1/2 and <<(P, T)>> is the total electronic quenching cross
section, given by
(5-3)
1/2
( i , T) = i (1 + m j / mi ) i (T) .
i
These modeling efforts however still require detailed information regarding the species
concentrations and temperature within the probed volume.
An alternative to modeling of the quenching rate and arguably superior is direct

measurement of the fluorescence decay time as performed by Kollner and Monkhouse
(1995) using point measurements and a picosecond laser.
A similar approach is
discussed by Cadou (1996) and also involves measurement of the fluorescence signal as
it decays, since it provides a direct measurement of the quenching rate. This requires an
extremely fast collection system, since the fluorescence signal decays on the order of a
few nanoseconds. While this is possible for single point measurements with use of a
photomultiplier tube, current multi-point detectors, such as intensified CCD cameras, do
not possess the speed required to perform this measurement.
103
The primary purpose of applying a quenching correction is to provide quantitative species

concentration information.
Due to limitations in modeling and current equipment
technologies, a quenching correction is a nontrivial task for this work. Previous work
with methane diffusion flames has shown uncorrected LIF measurements of OH
concentration to be within 10% of the actual value (Barlow and Collignon 1991).
However, due to differences in the constituents and geometry of the reacting flow of this
work, direct comparison is not possible. Since the primary purpose of this work is to
provide relative measurements, a fully quantitative measurement is not required.
Therefore, no attempt is made to correct for quenching effects in the flowfield.
5.1.2 Laser System
The PLIF system is based on an Nd:YAG laser (Continuum Powerlite 9010) operating at
10 Hz, pumping a tunable dye laser (Continuum ND6000), which in turn drives a
mixer/doubler system (U-oplaz) as in Figure 5-2. The Nd:YAG laser outputs 2000
mJ/pulse at 1064 nm (IR), and is equipped with a secondary harmonic generation system
to provide 1000 mJ/pulse at 532 nm (green). The 532 nm beam pumps the dye laser,
while excess energy at 1064 nm (energy not converted to 532 nm) is passed through a
delay line. The delay line allows the 1064 nm beam to coincide spatially and temporally
with the output of the dye laser for frequency mixing purposes. The mixer/doubler
system, shown in more detail in Figure 5-3, was custom designed in cooperation with Dr.
Sheng Wu of U-oplaz Technologies, for optimal energy conversion by special tuning of
the BBO crystals (Wu et al. 2000). Use of Rhodamine 590 as the dye laser in methanol
optimizes conversion efficiency near 564 nm (> 200 mJ/pulse), which is then doubled to
approximately 282 nm to excite the (1,0) band of OH (Dieke and Crosswhite 1962).
104
Energy in excess of 60 mJ/pulse is easily provided by this system, but this experiment did
not require operation at full power, providing approximately 30 mJ/pulse in the
measurement volume. This maintains fluorescence in the linear regime, and represents
an ideal compromise between systemic error (~15%) and SNR (~ 3.4%) over other laser
pumping options (Seitzman and Hanson 1993).
5.1.3 Optics
In order to take calibration shots of the laser beam profile, a method is employed which
allows the beam energy to be turned down without changing the actual beam energy
output (and also the beam characteristics) of the laser. This involves passing the beam
through a zero-order half-waveplate (U-oplaz, coated for 285 nm) mounted on a rotatable
stand (about the beam axis), which polarizes the beam to a particular orientation. The
beam then passes through a thin film plate polarizer (CVI, TFP-280-PW-2025-UV). This
allows the energy transmitted to vary from full power when the polarization is in line
with the waveplate, to minimum power (approximately 3-4% of full power) when the
polarizer is not aligned with the waveplate. Figure 5-4 gives more details of the optical
arrangement. After the polarizer, a portion of the beam (approximately 2%) is split using
a beamsplitter. Shot-to-shot laser energy is measured for each pulse with an energy
meter (Molectron J9LP). The beam is then narrowed using a plano-concave cylindrical
lens (radius of curvature = 100 mm), and spread into a sheet in the plane at 90 to the
converging plane by a plano-convex cylindrical lens (radius of curvature = 25.43 mm). It
should be noted that all beam steering is done using total-internal-reflection prisms, and
all optics are UV grade fused silica, coated with an antireflective coating which
minimizes reflections to less than 1% from 225 nm up to over 400 nm.
105
Bandpass
filters
ICCD camera
Quartz test
section
Burner
Sheet generating
cylindrical optics
PC based data
acquisition
Frequency
mixer/doubler
Tunable dye laser

Nd:YAG laser
Figure 5-2: PLIF system.
106
564nm
564nm /2
Wave Plate
HR 287nm
Dump
Pellin-Broca
Separation Prism
SHG
Output
Optional
1064nm
telescope
1064nm /2
Wave Plate
HR 287nm
HT 1064nm
Optional
mix
Indicates computer-controlled
precision rotation stages
Figure 5-3: Mixer/doubler system (Stages available but not used are indicated by
dashed lines).
beam
dump
rejected
polarization
282 nm UV laser
output from
mixer/doubler
spatial
filter
/2
waveplate
plano-convex
cylindrical lens
beamsplitter
thin film plate

polarizer
plano-concave
cylindrical lens
energy
meter
Figure 5-4: OH PLIF optics arrangement.
to
flame
107
5.1.4 ICCD Camera

The detector for the fluorescence signal is an intensified CCD camera (Roper
Scientific/Princeton Instruments ICCD-MAX), using a 512 x 512 Thomson CCD array,
operated with a gate width of 200 ns.
The photocathode used in the camera is a
handpicked DEP super-blue model, for maximum quantum efficiency in the UV. Due
to the requirement of high QE as well as fast gating, the microchannel plate (MCP) of the
camera is gated, since the thin UV sensitive coating on the photocathode does not allow
the intensifier to be gated quickly. Attached to the camera is a catadioptric (similar to
Cassegrain telescope designs) all-reflective F/1.2 UV lens with a focal length of 105 mm.
The lens provides exceptionally fast light throughout, as well as minimizing spherical and
chromatic aberrations. This results in a spatial resolution of 215 m x 215 m per pixel
at the focal plane with an image size of 11 cm2. A 2 mm thick UG5 Schott glass filter to
block light generated by the laser, and a 2 mm thick WG305 Schott glass filter to remove
light generated by flame luminosity and ambient sources filter the fluorescence signal. A
digital delay/pulse generator (Stanford Research Systems DG-535) controls camera
timing, which is synchronized to the laser pulse.
Particular benefits of this PLIF system include flexibility, exceptionally high energy
output, conversion efficiencies, and collection efficiencies. Other molecules of interest to
combustion can be readily probed using this system such as CH and NO, with much
higher energy levels than previous researchers (Hanson et al. 1990; Allen et al. (1995a)).
108
5.2 Experimental Procedure

The following steps outline the general procedures followed in conducting the
combustion dynamics experiments.
1. Laser setup.
Verify the calibration of the dye laser, either with a wavemeter or an optogalvanic (OG) cell.
Optimize the mixer/doubler tuning crystal angles for maximum energy

conversion.
2. Setup necessary gas flows.
Air jacket flow over acoustic drivers.
Nitrogen purge through ICCD camera.
3. Optics alignment.
Focus the ICCD camera on a card in the test section. Ensure the laser sheet is
passing through the probed volume cleanly. This is done with the aid of the
ICCD camera taking focusing images so scattered light from the laser sheet
can be minimized.
4. Beam profile calibration.
Minimize laser energy throughput using the waveplate. Set camera gain to 1,
and gate width to 10 ms.
Allow laser sheet to impinge on fluorescent card and acquire the beam profile
with the ICCD camera, placing card in three different positions (left, center,
right).
109
Return camera gain to 200, and gate width to 200 ns and waveplate to allow
maximum laser energy throughput.
5. Take background images.
Typically 200 images with no flame, but with the laser sheet passing through
the test section.
6. Set experimental conditions.
Light burner and set methane and carbon dioxide flow rates (see Appendix B
for more details).
Set acoustic driver power on controller, and activate drivers.
Allow laser sheet to pass through test section.
7. Perform experimental run (duration approximately three minutes).
Start LabView data acquisition program.
Start WinView camera imaging software, typically taking 300 images (limited
by system RAM).
8. End experimental run.
LabView and WinView routines end automatically. Save camera images to

hard drive (takes several minutes due to file size).
Turn off fuel and CO2 flows, extinguishing the flame.
Turn off acoustic drivers.
9. Repeat experiment.
Repeat steps 6-8 until more than 5000 images have been acquired at a
particular test condition.
10. Take post-run background images.
110
11. Take post-run beam profile calibration images.

12. Change experimental conditions.
Changes may include drive frequency, burner height, or burner configuration.
Repeat steps 4-11 until experimental session is complete.
13. Shutdown systems.
After the experiments have been performed, the raw data files are zipped and burned to
CD-ROM for archival purposes. The data is then ready for post-processing, described in
the next section.
5.3 Data Reduction

5.3.1 Phase Characterization
By taking advantage of the periodic forcing of the chamber, and assuming that the flame
responds accordingly in a periodic fashion, the PLIF images can be phase-binned and
averaged together, to generate the periodic response of the OH fluorescence in the flame.
As previously mentioned, the oscillating pressure used to phase-resolve the images is
acquired by a pressure transducer located 8 cm above the fuel spud, in the zone where the
flame is stabilized. The transducer signal is filtered about the fundamental driving
frequency using a phase preserving 4th order butterworth bandpass filter in Matlab, to
produce a clean signal with which to phase-bin. Each image is placed in an appropriate
bin, based on the position of the incident laser shot (and subsequent camera trigger)
relative to the rising edge zero crossing of the pressure signal. This process is illustrated
schematically in Figure 5-5 using 8 bins, while in the actual data analysis, 36 bins are
used. Since the hydroxyl radical is an intermediary of combustion and thus an indicator
111
for the reaction zone in the flame, this procedure yields a proportional measurement of
the heat release over a period of the acoustic driving cycle.
Incident Laser Shot
Phase reference
taken at rising
edge zero crossing.
Acoustic Chamber
Pressure
Record PLIF image from
incident laser shot
=0
Store image in
associated phase bin Image at Phase
Figure 5-5: Phase-binning procedure for OH PLIF images.
5.3.2 Image Processing

Due to the distributed nature of the flame under study and limitations on the ICCD
cameras field of view, multiple sets of images were taken at each test condition at
different heights. Each case contains a total of over 5000 images, phase-averaged into 36
equally spaced bins. Statistics indicate an even distribution among the bins, with well
112
over 100 images per bin. The averaged background is subtracted in each bin to eliminate
scattering effects from the laser; and corrections are made for variations in spatial and
shot-to-shot beam intensity. Images at the same phase but different heights are then
matched geometrically, and their intensities adjusted to match in the overlap region using
a least-squares minimization routine. The composite images are then smoothed using a
filter, using the weighting matrix given in Table 5-1. The weight is determined by the
inverse of the distance to the center pixel. Stronger smoothing is done in the y-direction,
the direction of the flow. Further details regarding the software written to perform these
processes can be found in Palm (in progress).
1/ 5
1/2
1/ 5
1/ 2
1
1/ 2
1/ 5
1
1
1
1/2
1/ 2
1
1/ 2
1/ 5
Table 5-1: Smoothing filter weighting matrix.
5.4 OH PLIF Results

5.4.1 Pressure and Heat Release Measurements
Phase-averaged images for 12 of the 36 bins are displayed in Figure 5-6 for a
representative case of the aerodynamically stabilized burner, driven at 32 Hz. The
relative change between images is more easily observed by noting the variation in
intensity over the lower right quadrant of each image. Spatial integration of the phase-
113
averaged OH PLIF images gives a global heat release at each phase angle. Plots of
chamber pressure and the computed global heat release, and their corresponding FFTs are
shown in Figure 5-7 through Figure 5-16.
It is evident from plots of the lowest frequency of 22 Hz (Figure 5-7 and Figure 5-8) that
the pressure signals contain much more harmonic content than the fundamental
frequency. Limitations of the response of the acoustic drivers at low frequencies account
for the excitation of higher harmonics. Data at the 27 Hz condition show a similar,
although largely attenuated effect. Once frequencies reach 32 Hz, the pressure traces are
relatively clean, and show almost no harmonics. These effects are common for both the
aerodynamically and bluff-body stabilized configurations.
In general, the FFTs of heat release show a response at the same driving frequency as the
excited acoustic mode. Figure 5-7 through Figure 5-10 (22 Hz & 27 Hz cases) show
higher harmonic content virtually identical in both the pressure and heat release. At
driving frequencies greater than 27 Hz, the heat release contains elevated levels of higher
harmonic content, which does not appear in the pressure traces. In both 32 Hz cases, the
additional frequency content other than the fundamental in the heat release traces is
minimal. However, the 37 Hz cases contain significant amounts of higher frequency heat
release content, particularly at the 2nd mode of the system at 74 Hz. This result is most
clearly evident at 55 Hz in Figure 5-15 and Figure 5-16, which show the ringing of higher
frequency modes over the fundamental mode of heat release at this driving frequency. In
114
30
60
90
120
150
180
210
240
270
300
330
Figure 5-6: OH PLIF images over a period of a sinusoidal pressure oscillation for
the aerodynamically stabilized burner at 32 Hz. The intensity scale is in number of
counts, and the x and y coordinates are in pixels.
115
Pressure
Pressure
0.015
1.4
Unfiltered
Filtered
1.2
0.01
Power Spectrum
psi
0.005
0.8
0.6
0.4
-0.005
0.2
-0.01
0.01
0.02
0.03
0.04
0.05 0.06
time (s)
0.07
0.08
0.09
0.1
20
40
60
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
Heat Release
0.07
1.1
Unfiltered
1st Mode
0.06
1.05
Power Spectrum
0.05
q`/qmean
0.95
0.04
0.03
0.02
0.9
0.01
0.85
0.01
0.02
0.03
0.04
0.05 0.06
time (s)
0.07
0.08
0.09
0.1
20
40
60
80
100
120
Frequency (Hz)
aerodynamically stabilized burner driven at 22 Hz.
116
-3
10
Pressure
x 10
Pressure
0.7
Unfiltered
Filtered
0.6
6
0.5
Power Spectrum
psi
2
0
-2
0.4
0.3
0.2
-4
0.1
-6
-8
0.01
0.02
0.03
0.04
0.05 0.06
time (s)
0.07
0.08
0.09
0.1
20
40
60
Heat Release
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
1.2
0.07
Unfiltered
1st Mode
0.06
1.1
0.05
Power Spectrum
1.15
q`/qmean
1.05
0.04
0.03
0.95
0.02
0.9
0.01
0.85
0.01
0.02
0.03
0.04
0.05 0.06
time (s)
0.07
0.08
0.09
0.1
20
40
60
80
100
120
Frequency (Hz)
stabilized burner driven at 22 Hz.
117
Pressure
Pressure
0.015
0.7
Unfiltered
Filtered
0.6
0.01
Power Spectrum
0.5
psi
0.005
0.4
0.3
0.2
-0.005
0.1
-0.01
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
0.08
20
40
60
Heat Release
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
1.15
0.1
Unfiltered
1st Mode
0.09
1.1
0.08
Power Spectrum
0.07
q`/qmean
1.05
0.06
0.05
0.04
0.03
0.95
0.02
0.01
0.9
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
0.08
20
40
60
80
100
120
Frequency (Hz)
118
-3
Pressure
x 10
Pressure
0.7
Unfiltered
Filtered
0.6
0.5
Power Spectrum
psi
2
0
-2
0.3
0.2
-4
0.1
-6
-8
0.4
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
0.08
20
40
60
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
Heat Release
0.08
1.15
Unfiltered
1st Mode
0.07
1.1
Power Spectrum
0.06
q`/qmean
1.05
0.05
0.04
0.03
0.02
0.95
0.01
0.9
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
0.08
20
40
60
80
100
120
Frequency (Hz)
Figure 5-10: Pressure and heat release traces and power spectrums for the bluffbody stabilized burner driven at 27 Hz.
119
-3
10
Pressure
x 10
Pressure
1.5
Unfiltered
Filtered
8
6
4
Power Spectrum
psi
2
0
-2
0.5
-4
-6
-8
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
20
40
60
Heat Release
140
160
180
200
140
160
180
200
Heat Release
1.25
0.5
Unfiltered
1st Mode
1.2
0.45
0.4
1.15
0.35
Power Spectrum
1.1
1.05
q`/qmean
80
100
120
Frequency (Hz)
0.95
0.3
0.25
0.2
0.15
0.9
0.1
0.85
0.8
0.05
0
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
20
40
60
80
100
120
Frequency (Hz)
120
Pressure
Pressure
0.01
1.5
Unfiltered
Filtered
0.005
Power Spectrum
psi
-0.005
0.5
-0.01
-0.015
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
20
40
60
Heat Release
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
1.25
0.35
Unfiltered
1st Mode
1.2
0.3
1.15
0.25
Power Spectrum
1.1
q`/qmean
1.05
1
0.95
0.2
0.15
0.1
0.9
0.05
0.85
0.8
0.01
0.02
0.03
0.04
time (s)
0.05
0.06
0.07
20
40
60
80
100
120
Frequency (Hz)
121
Pressure
Pressure
0.01
1.5
Unfiltered
Filtered
0.005
Power Spectrum
psi
-0.005
0.5
-0.01
-0.015
0.01
0.02
0.03
time (s)
0.04
0.05
0.06
20
40
60
Heat Release
140
160
180
200
140
160
180
200
Heat Release
1.25
0.35
Unfiltered
1st Mode
1.2
0.3
1.15
0.25
q`/ qmean
Power Spectrum
1.1
1.05
1
0.2
0.15
0.1
0.95
0.05
0.9
0.85
80
100
120
Frequency (Hz)
0.01
0.02
0.03
time (s)
0.04
0.05
0.06
20
40
60
80
100
120
Frequency (Hz)
122
-3
Pressure
x 10
Pressure
1.5
Unfiltered
Filtered
6
4
2
Power Spectrum
psi
0
-2
-4
0.5
-6
-8
-10
0.01
0.02
0.03
time (s)
0.04
0.05
0.06
20
40
60
Heat Release
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
1.25
0.35
Unfiltered
1st Mode
1.2
0.3
1.15
0.25
Power Spectrum
1.1
q`/qmean
1.05
1
0.95
0.2
0.15
0.1
0.9
0.05
0.85
0.8
0.01
0.02
0.03
time (s)
0.04
0.05
0.06
20
40
60
80
100
120
Frequency (Hz)
123
-3
Pressure
x 10
Pressure
1.5
Unfiltered
Filtered
6
4
2
Power Spectrum
psi
0
-2
-4
0.5
-6
-8
-10
0.005
0.01
0.015
0.02
time (s)
0.025
0.03
0.035
0.04
Heat Release
20
40
60
-3
1.05
Unfiltered
1st Mode
1.04
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
x 10
1.03
6
Power Spectrum
1.02
q`/qmean
1.01
1
0.99
5
4
3
0.98
2
0.97
0.96
0.95
0.005
0.01
0.015
0.02
time (s)
0.025
0.03
0.035
0.04
20
40
60
80
100
120
Frequency (Hz)
124
-3
Pressure
x 10
Pressure
1.4
Unfiltered
Filtered
1.2
4
1
Power Spectrum
psi
0
-2
-4
0.8
0.6
0.4
-6
0.2
-8
-10
0.005
0.01
0.015
0.02
time (s)
0.025
0.03
0.035
0.04
20
40
60
Heat Release
80
100
120
Frequency (Hz)
140
160
180
200
140
160
180
200
Heat Release
1.1
0.03
Unfiltered
1st Mode
1.08
0.025
1.06
1.04
Power Spectrum
0.02
q`/qmean
1.02
1
0.98
0.015
0.01
0.96
0.005
0.94
0.92
0.005
0.01
0.015
0.02
time (s)
0.025
0.03
0.035
0.04
20
40
60
80
100
120
Frequency (Hz)
125
this case, the aerodynamically stabilized case shows the 2nd and 3rd modes, while the
bluff-body case displays only significant content of the 3rd mode.
The FFTs of pressure and heat release contain amplitude and phase information of their
respective signals. This information can be extracted at a particular frequency, by inverse
FFT of the peak at the frequency under consideration. A representative plot is shown in
Figure 5-17, displaying the relationship (scaled for comparison) between the first mode
of heat release and pressure.
Figure 5-17 (a) displays the relative phases for an
aerodynamically stabilized burner subjected to acoustic forcing at 22 Hz, and Figure 5-17
(b) displays the relative phases for the bluff-body stabilized case. It can be noted that the
bluff-body case shows a larger phase difference between the pressure and heat release
than the aerodynamically stabilized case. This concept will be explored in more detail in
the next section with respect to Rayleighs criteria.
126
(a)
(b)
Figure 5-17: Phase relationship between the 1st mode of pressure and heat release
for the (a) aerodynamically stabilized and (b) bluff-body stabilized cases at 22 Hz.
Heat release traces have been scaled for ease of comparison.
127
5.4.2 Forced Rayleigh Index

As stated in section 4.5.3, the forced or frequency-driven Rayleigh is defined as
1
Rf =
(5-4)
p q
d ,
prms q
where prms is the root-mean-square of the amplitude of the driving pressure wave, and q is
the mean intensity of the heat release. p is redefined as the driving pressure amplitude,
and q becomes the fluctuation in heat release. The time dependence has been normalized
by the period of the driving acoustic wave, T, to give a nondimensional time . Again, Rf
can be applied globally to a system or locally to produce spatially resolved maps of
Rayleigh indices.
5.4.3 Global Rayleigh Results
Rf can be computed directly for the system with the phase relationship between heat
release and pressure. Since the heat release varies both spatially and temporally (while
the pressure is assumed to vary only temporally), it is important to define how the heat
release is calculated in the modified Rayleigh index, equation (5-4). For the global
results, the heat release is first calculated according to
Lx Ly
(5-5)
q( ) = q2 D ( , x, y ) dy dx ,
0 0
where q is the spatially integrated heat release of the spatially resolved quantity, q2 D . In
order to evaluate contributions from modes other than the driving frequency, Rf is
calculated in two ways: directly from the pressure and heat transfer global response; and
128
for a pressure signal that has been bandpass filtered about the fundamental driving
frequency.
The forced global Rayleigh index is plotted in Figure 5-18. In general, the bluff-body
(1st pressure mode filtered) stabilized configuration is less sensitive to changes in the
pressure field than the corresponding aerodynamically stabilized counterpart.
This
manifests itself as a frequency-driven global Rayleigh index with a lower magnitude.

The requirement of filtering the pressure about the primary excitation frequency is
justified, since the dynamic response of the flame is sought at a particular frequency,
without the additional harmonics introduced through inadequacies of the acoustic drivers.
This is especially apparent at a frequency of 22 Hz, which shows the discrepancies
between results for the aerodynamically stabilized burner when the first mode is filtered,
and when it is not.
1.5
Bluff-body
Aero
1st mode Bluff-body
1st mode Aero
Driving
0.5
0
Rf
Damping
Rf -0.5
-1
-1.5
-2
-2.5
20
25
30
35
40
Frequency [Hz]
45
50
55
Frequency [Hz]
Figure 5-18: Frequency driven global Rayleigh index.
129
5.4.4 Spatially Resolved Rayleigh Results

The pressure field generated in the chamber is a bulk mode (established in the previous
chapter), resulting in a relatively uniform pressure over the reaction zone at each phase.
This allows a 2-D map of Rayleighs criterion to be computed, using the assumption of
uniformity in pressure in the chamber at a particular instant in time. Note that the flame
may introduce its own pressure oscillations associated with the dynamical response the
flame has to the bulk pressure field. The 2-D map will give insight into which zones in a
particular configuration are more susceptible to acoustic oscillations.
Figure 5-19 through Figure 5-23 are contour plots of the 2-D forced Rayleigh index, Rf
for both the aerodynamically and bluff-body stabilized cases, at each of the five forcing
frequencies examined. Solid contours represent positive values for Rf (driving), while
negative values are indicated by the dashed contours (damping). For the first four
frequencies (22-37 Hz), the contour levels are [-20, 10, 4, 2, 2, 4, 10, 20]. Due to
significant differences in the dynamical response of the system at 55 Hz, contour levels
of [-3, 1.5, 1.5, 3] were used for clarity. In all cases, the pressure has been band pass
filtered about the fundamental of the driving frequency.
Comparison of Figure 5-19(a) and Figure 5-19(b) denotes the differences between the
aerodynamically stabilized and bluff-body stabilized cases at 22 Hz. The bluff-body tabs
appear to induce a stronger recirculation zone in the stabilization region at the base of the
flame, resulting in the flame stabilizing at a lower height and with more (negative)
intensity. This region is less susceptible to instability than the aerodynamically stabilized
130
case. In the positive region, the strong level +10 contours are smaller than in the bluffbody case, with the positive region in general being slightly smaller than the aerodynamic
situation.
Note it is possible to observe the quartz burner tube, which ends at
approximately 8 cm in height, before the flame diverges. In the region where the flame
diverges, there is a large negative Rf in both cases. These trends are also observed as the
driving frequency is increased to 27 Hz in Figure 5-20, but with the disparities in sizes of
the positive regions becoming more pronounced, and heights of the flame stabilization
zone becoming less pronounced. By Figure 5-21 at 32 Hz, the penetration of the positive
region in the central core of the flame has been greatly decreased, and differences in the
flame stabilization height have disappeared. The highest positive contours of Rf are
found at 32 Hz, but the diminished size of the positive zone, combined with the
appearance of larger negative zones do not produce as large global Rayleigh indices as
the 27 Hz case (Figure 5-18). As the driving frequency is increased to 37 Hz (Figure
5-22), the positive zone contributions to the frequency-driven Rayleigh index have
decreased further in size. Once the acoustic frequencies have reached 55 Hz in Figure
5-23, the character of the Rayleigh contour plot changes dramatically. The large positive
structures have vanished in the center of the flame and are replaced by large, relatively
low amplitude negative zones.
131
17.2 800
17.2 800
4
700
-4
-4
10
8.6 400
-2
300
10 10
4.3 200
0 0
-4
-4
100
-2
0-5 50
10
8.6 400
4.3 200
10
300
height [cm]
-4
4
-2
500
-4
-2
500
height [cm]
12.9 600
-2
12.9 600
-10
700
-2
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(a)
-10
100
0 0
-10
-4
-2
-4
-2
0-5 50
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(b)
burners at a driving frequency of 22 Hz.
132
17.2 800
17.2 800
2
700
700
2
4
10
10
8.6 400
4
4
2
-4
-4
4.3 200
-4
-2
100-2.5
150 200 250
0 300 3502.5400 450 5500
radius [cm]
(a)
-4
100
-2
0 0
0-5 50
-4-4
-2
0-5 50
10
300
-2
0 0
-4
-2
100
8.6 400
-2
4.3 200
-2
10
10
4
300
500
-4 -2
height [cm]
-10
-4
-2
500
12.9 600
-10
height [cm]
12.9 600
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(b)
133
17.2 800
17.2 800
-4
20
8.6 400
-4
-2
height [cm]
-2
300
-2
-10
10
10
-10
-4
500
-2
300
20
10
4.3 200
4.3 200
10
4
2
-4
100
-4
-2
-4
height [cm]
-2
-4
-4
8.6 400
-4
-2
12.9 600
-10
500
700
-2
12.9 600
-4
-4
-4
700
-2
0 0
0-5 50
100-2.5
150 200 250
0 300 3502.5400 450 5500
radius [cm]
(a)
100
0 0
-4
-2
0-5 50
-2
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(b)
134
17.2 800
17.2 800
-4
700
700
-4
-10
12.9 600
-4
-10
300
-4
0-5 50
100-2.5
150 200 250
0 300 3502.5400 450 5500
(a)
-2
10
300
4.3 200
10
22
2 4
-4
-4
-2 -4
radius [cm]
-10
-2
100
-2
-10
-2
100
4
2
4.3 200
-4
8.6 400
10
10
height [cm]
-4
-2
8.6 400
0 0
500
-4
-4
height [cm]
-2
-10
500
-2
-2
-4
12.9 600
-2
-2
0 0
0-5 50
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(b)
135
700
1.5
1.5
12.9 600
1.5
1.5
height [cm]
-3
-1.5
-3
300
-1.5
-3
-3
-1.5
-3
300
-3
-5
-1.5
4.3 200
8.6 400
-3
-1.5
-3
-1.5
-1.5
8.6 400
-1.5
-3
-1.5
height [cm]
500
1.5
-1.5
-1.5
500
1.5
-1.5
-1.5
-1.5
1.5
1.5
12.9 600
1.5
1.5
1.5
700
17.2 800
1.5
-5
17.2 800
4.3 200
-3
100
0 0
-1.5
-1.5
0-5 50
-1.5
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(a)
100
0 0
-1.5
-1.5
0-5 50
100-2.5
150 200 250
0 300 3502.5 400 450 5500
radius [cm]
(b)
136
The contour plots have been replotted into two sets of 2-D color plots (Figure 5-24 and
Figure 5-25) of Rayleigh indices, showing only the left side of the flames dynamical
response. This allows for ease of comparison as the driving frequency is changed. As
the forcing frequency increases from 22 Hz to 37 Hz, the size of the central hot zone
(positive local Rayleigh index) decreases and the cold zone (negative local Rayleigh
index) that appears above it travels down and increases in size. The large drop in the
global Rayleigh index (Figure 5-18) at 37 Hz corresponds to the appearance of larger
negative regions, particular in the central core of the flame. Other trends are similarly
observed between the global and 2-D Rayleigh indices. Although this does not hold at a
drive frequency of 55 Hz, comparison of the plots show that the 2-D Rayleigh indices are
of very low magnitude throughout the flame. The flame therefore seems to be relatively
insensitive to a driving frequency of 55 Hz. These trends are more clearly seen in an
axial Rayleigh plot (Figure 5-26), which is obtained by integrating the 2-D plot along the
radial direction at each height. This plot indicates that the magnitude of the Rayleigh
index is slightly lower for the bluff-body versus aerodynamically stabilized case.
The global Rayleigh index data (Figure 5-18) appears to indicate that the flame should
also be insensitive to a drive frequency of 22 Hz even more so than the 55 Hz case
(lower magnitude of Rf). However, comparison of the 22 Hz cases in Figure 5-24 and
Figure 5-25 denote that in this situation, correspondingly large negative regions balance
regions of large positive Rayleigh index. This point emphasizes the importance of
spatially resolved data of Rayleighs index. It is conceivable that a local flame region
responding in phase to a pressure fluctuation could drive an instability due to nonlinear
137
and geometric effects, if it were situated at an especially critical region (i.e., the point
in a Rijke tube), even if the global Rayleigh index indicated a stable state. Local regions
with Rf > 0 can also be identified and modified at the design stage, to improve the
stability margins of combustion systems.
138
22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
Figure 5-24: Aerodynamically stabilized 2-D Rayleigh plots.
22 Hz
27 Hz
32 Hz
37 Hz
Figure 5-25: Bluff-body stabilized 2-D Rayleigh plots.
55 Hz
139
12
Aero 22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
Bluff 22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
10
8
6
Rf
Driving
4
2
0
Frequency [Hz]
-2
Damping
-4
-6
0
10
12
14
16
18
20
Height [cm]
Figure 5-26: Axial Rayleigh index plot: Solid lines are the aerodynamically
stabilized burner, and dashed lines are the bluff-body stabilized case.
5.4.5 Combustion Response

The concept of a response function is well known in solid propellant combustion as a
modeling tool to quantify the coupling between the pressure and the burning rate. For
solid propellants, it is typically formulated as (m / m ) / ( p / p ) , where m represents the
mass flux,
( ) represents fluctuation quantities, and (~ ) denotes time-averaged quantities
(Culick, 1968). In this work, it is possible to measure a similar combustion response
140
function directly, which can be used to close the loop between combustor dynamics and
combustion dynamics (Figure 1-2). The combustion response function is defined as
(5-6)
CR =
/q)
(qrms
( prms / p )
In general, CR will be a complex quantity, since there is a phase difference between the
heat release and pressure. Again, this can be evaluated globally or for spatially resolved
regions, similar to the Rayleigh index, through judicious use of normalization values.
The global combustion response for both sets of burners is plotted in Figure 5-27. The
phase of the heat release has been defined such that it lags the pressure wave. The form
Figure 5-27: Global combustion response function.
141
of the global combustion response function is similar to the classic quasi-steady response
function used in solid propellants (Isella, 2001), containing a single resonant peak. A
simple scaling analysis of the magnitude of the fluctuations shows that for pressure
amplitudes on the order of p ~ 0.005 psi (Figure 5-7 through Figure 5-16) with a
combustion response magnitude of 200, the heat release fluctuation is approximately 7%
of the mean heat release rate.
The local combustion response is plotted in Figure 5-28 and Figure 5-29, displaying the
magnitude and phase respectively.
The magnitude plot (Figure 5-28) has been
normalized using the spatial mean of the heat release rate, as opposed to a temporal
mean. This is discussed further in Appendix D, where the alternate method is also
displayed. The plots are generated by performing an FFT in time for each spatial
location, extracting the fluctuating heat release and phase, and constructing the response
function.
The spatially resolved plots of magnitude show, in general, that the bluff-body stabilized
flame has a weaker response than the aerodynamically stabilized flame, which
corresponds to the global result (Figure 5-27). The phase plots (Figure 5-29) show
regions where the heat release is in phase (0 to 90 dark red, -270 to 360dark blue)
and out of phase (-90 to -270 orange to light blue).
142
22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
(a)
(b)
Figure 5-28: Combustion response magnitude (a) aerodynamically stabilized (b)
bluff-body stabilized.
143
22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
(a)
(b)
Figure 5-29: Combustion response phase (a) aerodynamically stabilized (b) bluffbody stabilized.
144
The additional information provided by the combustion response function can be used in
conjunction with the Rayleigh index to understand what is occurring in the flame during
driving or damping. Examining first the global plots, at 22 Hz the global Rayleigh index
(Figure 5-18) shows the bluff-body (first mode)3 to be considerably less driven than the
aerodynamic counterpart. The global combustion response at 22 Hz (Figure 5-27) shows
that while the magnitude of the response is similar, the heat release of the
aerodynamically stabilized burner is more in phase with the pressure oscillation than the
bluff-body burner. At 27 and 32 Hz, the phase relationship between the burners is nearly
identical, while the bluff-body has a weaker response to acoustic driving, and thus lower
Rayleigh indices. At 37 Hz, the aerodynamic burner has a lower Rayleigh index than the
bluff-body burner. While the magnitude of the response is lower for the bluff-body, it is
also slightly more in phase with the pressure, thus a higher Rayleigh index.
This
demonstrates the high sensitivity of the system to slight variations in phase. Finally at 55
Hz, the magnitude of the combustion response has dropped very low, such that there is
very little driving or damping, and the Rayleigh index drops close to zero.
A similar discussion will apply to the spatially resolved plots of Rayleigh index (Figure
5-24 and Figure 5-25) and combustion response (Figure 5-28 and Figure 5-29).
Considering primarily the phase plots, they generally display an anchoring region
centered at a height of 2 cm, followed by a strong driving section at the top of the burner
tube, and another damped region at the exit of the tube. The 55 Hz case shows extremely
For comparative purposes, the filtered first mode Rayleigh index is always used.
145
weak driving in the combustion response magnitude, as well as very few coherent
structures in phase.
5.5 Summary
A novel system is presented which is capable of measuring the combustion dynamics of a
flame under forced oscillatory conditions. The diagnostic used in this chapter is OH
PLIF, which provides temporally and spatially resolved measurements, and makes use of
the periodic forcing of the flame. The technique presented in this work can potentially be
used to directly measure the response of any optically accessible combustion system to an
acoustic field. It has been applied to a jet-mixed burner in two configurations: an
aerodynamically stabilized and a bluff-body stabilized design. Results are presented in
the form of spatially resolved and global Rayleigh indices, as well as global and spatially
resolved combustion response functions of the burner.
The importance of spatially
resolved data manifests itself in the development of more stable designs, as well as
improving predictive modeling capabilities.
A database of spatially and temporally
resolved data on instabilities is important to verify work done in numerical simulations
The 2-D contour and axially integrated plots of Rayleighs index indicate that the
dependence on frequency has a stronger impact on the dynamic response of the flame
versus burner configurations tested. Though geometric differences between the burners
are slight, the bluff-body design appears to be superior in terms of insensitivity to an
imposed acoustic field. The most dramatic difference between the two burner designs
occurs at the lowest frequency, 22 Hz, as illustrated in the global change in Rayleigh
146
index (Figure 5-18). The experimentally derived combustion response function indicates
that in the 22 Hz case, the change in Rayleigh index is due to primarily to a shift in phase
characteristics between the two burners. The Rayleigh index and combustion response
can be used in conjunction, to better understand the dynamics of the flame and acoustic
interaction.
These results agree with work by Chen et al. (1998), which demonstrate improved
general flame stability (i.e. flame anchoring, but not necessarily improvements with
respect to combustion instabilities) with the use of a bluff-body.
A more direct
comparison can be made to the work of Kendrick et al. (1999), where a bluff-body
stabilized system is shown to be superior at resisting the tendency to produce acoustic
oscillations than an aerodynamically stabilized system. This does not necessarily indicate
a configuration less prone to combustion instabilities since the aerodynamically stabilized
burner has a lower Rayleigh index at a drive frequency of 37 Hz using OH PLIF
diagnostics. Furthermore, assessment of the tendency for instabilities to appear must be
based on analysis of the complete system, comprising the combustion dynamics and the
dynamics of the combustor.
147
Chapter 6
Concluding Remarks
This work provides experimental measurements of a variety of devices involving thermoacoustic interactions pertaining to the general problem of combustion instabilities. We
begin with an electrically driven Rijke tube, perhaps the simplest demonstration of heatinduced pressure oscillations, continue on to a large scale industrial flare, and conclude
with detailed measurements of the combustion dynamics of two burners under forced
oscillatory conditions.
Measurements from the Rijke tube indicate the presence of hysteresis with respect to the
power input at high mass flow rates (over 3 g/s in this configuration). As the heater input
power is increased until the Rijke tube exhibits instability, when the power is reversed,
significantly less power continues to sustain the oscillation. A detailed stability map with
uncertainty of the stability boundary over a range of mass flow rates and heater power
148
levels is provided, at a heater location of x/L = . A precise definition of instability is

introduced combining both pressure amplitudes and frequencies of decomposed modes,
which allows for better determination of system instability in possibly ambiguous
situations.
A one-dimensional linear stability model is offered that accounts for
variations of heat release and associated time delays by incorporating numerical

simulations from Kwon and Lee (1985).
This model is based entirely on physical
properties of the experiment and uses no empirical fits.
The model qualitatively
reproduces the instability curve, but is unable to accurately match the stability boundary
over a wide range of mass flow rates, nor provide an explanation for hysteresis.
Data collected on the industrial flare emphasized the nonorthogonality of the acoustic
modes, with the quarter-wave mode occurring at 8.3 Hz, and the three-quarter-wave
mode located at 21.5 Hz. Sub-scale modeling efforts brought forth difficulties in scaling
unstable frequencies, since the acoustic losses scale with frequency.
A novel technique was demonstrated which can measure the combustion dynamics of a
flame in an acoustically excited environment.
Measurements of the combustion
dynamics of two versions (aerodynamically and bluff-body stabilized) of a partially

premixed jet burner were taken using two different techniques. Chemiluminescence
measurements offered greater convenience since a laser source is not required, while OH
PLIF measurements have finer spatial and temporal resolution. The method employed in
this work is useful in the design and prediction of how a particular burner will respond in
a real combustor environment.
149
Chemiluminescence measurements also provide insight into the flow visualization of

burners.
The evolution of the flame and flame base mean location and oscillation
amplitude were observed at each of the forcing conditions over an entire cycle. Forcing
at 55 Hz was virtually transparent to both burner configurations, and is considered to be a
different regime of operation. At frequencies below 37 Hz, the flame base position
decreased with the bluff-body stabilized burner when compared with the aerodynamically
stabilized system. At 37 Hz the opposite effect occurred.
Forced Rayleigh indices were computed using two techniques: heat release derived from
OH PLIF measurements, and heat release inferred from flame chemiluminescence.
Examination of 2-D spatially resolved Rf contour plots using OH PLIF (Figure 5-19 to
Figure 5-23) and the same plots computed using chemiluminescence (Figure C-6 to
Figure C-10), shows fairly consistent discrepancies due primarily to the line-of-sight
integration occurring with chemiluminescence. Most obvious is the hollow core in the
OH measurements, since the flame is not burning in the core region of the burner. There
are also differences in the relative sizes of the flame damping and driving zones.
Starting at the base of the flame, the OH results show a much smaller damped zone than
the chemiluminescence results. Axial plots show qualitative agreement in the general
shape of the response for both techniques, however the relative magnitudes differed in
various zones. Global Rayleigh indices for the first mode of pressure filtered show the
bluff-body burner superior in all cases in resisting oscillations in all cases except at 37
150
Hz. The opposite result is obtained for chemiluminescence global Rayleigh indices. Part
of this discrepancy is attributable to the fact that the chemiluminescence imaging
technique integrates across the flame.
Another factor is the contribution of other
chemiluminescent species that do not indicate zones of heat release, such as CO2. The
OH PLIF Rayleigh results agree with the chemiluminescent flow visualization of the
flame base; a lower flame base corresponds to a stiffer flame, which is more able to resist
coupling with the acoustic waves.
The combustion response of the flame is also computed, based on the OH PLIF
measurements. It is useful in conjunction with the Rayleigh index in explaining the
combustion dynamics of the flame. The high sensitivity of the Rayleigh index to the
phase of the combustion response is demonstrated, particularly for the 37 Hz case. These
measurements of the combustion dynamics of a flame are the first of its kind.
Directions for Future Work

The Rijke tube experiment offers an ideal test bed to develop theoretical models to
predict limit cycle amplitudes and stability boundaries. The experimental configuration
is easily modified to produce stability maps at other heater locations of interest, such as
x/L = 1/8, which, according to Rayleighs criteria, should preferentially drive the 2nd
mode of the system. In order to accurately model the response of the Rijke tube, more
detailed models of heat release from the grid and gas dynamics will be required.
Prediction of limit cycle amplitudes will also require additional modeling of nonlinear
151
processes. Nonorthogonality of the modes should be addressed, as temperature profiles

indicate a complex temperature distribution as the flow develops.
Although chemiluminescence does not provide accurate Rayleigh indices in this work, it
should be investigated further, using OH PLIF as a means to evaluate its accuracy.
Possible improvements would be to use a bandpass filter to eliminate sources of
chemiluminescence that do not indicate heat release. The sensitivity of the camera must
also be increased if this is done, possibly by addition of an intensifier. Other species of
interest can also be probed using PLIF techniques, such as CH and NO. Zones of NOx
production under an acoustic field can be identified and possibly minimized in future
burner designs to aid in the development of ultra low emissions engine systems.
Since a framework has been established for measuring combustion dynamics, a variety of
other burner designs can be evaluated. A simple burner, easily modeled would be useful
in verifying results from numerical and theoretical models. Additional improvements
would entail improvement of the test section, such as the use of flat windows to improve
optical access. The addition of other diagnostics such as PIV will improve visualization
of the flow field, more precisely mapping vorticity production and velocity responses to
the acoustic field.
152
Appendix A
Mechanical Drawings
153
Scale: 1" = 5"
Quartz
Tube
42"
12.87" O.D. (327 mm)

12.52" I.D. (318 mm)
6061-T6
Al Tube
0.25"
5.125"
front
cutout
2.5"
0.5"
NOTES:
1.
2.
3.
1.5" x 1.5" x
0.5" Al block
8"
4"
3.5"
0.125"
elbow
joint or weld
back
cutout
1 1/16"
hole
2.125"
Al plate or
disc
0.5"
12.625" I.D.
13" O.D.
24"
Al block is welded to the Al tube, after an appropriately sized hole is cut, to allow a bulkhead
fitting centered in the block.
Burner, plumbing, holes in quartz tube, and bulkhead fitting are not shown.
A cover for the holes should be constructed, which consists of a rolled sheet to match the Al tube, with
cutouts matching those on the tube. Appropriate slots (elbow joints, bulkhead) should also be cut, to
allow the cover to rotate around the tube on the platform, from a fully open to a fully closed position.
It is preferable if both slots can be covered independently, but not required.
Figure A-1: 1/12th scale flare model: chamber section.
154
Plan
Scale: 1" = 3"
0.125"
2"x1/8"x1" Shield
Supports, with grooves
to maintain shield
positions, see Detail B
1" Al tube,
0.125" wall
R = 4.25"
3.11" 1/32"
3.94" 1/32"
7 " O.D.
0.2495" reamed
hole for spuds
10" I.D.
Thin walled stainless
sheets, with important
dimensions as
indicated
0.36"
0.70"
Burner Section A
Scale: 1" = 3"
10" I.D.
7.5" O.D.
2.875"
0.46"
0.25" O.D.
0.4"
Build a simple removable
support if necessary
Spuds (40) are 1/8" I.D.,

1/16" wall stainless steel
tubing, press fit into
ring. Approx. 0.1"
protrudes into ring,
exactly 0.4" rises above.
Re-drill centers if
necessary, after cutting.
8.5"
Seal end of tube,
and drill 4 - "
radial holes
3.125"
" tube
Figure A-2: 1/12th scale flare model: burner section.
Connect
to
bulkhead
fitting
155
Cross-Section, Section B
Scale 1" = 1"
P-NPT 1/8"
thread
0.49"
-20 UNF-2A
See DWG No. 7002
and DWG No. 112-1040-90
Mounting Hole Preparation
0.56"
0.11"
0.177"
0.05"
0.81"
1.66"
0.26"
1.75" brass
hex bar stock
20
0.0625"
hole
0.25"
0.125"
0.66"
1"
0.188"
0.0625"
hole
0.5" dia
1.25" dia
1.75" hex
Plan, Bottom View
Plan, Top View

0.49"
20
decline
Scale: 1" = 1"

-20 UNF-2A
See DWG No. 7002
Mounting Hole
Preparation
0.5"
B
1/16" holes
P-NPT
1/8" thread
Figure A-3: Instrumentation boss.
156
Appendix B
Gas Mixture Viscosity
Calculation
157
For a gas mixture, the mixture viscosity can be calculated using the following equation
(Kanury, 1975):
i i
mixture =
(B-1)
i =1
jij
j =1
where
1 M i
ij =
1+
8 M j
1 / 2
1 / 2 M 1 / 4
1 + i j
j M i
and i is the mole fraction of species i, Mi is the molecular weight of species i, i is the
viscosity of species i, and n is the total number of species.
In the experiments measuring combustion dynamics, a binary gas mixture is used. The
mixture viscosity is important since the flow is premixed and measured using a single
laminar flow element. A quadratic fit for the viscosity of several gases, valid from 5 C
45 C is performed, such that
i = a2 T2 + a1 T + a0,
(B-2)
where i is the viscosity of species i in micropoise (1 gm/cm/sec = 106 micropoise) and T

is in C. The constants a2, a1, and a0 are given in Table B-1.
Species
CH4
CO2
N2
a2
-0.005
-0.010
-0.010
a1
0.710
1.060
1.100
a0
96.499
130.500
160.000
Table B-1: Coefficients for viscosity quadratic fit,

valid for P = 1 atm, 5 C < T < 45 C.
158
Appendix C
Chemiluminescence
Rayleigh Indices
159
This appendix contains the spatially resolved chemiluminescent forced Rayleigh indices.
Figure C-1 to Figure C-5 show the axial profiles, with the aerodynamic and bluff-body
burners plotted together. The axial profiles are obtained by integrating the averaged 2-D
flame contours at each height and phase, and multiplying with the bulk averaged and
filtered (about the driving mode) pressure. The 2-D plots (Figure C-6 to Figure: C-10)
are obtained by direct evaluation of Rayleighs criterion with the averaged 2-D contours.
2
aerodyn
bluff-body
1
0
-1
Axial Rf
-2
-3
-4
-5
-6
-7
-8
10
12
14
16
18
Height [cm]
Figure C-1: Chemiluminescence axial forced Rayleigh indices at 22 Hz.
160
2
aerodyn
bluff-body
1
Axial Rf
-1
-2
-3
-4
-5
-6
10
12
14
16
18
Height [cm]
4
aerodyn
bluff-body
2
Axial Rf
-2
-4
-6
-8
10
12
14
16
18
Height [cm]
161
2
aerodyn
bluff-body
0
Axial Rf
-2
-4
-6
-8
-10
10
12
14
16
18
Height [cm]
3
aerodyn
bluff-body
2
1
0
Axial Rf
-1
-2
-3
-4
-5
-6
10
12
14
16
18
Height [cm]
162
(a)
(b)
14
14
-6
-1 2
-14
4
-2
-6
10
0
6
0
0
Height [cm]
Height [cm]
10
-1 0
-8
-8
12
-1
-1
12
-4
-1 8
16
10
16
-4
-4
-6
-8
-1 0
-4
-8
-6
0
0
-5
Radius [cm]
0
-5
Radius [cm]
Figure C-6: Chemiluminescence 2-D forced Rayleigh indices at 22 Hz for (a)

aerodynamically and (b) bluff-body stabilized burners.
163
(a)
(b)
16
16
-2
-2
14
14
12
12
-4
8
0
-8
-6
-6
-8
10
Height [cm]
Height [cm]
-1
-4
-2
10
-2
6
2
2
0
0
-5
-6
-4
Radius [cm]
0
-5
-2
-4
-2
Radius [cm]
Figure C-7: Chemiluminescence 2-D forced Rayleigh indices at 27Hz for (a)
164
(a)
(b)
16
16
-6
14
-2
14
-2
0
-4
-4
12
-4
-1
10
Height [cm]
-2
-2
Height [cm]
-8
-8
-8
-2
-1 0
10
-1 2
-4
12
4
6
4
4
-6
-4
-6
-6
-2
-4
2
0
0
-5
-2
Radius [cm]
0
-5
Radius [cm]
Figure C-8: Chemiluminescence 2-D forced Rayleigh indices at 32Hz for (a)
165
(b)
(a)
6
4
-2
-6
-1 0
14
-12
-8
-2
14
-10
16
16
12
-2
-6
12
-6
-6
Height [cm]
-1 2
-6
-2
10
-4
-1 2
0
0
-1 0
-1
-8
Height [cm]
10
-2
-6
-4
-4
-6
0
-5
Radius [cm]
0
-5
Radius [cm]
Figure C-9: Chemiluminescence 2-D forced Rayleigh indices at 37 Hz for (a)

166
(a)
(b)
16
16
-2
-4
14
6
0
14
-2
-4
-1 2
-2
12
-2
-2
-2
8
-6
-4
10
Height [cm]
Height [cm]
10
-2
0
12
-2
0
8
-6
-2
-2
-4
-2
-2
2
0
0
-5
Radius [cm]
0
-5
Radius [cm]
Figure: C-10: Chemiluminescence 2-D forced Rayleigh indices at 55 Hz for (a)

167
Appendix D
Locally Normalized
Combustion Response
168
The magnitude of the local combustion response can be plotted in two ways:
using the a spatial heat release mean, or
using a local, temporal mean for the heat release.
Section 5.5 uses the former method, as it was found to be more useful, however it was not
evident that this would be the case.
The main problem that arises when using the local temporal mean to normalize the heat
release is at the edges of the flame. The edges of an oscillating flame are continually
fluctuating in space, as well as time. As a result, a particular spatial location near the
edge may or may not contain a flame at any particular instant. While the mean heat
release rate will be non-zero (compared to a region outside the flame zone), the temporal
mean will be very low due to the periods of time when there is no flame present. This
causes the q / q term in the combustion response function to be extremely large
compared to other interior regions of the flame.
This phenomenon is displayed in Figure D-1, using a log scale for the magnitude. The
same process that magnifies the edge of the flame also enhances any inherent noise in the
image field. This is seen in Figure D-1 as the fingers extending from to the left.
Although it was determined that these plots are not practically useful, they are presented
here to provide guidance in future works.
169
22 Hz
27 Hz
32 Hz
37 Hz
55 Hz
(a)
(b)
Figure D-1: Locally normalized combustion response magnitude (intensity is on a

log scale) (a) aerodynamically stabilized (b) bluff-body stabilized
170
References
Allen, M.G., McManus, K.R., Sonnenfroh, D.M., and Paul, P.H., (1995) Planar laserinduced fluorescence imaging measurements of OH and hydrocarbon fuel fragments in
high-pressure spray-flame combustion, Applied Optics, Vol. 34, No. 27.
Allen, M.G., McManus, K.R., and Sonnenfroh, D.M., (1995a) PLIF Imaging in Spray
Flame Combustors at Elevated Temperatures, AIAA Paper 95-0172.
Allen, M.G., Parker, T.E., Reinecke, W.G., Legner, H.H., Foutter, R.R., Rawlins, W.T.,
and Davis, S.J. (1993) Fluorescence Imaging of OH and NO in a Model Supersonic
Combustor, AIAA Journal, Vol. 31, No. 3.
Anderson, T.J., Kendrick, D.W., and Cohen, J.M. (1998) Measurement of
Spray/Acoustic Coupling in Gas Turbine Fuel Injectors, presented at the 36th
Aerospace Sciences Meeting & Exhibit, Reno, NV, AIAA 98-0718.
171
Barlow, R.S., and Collignon, A., (1991) Linear LIF Measurements of OH in

Nonpremixed Methane-Air Flames: When are Quenching Corrections Unnecessary,
AIAA Paper 91-0179.
Barlow, R.S., Dibble, R.W., Chen, J.-Y., and Lucht, R.P. (1990) Effect of Damkohler
Number on Superequilibrium OH Concentration in Turbulent Nonpremixed Jet
Flames, Combustion and Flame, Vol. 82, pp. 235-251.
Battles, B.E. and Hanson, R.K. (1995) Laser-Induced Fluorescence Measurements of
NO and OH Mole Fraction in Fuel-Lean, High-Pressure (1-10 atm) Methane Flames:
Fluorescence Modeling and Experimental Validation, J. Quant. Spectrosc. Radiat.
Transfer, Vol. 54, No. 3, pp. 521-537.
Bjerken, L. (1999) Personal Communication.
Burnley, V.S., and Culick, F.E.C. (2000) Comment on Triggering of Longitudinal
Combustion Instabilities in Rocket Motors: Nonlinear Combustion Response, J. of
Prop. Power, Vol. 16, No. 1, pp. 164-166.
Broda, J.C., Seo, S., Santoro, R.J., Shirhattikar, and Yang, V. (1998) An Experimental
Study of Combustion Dynamics of a Premixed Swirl Injector, Twenty-Seventh
Symposium (International) on Combustion, The Combustion Institute.
Cadou, C.P. (1996) Two-Dimensional, Time-Resolved Temperature Measurements in a
Resonant Incinerator using Planar Laser-Induced Fluorescence, Ph.D. Thesis,
University of California (Los Angeles).
Cadou, C.P., Logan, P., Karagozian, A.R., and Smith, O.I. (1991) Laser diagnostic
techniques in a resonant incinerator, Environ. Sensing Combust. Diagn., SPIE, Vol.
1434, pp. 67-77.
172
Cadou, C.P., Smith, O.I., and Karagozian, A.R. (1998) Transport Enhancement in
Acoustically Excited Cavity Flows, Part 2: Reactive Flow Diagnostics, AIAA Journal,
Vol. 36, No. 9, pp. 1568-1574.
Cattolica, R. (1981) OH rotational temperature from two-line laser-excited
fluorescence, Applied Optics, Vol. 20, No. 7, pp. 1156-1166.
Cessou, A. and Stepowski, D. (1996) Planar Laser Induced Fluorescence Measurement
of [OH] in the Stabilization Stage of a Spray Jet Flame, Combust. Sci. and Tech., Vol.
118, pp. 361-381.
Chen, T.Y., Hegde, U.G., Daniel, B.R., and Zinn, B.T. (1993) Flame Radiation and
Acoustic Intensity Measurements in Acoustically Excited Diffusion Flames, J. Propul.
Power, Vol. 9, No. 2.
Chen, Y.C., Chang, C.C., Pan, K.L., and Yang, J.T. (1998) Flame Lift-off and
Stabilization Mechanisms of Nonpremixed Jet Flames on a Bluff-body Burner,
Combustion and Flame, Vol. 115, pp. 51-65.
Christo, F.C., Fletcher, D.F., and Joseph, S.D. (1998) Computational fluid dynamics
modelling of a landfill gas flare, Journal of the Institute of Energy, Vol. 71, pp/ 145151.
Clanet, C., Searby, G., and Clavin, C. (1999) Primary acoustic instability of flames
propagating in tubes: cases of spray and premixed gas combustion, J. Fluid Mech.,
vol. 385, pp.157-197.
Collyer, A.A., and Ayres, D.J. (1972) The generation of sound in a Rijke tube using two
heating coils, J.Phys.D:Appl.Phys., Vol. 5.
173
Correa, S.M., (1992) A Review of NOx Formation Under Gas-Turbine Combustion

Conditions, Combust. Sci and Tech., Vol. 87, pp. 329-362.
Culick, F.E.C. (1968) A Review of Calculations for Unsteady Burning of a Solid
Propellant, AIAA Journal, Vol. 6, No. 12, pp. 2241-2255.
Culick, F.E.C., (1976) Nonlinear Behavior of Acoustic Waves in Combustion
Chambers, Parts I and II, Acta Astronautica, Vol. 3, pp. 714-757.
Culick, F.E.C. (1987) A Note on Rayleighs Criterion, Combust. Sci. and Tech., Vol.
56, pp. 159-166.
Culick, F.E.C. (1999) Combustor Dynamics: Fundamentals, Acoustics, and Control, A
Short Course of Lectures, United Technologies Research Center.
Culick, F.E.C. and Jahnke, C. (1994) An Application of Dynamical Systems Theory to
Nonlinear Combustion Instabilities, J. Prop. and Power, Vol. 10, No. 4, pp. 508-517.
Crosley, D. (1993) Collisional effects on laser-induced fluorescence flame
measurements, Optical Engineering, Vol. 20, No. 7.
Dieke, G.H. and Crosswhite, H.M. (1962) The Ultraviolet Bands of OH, J. Quant.
Spectrosc. Radiat. Transfer, Vol. 2, pp. 97-199.
Dyer, M.J. and Crosley, D.R. (1982) Two-dimensional imaging of OH laser-induced
fluorescence in a flame, Optics Letters, Vol. 7, No. 8.
Eckbreth, R.C. (1988) Laser Diagnostics for Combustion Temperature and Species,
Abacus Press, Cambridge.
Fric, T.F., (1993) Effects of Fuel-Air Unmixedness on NOx Emissions, Journal of
Propulsion and Power, Vol. 9, No. 5, pp. 708-713.
174
Hanson, R.K., Seitzman, J.M., and Paul. P.H. (1990) Planar Laser-Fluorescence
Imaging of Combustion Gases, Applied Physics B, Vol. 50, pp. 441-454.
Heberle, N.M., Smith, G.P., Jeffries, J.B., Crosley, D.R., and Dibble, R.W., (2000)
Simultaneous laser-induced fluorescence and Rayleigh scattering measurements of
structure in partially premixed flames, Appl. Phys. B, Vol. 71, pp. 733-740.
Howe, M.S. (1998) Acoustics of Fluid-Structure Interactions, Cambridge University
Press, Cambridge.
Hurle, I.R., Price, R.B., Sugden, T.M., and Thomas, A., (1968) Sound Emission from
Open Turbulent Premixed Flames, Proc. Roy. Soc., Vol. 303, pp. 409-427.
Isella, G.C. (2001) Modeling and Simulation of Combustion Chamber and Propellant
Dynamics and Issues in Active Control of Combustion Instabilities, Ph.D. Thesis,
California Institute of Technology, Pasadena, California.
Isella, G., Seywert, C., Culick, F.E.C., and Zukoski, E.E. (1997) A Further Note on
Active Control of Combustion Instabilities Based on Hysteresis Combust. Sci. and
Tech., Vol. 126, pp. 381-388.
John Zink Corp. (1988) Biogas Flares presented at the GRCDA 11th Annual
International Landfill Symposium, Houston, Texas.
Kanury, A.M., (1975) Introduction to Combustion Phenomena, Gordon and Breach
Science Publishers, New York.
Kappei, F., Lee, J.Y., Johnson, C.E., Lubarsky, E., Neumeier, Y., and Zinn, B.T. (2000)
Investigation of Oscillatory Combustion Processes In Actively Controlled Liquid Fuel
Combustor, presented at the 36th Joint Prop. Conf., Huntsville, Al, AIAA 2000-3348.
175
Katto, Y., and Sajiki, A. (1977) "Onset of Oscillation of a Gas-Column in a Tube Due to
the Existence of Heat-Conduction Field (A Problem of Generating Mechanical Energy
from Heat)", Bulleton of the JSME, Vol. 20, No. 147, pp. 1161-1168.
Kendrick, D.W., (1995) An Experimental And Numerical Investigation Into Reacting
Vortex Structures Associated With Unstable Combustion, Ph.D. Thesis, California
Institute of Technology, Pasadena, CA.
Kendrick, D.W., Anderson, T.J., Sowa, W.A., and Snyder, T.S. (1999) Acoustic
Sensitivities of Lean-Premixed Fuel Injectors in a Single Nozzle Rig, J. Eng. Gas
Turbines and Power, Vol. 121, Iss. 3, pp. 429-436.
Kwon, Y-P., and Lee, B-H. (1985) Stability of the Rijke thermoacoustic oscillation, J.
Acoust. Soc. Am., Vol. 74, No. 4, pp. 1414-1420.
Laufer, G., (1996) Introduction to Optics and Lasers in Engineering, Cambridge
University Press, Cambridge, UK.
Lucht, R.P., Laurendeau, N.M., and Sweeney, D.W.,(1982) Temperature measurement
by two-line laser-saturated OH fluorescence in flames, Applied Optics, Vol. 21, No.
20, pp. 3729-3735.
Madarame, H. (1981a) 1st Report: Oscillations Induced by Plane Heat Source in Air
Current, Bulletin of the JSME, Vol. 24, No. 195, Sept. 1981.
Madarame, H. (1981b) 2nd Report: Oscillations induced by Interference of Heat
Sources. Bulletin of the JSME, Vol. 24, No. 198, Dec. 1981.
Maling, G.C. (1963) Simplified Analysis of the Rijke Phenomenon, JASA, Vol. 35, No.
7, pp. 1058-1060.
176
Masayuki, T., Berg, P.A., Harrington, J.E., Luque, J., Jeffries, J.B., Smith, G.P., and
Crosley, D.R., (1998) Collisional Quenching of CH(A), OH(A), and NO(A) in Low
Pressure Hydrocarbon Flames, Combustion and Flame, Vol. 114, pp. 502-514.
Masri, A.R., Dibble, R.W., and Barlow, R.S., (1996) The Structure of Turbulent Nonpremixed Flames Revealed by Raman-Rayleigh-LIF Measurements, Prog. Energy
Combust. Sci., Vol. 22, pp. 307-362.
McManus, K., Yip, B., and Candel, S., (1995) Emission and Laser-Induced
Fluorescence Imaging Methods in Experimental Combustion, Experimental Thermal
and Fluid Science, Vol. 10, pp. 486-502.
Murray, R.M. (1995) Sparrow Reference Manual, California Institute of Technology.
Palm, S.L. (in progress) Dynamic Response of Premixed and Non-Premixed Flames
Under Acoustic Forcing, Ph.D. Thesis, California Institute of Technology, Pasadena,
CA.
Palmer, J.L., and Hanson, R.K., (1996) Temperature imaging in a supersonic free jet of
combustion gases with two-line OH fluorescence, Applied Optics, Vol. 35, No. 3, pp.
485-499.
Paul, P.H. (1994) A Model for Temperature Dependent Collisional Quenching of OH
A2 +, J. Quant. Spectrosc. Radiat. Transfer, Vol. 51, No. 3, pp. 511-524.
Paul, P.H., and Dec, J.E. (1994) Imaging of reaction zones in hydrocarbon-air flames by
use of planar laser-induced fluorescence of CH, Optics Letters, Vol. 19, No. 13.
Paul, P.H., Gray, J.A., Durant, J.L., and Thoman, J.R., (1993) A Model for
Temperature-Dependent Collisional Quenching of NO A2 +, Appl. Phys. B, Vol. 57,
pp. 249-259.
177
Paul, P.H., Gray, J.A., Durant, J.L., and Thoman, J.R. (1994) Collisional Quenching
Corrections for Laser-Induced Fluorescence Measurements of NO A2 +, AIAA
Journal, Vol. 32, No. 8.
Poinsot, T.J., Trouve, A.C., Veynante, D.P., Candel, S.M., and Esposito, E.J., (1987)
Vortex-driven acoustically coupled combustion instabilities, J. Fluid Mech., Vol.
177, pp. 265-292.
Poinsot, T., Bourienne, F., Candel, S., and Esposito, E., (1989) Suppression of
Combustion Instabilities by Active Control, Journal of Propulsion and Power, Vol. 5.,
No. 1, pp. 14-20.
Poncia, G. (1998) A Study on Thermoacoustic Instability Phenomena in Combustion
Chambers for Active Control, Ph.D. Thesis, Politecnico di Milano.
Pun, W., Palm, S.L., and Culick, F.E.C. (2000) PLIF Measurements of Combustion
Dynamics in a Burner under Forced Oscillatory Conditions AIAA-2000-3123,
presented at the 36th Joint Propulsion Conference, Huntsville, AL.
Putnam, A.A., and Dennis, W.R. (1954) Burner Oscillations of the Gauze-Tone Type,
J. Acoust. Soc. Am., Vol. 26, No. 5, pp. 716-725.
Qin, W., Egolfopoulos, F.N., and Tsotsis, T.T. (2001) A Detailed Study of the
Combustion Characteristics of Landfill Gas, Chemical Engineering Journal, Vol.
3773, pp. 1-16.
Raun, R. L., Beckstead, M.W., Finlinson, J.C., and Brooks, K.P. (1993) A Review of
Rijke Tube Burners and Related Devices, Prog. Energy Combust. Sci., Vol. 19, pp.
313-364.
Rayleigh, J.W.S. (1945) The Theory of Sound Vol. II, Dover Publications, New York.
178
Rijke, P. (1859) Notiz ber eine neie Art, die Luft in einer an beiden Enden offenen
Rhre in Schwingungen zu versetzen, Annal der Physik, Vol. 107, pp. 339-343.
Rosfjord, T. (1995) Lean-Limit Combustion Instability United Technologies Research
Center.
Samiengo, J.M., Yip, B., Poinsot, T., and Candel, S., (1993) Low-Frequency
Combustion instability Mechanisms in a Side-Dump Combustor, Combustion and
Flame, Vol. 94, Iss. 4, pp. 363-380.
Sankar, S.V., Jagoda, J.I., and Zinn, B.T., (1990) Oscillatory Velocity Response of
Premixed Flat Flames Stabilized in Axial Acoustic Fields, Combustion and Flame,
Vol. 80, pp. 371-384.
Schadow, K., Yang, V., Culick, F.E.C., Rosjford, T., Sturgess, G., and Zinn, B., (1996)
Active Combustion Control for Propulsion Systems, AGARD Workshop Report,
May 6-9, 1996, Athens, Greece.
Seitzman, J.M., and Hanson, R.K., (1993) Comparison of Excitation Techniques for
Quantitative Fluorescence Imaging of Reacting Flows, AIAA Journal, Vol. 31, No. 3,
pp. 513-519.
Seywert, C. (2001) Combustion Instabilities: Issues in Modeling and Control Ph.D.
Thesis, California Institute of Technology, Pasadena, California.
Shih, W.P, Lee, J.G., and Santavicca, D.A., (1996) Stability and Emissions
Characteristics of a Lean Premixed Gas Turbine Combustor, Twenty-Sixth
Symposium (International) on Combustion, The Combustion Institute.
179
Smith, G.P., Luque, J., Jeffries, J.B., Crosley, D.R. (2001) Rate Constants for Flame
Chemiluminescence, 2nd Joint Meeting of the U.S. Sections of the Combustion
Institute, Oakland, CA.
Sterling, J.D., (1987) Longitudinal Mode Combustion Instabilities in Air Breathing
Engines, Ph.D. Thesis, California Institute of Technology, Pasadena, CA.
Sterling, J.D. and Zukoski, E.E., (1991) Nonlinear Dynamics of Laboratory Combustor
Pressure Oscillations, Combust. Sci. and Tech., Vol. 77, pp. 225-238.
Swenson, G., Pun, W., and Culick, F.E.C., (1996) Nonlinear Unsteady Motions and
NOx Production in Gas Turbine Combustors, 11th Symposium on Combustion and
Explosion, Russian Academy of Sciences, Chernogolovka, Russia, November.
Venkataraman, K.K., Preston, L.H., Simons, D.W., Lee, B.J., Lee, J.G., and Santavicca,
D.A. (1999) Mechanism of Combustion Instability in a Lean Premixed Dump
Combustor, J. Propul. Power, Vol 15, Iss. 6, pp. 909-918.
Wu, S., Blake, G.A., Sun, S., and Ling, J., A multicrystal harmonic generator that
compensates for thermally induced phase mismatch, Optics Communications, Vol.
173, pp. 371-376.
Yip, B., Miller, M.F., Lozano, A., and Hanson, R.K. (1994) A combined OH/acetone
planar laser-induced fluorescence imaging technique for visualizing combusting flows,
Experiments in Fluids, Vol. 17, pp.330-336.
Zsak, T.W. (1993) An Investigation of the Reacting Vortex Structures Associated with
Combustion, Ph.D. Thesis, California Institute of Technology, Pasadena, CA.

Punthesis PDF

Uploaded by

Punthesis PDF

Uploaded by

Measurements of

In Partial Fulfillment of the Requirements

California Institute of Technology

In particular, I would like to

A one-dimensional model of the stability boundary with linear

acoustics is shown to have qualitative agreement with experimental data.

In the experiments, a test chamber is

acoustically excited by a pair of low-frequency drivers. The response of the flame is

This is the first known

measurement of the combustion dynamics of a flame over a range of frequencies.

The burner is operated in two

configurations; with an aerodynamically stabilized flame and with a flame stabilized by

Acknowledgements ...................................................................................... iii

CHAPTER 2 THE RIJKE TUBE .............................................................13

2.1.3 Pressure Transducers......................................................................................... 16

CHAPTER 3 FULL-SCALE FLARE AND SUB-SCALE MODEL ....44

CHAPTER 4 CHEMILUMINESCENCE MEASUREMENTS ...........62

4.1 EXPERIMENTAL SETUP ........................................................................................... 63

CHAPTER 5 OH PLIF MEASUREMENTS ...........................................97

5.1.2 Laser System ................................................................................................... 103

CHAPTER 6 CONCLUDING REMARKS .......................................... 147

Appendix A Mechanical Drawings ........................................................ 152

Appendix B Gas Mixture Viscosity Calculation.................................. 156

Appendix C Chemiluminescence Rayleigh Indices ............................. 158

Appendix D Locally Normalized Combustion Response.................... 167

REFERENCES ......................................................................................... 170

Figure 3-3: Flare bulk temperature profile........................................................................ 49

Figure 4-10: Chemiluminescence contour plots at 27 Hz for (a) aerodynamically

Figure 5-2: PLIF system. ................................................................................................ 105

Figure C-10: Chemiluminescence 2D forced Rayleigh indices at 22 Hz for (a)

From document EPA-456/F-98-006 (September 1998)

combustors operating under these conditions, as shown in Figure 1-1.

combustion instability refers to the presence of a large amplitude pressure oscillation

Lean premixed combustion systems are especially susceptible to combustion instabilities,

Figure 1-2: Combustor system.

To provide accurate measurements of nonlinear instability phenomena of

To develop tools for evaluation of the combustion response of a burner in a

To evaluate the use of chemiluminescence and OH PLIF as indicators for the

1.3 Rayleighs Criterion

together without mass sources, we arrive at

are mean quantities and

represent fluctuating quantities. Assuming small

changes in a correspondingly low mean flow, orthogonality between u and p , and

1.4 Previous Work

1.4.3 Combustion Dynamics of Unsteady Flames

Two techniques that can be used to perform these measurements are

chemiluminescence and planar laser-induced fluorescence (PLIF).

Chemiluminescence of various radicals of combustion, in particular CH, is an excellent

chemiluminescence is on the order of approximately 100 s, versus 100 ns when

Poinsot et al. (1987) (440-590 Hz)

Cadou et al. (1991) (43 Hz)

Sterling and Zukoski (1991) (188 Hz)

Shih et al. (1996) (400 Hz)

Broda et al. (1998) (1750 Hz)

Cadou et al. (1998) (328 Hz)

Kendrick et al. (1999) (235 Hz, 355 Hz)

Venkataraman et al. (1999) (490 Hz)

Kappei et al. (2000) (370-460 Hz)

Chen et al. (1993) (300 Hz, 400 Hz)

Cadou et al. (1998) (360 Hz, 420 Hz)

Table 1-1: Previous work in oscillating flames.

Most experimental work characterizing various combustor configurations has been

Work by Chen et al. (1993) with

A more generalized body of work is required to provide a scientific

understanding of guidelines that will be useful in designing stable combustion systems.

2.1 Experimental Setup

Figure 2-1: Electrical Rijke tube experimental setup.

2.1.1 Heat Source

The power supplies are computer controlled using a software-

Table 2-1 lists some of the important characteristics of the pressure

> 250 kHz

2.1.5 Data Acquisition System

A Computer Boards CIO-

Cold junction compensation

Flow rate (LFE pressure drop)

Table 2-2: Data acquisition analog measurements.