DIODE-LASER ABSORPTION SENSORS FOR COMBUSTION...
Transcript of DIODE-LASER ABSORPTION SENSORS FOR COMBUSTION...
DIODE-LASER ABSORPTION SENSORS FOR
COMBUSTION CONTROL
by
Xin Zhou
July 2005
TSD Report 161
High Temperature Gasdynamics Laboratory Mechanical Engineering Department
Stanford University Stanford, California 94305
ii
Copyright by Xin Zhou 2005
All Rights Reserved
iii
Abstract
Combustion is the most widely used energy conversion technique in the world. Measurements of
important combustion parameters are critical to understand combustion processes, improve
combustion efficiency and reduce the production of pollutants such as nitrogen oxides. Diode-
laser absorption sensors offer significant opportunities and advantages for in situ measurements
of multiple combustion parameters such as temperature and species concentration due to their
high sensitivity, high spectral resolution, fast time response, robustness and non-intrusive
character. The overall objective of this thesis is to design and develop time-resolved and real-time
tunable diode laser sensors with the potential for combustion control.
A crucial element in the design of a tunable-diode-laser optical-absorption-based sensor is the
selection of optimum transitions. Water vapor is present in ambient air as well as a primary
hydrocarbon combustion product, and thus provides a ubiquitous target for absorption-based
sensors. There are nearly half a million possible water vapor absorption transitions cataloged
between 1 and 2 µm, and an important part of this thesis is the development of a design-rule
approach for absorption transition selection. The strategy and spectroscopic criteria for selecting
optimum wavelength regions and absorption line combinations are developed for two-line
thermometry. The development of this design-rule approach establishes a new paradigm to
optimize tunable diode laser sensors for target applications.
The water vapor spectrum in the 1-2 µm near-infrared region is systematically analyzed to find
the best absorption transition pairs for sensitive measurement of temperature in the target
combustion environment using a single tunable diode laser. The use of a single laser capable of
tuning over two or more water lines can offer advantages over wavelength-multiplexing
techniques and make the system compact, rugged, low cost and simple to operate. Two sensors
are developed in this work. The first sensor is a 1.8 µm, single-laser temperature sensor based on
direct absorption scans. Successful time-resolved measurements in a variety of laboratory and
practical devices are presented and used to identify potential improvements, and design rules for a
second-generation sensor are developed based on the lessons learned. The second generation
sensor is a 1.4 µm, single-laser temperature sensor using water vapor absorption detected by
wavelength-modulation spectroscopy (WMS), which facilitates rapid data analysis and a 2 kHz
real-time data rate in the combustion experiments reported here. As part of the sensor
development, fundamental spectroscopic parameters for the selected transitions are measured to
iv
improve the available databases. Demonstration experiments in a heated cell and a forced
Hencken burner confirm the sensitivity and accuracy of the sensors. The first application of TDL
thermometry to a liquid-fuel swirl-stabilized spray combustor also is presented, illustrating the
potential for noninvasive temperature measurements in harsh, practical environments such as gas
turbine combustors.
The ability of the 1.4 µm temperature sensor to predict the approach to the lean blowout (LBO)
limit is investigated, and active control of thermoacoustic instabilities is successfully
demonstrated in a practical swirl-stabilized flame. These results illustrate the potential of this
sensor for active combustion control.
v
Acknowledgements
I am deeply indebted to my advisor, Professor Ron Hanson, for his unwavering support. His
comments have been of greatest help at all times. Without his numerous encouragements,
inspiration, many insightful ideas and suggestions, this work would not be possible. I wish to give
special thanks to Dr. Jay Jeffries for his time, patience, help and great guidance in all the time of
research and writing of this thesis. I would also like to acknowledge the members of my
committee: Prof. Christopher F. Edwards, Prof. Craig T. Bowman and Prof. Piero Pianetta, for
their time and advice.
I would like to thank Prof. Ephraim Gutmark and Dr. Guoqiang Li from University of Cincinnati,
Dr. Tom Jenkins from MetroLaser for their cooperation and help in my research.
I would like to acknowledge the senior students and my colleagues who have provided help and
support during my research work: Micheael Webber, Jian Wang, Scott Sanders, Suhong Kim,
Jonathan Liu, Dan Mattison, Lin Ma, Kent Lyle, Xiang Liu, Ning Xu, Hejie Li, Adam Klingbeil,
Greg Rieker and all members of Hanson group. It is a pleasure to acknowledge the great moments
I spent with my wonderful friends: Liqiang, Xiaojun, Yue, Shuhuai, Xuejiao, Woo Kyung.
I would like to thank my wife, Yuhua, for her love, support, patience, and confidence in me. She
has made my life joyful even during difficult times. I am extremely grateful to my parents
Chunying Wang and Xuezhi Zhou for their constant support, understanding, encouragement and
love all through my life.
This work was supported by the ONR via the University of Cincinnati and the Global Climate
Energy Program at Stanford.
vi
vii
Table of Contents
Abstract………………………………………………………………………….………….. iii
Acknowledgements..…………………………………………………………….………….. v
Table of Contents……………………………………………………………….………….. vii
List of Tables………………………………………………………………………………... x
List of Figures……………………………………………………………………………….. xi
Chapter 1. Introduction……………………………………………………………….. 1
1.1 Motivation and objectives………………………………………………….. 1
1.2 Scope of the current work………………………………………………….. 3
1.3 Organization of the thesis…………………………………………………... 4
Chapter 2. Theory of absorption spectroscopy…..…………………………………... 5
2.1 Beer-Lambert law…………………….. …………………………………… 5
2.2 Lineshape function…………………………………………….. ………….. 6
2.2.1 Gaussian lineshape function……………………………………. 6
2.2.2 Lorentzian lineshape function………………………………….. 7
2.2.3 Voigt lineshape function………………………………………... 9
2.3 Diode-laser absorption spectroscopy techniques…………………………… 11
2.3.1 Direct absorption spectroscopy…………………………………. 12
2.3.2 Modulation spectroscopy……………………………………...... 22
Chapter 3. Development of design rules for absorption-based sensors…………….. 30
3.1 Spectroscopic database (HITRAN) ………………………………………... 30
3.2 Design rules of selecting optimum transitions for 2-line T sensor…..……... 32
3.3 Selection of optimum transitions for high pressure T sensor…. ………….. 36
3.3.1 Motivation………………………………………………………. 38
3.3.2 Line selection criteria…………………………………………… 40
3.3.3 Summary……………………………………………………..…. 50
3.4 Multiplexing technique…………………………………………………….. 50
3.4.1 Wavelength Division Multiplexing (WDM) …………………… 51
3.4.2 Time Division Multiplexing (TDM) …………………………… 52
3.4.3 Frequency Division Multiplexing (FDM) ……………………… 53
viii
Chapter 4. Temperature sensing using H2O transitions near 1.8 µm……………… 55
4.1 Water, H2O…….…………………………………………………………... 55
4.2 Development of single-laser T sensor (direct absorption) ………………… 57
4.2.1 Selection of water line pairs…………………………………….. 57
4.2.2 Spectroscopy experiments, results and discussions……..……… 63
4.3 Combustion Demonstration………………………………………………… 69
4.3.1 Temperature and concentration measurements………………..… 69
4.3.2 Identification of acoustic instabilities…………………………… 73
4.3.3 Closed-loop control of mean temperature……………………….. 74
4.3.4 Time-resolved measurements in a swirl-spray combustor……... 77
4.4 Summary……………………………………………………………………. 82
Chapter 5. Temperature sensing using H2O transitions near 1.4 µm……….…….. 85
5.1 Motivation……………………………………………………………….… 85
5.2 Development of single-laser T sensor (2f) ………………………………… 87
5.2.1 Line selection……………………………………………………. 87
5.2.2 Spectroscopic verification……………………………………….. 94
5.2.3 2f temperature sensor validation..……………………………….. 103
5.2.4 Real-time capabilities…………...……………………………….. 107
5.3 Combustion Demonstration……………………………...…………………. 111
5.3.1 Identification of acoustic instabilities…………………………… 111
5.3.2 Real-time measurements in a swirl-spray combustor…………… 113
5.3.3 Comparison with 1.8 µm sensor…………………………..…….. 117
5.4 Summary……………………………………………………………………. 120
Chapter 6. Application of fast temperature sensor to combustion control….…….. 121
6.1 Motivation………………………...…….………………………………….. 121
6.2 Swirl-stabilized combustor…. ……………………………………………... 122
6.3 Lean blowout (LBO) prediction…………….……………...…….………… 124
6.3.1 Experimental setup………………………………………………. 124
6.3.2 Results and discussions…………………….……………………. 126
6.4 Combustion instability control………….………………………………….. 139
ix
6.4.1 Experimental setup……………………………………………… 139
6.4.2 Results and discussions….……...……………………………….. 140
6.5 Summary…………………………………………………………………… 145
Chapter 7. Conclusions and future work…………………………………………... 147
7.1
Summary of the use of design rules to identify the optimum transitions for
IC engine applications.……………………………………………………...147
7.2
Design of a single laser absorption sensor for temperature measurements
using direct absorption……………………..……………………………….147
7.3
Design of a single laser absorption sensor for temperature measurements
using WMS………………………………………………………………….149
7.4 Investigation of the 1.4 µm WMS T sensor for combustion control……….. 150
7.5 Potential Plan for future work……………………………………….. 151
Appendix Architecture of the real-time WMS sensor...…..……………………...… 153
References…………………………………………………………………………………… 161
x
List of Tables Chapter 3
Table 3.1 The 14 candidate lines………………………………………………………. 44
Table 3.2 The 16 attractive line pairs………………………………………………….. 46
Chapter 4
Table 4.1 Fundamental vibrations, frequencies, types and description for H2O………. 56
Table 4.2 Assignments of the NIR vibrational absorption spectrum of water………… 57
Table 4.3 Line selection result using the selection criteria in the near-infrared region
based on HITEMP…………………………………………………………... 60
Table 4.4 Candidate H2O line intensity pairs for measurements of temperature and
water concentration in the near-infrared region based on HITEMP………… 60
Table 4.5 Spectroscopic data for the selected H2O line pair…………………………… 67
Chapter 5
Table 5.1 Candidate H2O line intensity pairs for measurements of temperature and
water concentration in the 1-2 µm region based on HITTRAN2004……….. 91
Table 5.2 Line selection result using the selection criteria in the near-infrared region
based on HITRAN2004……………………………………………………... 91
Table 5.3 Spectroscopic data for the selected H2O line pair…………………………… 103
Chapter 6
Table 6.1 Different inter, intermediate and outer swirler configurations……………… 124
xi
List of Figures Chapter 2
Figure 2.1 Schematic of typical absorption measurements…………................................ 5
Figure 2.2 Comparison of Gaussian, Lorentzian and Voigt lineshape on a normalized
frequency and intensity scale………..…………………………......................11
Figure 2.3 Schematic of typical scanned-wavelength direct-absorption measurements… 12
Figure 2.4 Schematic of typical direct-absorption measurements………………………. 13
Figure 2.5 Etalon signal vs. time………………………………………………………… 14
Figure 2.6 Calculated Doppler width as a function of temperature for one H2O
transition……………………………………………………………………...15
Figure 2.7 Two different temperature-dependent transitions……………………………. 17
Figure 2.8 Line strength as a function of temperature. Temperature is inferred from the
ratio of integrated areas for two different transitions……………………..….18
Figure 2.9 Line strength ratio sensitivity (E”1-E”2= 1500 cm-1) as a function of
temperature…………………………………………………………………...19
Figure 2.10 Schematic of typical fixed-wavelength direct-absorption measurements…… 21
Figure 2.11 Fixed-wavelength two line technique……………………………………….. 22
Figure 2.12 A typical arrangement for the WMS technique……………………………… 23
Figure 2.13 The Voigt profile and its first three harmonic signals vs. normalized
frequency……………………………………………………………………..25
Figure 2.14 Second harmonic line shape and line center peak height for different “m”… 26
Figure 2.15 Comparison of 2f line shape with and without intensity modulation…. 27
Figure 2.16 Temperature inferred from 2f peak ratio of two different temperature-
dependent transitions…………………………………………………………28
Chapter 3
Figure 3.1
Typical high EGR and Super-charged intake cycles in internal combustion
engine and representative water spectra under two limiting conditions
during the cycle………………………………………………………………
38
Figure 3.2 Survey spectra of H2O at 300 K in the 1~8 µm region based on the HITRAN
database………………………………………………………………………39
xii
Figure 3.3
The simulated 2f signals for nine of the fourteen lines (1, 2, 5, 8, 10, 11, 12,
13, and 14) for the compression portion of the high EGR (panels a and b)
and the supercharged (panels c and d) cycles using a modulation amplitude
of 0.8cm-1……………………………………………………………………..
43
Figure 3.4
High EGR compression cycle: (a) Simulated 2f ratio, (b) temperature
sensitivity, and (c) temperature uncertainty for the line pair 2 and 5 as a
function of pressure/temperature. (a=0.8cm-1)……………………………….
48
Figure 3.5
Super-charged intake compression cycle: (a) Simulated 2f ratio, (b)
temperature sensitivity, and (c) temperature uncertainty for the line pair 5,
11, 12, 13 and 15 as a function of pressure/temperature. (a=0.8cm-1)………..
49
Figure 3.6 Schematic of the wavelength division multiplexing…………………………. 51
Figure 3.7 Grating separates the colors in incident light………………………………... 51
Figure 3.8 Schematic of Time Division Multiplexing…………………………………... 52
Figure 3.9 Schematic of the Frequency Division Multiplexing…………………………. 53
Chapter 4
Figure 4.1 The structure of water molecule and its three fundamental vibrations……… 56
Figure 4.2 Survey spectra of H2O at 1000 K in the near-infrared region based on the
HITEMP database…………………………………………………………… 56
Figure 4.3
Expanded view of absorption spectra for the selected H2O line pairs in the
near-infrared region based on the HITEMP database; evaluated for P=1
atm, 10% H2O, 90% air……………………………………………………...
61
Figure 4.4 Calculated temperature sensitivity of line strength ratio as a function of
temperature for line pair 2, 5 and 10 based on the HITEMP database……… 62
Figure 4.5 Experimental schematic of the measurement system for determining
spectroscopic parameters……………………………………………………. 64
Figure 4.6 Reduced H2O line-shape (line pair #10) recorded in a static cell at T=944
K, PH2O=17.44 Torr………………………………………………………….. 66
Figure 4.7 Line strength of the transitions contributing to line pair #10 near 1.8 µm at
1000 K based on HITEMP parameters……………………………………… 66
Figure 4.8
Calculated and measured line strengths for the components of line pair #10
as a function of temperature. “Line 2” is the high temperature transition at
5553.86 cm-1; “Line 1” is the low temperature transition at 5554.18 cm-1….
67
xiii
Figure 4.9
The ratio of peak absorbance coefficients, Rpeak(line pair #10), calculated as
a function of temperature for various values of water mole fraction at 1 atm
(based on HITEMP database [Rothman 1998])……………………………...
68
Figure 4.10 The ratio of line strength and peak absorbance coefficients and their
sensitivity to temperature versus temperature for the line pair #10………….69
Figure 4.11 Schematic diagram of the measurement system applied to the Hencken
burner……………………………………………………………………….. 70
Figure 4.12 Measured temperatures in the burned-gas region above a C2H4-air flame in
a 5 cm ×5 cm Hencken burner……………………………………………….71
Figure 4.13 Measured temperatures and its power spectrum in the burned region above
the C2H4-air flame……………………………………………………………73
Figure 4.14 Experimental schematic of the measurement system applied to the Hencken
burner………………………………………………………………………...74
Figure 4.15 Block diagram showing the strategy used for closed-loop control of the
mean temperature…………………………………………………………….75
Figure 4.16 The temperature response to a desired set-point temperature………………. 76
Figure 4.17 The response time of the closed-loop control system………………………. 76
Figure 4.18 Experimental schematic of the measurement system applied to the swirl
spray combustor…………………………………………………………….. 78
Figure 4.19 Reduced line-shapes for gas and liquid fuel………………………………… 79
Figure 4.20 Measured temperatures and its power spectrum in the burned region above
the Propane-air flame (unforced (a) and forced flow (b)).…………………..80
Figure 4.21 Four sensor positions investigated: 1. Top of flame 2. Under flame 3.
Above flame 4. Diagonal…………………………………………………… 81
Figure 4.22 Power spectrum at four sensor positions investigated (Propane)….………. 81
Chapter 5
Figure 5.1 Linestrength of H2O in the 1 to 2 µm spectral region at 1000 K (from
HITRAN 2004 database)…………………………………………………….87
Figure 5.2 Linestrength scaled by values at room temperature as a function of
temperature for H2O lines with various lower state energies………………..88
Figure 5.3
Expanded view of absorption spectra for the four selected H2O line pairs in
the 1.4 µm region based on the HITRAN2004 database; evaluated for P=1
atm, 10% H2O, 90% air……………………………………………………...
92
xiv
Figure 5.4
Expanded view of absorption spectra for the selected H2O line pairs in the
1.8µm region based on the HITRAN2004 database; evaluated for P=1 atm,
10% H2O, 90% air…………………………………………………………...
93
Figure 5.5 Experimental schematic of the measurement system for determining
spectroscopic parameters……………………………………………………. 95
Figure 5.6 Sample data (50-scan average) obtained from cell experiment at T = 951 K,
PH2O = 15.47 Torr……………………………………………………………. 96
Figure 5.7 Reduced H2O lineshape recorded in the cell at T = 951 K, PH2O = 15.47
Torr. The low T line is the line with the smaller value of lower-state E”…... 96
Figure 5.8 Line strength of the selected transitions at 1000 K based on HITRAN2004
parameters revealing that the high T line is composed of two lines………… 97
Figure 5.9
Measured integrated absorbance area vs. H2O pressure at T=951K for the
“Low T Line”, “High T Line 1” and “High T Line 2”. The line strength can
be calculated from the slope…………………………………………………
98
Figure 5.10
Measured collision width vs. H2O pressure at T=951K for the “Low T
Line”, “High T Line 1” and “High T Line 2”. The self-broadening
coefficient can be calculated from the slope…………………………………
98
Figure 5.11 Calculated and measured line strengths for the “Low T Line” as a function
of temperature……………………………………………………………….. 100
Figure 5.12 Calculated and measured line strengths for the “High T Line 1” and “High
T Line 2” as a function of temperature……………………………………… 100
Figure 5.13 Calculated and measured self-broadening coefficients for the “Low T Line”
as a function of temperature………………………………………………… 101
Figure 5.14 Calculated and measured self-broadening coefficients for the “High T
Line” as a function of temperature………………………………………….. 101
Figure 5.15 Calculated and measured air-broadening coefficients for the “Low T Line”
as a function of temperature………………………………………………… 102
Figure 5.16 Calculated and measured air-broadening coefficients for the “High T Line”
as a function of temperature………………………………………………… 102
Figure 5.17 Schematic of single-laser scanned-wavelength method…………………….. 103
Figure 5.18 Arrangement for the 2f sensor validation experiments……………………… 105
Figure 5.19
Comparison of measured 2f peak ratio with simulated 2f peak ratio (top);
Comparison of measured temperature with thermocouple temperature.
(bottom)……………………………………………………………………...
106
xv
Figure 5.20
Comparison of measured 2f peak ratio with simulated 2f peak ratio (top);
Comparison of measured temperature with thermocouple temperature.
(bottom)……………………………………………………………………...
107
Figure 5.21 Complete hardware-software framework…………………………………… 108
Figure 5.22 Diagram of the bias-tee……………………………………………………… 108
Figure 5.23 The trigger signal (top) and the laser transmission signal (bottom)………… 109
Figure 5.24 Timing diagram: timing trigger signal (Top), the second-harmonic signal
(Middle) and the acquired signal (Bottom)………………………………….110
Figure 5.25 Schematic diagram of the measurement system applied to a Hencken burner 111
Figure 5.26 Measured temperature and its power spectrum in the burned region above
the C2H4-air flame……………………………………………………………112
Figure 5.27 Schematic diagram of the measurement system applied to the swirl-
stabilized spray combustor…………………………………………………..113
Figure 5.28 Reduced H2O 2f line shapes (single scan) recorded in gas fuel (propane)
and liquid fuel (ethanol)……………………………………………………..114
Figure 5.29 Measured acoustic signal and its power spectrum in the burned region
above the propane-air flame…………………………………………………115
Figure 5.30 Measured temperature and its power spectrum in the burned region above
the propane-air flame……………………………………………………….. 115
Figure 5.31 Measured acoustic signal and its power spectrum in the burned region
above the ethanol-air flame………………………………………………….116
Figure 5.32 Measured temperature and its power spectrum in the burned region above
the ethanol-air flame…………………………………………………………116
Figure 5.33 Calculated spectroscopic features for water line pairs in the 1.4 µm and 1.8
µm sensors based on HITRAN; XH2O = 10%..................................................117
Figure 5.34 Comparison of the measurement strategy of the 1.4 µm sensor and 1.8 µm
sensor………………………………………………………………………...119
Chapter 6
Figure 6.1 Schematic of swirl-stabilized combustor……………………………………. 123
Figure 6.2 Scheme of the experimental setup…………………………………………... 124
Figure 6.3 Raw data with/without beam steering noise………………..……………….. 125
Figure 6.4 Reduced H2O 2f line shape (single scan) recorded in gas fuel (propane)…... 126
xvi
Figure 6.5 The blowout process for the first set of experiments (air flow rate = 27.3
SCFM)………………………………………………………………………. 127
Figure 6.6
Microphone and 2f sensor result during the lean blowout process for the
first set of experiments. (Air flow rate = 27.3 SCFM). (a) Laser beam:
20mm height from the injector (b) Laser beam: 50mm height from the
injector……………………………………………………………………….
129
Figure 6.7 The blowout process for the second set of experiments (air flow rate=38.7
SCFM)............................................................................................................. 132
Figure 6.8 The blowout process for the third set of experiments (air flow rate= 52.0
SCFM)………………………………………………………………………. 132
Figure 6.9 The blowout process for the fourth set of experiments (air flow rate=
63.9 SCFM) ……………………………………………………...…… 132
Figure 6.10
Microphone and 2f sensor result during the lean blowout process for the
second set of experiments. (Air flow rate = 38.7 SCFM). (a) Laser beam:
20mm height from the injector (b) Laser beam: 50mm height from the
injector……………………………………………………………………….
134
Figure 6.11
Microphone and 2f sensor result during the lean blowout process for the
third set of experiments. (Air flow rate = 52.0 SCFM). (a) Laser beam:
20mm height from the injector (b) Laser beam: 50mm height from the
injector……………………………………………………………………….
136
Figure 6.12
Microphone and 2f sensor result during the lean blowout process for the
fourth set of experiments. (Air flow rate = 63.9 SCFM). (a) Laser beam:
20mm height from the injector (b) Laser beam: 50mm height from the
injector……………………………………………………………………….
138
Figure 6.13 Scheme of the experimental setup…………………………………………... 139
Figure 6.14 2f peak ratio and its power spectra before and after control were applied on
the swirl-spray combustor (Propane/Air)…………………………………… 141
Figure 6.15 Microphone signal and its power spectra before and after control were
applied on the swirl-spray combustor……………………………………….. 142
Figure 6.16 Control performances versus controller time-delay………………………… 144
Figure 6.17 Control performances versus amplifier gain………………………………… 145
Appendix
Figure A.1 A simplified block diagram…………………………………………………. 153
xvii
Figure A.2 Flow chart for the data acquisition and analysis program…………………... 154
1
Chapter 1: Introduction 1.1 Motivation and objectives Combustion is the most widely used energy conversion technique in the world. Therefore
improvements in combustion efficiencies or reduction in harmful combustion emissions could
have an enormous impact on the world’s atmosphere. Measurements of important combustion
parameters are critical to understand the combustion process, improve combustion efficiency and
reduce the production of pollutants such as the nitrogen oxides. Numerous combustion diagnostic
techniques have been developed for measurements of temperature and species concentration, and
fall largely into two strategies: physical probing methods and optical methods. Compared with the
traditional physical probing methods, optical techniques offer significant advantages for non-
intrusive investigation of combustion process. The optical diagnostic methods include passive
emission measurements and active laser-based optical diagnostics. These methods include
traditional scattering techniques of Rayleigh, Mie, and Raman, as well as linear wavelength-
resonant processes such as absorption and laser-induced fluorescence, and non-linear techniques
such as wave-mixing, or coherent anti-strokes Raman scattering, etc. All of these laser-based
measurement schemes have contributed to modern combustion diagnostics. [Kohse-Höinghaus
2002] The continuous (or CW) laser measurement methods can be used for time-resolved
diagnostics and have the potential for real-time measurements. CW optical absorption can be
wavelength-tuned to provide a robust real-time measurement with sufficient data rate for the
potential use for combustion control. This thesis concentrates on the design of such time-resolved
and real-time tunable diode laser sensors with the potential for combustion control.
Modern combustion applications can be improved with real-time control. For many years, there
has been much concern about the reduction of NOx emissions from combustion [Martin 1990;
John 1997]. Such concern has led to the investigation of a number of schemes to reduce NOx
emissions. Lean premixed combustion is one of the effective approaches to reduce NOx
emissions because of lower flame temperatures [Martin 1990]. Unfortunately, this approach has
two major drawbacks: First, it is susceptible to lean blowout (LBO) of combustion, which could
lead to significant potential safety hazards, energy loss, and substantial costs of a power plant
shut-down [Muruganandam 2005], and it is thus important to precisely predict the lean blowout
limit. Second, it is susceptible to thermoacoustic combustion instabilities as a result of the
coupling of heat release to acoustic oscillations [Paschereit 1998; Mcmanus 1993; Docquier
Chapter 1
2
2002], which can lead to decreased combustion efficiency, increased noise pollution, and serious
system performance degradation.
Blowout and instability control are two major areas of concern. Temperature is a fundamental
parameter of combustion systems and a good physics-based control variable [Furlong 1996],
because it is a measure of combustion heat release and determines the overall thermal efficiency.
To meet the requirements of an active control strategy, the sensor system must be able to
determine the states of the combustion system rapidly and accurately [Docquier 2002]. Since
combustion instabilities usually occur with frequencies less than 500 Hz, a kHz real-time speed is
required to provide an effective feedback control signal. A traditional thermocouple sensor has
difficulty meeting such fast time response requirements. Sensor systems based on absorption
spectroscopy techniques are attractive for such applications due to their high sensitivity, high
spectral resolution, fast time response, robustness and non-intrusive character. Furthermore, the
line-of-sight measurements can provide an additional advantage by evaluating a flow-field-
averaged quantity. The present work is aimed at the design of a diode-laser absorption sensor
system for temperature that might be used for blowout and instability control in a practical
combustion control system, e.g., for gas turbine combustors.
There are many reasons that tunable semiconductor diode lasers (TDL) are nearly ideal sources
for CW absorption sensor applications including compact and rugged packaging, low cost,
compatibility with optical fiber and relative ease of use. Improvements in the performance,
reliability and wavelength availability of tunable diode lasers offer the potential to increase
detection sensitivity and accuracy of TDL sensors. Building upon the concepts of time-division
multiplexing (TDW) and wavelength-division multiplexing (WDM), sensors using multiple diode
lasers have been demonstrated in various environments [Allen 1998; Philippe 1993; Webber 2000;
Sanders 2000]. These multiplexing methods have some disadvantages with regard to system
complexity and cost, and chapters 4 and 5 will illustrate the design of a sensor with a single diode
laser.
A good sensor design can lead to remarkably improved sensor performance and accuracy. An
important step in the design of an absorption-based sensor is the line selection. The HITRAN
(High Resolution Transmission Molecular Absorption Database) database [Rothman 2003]
consists of important spectroscopic parameters of specific spectral lines, and has been widely
used to simulate absorption spectra and line selection for molecules present in the atmosphere.
Introduction
3
One goal of this work is to develop the strategy and spectroscopic criteria for selecting optimum
wavelength regions and absorption line combinations for sensor design. The development of such
design rules in this work makes it possible to produce absorption sensors that can achieve
accuracy and performance for a wide range of applications. Chapter 3 illustrates the use of
HITRAN for sensor design and chapter 4 and 5 extends these design rules for single-laser sensors.
The sensors developed in chapters 4 and 5 are then applied to real-time blowout prediction and
instability control. Chapter 4 introduces a 1.8 µm single-laser temperature sensor based on
wavelength-scanned direct-absorption measurement of two adjacent H2O lines. This sensor
system has the desired flexibility, sensitivity, speed and accuracy to be a useful tool for
fundamental and applied combustion monitoring. However, demonstration measurements show
this sensor has limitations; for example, it can not provide a kHz real-time rate due to its
relatively complex data reduction strategy. From the lessons learned with the 1.8 µm single-laser
temperature sensor, the design rules were modified leading to a new 1.4 µm single-laser
temperature sensor (chapter 5) based on a combination of scanned-wavelength and wavelength
modulation spectroscopy (WMS) with 2f detection. A real-time temperature readout rate of 2 kHz
was achieved for the 1.4 µm WMS sensor. This new sensor system offers significant
opportunities and advantages for in situ measurements of temperature for combustion control.
Finally, in chapter 6 a multiple-swirl spray combustor is used for sensor evaluation, leading to the
demonstration of TDL thermometry in a liquid-fuel swirl-stabilized spray flame. The 1.4 µm
WMS sensor is then successfully applied to lean blowout prediction and thermoacoustic
instability control in the swirl-stabilized combustor providing solid evidence that TDL sensing is
especially promising for use in combustion control applications.
1.2 Scope of the current work
The overall objective of this thesis is to develop strategies for sensor design and to extend the
application domains of absorption gas sensing to practical combustion environment e.g. a liquid-
fuel swirl-stabilized combustor. Specific objectives of this thesis are the following:
1. Investigate the water vapor spectrum in the 1~2 µm near-infrared region, develop the
strategy and spectroscopic criteria for selecting optimum wavelength regions and
absorption line combinations. The design rules developed in this work should prove
useful to those interested in temperature sensing using absorption spectroscopy.
Chapter 1
4
2. Explore the use of design rule TDL sensor development by using HITRAN to design and
choose two laser colors to monitor temperature during the compression stroke in an
internal combustion engine. This sensor design project illustrates the concept of design
rules for absorption sensor development.
3. Use the design rule concept to develop a robust temperature sensor for real-time
combustion sensing and control. Two generations of sensors have been investigated for
combustion diagnostics in practical combustion environments, e.g. a liquid-fuel swirl-
stabilized spray combustor. The first generation sensor near 1.8 µm is based on
wavelength-scanned direct absorption, and the second generation sensor near 1.4 µm uses
scanned wavelength modulation spectroscopy. Measurements of pertinent fundamental
spectroscopic parameters for both sensors were made using a heated cell. The
performance of each sensor has been investigated and validated in the laboratory and
demonstrated in liquid-fueled swirl-spray flames.
4. Demonstrate the utility of the second generation TDL sensor for use in a real-time closed-
loop control in a practical combustion system for lean blowout and acoustic instabilities
in a swirl-stabilized combustor.
1.3 Organization of the thesis The fundamentals of high-resolution absorption spectroscopy are introduced in Chapter 2, where
both direct absorption and wavelength-modulation spectroscopy are discussed. Chapter 3
introduces the characteristics of the HITRAN database, and then discusses the design-rule
strategy and spectroscopic criteria for selecting optimum wavelength regions and absorption line
combinations. The concepts developed in chapter 3 make it possible to design absorption sensors
that can achieve high accuracy and performance for a wide range of applications. Chapter 4
presents the development of a first-generation single-laser temperature sensor based on the direct
absorption technique. The sensor design, fundamental parameter measurements, and laboratory
spray combustor demonstrations are presented in detail. In Chapter 5, a second generation
wavelength-scanned WMS sensor is developed with design rules evolved from the lessons
learned from the investigations described in chapter 4, and a direct comparison of these two
sensors is made. Chapter 6 presents the application of the second-generation sensor to combustion
control in a swirl-stabilized spray combustor. Chapter 7 summarizes the thesis and suggests future
work. An appendix summarizes the hardware and software architecture of the real-time second-
generation wavelength-scanned WMS temperature sensor.
5
Chapter 2: Theory of absorption spectroscopy Diode-laser absorption spectroscopy techniques have become one of the most powerful tools for
gas sensing applications. [Arroyo 1993; Baer 1994; Nagali 1997; Kohse-Höinghaus 2002] Sensor
systems based on absorption spectroscopy techniques can offer significant opportunities and
advantages for in situ measurements of multiple flowfield parameters such as temperature,
pressure, velocity and density due to their high sensitivity, high spectral resolution, fast time
response, robustness and non-intrusive character. [Allen 1998; Philippe 1993] Diode-laser
absorption spectroscopy techniques usually fall into one of two categories: direct absorption
spectroscopy and modulation spectroscopy. This chapter will cover the theoretical aspects and the
fundamental principles of both techniques.
2.1 Beer-Lambert law
gas
I0(ν) I(ν)
L
Figure 2.1 Schematic of typical absorption measurements.
The fundamental theoretical principle of absorption spectroscopy is the Beer-Lambert law. This
law describes the relationship between transmitted and incident spectral intensities when the laser
beam passes through a uniform gaseous medium (Fig. 2.1). The equation of Beer-Lambert law is
simple and straightforward:
exp( )v vo v
IT k LI
≡ = ⋅
(2.1)
where Tν is the fractional transmission; I and Io are the transmitted and incident laser intensities;
kν [cm-1] is the spectral absorption coefficient; L[cm] is the path length. The product kν L
Chapter 2
6
represents the spectral absorbance, αν. The spectral absorption coefficient kν [cm-1] comprising Nj
overlapping transitions in a multi-component environment of K species can be expressed as
,, 0,1 1
( ) ( )jNK
j i jv i j ij i
P Tk SX Φ ν ν= =
= −∑ ∑ (2.2)
where P [atm] is the total pressure, Xj is the mole fraction of species j, Si,j [cm-2 atm-1] and Φi,j [cm]
are the linestrength and lineshape of a particular transition i of the species j, respectively.
The area underneath lineshape function Φi,j is normalized to unity so that
, ,0( ) ν 1i j iv v dΦν − ≡∫ (2.3)
2.2 Lineshape function
Broadening mechanisms can be grouped into homogeneous broadening which affects all
molecules in the same way, and inhomogeneous broadening which has different effects on some
groups of molecules. Due to different broadening mechanisms, a specific absorption transition
may be a convolution of multiple lineshape functions. The lineshape function provides important
information since it is a function of many parameters such as pressure, species concentration and
temperature. Three significant lineshape functions (Gaussian, Lorentzian and Voigt) and the
broadening mechanisms they describe are discussed in the following sections.
2.2.1 Gaussian lineshape function
The Gaussian lineshape function arises from inhomogeneous broadening mechanisms such as
Doppler broadening, which is caused by random thermal motion of the absorber molecules.
Statistical mechanics implies that the distribution of molecular speeds within a dilute gas is
Maxwellian; the Doppler lineshape is then described by a classical bell-shaped Gaussian curve:
∆−
−∆
=ΦDD
D ννν
πνν 0
2
2ln4exp2ln2)( (2.4)
Theory of absorption spectroscopy
7
where Dν∆ is the full width at half maximum (FWHM) of the lineshape, called the Doppler
width, and can be calculated using
70 02
8 ln 2 7.1623 10DkT Tmc M
ν ν ν−∆ = = × (2.5)
where 0ν [cm-1] is the linecenter frequency, T [K] is the temperature, and M [a.m.u] is the
molecular weight of the absorber species. The higher the temperature of the gas, the bigger the
Doppler width, thus the broader the line. Doppler width can provide a measure of gas temperature.
The peak height of the Gaussian lineshape function is
02 ln 2( )D
D
νν π
Φ =∆
(2.6)
2.2.2 Lorentzian lineshape function
Natural lifetime broadening and collisional broadening are the dominant homogeneous
broadening mechanisms and produce Lorentzian lineshapes. Natural broadening stems from the
uncertainty in energy of the states with finite lifetime involved in the absorption transition. As
described by the Heisenberg Uncertainty Principle
12
νπτ
∆ ≥ (2.7)
The energy of a photon can not be precisely known due to the finite lifetime of the excited state.
If the absorption line is damped only by the natural lifetime of the energy states, this is termed
“natural” broadening. Natural broadening leads to a Lorentzian lineshape function
( )2
20
1 2( )
2
n
nn
ν
νπ νν ν
∆
Φ =∆ − +
(2.8)
Chapter 2
8
where nν∆ [cm-1] is the “natural” width (FWHM) and 0ν [cm-1] is the linecenter frequency. In
most cases, natural broadening can be neglected due to the relatively long lifetime of the energy
levels.
Collisional broadening is another important homogeneous broadening mechanism. It is produced
by collisions of the emitting or absorbing particle with other particles. Based on two key
assumptions: a.) collisions are binary; b) collision duration is negligible compared to time
between the collisions, the collisional broadening lineshape takes the form of a Lorentzian profile,
( )2
20
1 2( )
2
c
cc
ν
νπ νν ν
∆
Φ =∆ − +
(2.9)
where cν∆ [cm-1] is the collisional width (FWHM) and 0ν [cm-1] is the linecenter frequency.
When collisions occur between different species, we call the process foreign-gas broadening.
When collisions take place between same species, we call it self-broadening. In the limit of
binary collisions, the collision width is proportional to pressure at constant temperature and the
total collision width in a multi-component environment is given by
2 )(C j jj
P Xν γ∆ = ∑ (2.10)
where Xj is the mole fraction of component j, and γj [cm-1 atm-1] is the collisional broadening
coefficient due to perturbation by the jth component. (Note, γj is the half-width at half maximum
per atm of pressure of the partner j) The temperature dependence of γj can be expressed as:
00( ) ( )
j
j j
nTT TT
γ γ =
(2.11)
where T0 is the reference temperature and nj represents the corresponding coefficient of
temperature dependence.
Theory of absorption spectroscopy
9
The peak height of the Lorentzian lineshape function is
02( )c
c
νν π
Φ =∆
(2.12)
2.2.3 Voigt lineshape function
Doppler broadening often dominates at low pressure, and collisional broadening becomes
predominant at high pressure. In general, the overall broadening is a combination of natural,
collisional and Doppler broadening. If the broadening mechanisms are independent, the lineshape
is given by the Voigt profile.
( ) ( ) ( )V D Cu u duν ν+∞
−∞
Φ = Φ Φ −∫ (2.13)
Defining the Voigt “a” parameter as
ln 2 C
D
a νν∆
=∆
(2.14)
with the nondimensional line position w as
02 ln 2( )
D
wν ν
ν−
=∆
(2.15)
and the integral variable y as
2 ln 2
D
uyν
=∆
(2.16)
then the Voigt function becomes
Chapter 2
10
2 2
02 2 2 2
2 ln 2 exp( ) exp( )( ) ( )( ) ( )V D
D
a y a ydy dya w y a w y
ν νν π π π
+∞ +∞
−∞ −∞
− −Φ = = Φ
∆ + − + −∫ ∫ (2.17)
The Voigt functional form has been the basis for most quantitative analysis in absorption
spectroscopy. Unfortunately, a simple analytic form is not available and so practical systems have
adopted different numerical approximations to the true Voigt function.
The line width (FWHM) of the Voigt lineshape can be estimated using: [Olivero 1977]
( )2 20.5346 0.2166V C C Dν ν ν ν∆ = ∆ + ∆ + ∆ (2.18)
The peak height of Voigt lineshape function can be expressed by: [Mayinger 2001]
01( )V
CED
β βνπ γγ π
−Φ = + ⋅
(2.19)
with( )
ED
C ED
γβγ γ
= +
and ln 2
DED
γγ = where γD and γC are the half-width at half maximum
line widths (HWHM) of the Gaussian and Lorentzian shapes respectively.
Theory of absorption spectroscopy
11
1.0
0.8
0.6
0.4
0.2
0.0
Φ(ν
)/Φ(ν
0)
-10 -5 0 5 10(ν−ν0)/∆ν
Gaussian profile Lorentzian profile Voigt profile (a=0.1) Voigt profile (a=1) Voigt profile (a=10)
Figure 2.2 Comparison of Gaussian, Lorentzian and Voigt lineshape on a normalized
frequency and intensity scale.
Figure 2.2 shows a comparison of Gaussian, Lorentzian and Voigt lineshape on a normalized
frequency and intensity scale. As can be observed in Figure 2.2, the Gaussian, Voigt (“a”=1) and
Lorentzian lineshape will reach 1% of the peak at 1.3 line widths, 3.9 line widths and 5.2 line
widths from the center, respectively. The Voigt profile shows a Lorentzian-like behavior in the
line wings and a Gaussian-like behavior in the line center. For small Voigt “a” parameter, a → 0,
the Voigt profile grows into the Gaussian profile. For large Voigt “a” parameter, a Lorentzian
profile is recovered.
2.3 Diode-laser absorption spectroscopy techniques
Many highly innovative spectroscopic techniques have been and continue to be demonstrated for
combustion using a wide variety of lasers [Kohse-Höinghaus 2002]. This section focused on two
of the most widely used absorption spectroscopy techniques based on tunable diode lasers: direct
absorption spectroscopy and wavelength modulation spectroscopy. Though somewhat different in
principle, these two techniques both can be used to measure important parameters such as
temperature and species concentration in practical combustion applications. This section will
Chapter 2
12
cover the principle, experimental methods, and the capabilities and limitations of these two
techniques. In addition, the relative merits of the direct absorption technique and the modulation
spectroscopy technique are discussed.
2.3.1 Direct absorption spectroscopy
Basically there are two different experimental methods which have been used for direct
absorption spectroscopy: scanned- and fixed-wavelength direct-absorption techniques. Details of
these two methods for temperature and species concentration measurement will be discussed in
the following subsection.
A. Scanned-wavelength direct absorption spectroscopy
Scanned-wavelength is the most common method in direct absorption spectroscopy. The laser
frequency is tuned over the selected absorption transition and the measured line shape is analyzed
to obtain important information such as absorption line strength and broadening coefficients.
From these spectroscopic measurements practical information can be inferred such as gas
temperature, species concentration, gas velocity, and gas pressure.
Figure 2.3 Schematic of typical scanned-wavelength direct-absorption measurements.
Theory of absorption spectroscopy
13
6
5
4
3
2
1
0
Sig
nal [
V]
1.00.80.60.40.20.0Time [ms]
Regions used to fit baseline
Transmitted signal Baseline
Region to monitor background signal
Figure 2.4 Schematic of typical direct-absorption measurements.
A typical experimental schematic of direct-absorption measurements is shown in Figure 2.3. The
laser controller, which includes temperature and current controllers, drives the diode laser. A
function generator is used to ramp the laser injection current and thus tune the wavelengths of the
laser over the desired absorption features. Figure 2.4 presents typical transmitted laser intensity
and corresponding absorbance. Note that laser current is intentionally tuned below laser threshold
to measure the background signal. The background signal comes from the detector background
signal, room lights, flame emission and interference signals. This signal should be relatively
constant and small, and is subtracted before we determine absorbance. The incident laser intensity
is obtained by fitting the regions without absorption to a low-order polynomial. As indicated
above, the wing of the Voigt line shape will decrease to 1% of the peak absorbance at 4 line
widths from center frequency. This can be used as “rule of thumb” to select the regions for
baseline fitting.
Chapter 2
14
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Sig
nal [
V]
1.00.80.60.40.20.0Time [ms]
Etalon signal
FSR
Figure 2.5 Etalon signal vs. time.
To transform the transmitted intensity data from the time-domain into the laser wavelength-
domain, a solid etalon with a known free spectral range (FSR) is used (shown in figure 2.5). Such
an etalon is the simplest form of a Fabry-Perot interferometer, in which a beam of light undergoes
multiple reflections between two reflecting surfaces, and whose resulting optical transmission (or
reflection) is periodic in wavelength. The FSR is defined by the relation [Born 1975]
1 1( )2
FSR cmnd
− ≡ (2.20)
where n is the index of refraction of the material between the mirrors, and d is the distance
between the two parallel mirrors. Since the peak-to-peak separation in the etalon trace is constant
(the FSR), it allows a simple transformation between relative frequency and time.
Temperature Measurement
1. Doppler line width
As already mentioned above, the Doppler width can provide a measure of gas temperature.
Temperature is expressed as:
Theory of absorption spectroscopy
15
2
707.1623 10
DT M νν−
∆= ×
(2.21)
where Dν∆ is the Doppler width, 0ν [cm-1] is the linecenter frequency, and M [a.m.u] is the
molecular weight of the absorber species. Figure 2.6 shows the calculated Doppler width as a
function of temperature for a H2O transition near 1.4um, the resulting curves for other transitions
are similar to this one.
0.05
0.04
0.03
0.02
Dop
pler
Wid
th [c
m-1
]
200015001000500Temperature [K]
Figure 2.6 Calculated Doppler width as a function of temperature for one H2O transition.
The uncertainty of T, represented by the standard deviation σT, can be estimated using Equation
(2.22)
// / ( ) 2
/ / /D
D
TT D
D D D
T dT dT T
ν
ν
σ σσ νσ ν ν ν
∆
∆
∆= ≅ =
∆ ∆ ∆ (2.22)
According to Equation (2.22), the fractional uncertainty in T is twice that of the measured
Doppler width.
As discussed earlier, the line shape is dominated by collisional broadening at a pressure of one
atmosphere. From these Voigt and Lorentzian lineshape functions, it may be difficult in practice
Chapter 2
16
to extract the Doppler width accurately. Therefore, this method is applicable only in cases where
the Doppler width is dominant. For H2O in the NIR, where airγ ~0.05cm-1/atm at 296K, this
would require pressure less than 0.04 atm (30 Torr) for a Doppler width 5 times the collisional
width at 296K. The Doppler-width method works best for low pressure conditions. Hence, for the
applications addressed in this thesis (atmospheric pressure or above), the use of lineshape to infer
temperature is not discussed further.
2. Two-line technique
The two-line technique is the most widely used method for temperature measurement in scanned-
wavelength spectroscopy. The gas temperature is obtained by comparing the line strength of two
different transitions which have different temperature dependence.
The temperature-dependent linestrength [cm-2atm-1] can be expressed in terms of the known
linestrength at a reference temperature T0: 1"
0 0 00
0
0
0
( ) 1 1( ) ( ) exp 1 exp 1 exp( )
Q T T hchcE hcS T S TkTQ T T k T T kT
ν ν−
− − = − − − − (2.23)
where Q(T) is the molecular partition function, h [J sec] is Planck’s constant, c [cm/s] is the speed
of light, k [J/K] is Boltzmann’s constant, and E” [cm-1] is the lower-state energy.
The partition function is determined over a range of temperatures using the following polynomial,
which represents a best fit of a summation over all calculated energy levels:
2 3( )Q T a bT cT dT= + + + (2.24)
The coefficients of the polynomial expression for various species (including H2O) are included in
the HITRAN database. [Rothman 1998]
The temperature can be inferred from the measured ratio of integrated absorbance for two
different temperature-dependent transitions (shown in figure 2.7 and 2.8). Because the two
integrated absorbances are obtained with the same partial pressure of water and same path length,
the ratio of these two integrals reduces simply to the ratio of line strengths, which by Equation
(2.23) is given by
Theory of absorption spectroscopy
17
( )1
2
1 " "0 11 11 2
2 2 0 2 02
( ) ( , )( ) 1 1exp[ ]( ) ( , )( )
abs
abs
P L S T d S TA S T hcR E EA S T S T k T TP L S T d
ν
ν
ν ννν
Φ = = = = − − − Φ
∫∫
(2.25)
where Pabs [atm] is the partial pressure of the absorbing species, φν [cm] is the line-shape function
of a particular transition, S(T0, νi) is the line strength of the transition centered at νi (cm-1), for the
reference temperature T0; E” is the lower state energy (cm-1), and T is the gas temperature (K).
Note that if the two transitions are sufficiently close to each other in frequency, the last ratio term
of Equation (2.23) may be approximated by 1.
Abs
orba
nce
Frequency
A1
A2
S1(T)
S2(T)=
Integrated Area Ratio Line Strength Ratio
Figure 2.7 Two different temperature-dependent transitions
Chapter 2
18
12x10-3
10
8
6
4
2
0
Line
str
engt
h S
[cm
-2/a
tm]
30002500200015001000500Temperature [K]
5
4
3
2
1
0
Line Strength R
atioS1
S2
Ratio=S1/S2
Figure 2.8 Line strength as a function of temperature. Temperature is inferred from the ratio of
integrated areas for two different transitions
Thus, the temperature of the gas can be obtained using the relation
" "2 1
" "2 01 2 1
2 1 0 0
( )
( ) ( )ln ln( )
hc E EkTS TA hc E E
A S T k T
−=
−+ +
(2.26)
where T0 is the reference temperature for the line strengths. The quantity hc/k has a numerical
value of 1.438 cm K. A1 and A2 are the integrated areas of the absorption lines. E” is the lower
state energy for the given line.
Theory of absorption spectroscopy
19
8
7
6
5
4
3
2
1
0
Sen
sitiv
ity (
dR/R
)/(d
T/T
)
2500200015001000500Temperature [K]
T=1.44*(E"1-E"2)
Figure 2.9 Line strength ratio sensitivity (E”1-E”2= 1500 cm-1) as a function of temperature.
The relative sensitivity of this ratio to temperature can be obtained by differentiating Equation
(2.25): " "1 2( )/ ( )
/E EdR R hc
dT T k T−
=
(2.27)
It can also be seen from this equation that a line pair with a high lower state energy difference is
desired to have high temperature sensitivity. However, in practice this is limited by two practical
issues. First, because temperature is determined from the ratio of the measured absorbance on two
absorption lines, equal SNR is desired for both measurements. Lines with very high E” have very
small absorbance and the ability to measure the absorbance becomes the upper limit on E”.
Second, lines with small E” have large absorbance in cold boundary layers and this becomes a
practical lower limit on E”.
Line selection is a crucial step in the sensor design. The strategy and spectroscopic criteria for
selecting optimum absorption transitions will be discussed in chapter 3.
Chapter 2
20
Species Concentration Measurement
As mentioned above, the integrated absorbance is proportional to partial pressure; thus, species
concentration can be obtained:
( )( )A AX
P L S TPL S T dν ν= =
⋅ ⋅Φ∫ (2.28)
where P[atm] is the total pressure, L[cm] is the path length, S(T) [cm-2/atm] is the line strength
and A[cm-1] is the integrated area.
There are two important advantages to note regarding the scanned-wavelength method. The first
is that the entire line shape can be well resolved. This is important for four reasons. First, fitting
the entire line shape eliminates line broadening effects, so the ratio of integrated areas is reduced
to a ratio of line strengths which is only a function of temperature; this makes scanned-
wavelength thermometry more robust in hostile environments where gas composition and
pressure change rapidly with time. Second, many fundamental spectroscopic parameters such as
line strength and broadening coefficients can be inferred from the line shape. Thus the scanned-
wavelength method is usually employed to measure and validate the fundamental spectroscopic
parameters in a database. Third, the spectrally resolved line shape can distinguish the contributing
absorbance from nearby transitions. This method is not only applicable to isolated line conditions,
but also well suitable to the situation where overlap occurs. Therefore this method offers the
potential to measure multi-transitions using a single scan. In addition, the absolute species
concentration can be obtained without any calibration. Fourth, the baseline is obtained by fitting
the regions outside the absorption line (i.e., without absorption) to a low-order polynomial, thus
eliminating the need of an extra reference laser beam in the setup.
The second important advantage to note is that the scanned-wavelength method can distinguish
background noise and non-resonant attenuation from resonant gas absorption such as optical
window attenuation and beam steering. The signal interference from emission can be assessed by
scanning the laser below the current threshold. Etalon noise from optical components can be seen
by comparing the transmitted signal to the laser intensity.
However, the scanned-wavelength method also has disadvantages. First, the baseline fit may
become difficult at high pressure conditions where the line is broadened and blended with
neighbors and the limited laser scan range precludes collecting a precise baseline signal. Second,
Theory of absorption spectroscopy
21
the laser scan range decreases with scan rate, and the need to maintain an acceptable scan range
restricts the resulting frequency bandwidth. Furthermore, the time-consuming nonlinear Voigt fit
will limit the measurement bandwidth.
B. Fixed-wavelength direct absorption spectroscopy
Gas MediaCombiner
Demultiplexer Detector
Figure 2.10 Schematic of typical fixed-wavelength direct-absorption measurements.
A typical arrangement for the fixed-wavelength technique is shown in figure 2.10. Compared
with the scanned-wavelength method, an additional non-resonant reference laser beam is needed
to infer the baseline to account for non-resonant losses from beam steering and window fouling.
A combiner is used to multiplex two laser beams together, so they can pass through the same
measurement path. A demultiplexing technique is also needed to separate the multiplexed laser
beams. Such (de)multiplexing techniques including wavelength division (de)multiplexing, time-
division (de)multiplexing and frequency-division (de)multiplexing will be discussed in detail in
section 3.4.
Temperature Measurement
In the fixed-wavelength technique, the laser wavelength is usually fixed at the transition’s center
frequency. The temperature can be inferred from the measured ratio of peak absorbance for two
different temperature-dependent transitions (shown in figure 2.12). The ratio of two peak
absorbances is given by
1 1
2 1
1 11
2 2 2
( ) ( )( ) ( )
abs
abs
P L S T S THRH P L S T S T
ν ν
ν ν
Φ Φ= = =
Φ Φ (2.29)
where φν [cm] is the line-shape function of a particular transition. The peak height ratio is not
only a function of temperature, but also pressure and mole fraction.
Chapter 2
22
Abs
orba
nce
Frequency
H1
H2
S1(T)Φν1(X, P, T)
S2(T)Φν2(X, P, T)
=
Peak Height Ratio Absorption Coefficient Ratio
H1
H2
=kν1
(T, X, P)
kν2(T, X, P)
Figure 2.11 Fixed-wavelength two line technique
Species Concentration Measurement
( )HX
P L S Tν
=⋅ ⋅Φ ⋅
(2.30)
where H is the peak absorption. The peak height of Gaussian, Lorentzian and Voigt line shapes
are given by Eqn. (2.6), (2.12) and (2.19), respectively. Thus, the relationship between peak
height and species concentration is more complicated than using integrated absorption via the
scanned-wavelength method.
2.3.2 Modulation spectroscopy
Modulation spectroscopy is a widely used technique for sensitive trace-species detection
[Philippe 1993; Dharamsi 1996; Schilt 2003], and can be used to significantly reduce 1/f noise by
shifting detection to higher frequencies. Modulation spectroscopy is also classified into two
Theory of absorption spectroscopy
23
categories: wavelength-modulation spectroscopy (WMS) and frequency-modulation spectroscopy
(FMS), according to the relative magnitude of modulation frequency and transition half-width
frequency.
WMS utilizes a modulation frequency less than the half-width frequency of the transition
lineshape. FMS, on the contrary, uses a modulation frequency larger than the half-width
frequency of the transition lineshape. WMS and FMS provide a substantial sensitivity
enhancement compared to direct absorption methods previously discussed.
WMS with detection at 2f is chosen and discussed in this thesis. Numerous descriptions of FMS
can be found in the literature. [Cooper 1987; Trans 1984; Avetisov 1996; Reid 1981; Dharamsi
1996]
Wavelength-modulation spectroscopy (WMS)
Figure 2.12 A typical arrangement for the WMS technique.
A typical arrangement for the measurement by the WMS technique is shown in figure 2.12. A
combination of slow ramp and a fast sinusoidal modulation is used to drive the diode laser. The
laser output frequency can be expressed by:
( )ν( ) ν(t) cos 2 mt a f tπ= + (2.31)
Where ν (t) [cm-1] is the laser center frequency, a [cm-1] is the modulation amplitude and fm [Hz]
is the modulation frequency. The transmission coefficient can be expanded in a Fourier cosine
series:
0( cos(2 )) (ν, )cos( 2 )
n
m n mn
a f t H a n f tτ ν π π=+∞
=+ = ∑ (2.32)
Ramp
Measurement path length
Detector signals to lock-in
Diode laser Controller
Computer
Function generator
Diode Laser CollimatorModulate at f
Lock-in amplifier
Reference signalFunction generator
Harmonic Signal
+
Chapter 2
24
where (ν, )nH a is the nth Fourier coefficient of the transmission coefficient, and they are given by
01(ν, ) ( cos )
2H a a dπ
π τ ν θ θπ
+−= +∫ (2.33)
1(ν, ) ( cos ) cos( )nH a a n dππ τ ν θ θ θ
π+−= + ⋅ ⋅∫ (2.34)
For weak transitions,
( ) 0.05S P X Lφ ν⋅ ⋅ ⋅ ⋅ ≤ (2.35)
The transmission coefficient can be approximated thus:
( )
0
( )( ) [1 ( ) ]( )
LI e S P X LI
α νντ ν φ νν
−= = ≈ − ⋅ ⋅ ⋅ ⋅ (2.36)
And the nth harmonic Fourier coefficient simplifies as
(ν, ) ( cos ) cos( )nS P X LH a a n dπ
π φ ν θ θ θπ
+−
⋅ ⋅ ⋅= − + ⋅ ⋅∫ (2.37)
The analytical expression of the second Fourier coefficient of Gaussian and Lorentzian line shape
are given by [Reid 1981]
(ν, ) ( cos ) cos( )nS P X LH a a n dπ
π φ ν θ θ θπ
+−
⋅ ⋅ ⋅= − + ⋅ ⋅∫ (2.38)
2 (ν, ) ( cos ) cos(2 )S P X LH a a dππ φ ν θ θ θ
π+−
⋅ ⋅ ⋅= − + ⋅ ⋅∫ (2.39)
A Voigt profile and the corresponding 1f, 2f, 3f WMS line shapes are shown in figure 2.13 with a
relative ordinate scale. There are three important points to note: First, the Nth-harmonic line shape
has N+1 turning points; thus the existence of more than N+1 turning points will imply that there
are more than one transition in the spectral region scanned. [Dharamsi 1996] Second, the (N+1)th
harmonic signal has a smaller magnitude than Nth harmonic signal; hence lower harmonics are
usually employed in practice due to their relatively strong signal. Third, the line shapes of even-
numbered harmonics are symmetric about the line center, while the line shapes of odd-numbered
harmonics are asymmetric.
Theory of absorption spectroscopy
25
-10 -5 0 5 10Normalized Frequency (ν−ν0)/∆ν
Mag
nitu
de (
a.u.
)
-10 -5 0 5 10
Voigt Profile 1f Signal
2f Signal 3f Signal
Figure 2.13 The Voigt profile and its first three harmonic signals vs. normalized frequency
Detection using the signal at the second harmonic (2f) of the modulation frequency is the most
frequently applied method [Philippe 1993], for two reasons: first, the 2f line shape is symmetric
and peaks at line center due to the nature of even function. Second, as mentioned above, the 2f
provides the strongest signal of the even-numbered harmonics.
According to Eqn. (2.37), the second harmonic Fourier coefficient is given by
2 (ν, ) ( cos ) cos(2 )S P X LH a a dππ φ ν θ θ θ
π+−
⋅ ⋅ ⋅= − + ⋅ ⋅∫ (2.40)
The 2f signal depends not only on the transition parameters such as line strength but also depends
on the modulation amplitude “a”. Before further discussion, we first introduce the modulation
depth m, which is an important dimensionless parameter widely used in WMS.
/ 2am
ν=
∆ (2.41)
where a [cm-1] is the modulation amplitude and ∆ν/2 [cm-1] is the half width of the transition.
Chapter 2
26
20x10-3
15
10
5
0
-5
-10
-15
Sig
nal [
a.u.
]
-4 0 4Normalized Frequency (ν−ν0)/∆ν
54321m
m=1 m=2.2 m=3 m=4
Peak at m=2.2
Figure 2.14 Second harmonic line shape and line center peak height for different “m”.
2f line shapes of a Voigt profile are shown in figure 2.14 (left) for 1 4m≤ ≤ . The maximum
amplitude of the 2f signal occurs at line center. The line shape becomes wider as modulation
depth increases. The line center peak height as a function of modulation depth “m” is shown in
figure 2.14 (right). The peak height is maximum at m=2.2, which optimizes the signal-to-noise
ratio.
So far in this section it has been assumed that laser intensity is independent of frequency
modulation. However, for diode lasers it is convenient to modulate the laser wavelength by
modulating the inject current. Injection current modulation also modulates the laser intensity. For
small modulation amplitude, one can assume this intensity modulation is small, however at larger
modulation amplitude, a more detailed treatment is necessary. Philippe and Hanson [Philippe
1993] developed the first model to take into account intensity modulation of the emitted light. A
generalized and analytical theory of Lorentzian profile has been developed and can be found in
refs [Schilt 2003; Kluczynski 1999; Kluczynski 2001; Liu 2004]
Theory of absorption spectroscopy
27
Following the previous development of Eqn (2.32), the laser output intensity can be modified to
include intensity modulation
( )0 0 0I ( ) ( ) cos 2 mt I i f tν π φ= + + (2.42)
where I0(t) is the laser output intensity, Φ is the phase shift between intensity and frequency
modulation, and i0 is the intensity amplitude around the average laser intensity 0I at a given ν .
The final expression for 2f signal at ν is given by [Philippe 1993]
0 02 3 0 2 1S ( ) ( , ) ( ) ( , ) ( , )
2 2i iH I H Hν ν ν ν ν ν ν ν= − ∆ + ∆ − ∆ (2.43)
20x10-3
15
10
5
0
-5
-10
Sig
nal [
a.u.
]
-10 -5 0 5 10Normalized Frequency (ν−ν0)/∆ν
2f with IM 2f without IM
i0/I0=10%
Figure 2.15 Comparison of 2f line shape with and without intensity modulation.
Figure 2.15 shows a comparison of 2f line shape with and without intensity modulation. The 2f
line shape becomes asymmetric due to intensity modulation. Since 1f and 3f both are odd
functions, which are zero at line center, there is little difference of the 2f signal at line center from
intensity modulation. Thus, for an isolated transition at line center, these modulation effects are
Chapter 2
28
negligible. For this reason, in the experiments reported in this thesis, the line center peak height is
measured and the modulation effects are ignored.
Temperature Measurement
2f P
eak
Rat
ioTemperature
2f S
igna
l
Wavelength
Ratio of 2f peak heightyields gas temperatureλ1
λ2 2f peak height
Figure 2.16 Temperature inferred from 2f peak ratio of two different temperature-dependent
transitions
Similar to the direct absorption technique, WMS uses a two-line signal ratio for temperature
measurements (figure 2.17).
1 11 2 1 1 12
2 2 2 2 2 2 2
( cos ) cos 2( ) ( ) ( ) ( )( , , , , )
( ) ( ) ( ) ( ) ( cos ) cos 2f
a dI H I S TR f T X P a laser
I H I S T a d
π
ππ
π
φ ν θ θ θν ν νν ν ν φ ν θ θ θ
+
−+
−
+ ⋅ ⋅= = =
+ ⋅ ⋅
∫∫
(2.44)
As can be seen in the above equation, the 2f peak height ratio is not only a function of
temperature, but also a function of species concentration, modulation amplitude and pressure
through the effects of line shape function. This obviously complicates the temperature
determination. A numerical simulation is used to provide the relationship between temperature
and 2f peak height ratio. This necessitates accurate knowledge of the spectroscopic parameters.
As pointed out earlier, a major limitation of temperature measurement by WMS is the
dependence of multiple parameters. In some cases, this can be eliminated by the use of suitable
modulation amplitude [Liu 2004] or an appropriate line pair. Line selection will be detailed in
chapter 3. In this section, we will discuss the selection of modulation amplitude. As indicated in
figure 2.14, the 2f peak height is most insensitive to modulation depth “m” (thus line width at
Theory of absorption spectroscopy
29
fixed modulation amplitude) when m is near 2.2. Therefore the 2f peak height ratio can be
accurately reduced to line strength ratio [Liu 2004] with the optimized modulation amplitudes
(where m~2.2).
It also should be noted that, for simplicity, many hardware-related parameters are not yet taken
into account. Values for a number of parameters related to the instrument hardware are required
in addition to measurements of the 2f signal itself, including the laser intensity, detector setting,
signal amplification, lock-in setting etc. The usual approach is to calibrate the WMS sensor at a
reference condition to eliminate the dependence of hardware-related parameters.
Species Concentration Measurement
WMS with 2f detection is often used to measure trace species concentration. For precise work,
the 2f sensor may have to be calibrated by the user, which requires careful work. Calibration is
generally performed in either of two ways: from direct absorption, or, from a known gas
concentration.
If the measured species concentration is sufficiently low, the lineshape function become
insensitive to species concentration, and then the species concentration will be directly
proportional to 2f peak height:
2 (ν, )( cos ) cos(2 )
H aXS P L a dπ
π
πφ ν θ θ θ+
−
⋅= −
⋅ ⋅ ⋅ + ⋅ ⋅∫ (2.45)
If calibration and measurement are made at same temperature, the relationship between measured
and calibrated species concentrations is straightforward:
2
2
(ν, )(ν, )
measuremeasure Calib
calib
H aX XH a
= (2.46)
If they are different, numerical integration is necessary to correct for the difference:
2
2
( ) [ ( cos ) cos(2 ) ](ν, )(ν, ) ( ) [ ( cos ) cos(2 ) ]
calib
measure
calib Tmeasuremeasure Calib
calib measure T
S T a dH aX XH a S T a d
ππππ
φ ν θ θ θ
φ ν θ θ θ
+−+−
⋅ + ⋅ ⋅∫= ⋅
⋅ + ⋅ ⋅∫ (2.47)
As previously noted, the 2f peak height is used for the temperature and concentration
measurements. Thus, it is not always necessary to obtain the entire 2f line shape. Similar to the
Chapter 2
30
direct absorption technique, a fixed-wavelength technique can also be used with WMS.
Compared with scanned-wavelength WMS, fixed-wavelength WMS can achieve much faster
measurement response.
The fundamental principles for direct absorption spectroscopy and wavelength modulation
spectroscopy are described in this chapter. Sensor design for both techniques will be discussed in
chapter 3. The work in chapter 4 will illustrate the use of scanned-wavelength direct absorption
for temperature sensing in combustion flame, and the work in chapter 5 will incorporate scanned-
wavelength WMS to provide real-time temperature measurements. Chapter 6 will take advantage
of the real-time capability of this scanned-wavelength modulation spectroscopy sensor for
combustion control applications.
31
Chapter 3: Development of design rules for absorption-based
sensors
An important step in the sensor design is the line selection. A proper selection of the line can
optimize the accuracy and performance of the sensor. The HITRAN (High Resolution
Transmission Molecular Absorption Database) database consists of quantitative spectroscopic
parameters for many of the small molecular constituents of atmosphere, and provides an
important tool for sensor design. This chapter first introduces the characteristics of the HITRAN
database, and then discusses the general strategy and spectroscopic criteria for selecting optimum
wavelength regions and absorption line combinations for a two-line temperature sensor. The
concepts developed here make it possible to design an absorption sensor that can achieve
accuracy and performance goals for a wide range of applications. As an example, the optimum
transitions are selected for a high-pressure temperature (T) sensor using the ratio of two
absorption lines for practical application during compression in an internal combustion engine. In
the final section, several important multiplexing techniques are presented.
3.1 Spectroscopic database (HITRAN)
The HITRAN database [Rothman 2003] began in the 1960s as an effort to model atmospheric
transmission of light by the Air Force Cambridge Research Laboratories (AFCRL). This database
has grown over the past forty years, and the latest HITRAN EDITION 2004 includes 1,789,569
spectral lines and 34,001,811 spectroscopic parameters for 39 molecules. The most important
spectral parameters contained are line position, line intensity, lower state energy, air-broadened
halfwidth, self-broadened halfwidth, temperature-dependence coefficient for air-broadened
halfwidth, air-pressure shift and quantum numbers. These parameters are frequently used by
many researchers to perform spectroscopic theoretical modeling and simulate practical
experiments.
The following table describes the format of spectroscopic parameters in the HITRAN 2004
database: [Rothman 2003]
M I ν S A γ-air γ-self E” n δ 1 1 7154.354 1.55E-23 1.37E-03 .0321 .1770 1789.0428 .53 .01459
V’ V” Q’ Q” Ierr Iref * (flag) g’ g” 1 0 1 0 0 0 8 8 0 8 8 1 354543 311930 3 0.0 0.0
Chapter 3
32
Where:
M Molecule number (HITRAN chronological assignment) I Isotopologue number (Ordering by terrestrial abundance) ν Vacuum wavenumber [cm-1] S Intensity [cm-1/(molecule-cm-2)] at standard 296 K A Einstein-A coefficient [s-1]
γ-air Air-broadened halfwidth (HWHM at 296K) [cm-1/atm] γ-self Self-broadened halfwidth (HWHM at 296K) [cm-1/atm]
E” Lower-state energy [cm-1] n Temperature-dependence coefficient (for γ-air) δ Air-pressure shift [cm-1/atm]
V’ Upper-state “global” quanta V” Lower-state “global” quanta Q’ Upper-state “local” quanta Q” Lower-state “local” quanta Ierr Uncertainty indices (Accuracy for ν, S, γ-air, γ-self, N, δ) Iref Reference indices (References for ν, S, γ-air, γ-self, N, δ)
* (flag) Flag (Pointer to program and data for the case of line mixing) g’ Statistical weight of the upper state g” Statistical weight of the lower state
Note that the units for intensity in HITRAN are cm-1/(molecule-cm-2) at the standard HITRAN
temperature (296 K). This intensity unit [cm-1/(molecule-cm-2)= cm-2/(molecule-cm-3)] is the line
strength [in units of cm-2] normalized by concentration [in units of molecule-cm-3]. For many
applications, it is more convenient to use the unit cm-2/atm, which is the line strength [in units of
cm-2] normalized by pressure [in units of atm]. The relationship between these two units is
1 2 3
2 [ /( )] [ ][ / ][ ]
i
i
S cm molecule cm n molecule cmS cm atmp atm
− − −− − −
= (3.1)
This can be further simplified by the ideal gas equation to
21 3 1 2
2 7.339 10 [( ) / ] [ /( )][ / ][ ]
molecule cm K atm S cm molecule cmS cm atmT K
− − −− × − −
= (3.2)
This equation can be used to convert the HITRAN entry for S(296K) to S(cm-2/atm) at 296 K.
Calculations of S(T), cm-2/atm, are then done with Eqn. (2.23).
Development of design rules for absorption-based sensors
33
3.2 Design rules of selecting optimum transitions for 2-line T sensor
An important step in sensor design is the line selection. Senor performance can be greatly
improved by selecting optimum transitions. As mentioned above, the HITRAN database contains
1,789,569 transitions; it is a tedious and impractical process to pick a transition line-by-line using
a manual method. An efficient and systematic computer-based method is developed and used here
to select the appropriate transitions. The strategy and spectroscopic criteria for selecting optimum
absorption transitions are discussed in this section. To elucidate useful design rules and concepts,
this section will concentrate on the transition selections for absorption spectroscopy thermometry
based on the two-line absorption technique. The design rules discussed here should prove useful
to those interested in temperature sensing using absorption spectroscopy. However, it should be
pointed out that the selection of transitions for two-line ratio thermometry can be complicated by
many interrelated factors which determine the final sensor performance of a particular line pair.
Among the most important factors that must be considered in the selection of transitions are: (a)
absorption strength, (b) appropriate spectral separation, (c) temperature sensitivity, (d) lack of
interference from nearby transitions, and (e) effects of nonuniformities such as boundary layers. It
should be stressed that the interaction among all these factors has a considerable influence on the
selection process, and thus far no single figure of merit is derived to simplify this step. Thus, the
optimum transitions for the line pair are chosen case-by-case.
There are multiple different spectroscopic criteria one must consider to choose a line pair.
Understanding the definitions of the various spectroscopic parameters and how they affect the
sensor performance will greatly simplify the selection process. Because the two integrated
absorbances are obtained with the same partial pressure of water and same path length, the ratio
of these two integrals reduces simply to the ratio of line strengths, which is given by Equation
(2.25) The relative sensitivity of this ratio to temperature can be obtained by Equation (2.27):
The following selection criteria are developed:
Criterion 1: Both lines need sufficient absorption over the selected temperature
range.
The peak absorption of the transition is
Chapter 3
34
, ,( )v peak i abs v peakS T P x Lα φ= ⋅ ⋅ ⋅ ⋅ (3.3)
where P[atm] is the total static pressure, Si(T) [cm-2/atm] is the line strength, L [cm] is the path
length, φν,peak is the peak value of line-shape function, and xabs is the mole fraction of absorbers.
An empirical approximation to the Voigt profile [Whiting 1976] is used here to calculate the peak
value of the line-shape function.
We assume a noise level (NL), and a desired signal/noise ratio (SNR), which requires that the
peak absorption be greater than (NL)*(SNR). In addition, the peak absorption must be less than
about 0.8 to avoid experimental difficulties associated with “optically-thick” measurements.
For a path length of L (cm) and absorber concentration range between xabs,min and xabs,max at a
pressure of P (atm),
, ,min ,( ) ( ) ( )v peak i abs v peakS T P x L NL SNRα φ= ⋅ ⋅ ⋅ ⋅ ≥ ⋅ (3.4)
, ,max ,( ) 0.8v peak i abs v peakS T P x Lα φ= ⋅ ⋅ ⋅ ⋅ ≤ (3.5)
so that the constraint on the product of line strength and line-shape function becomes
,,min ,max
( ) ( ) 0.8( )i v peakabs abs
NL SNR S TP x L P x L
φ⋅≤ ⋅ ≤
⋅ ⋅ ⋅ ⋅ (3.6)
in the temperature range Tmin ~ Tmax K.
Criterion 2: The absorption ratio should be single-valued with temperature and the
line strengths of the two lines should be similar.
The absorption ratio is best determined if the measurement uncertainty is similar for the two
absorption transitions. A line strength ratio between R=0.2 and R=5 is thus imposed. Although
the limits of R are somewhat arbitrary, this criterion ensures that the measured absorbance using
the two transitions have similar signal-to-noise ratio (SNR).
Development of design rules for absorption-based sensors
35
Criterion 3: The two lines should have sufficiently different lower state energy E″ to
yield an absorption ratio that is sensitive to the probed temperature.
From Equation (2.25), the line strength ratio can be obtained from the ratio of the integrated
absorbance area for two transitions.
1 1
2 2
( )( )( )
S T AR TS T A
= =
(3.7)
This ratio is a function of two integrated absorbance areas A1 and A2, which are measured from
the best Voigt fit to the line-shape profile. The uncertainty of R, represented by the standard
deviation σR, is then calculated using the error propagation equation, [Bevington 1992]
1 2 1 2
2 22 2 2 2
1 2 1 2
2R A A A AR R R RA A A A
σ σ σ σ ∂ ∂ ∂ ∂
≅ + + ∂ ∂ ∂ ∂ (3.8)
The partial derivatives in Equation (3.8) are represented as follows:
11 AR
AR
=∂∂
22 AR
AR
−=∂
(3.9)
Assuming the integrated absorbance areas A1 and A2 are uncorrelated; dR/R can be estimated
using Equation (3.10),
1 2
2 2
1 2
A ARdRR R A A
σ σσ ≈ ≅ +
(3.10)
The sensitivity of line strength ratio to temperature is obtained from Equation (2.27). It is
generally desirable that the temperature sensitivity be as high as possible, resulting in a more
accurate sensor.
If the integrated absorbance can be determined within X%, the criteria to obtain a temperature
accuracy of Y% in the temperature range of Tmin - Tmax K, constrains the minimum lower state
energy difference
Chapter 3
36
" " "1 2 max
/ %* 2 1* */ % 1.4388i
dR R k XE E E T TdT T hc Y
∆ = − ≥ = (cm-1) (3.11)
It can also be seen from this equation that a line pair with a large lower state energy difference is
desired to have high temperature sensitivity.
Criterion 4: The two lines should be free of significant interference from nearby
transitions.
Some transitions will be rejected because of the appearance of adjacent features which can
produce interference. The remaining transitions are regarded as the promising features for
temperature measurement.
Recall that the HITRAN/HITEMP database contains 1,789,569 transitions for all gases, and thus
it is possible that numerous transitions can meet the selection criteria above. This has led to the
development of the following two criteria, either (or both) of which can be applied to find the
most promising line pair.
Criterion 5: The two lines are desired to have same lineshape function.
As was described in the previous chapter, lineshape function is a very important factor in
spectroscopy experiment. In the fixed-wavelength direct absorption and WMS schemes, the peak
ratio is not only the function of temperature, but also depends on pressure and mole fraction due
to the effect of the lineshape function. This leads to complications and potentially added
uncertainty for the measurements. To overcome this difficulty, two lines which have enarly
identical lineshape function can be selected. More specifically, we can examine
HITRAN/HITEMP database to select two lines which have nearly the same air-broadened
halfwidth, self-broadened halfwidth and temperature-dependence coefficients. Thus effects of the
lineshape function can be cancelled in Eqn. (2.29) and (2.44), and the ratio is simplified to a
function of temperature alone.
Development of design rules for absorption-based sensors
37
Criterion 6: The two lines are close enough to be scanned by a single laser.
Two-line absorption can be completed using two separate lasers multiplexed onto a single optical
path. Such multiplexing techniques will be discussed in the 3.4 section. However, these dual (or
multi)-laser methods have some disadvantages with regard to system complexity and cost. A
single-laser concept can offer a number of essential advantages. Using a single-laser allows a
much easier experimental setup and less cost with similar performance compared to the
multiplexing technique. For a single-laser sensor, the line selection process includes criteria to
insure the selected lines are close enough together in wavelength to be encompassed by a single-
laser scan, yet far enough apart to avoid overlap.
In this section, various criteria for line selection have been suggested. From this screening, the
transitions most applicable to temperature sensing emerge. It is clear that studies with different
criteria may yield different line pair choices, and it is hoped that presentation of the current
selection guideline logic will facilitate future investigations.
Chapter 4 and 5 will present the developments of single-laser temperature sensors using the
design rules developed here. In the following sections, the design rules are extended to high
pressure applications using a wavelength-multiplexed strategy.
3.3 Selection of optimum transitions for high pressure T sensor
Although the discussion in section 3.2 is based on direct absorption, the design rules can be easily
extended to other measurement strategies. In this section, the design rules are refined for the
selection of optimum H2O line pairs for a specific application, namely temperature measurements
in the compression phase of an IC engine (e.g., approximately 300-1100 K, 0.5atm-50atm). At
present we limit the selection to transitions in the 1.25-1.65 µm region where telecommunication
lasers and fibers are currently available. The water vapor spectrum in the 1.25-1.65 µm region is
systematically analyzed to find the best absorption transitions for sensitive measurement of in-
cylinder gas temperature over short path lengths for an internal combustion engine application.
The strategy and spectroscopic criteria are discussed for selection of optimum wavelength regions
and absorption line combinations for the time-varying pressures and temperatures expected
during the compression portion of an engine cycle. We have identified 14 water transitions in this
spectral region as promising for this target application. Based on these findings, 16 potential line
Chapter 3
38
pairs of H2O were considered for a wavelength-modulated absorption sensor for in-cylinder gas
temperature during the compression stroke. As part of the sensor development effort, the
expected performance is modeled for a variety of engine cycles.
3.3.1 Motivation
The IC engine is the most common source for vehicle power, and thus improvements in engine
performance to reduce pollutant emissions or increase fuel economy can have a tremendous
impact on the environment and the world’s energy resources. New engine concepts involve lower
temperature combustion to reduce NOx emissions and novel ignition processes to improve fuel
economy. Development of these new concepts would be facilitated by crank angle-resolved in-
cylinder temperature measurements made across a short path-length via an intrusive optical probe.
Spectroscopy-based sensors are attractive for this application, owing to their potentially fast time
response and species specificity, but must be designed to take into account the time-varying
pressure and temperature of typical IC engine cycles. For example, the variable pressure
broadening complicates absorption measurements and introduces varying degrees of overlap of
absorption transitions with neighboring transitions.
1200
1000
800
600
400
Tem
pera
ture
[K]
5040302010Pressure [atm]
72207210720071907180Frequency [cm
-1]
Super-charged intake High EGR
Abs
orba
nce
Figure 3.1 Typical high EGR and Super-charged intake cycles in internal combustion engine and
representative water spectra under two limiting conditions during the cycle.
Development of design rules for absorption-based sensors
39
Figure 3.1 presents two extreme examples of IC engine compression cycles, one with high
temperatures from large exhaust gas recirculation (High EGR) and one with very high pressure
from a super-charged intake. The inset in Figure 3.1 illustrates the pressure broadened spectrum
of a segment of the water vapor spectrum at the limiting pressure values of the super-charged
cycle, i.e. at 1 and 50 atm. The well-resolved water vapor spectrum at 1 atm becomes a blended,
highly overlapped spectrum at 50 atm without the ability to measure the zero-absorption base line
by scanning off-resonance. Second, the speed of an IC engine cycle can be very rapid (>2400
rpm), and a temperature measurement requires a sensor bandwidth of >15kHz to resolve one
degree of crank angle.
10-4
10-3
10-2
10-1
100
101
102
S[cm
-2/a
tm]
87654321Wavelength [µm]
H2O @ 300K
Telecommunications lasers available
1.25-1.65 µm(mature technology)
Quantum Cascade Lasers
QC lasers available5.0-5.4µm and near 7µm(developing technology)
ν2ν1
2ν2
ν2+ν3ν1+ν2
2ν12ν3
ν1+ν3
ν3ν2
ν1
2ν2
ν2+ν3ν1+ν2
2ν12ν3
ν1+ν3
ν3
Figure 3.2. Survey spectra of H2O at 300 K in the 1~8 µm region based on the HITRAN database.
Water is an attractive species for TDL thermometry, as it is naturally present in humid air, it is
one of the primary hydrocarbon combustion products, and it has a strong absorption spectrum.
Figure 3.2 graphically depicts the line strengths of water over a range of wavelengths from 1 to 8
µm at a temperature of 300 K. The many strong absorption transitions provide numerous options
for measurements in combustion environments. Although mid-infrared (MIR) transitions in the
fundamental vibration bands are 10 times stronger than the combination and overtone vibrational
transitions in the near infrared (NIR), laser and fiber technology is less mature in the MIR than in
the NIR. Accordingly, in the current work we limit the choice of transitions to the 1.25~1.65 µm
region (7788 H2O transitions in HITRAN 2004) where the telecommunication lasers and fiber
optics technology are well-developed and readily available. Telecommunication laser and fiber
technologies offer many attractive features, including advanced laser performance, simple
installation, easy laser beam alignment, improved ruggedness and flexibility, and reduced overall
system cost.
Chapter 3
40
There are two alternative strategies for fast time-resolved absorption measurements: a
wavelength-agile (wide-scan tuning at high speed) strategy and a wavelength-multiplexed
strategy. The wavelength-agile strategy is described in the literature (see papers 4-8), and Sanders
et al [Sanders 2003] have recently applied wavelength-agile strategy to IC engine compression
with some success. Here we examine the alternative strategy, wavelength-multiplexed absorption
for near-real-time temperature sensing.
The combination and overtone vibrational absorption transitions in the NIR have only a few
percent absorption across the diameter of a modern IC engine cylinder for intake air with a
natural range of humidity, and proportionally less absorption for measurements made over
reduced lengths (e.g. 1 cm) to provide spatially resolved results. Therefore, we examine the use of
wavelength modulation spectroscopy (WMS) to improve the signal-to-noise ratio (SNR) of the
NIR absorption measurements. WMS is an extremely sensitive technique which has been
successfully demonstrated in gas-sensing applications. [Liu 2004; Reid 1981; Silver 1999; Hovde
2001; Aizawa 2001; Bullock 1997] The time-varying pressure in an IC engine cycle produces a
time-varying absorption line width, and hence the WMS absorption measurements will have a
time-varying absorption modulation depth. The target transitions also will have time-varying
blending with their neighbors. Thus, a simulation computer program is developed to predict the
time-varying WMS signal as a function of crank angle for a specific variation of temperature and
pressure during the compression stroke. The WMS measurement principles needed for the
variable pressure IC engine application are examined in chapter 2.
3.3.2 Line selection criteria
The following selection criteria are developed:
Criterion 1: Both lines need sufficient absorption over the entire cycle.
The peak absorbance of a single H2O transition is
2, ,( )v peak i H O v peakS T P x Lα φ= ⋅ ⋅ ⋅ ⋅ (3.19) where , peakνφ is the peak value of line-shape function and XH2O is the mole fraction of water. An
empirical approximation to the Voigt profile [Whiting 1976] is used to calculate the peak value of
the line-shape function.
Development of design rules for absorption-based sensors
41
We assume a minimum detectable absorbance of 2 × 10-4, which together with a desired
signal/noise ratio of 10 requires that the peak absorbance be greater than 2×10-3. Anticipating the
need for spatially resolved measurements, we assume a typical desired path length of L = 1 cm
with intake air at 50% relative humidity, i.e.,
2
3, ,( ) 1 2 10peak i H o peakS T P x cmν να φ −= ⋅ ⋅ ⋅ ⋅ ≥ × (3.20)
in the entire compression cycle with a temperature range of about 300 – 1100 K and a pressure
range of about 0.5 atm – 50 atm. Measurements over longer path lengths have larger absorbance
and can be performed with the same transitions.
A search of the HITRAN2004 database [Rothman 2003] reveals a total of 139 H2O transitions
that provide sufficient absorption in the 1.25-1.65 µm NIR region, over the range of temperature
and pressure considered.
Criterion 2: The lines should be free of significant interference from nearby
transitions.
For applications such as the compression in IC-engines, the time varying pressure and
temperature produces time-varying absorption line strengths and line-shape functions. Thus, the
degree of interference from neighboring transitions varies with time (crank angle). Of the 139
candidate transitions, 107 are rejected because of significant interference from adjacent features.
For the P/T engine cycles shown in Fig. 3.1, collisional broadening dominates the line-shape. The
collisional broadening of a typical water vapor transition is airγ ~0.05cm-1/atm at 300K; thus, if
the candidate line has a neighboring transition within 2.5 cm-1, there will be significant overlap at
the highest pressures. The degree of interference depends on the relative line strengths during the
cycle. The simplest selection model would only retain lines which are completely isolated (no
neighbors within 2.5 cm-1; however this design rule is too restrictive. For example, it is possible
to have constructive interference. If the neighboring line has an E” nearly the same as the
candidate transition, the pressure broadening can increase the total absorption of the pressure-
blended lines. On the other hand, if the neighboring lines have quite different E” the pressure-
blended absorption feature has a very different temperature dependence. Therefore, we examined
simulated spectra for all of the candidate transitions at selected P/T points in the example engine
compression cycle and excluded candidates with significant interference from neighbor
Chapter 3
42
transitions, with one exception: 1) if the two lines are close enough together (~0.4 cm-1) that they
blend into a single feature at modest pressures and 2) if the lower state energy of the neighbor is
similar the candidate line is not eliminated for interference. An example of such blended lines
will be discussed in detail below. Criterion 2 reduces the potential number of candidates to 32
lines.
Criterion 3: The 2f signal of absorption lines should have high SNR over the
compression cycle.
The strongest 2f signal is obtained when modulation amplitude “a” is equal to 2.2 times the half-
width of the absorbance line. Since the modulation depth “m” will be much less than 2.2 for high
pressure applications, it is desirable to increase “a” to improve SNR. Our past laboratory
experiments [Liu 2004] determine that “a” values as large as 0.8 cm-1 can be achieved with
modern telecommunication diode lasers at the needed modulation frequency (~ 80 kHz). Hence
the simulations are based on fixed modulation amplitude “a” = 0.8 cm-1. For this value of “a”, the
2f signal strength is sacrificed at the low pressure end due to the over-modulation (m>2.2) in an
attempt to extract useful signal at the highest pressures due to under-modulation. Thus, a proper
choice of “a” is essential for the sensor performance and should be matched with the target engine
cycle.
The 2f signals of the 32 candidate lines are calculated over the high EGR and super-charged
intake compression cycle using a modulation amplitude a=0.8cm-1. From past laboratory
experiments [Liu 2004] we determine an optimistic detection limit to be about 5×10-5 in the units
of Equation (3.21) with a laser power of 5 mW, which is used in the simulation.
2(ν, ) cos(2 )( ) ( cos )i ii
P X LH a dS T a θ θ
πφ ν θ
ππ
⋅ ⋅=− ⋅ ⋅+∑
+−∫ (3.21)
A desired SNR ≥ 10 is imposed during the compression intake. There are 14 candidate lines
(listed in table 3.1) which satisfy criteria 1-3. Lines 1 and 2 (and lines 11 and 12) are only
separated by 0.33 cm-1 (0.18 cm-1). These two pairs of transitions are examples of the pressure-
blended lines with similar E” mentioned above. The other ten candidate transitions in Table 3.1
are all well-isolated individual transitions which could be used in pressure-broadened applications.
Figures 3.3 illustrate the simulated 2f signals for nine of the fourteen lines (1, 2, 5, 8, 10, 11, 12,
13, and 14) for the compression portion of the high EGR (panels a and b) and the super-charged
Development of design rules for absorption-based sensors
43
(panels c and d) cycles using a modulation amplitude of 0.8cm-1. Note that lines 1&2 in panels a
and c have the lowest internal energy, and thus have the largest signals at low temperature. The
highest temperatures in the compression stroke of the IC engine correspond to the largest
pressures, where these two lines have been collision-broadened into a single feature. Note that
without this broadening the signal size at the highest temperatures would be too small to retain
these low E” lines as potential candidates. The simulated signals from the collisional blending of
a pair of higher E” lines (11 and 12) is also illustrated in Fig 3.3 panels a and c. These lines do
not have significantly different signals than for the isolated line examples shown in Fig. 3.3
panels b and d.
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
he
ight
800700600500400Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
800700600500400Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
a
dc
b25x10
-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
he
ight
800700600500400Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
800700600500400Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
he
ight
800700600500400Temperature [K]
(1) 1405 nm (2) 1405 nm (11) 1429 nm (12) 1429 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
1000800600Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
High EGR
a = 0.8 cm-1
SNR = 10
25x10-3
20
15
10
5
0
2f p
eak
heig
ht
800700600500400Temperature [K]
(5) 1388 nm (8) 1350 nm (10) 1392 nm (13) 1347 nm (14) 1345 nm
Super-charged intake
a = 0.8 cm-1
SNR = 10
a
dc
b
Figure 3.3 The simulated 2f signals for nine of the fourteen lines (1, 2, 5, 8, 10, 11, 12, 13, and 14)
for the compression portion of the high EGR (panels a and b) and the supercharged (panels c and
d) cycles using a modulation amplitude of 0.8cm-1.
Chapter 3
44
Criterion 4: Minimum temperature sensitivity of 2 line ratio.
The 14 candidate lines in Table 3.1 are combined into line pairs for two-line ratio thermometry as
shown in Eqn. (2.44). The temperature sensitivity is best for the largest difference
in " "1 2"E E E∆ = − . [Zhou 2003] The minimum "E∆ we accept is 300 cm-1. This reduces the
number of potential line pairs from 91 line pairs to 54 line pairs as indicated in Table 3.1. The
collisionally blended lines 1&2 and 11&12 are each considered as a single feature, reducing the
number of potential line pairs to 37.
Table 3.1 The 14 candidate lines.
Line E” [cm-1] λ [nm] f [cm-1] Potential line pairs
(involving this line as Low E” line)
1 399 1404.94 7117.75 9 2 447 1405.00 7117.42 3 586 1409.51 7094.68 6 4 649 1412.32 7080.57 4 5 742 1388.14 7203.89 4 6 744 1414.13 7071.48 4 7 782 1417.59 7054.23 3 8 920 1350.42 7405.11 3 9 920 1418.91 7047.68 3
10 1045 1391.67 7185.60 1 11 1294 1428.98 6997.99 12 1327 1428.95 6998.17 13 1327 1346.59 7426.14 14 1558 1344.88 7435.62
Criterion 5: The 2f signal ratio should be single-valued with temperature.
From Equation (2.44), the ratio of 2f signals is closely related to ratio of the individual line
strengths; note that this simple picture neglects blending of neighboring features, and thus a more
complex simulation is required for quantitative comparison. Simulations of the 2f signals were
calculated by our computer model including pressure broadening assuming air colliders. The
quantitative spectroscopy is taken from the 2004 version of the HITRAN database [Rothman
Development of design rules for absorption-based sensors
45
2003] for the candidate lines (and their neighbors) and calculated for the P & T expected in the
extreme compression cycles (high EGR and super-charged intake) shown in figure 3.1. Ratios of
these signals versus temperature (and pressure) during these candidate cycles were examined to
estimate performance as a temperature sensor. Based on these simulations, 2 line pairs are
rejected due to the multi-valued behavior with temperature. There remain a total of 35 line pairs
which satisfy criteria 1-5.
Criterion 6: The 2f signal ratio should have good temperature sensitivity and
measurement accuracy.
Using the spectroscopic parameters data in HITRAN, the 2f WMS signals were calculated for the
two compression cycles to evaluate the potential expected uncertainty of sensors based on the 35
line pairs. We estimate the temperature uncertainty as follows:
First, the 2f peak ratio can be obtained from
(2 ) "(2 ) "
peak
peak
f HHighE LineRf LLowE Line
= =
(3.22)
Second, the uncertainty of R, represented by the standard deviation Rσ , is then calculated
[Bevington 1992]:
2 22 2 2 22R H L HL
R R R RH L H L
σ σ σ σ∂ ∂ ∂ ∂ ≅ + + ∂ ∂ ∂ ∂ (3.23)
Where Hσ and Lσ are the estimates for potential measurement uncertainties of 2f peak heights
(high E” line & low E” line). We assume Hσ = Lσ = Mσ , where Mσ is the measured noise
floor (5×10-5 in units defined earlier). The partial derivatives in Equation (3.23) are represented
as follows:
R RH H
∂=
∂
R RL L
∂= −
∂ (3.24)
Assuming the 2f peak heights H and L are uncorrelated; The measurement uncertainty Rσ
becomes,
21 1 21 1H RR M ML L Lσ σ σ = + = +
(3.25)
Chapter 3
46
Thus the measured temperature uncertainty Tσ can be estimated
1 1 21/ / (2 ) "peak
R RT MdR dT dR dT f LowE Line
σσ σ≈ = +
(3.26)
The potential uncertainty of sensors based on the promising 35 line pairs are simulated using Eqn.
(3.26) for both high EGR and super-charged intake cycles. If we require the maximum
temperature uncertainty to be less than 30 K and average temperature uncertainty less than 10 K
during both high EGR and super-charged intake cycle, the potential line pairs are reduced to 16,
as summarized in table 3.2. These are regarded as the most promising water vapor features for
temperature measurement in the compression phase of the IC engine using the selection criteria
(1-6) noted above. Obviously, studies with different criteria may yield other line pair choices.
However, it is hoped that presentation of the current selection guidelines will facilitate future
investigations of other H2O transitions and other species.
Table 3.2 The 16 attractive line pairs.
Super-charged intake High EGR Line pair
Low E” Line
High E” Line Avg.
Tσ [K] Max Tσ [K] Avg. Tσ [K]
Max Tσ [K]
1 Line 1 Line 5 6.0 29.5 3.8 11.7 2 Line 1 Line 8 4.1 13.7 3.3 7.6 3 Line 1 Line 10 4.9 21.7 3.0 7.9 4 Line 1 Line 11 3.5 12.0 2.4 6.1 5 Line 1 Line 13 3.3 11.5 2.3 5.1 6 Line 3 Line 8 7.2 20.0 7.0 27.8 7 Line 3 Line 11 4.7 15.1 3.4 7.8 8 Line 3 Line 13 4.1 11.7 3.2 6.4 9 Line 4 Line 14 6.6 16.7 5.1 9.5 10 Line 5 Line 11 6.2 14.3 4.5 7.9 11 Line 5 Line 13 5.5 15.7 4.2 8.3 12 Line 5 Line 14 5.5 26.8 3.7 10.2 13 Line 8 Line 14 6.6 17.1 7.2 12.3 14 Line 9 Line 11 6.4 16.2 5.0 9.7 15 Line 9 Line 13 5.9 17.8 5.0 12.0 16 Line 9 Line 14 5.7 29.1 4.1 12.7
Development of design rules for absorption-based sensors
47
Based on the simulation results, these 16 line pairs have good temperature sensitivity and
accuracy for the entire compression cycle, and hence they show good potential for temperature
measurements during the compression phase of an internal combustion engine. Our results also
show that there is no single line pair which is best for all the conditions. Therefore, the optimum
line pair should be chosen case-by-case. For example, if we pick the high temperature end of high
EGR cycle (2 atm < P < 30 atm & 600 K < T <1050 K), the line pairs 5 and 12, are the most
promising candidates. Figure 3.4 shows in panel (a) the ratio of the 2f signals for these two line
pairs during compression for the high EGR cycle. The temperature sensitivity from this ratio is
nearly constant for the entire compression stroke as illustrated in panel (b) of Fig. 3.4. The
calculated uncertainty in temperature using the evaluation criteria above shows both line pairs are
quite promising for P and T values near the middle of the compression stroke. The uncertainty
becomes larger at the low T portion of the cycle where the high E” line has very low signals and
at the high P/T portion of the cycle where the low E” transition has low signal. As mentioned
above, the 2f signal at the low pressure end of the cycles is sacrificed by using large “a” (0.8 cm-1)
to achieve best performance at the high pressure portion of the compression cycles. Instead of the
fixed “a” method, other approaches such as the use of variable “a” or use of additional lines (e.g.
three rather than two) may be desired to improve sensor accuracy and performance during the
entire cycles. However, for the high EGR case, these two line pairs (5 and 12) have the optimum
performance.
If we pick the compression stroke for the super-charged intake cycle (10 atm < P < 50 atm & 600
K < T < 850 K), there are five lines pairs (5, 11, 12, 13, 15) which have nearly the same modeled
performance, as shown in figure 3.5. These line pairs have predicted temperature uncertainties
less than 10 K during compression. Here the simulations show quite similar behavior for the
signal ratio, temperature sensitivity, and predicted uncertainty, limited by signal for the over-
modulated low temperature portion of the cycle and similarly limited by signal at the under-
modulated high pressure portion of the cycle.
Chapter 3
48
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Rat
io
1000900800700600500Temperature [K]
Line pair 5 Line pair 12
20x10-3
15
10
5
0
Sen
sitiv
ity (
dR/R
)/dT
1000800600Temperature [K]
Line pair 5 Line pair 12
20
15
10
5
0
σ T [K
]
1000900800700600500Temperature [K]
Line pair 5 Line pair 12
Figure 3.4. High EGR compression cycle: (a) Simulated 2f ratio, (b) temperature sensitivity, and (c) temperature uncertainty for the line pair 2 and 5 as a function of pressure/temperature. (a=0.8cm-1)
a
b
c
Development of design rules for absorption-based sensors
49
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Rat
io
800700600500400Temperature [K]
Super-charged intake
a = 0.8 cm-1 Line pair 5
Line pair 11 Line pair 12 Line pair 13 Line pair 15
20x10-3
15
10
5
0
Sen
sitiv
ity (d
R/R
)/dT
800700600500400Temperature [K]
Super-charged intake
a = 0.8 cm-1
Line pair 5 Line pair 11 Line pair 12 Line pair 13 Line pair 15
40
30
20
10
0
σ T [K
]
800700600500400Temperature [K]
Super-charged intake
a = 0.8 cm-1
Line pair 5 Line pair 11 Line pair 12 Line pair 13 Line pair 15
a
b
c
Figure 3.5. Super-charged intake compression cycle: (a) Simulated 2f ratio, (b) temperature sensitivity, and (c) temperature uncertainty for the line pair 5, 11, 12, 13 and 15 as a function of pressure/temperature. (a=0.8cm-1)
It is tempting to use simulations to further reduce the number potential line pairs in Table 3.2.
However, the quantitative uncertainty in the HITRAN database limits the value of further down-
selection. Even though HITRAN 2004 has significantly improved the quantitative spectroscopy
Chapter 3
50
for the 2ν1, 2ν3, and ν1+ν3 overtone and combination bands of H2O, important differences
between database and measured data for line strength and pressure broadening persist [Liu 2005].
Therefore further reduction of the number of candidate line pairs in Table 3.2 awaits confirmation
of the quantitative spectroscopic data for the candidate lines and their neighbors. It should also be
noted that intensity modulation effects, pressure shifts of line position, and optimum selection of
modulation depth, are not considered in the current study, and these may also impact selection of
optimum pairs for specific applications.
It is important to emphasize that this chapter concentrates on the temperature measurement
technology based on two-line water vapor, fixed “a” 2f spectroscopy. Building on this research,
however, temperature strategies based on more than two lines, variable “a” and other species can
be easily derived for other applications (such as fired engine cycles with higher T). The line
selection criteria and logic developed here should prove useful to those interested in temperature
sensing using absorption spectroscopy.
3.3.3 Summary
The optimum selection of the H2O lines for TDL absorption-based temperature measurements in
an internal combustion engine was investigated. The strategies and criteria to select optimum
water features in the 1.25-1.65 micron wavelength range have been detailed. Systematic
examination of models of the water vapor spectrum yield 14 candidate NIR water transitions
which can be combined into 16 attractive line pairs for in-cylinder gas temperature measurements
during compression for internal combustion engines. Simulations show that these line pairs have
good potential for TDL thermometry during the compression cycle for internal combustion
engines.
3.4 Multiplexing techniques
As discussed earlier, the two-line thermometry technique can be accomplished with the
combination of two laser beams along the same measurement path. There are three commonly
used techniques for such architecture: Wavelength Division Multiplexing (WDM), Time Division
Multiplexing (TDM) and Frequency Division Multiplexing (FDM), each with its individual
advantages and limitations. These techniques will be described briefly here, mainly to serve as a
guide for optimum selection for a given application.
Development of design rules for absorption-based sensors
51
3.4.1 Wavelength Division Multiplexing (WDM)
λ2
Measurementpath length
Diode Laser
Collimator
λ1 λ1
λ2
Grating
Combiner Detector
Figure 3.6 Schematic of the wavelength division multiplexing.
An arrangement quite frequently used for wavelength division multiplexing technique is shown in
figure 3.6. The system consists of two lasers operating at different wavelengths. They are
combined using a 1×2 single mode fiber coupler. The output beam is directed through the
measurement path by collimating lens. The two different wavelengths are spatially separated by a
diffraction grating.
m = -2 m = -1
m = -1 m = 0
m = 1
m = 1 m = 2
m = 2
α β
m = -2
m = 3 m = 3
Grating
Incident beam
Figure 3.7 Grating separates the colors in incident light.
A typical diffraction grating consists of a large number of parallel, closely spaced slits (or
grooves). The maximum of the laser intensity is obtained when the reflected beams from different
slits are in phase which occurs when
sin sind d mα β λ+ = (3.27)
Where α is the incident angle, β is the reflection angle, d is the slit separation, m is the order of
diffraction and λ is the wavelength. It can be seen that the output angle will depend on
wavelength. So the laser beams with different wavelength will propagate in different directions,
Chapter 3
52
as shown in figure 3.7. Order zero is the reflected beam. The first order (positive or negative) is
usually employed for detection.
With a given angle of incidence, the angle α, the change of diffraction angle β corresponding to a
small change in wavelength can be obtained by differentiating Eqn (3.27)
cosm
dδβδλ β
=⋅
(3.28)
Therefore, the wavelengths in WDM should be chosen based on the grating specifications and
distance between grating and detector. The main advantage of WDM is its simplicity. However,
WDM can not be used for a large number of wavelengths due to the possible overlap between
different orders reflection or when the wavelengths are too close. In principle this difficulty may
be overcome by Time Division Multiplexing (TDM) and Frequency Division Multiplexing
(FDM).
3.4.2 Time Division Multiplexing (TDM)
5
4
3
2
1
0
Sig
nal [
V]
20x103
151050Time [us]
5
4
3
2
1
0
Laser 2
Laser 1
Figure 3.8 Schematic of Time Division Multiplexing.
Development of design rules for absorption-based sensors
53
The concept of TDM is relatively simple; two signals are alternated in time, shown in figure 3.8
for two lasers. An asset of TDM is its flexibility and relative simple setup. The absorption signals
at two different wavelengths can be measured using one detector. At high repetition rate, however,
accurate timing circuits are usually required for TDM. Another limitation of TDM is that the two
absorption signals are not measured simultaneously; this could lead to measurement uncertainty if
the experimental conditions change rapidly. For WMS sensors, a different scheme, known as
frequency-division multiplexing, is preferred.
3.4.3 Frequency Division Multiplexing (FDM)
λ2
Measurement path length Diode Laser
Collimator
λ1 Combiner Detector
Modulate at f1
Lock-in amplifier Lock-in amplifier
Detector signals to lock-in
Modulate at f2
Figure 3.9 Schematic of the Frequency Division Multiplexing.
A typical setup using Frequency Division Multiplexing (FDM) is shown in figure 3.9. FDM is
commonly used in modulation spectroscopy where each laser signal can be modulated at a
different frequency. The optimum modulation frequencies are chosen based on the minimum
interference between different harmonic signals. The output signal from the detector is fed into a
pair of lock-in amplifiers set to detect 2f1 and 2f2 signal respectively.
FDM offers two advantages. The wavelength separation needed for WDM is not required for
FDM, and simultaneous measurements of two absorption signals can be made with a single
detector. The major limitation of FDM is that it is not applicable for direct absorption
spectroscopy.
Chapter 3
54
55
Chapter 4: Temperature sensing using H2O transitions near 1.8 µm
In this chapter, the water vapor spectrum in the 1-2 µm near-infrared (NIR) region is
systematically analyzed to find the best absorption transitions for sensitive measurement of H2O
concentration and temperature in combustion environments using a single tunable diode laser
with typical distributed feedback (DFB) single-mode scanning range (1 cm-1). The use of a single
laser, even with relatively narrow tuning range, can offer distinct advantages over wavelength-
multiplexing techniques. The strategy and spectroscopic criteria for selecting optimum
wavelength regions and absorption line combinations are then discussed. It should be stressed that
no single figure of merit has been derived to simplify the selection process, and the optimum line
pair is chosen case-by-case. Our investigation reveals that the 1.8 µm spectral region is especially
promising, and we have identified 10 of the best water line pairs in this spectral region for
temperature measurements in flames. Based on these findings, a pair of H2O transitions near 1.8
µm was selected as an example for the design and development of a single-laser sensor for
simultaneously measuring H2O concentration and temperature in atmospheric-pressure flames. As
part of the sensor development effort, fundamental spectroscopic parameters including the line
strength, line-center frequency, and lower state energies of the probed transitions were measured
experimentally to validate and improve the HITRAN database values. We conclude with
demonstration results in a steady and a forced atmospheric-pressure laboratory combustor.
4.1 Water, H2O
Water is an attractive combustion species to monitor, as it is one of the primary products of
hydrocarbon combustion and an excellent indicator of overall combustion efficiency, while
temperature, as a fundamental parameter of combustion systems, determines the overall thermal
efficiency. Simultaneous measurements of H2O concentration and temperature thus hold high
potential for combustion sensing and control.
H2O is a non-linear and triatomic molecule, which consists of two light hydrogen atoms attached
to a relative heavy oxygen atom. Although the structure of a water molecule is simple, its
absorption spectrum is relatively complicated. The structure of water molecule and its allowed
three fundamental vibration modes are shown in figure 4.1 and listed in table 4.1.
Chapter 4
56
H H
O
H H
O
H H
O
(a) Symmetric stretch, ν1(b) Symmetric bend, ν2 (c) Antisymmetric stretch, ν3
Figure 4.1 The structure of a water molecule and its three fundamental vibrations.
Table 4.1 Fundamental vibrations, frequencies, types and description for H2O.
[Banwell 1994]
Vibration Frequency [cm-1] Type Description
ν1 3651.7 || Symmetric stretch
ν 2 1595.0 || Symmetric bend
ν 3 3755.8 ⊥ Antisymmetric stretch
10-5
10-4
10-3
10-2
10-1
100
Line
Stre
ngth
[cm
-2/a
tm]
2.01.91.81.71.61.51.41.31.21.11.0Wavelength [µm]
2ν1, ν1+ν3, 2ν3
ν1+ν2, ν2+ν3
T=1000K HITRAN/HITEMP
2ν1+ν2, ν1+ν2+ν3, ν2+2ν3
Figure 4.2 Survey spectra of H2O at 1000 K in the near-infrared region based on the
HITEMP database.
Temperature sensing using H2O transitions near 1.8 µm
57
Figure 4.2 is the graphical depiction of the near-infrared line strengths of water over a range of
wavelengths from 1 to 2 µm at a temperature of 1000 K. The assignments of the NIR vibrational
absorption spectrum are given in table 4.2.
Table 4.2 Assignments of the NIR vibrational absorption spectrum of water
Wavelength
[nm]
Frequency
[cm-1] Assignment Note
1200 8330 aν1 + ν2 + bν3; a+b=2 1470 6800 aν1 + bν3; a+b=2
1900 5260 aν1 + ν2 + bν3; a+b=1
a and b are integers (0, 1, 2…)
4.2 Development of single-laser T sensor
The primary goal of this work is to elucidate useful design rules for the selection of the optimum
transitions for a robust, single-diode-laser sensor system for real-time measurements of
temperature and water vapor mole fraction in combustion gases at elevated temperature. (The
temperature range of interest is 1000 to 2500 K.) By analyzing the NIR water spectrum in the 1-2
micron range using the criteria we developed in the previous chapter, the 10 most promising (by
our criteria) NIR single-laser water transition pairs are suggested for temperature measurements
in flames; all these pairs lie near 1.8 micron. One of the optimum transition pairs, located near
1.8005 micron (5554 cm-1), is chosen and utilized to develop a prototype single-laser sensor
system. Fundamental spectroscopic measurements of the selected H2O transitions are used to
validate the HITRAN/HITEMP database. [Rothman 1998] The updated spectroscopic parameters
(line strengths, line-center positions, and lower state energies) form the theoretical basis for future
applications of this diode-laser sensor system. Subsequent to the spectroscopic efforts, the utility
of this new sensor is demonstrated in a small-scale laboratory combustor. We conclude that this
system has the desired flexibility, high speed and accuracy to be a useful tool for fundamental and
applied combustion monitoring.
4.2.1 Selection of water line pairs
There have been several previous studies that discuss the selection of transitions for absorption-
based thermometry. Chang et al. [Chang 1991] investigated candidate line pairs for NO
thermometry near 226 nm in the UV region, and Arroyo et al. [Arroyo 1994] identified a useful
Chapter 4
58
H2O line pair near 1.38 µm for H2O thermometry. Nagali and Hanson [Nagali 1997] investigated
a diode laser sensor for monitoring water vapor in high-pressure combustion gases in the 1.3~1.4
µm region. However, a systematic analysis of the broader NIR (1-2 µm) H2O spectrum aimed at
thermometry has not been reported. The design rules discussed here (and in chapter 3) should
prove useful to those interested in temperature sensing using absorption spectroscopy.
A primary objective of this chapter is to discuss the selection of optimum H2O line pairs for
absorption measurements of temperature and water mole fraction in representative combustion
environments (1000~2500 K, P≈1 atm). At present we limit the selection to transitions accessible
within the tuning range of single diode lasers currently available (380nm<λ<2400nm). Following
the selection criteria developed in chapter 3, the line selection process is as follows:
Criterion 1: Both lines need sufficient absorption over the selected temperature
range.
We assume a minimum detectable absorbance of 10-4, which together with a desired signal-to-
noise ratio (SNR) of 10 requires that the peak absorption be greater than 10-3. In addition, the
peak absorption must be less than about 0.8 to avoid experimental difficulties associated with
“optically-thick” measurements.
For a pathlength of 5 cm and a combustion product water vapor mole fraction between 0.01 and
0.3 at a pressure of 1 atm,
3
,,, 105%11)()(2
−≥⋅⋅⋅⋅=⋅⋅⋅⋅= peakipeakoHipeak cmatmTSLxPTS ννν φφα (4.1)
8.05%301)()( ,,, 2≤⋅⋅⋅⋅=⋅⋅⋅⋅= peakipeakoHipeak cmatmTSLxPTS ννν φφα (4.2)
so that the constraint on the product of line strength and line-shape function becomes 11
,11 02.0)(53.0 −−−− ⋅≥⋅≥⋅ atmcmTSatmcm peaki νφ (4.3)
in the temperature range 1000 ~ 2500 K.
A total of 856 transitions in the HITEMP [Rothman 1998] database meet the absorption strength
criterion in the 1-2 µm NIR region.
Temperature sensing using H2O transitions near 1.8 µm
59
Criterion 2: The absorption lines lie within a single laser scan, and do not overlap
significantly at atmosphere pressure.
A typical rapid-tuning range of a single-mode DFB diode laser is ~1 cm-1. Hence we require that
the spectral separation of the line pairs must lie between 0.1 and 0.6 cm-1. If line spacing is larger
than 0.6 cm-1, ambiguity in the baseline fit will result in unacceptable uncertainty in the
measurements. If line spacing is smaller than 0.1 cm-1, the two lines will overlap at atmosphere
pressure. Criterion 2 reduces the potential candidates to 339 line pairs.
Criterion 3: The absorption ratio should be single-valued with temperature and the
line strengths of the two lines should be similar.
The absorption ratio is best determined if the measurement uncertainty is similar for the two
absorption transitions. In addition, if one transition is much stronger, the wing of the strong
transition will have obvious influence on the measurement of the weak transition. A line strength
ratio between R=0.2 and R=5 is thus imposed. Although these limits of R are somewhat arbitrary,
this criterion ensures that these two transitions have similar SNR ratio [Nagali 1997]. There are
285 line pairs which satisfy criteria 1-3.
Criterion 4: The two lines should have sufficiently different lower state energy E″ to
yield an absorption ratio that is sensitive to the probed temperature.
If the integrated absorbance can be determined within 4%, a temperature accuracy of 5% in the
temperature range of 1000 ~ 2500 K requires the lower state energy of the two transitions to
differ by at least 2000 cm-1.
1"2
"1
" 20004388.11*2500*
05.02*04.0
// −==≥−=∆ cm
hckT
TdTRdREEEi
(4.4)
There are total 24 line pairs that satisfy criteria 1-4, and they all have good temperature sensitivity
in temperature range 1000 ~ 2500 K.
Chapter 4
60
Criterion 5: The two lines should be free of significant interference from nearby
transitions.
Fourteen of the promising pairs of transitions are rejected because of the appearance of adjacent
interference features when the candidate line pairs are examined for 1000<T<2500K. The
remaining 10 line pairs are regarded as the most promising water vapor features for temperature
measurement in combustion environments using the selection criteria noted above, including a
single laser with a ~1 cm-1 scan range. Table 4.3 summarizes the line selection results.
Table 4.3. Line selection result using the selection criteria in the near-infrared region based on HITEMP.
Transitions between 1 µm and 2 µm 447207
Transitions, satisfying 1 856 Line pairs, satisfying 1,2 339
Line pairs, satisfying 1, 2, 3 285 Line pairs, satisfying 1, 2, 3, 4 24
Line pairs, satisfying 1, 2, 3, 4, 5 10
Table 4.4 Candidate H2O line intensity pairs for measurements of temperature and
water concentration in the near-infrared region based on HITEMP.
Line pair λ [nm] ν [cm-1]
103 S @1000K
[cm-2atm-1] E” [cm-1] deltE”
[cm-1] Sensitivity
Rank Line
Spacing [cm-1]
Notes
1881.23 5315.670 2.1058 2552.881 1881.03 5316.238 3.7153 95.18 2457.70 3 0.568 A, B, C
1863.29 5366.847 1.9957 2433.802 1863.17 5367.196 3.2740 416.21 2017.59 10 0.349 B, C, D
1839.95 5434.922 42.2922 173.37 3 1839.88 5435.150 10.0224 2337.67 2164.31 5 0.228 A, C, D
1822.75 5486.214 10.2280 503.97 4 1822.60 5486.680 2.9274 2552.86 2048.89 7 0.466 B, C
1818.83 5498.025 2.8254 3391.175 1818.78 5498.203 6.4334 610.11 2781.06 1 0.178 A, C
1818.78 5498.203 6.4334 610.11 6 1818.70 5498.427 5.0516 2630.22 2020.11 8 0.224 C
1818.70 5498.427 5.0516 2630.227 1818.51 5498.997 19.2885 610.34 2019.88 9 0.570 B, C
1812.26 5517.987 3.4634 3135.808 1812.16 5518.291 8.1060 661.55 2474.25 2 0.304 A, B, C, D
1810.62 5522.964 4.6316 2818.429 1810.46 5523.455 7.3145 757.78 2060.64 6 0.491 B, C
1800.57 5553.797 2.6036 3314.8810 1800.45 5554.175 9.3542 982.91 2331.97 4 0.378 A, B,C, D, E
Notes: A: good sensitivity B: large separation, may be used at higher pressure
Temperature sensing using H2O transitions near 1.8 µm
61
C: interfering absorption by room air, need purge D: isolated from nearby interference E: verified experimentally
14x10-3
12
10
8
6
4
2
0
Spec
tral a
bsor
ptio
n co
effic
ient
[cm
-1]
5316.85316.0
Frequency [cm-1]
1881.25 1881.00
5367.65367.0
1863.25 1863.00
5435.65435.05434.4 5486.85486.45486.0
1840.001839.75 1822.751822.50
Wavelength [nm]
T=296K T=1000K T=2000K
P=1atmX=0.1
1 2 3 4
14x10-3
12
10
8
6
4
2
0
Spec
tral a
bsor
ptio
n co
effic
ient
[cm
-1]
5499.05498.0
Frequency [cm-1]
1818.75 1818.50
5518.85518.0
1815.0 1812.0
5524.05523.0
1810.50 1810.25
Wavelength [nm]
5554.55554.05553.5
1800.50 1800.30
T=296K T=1000K T=2000K
P=1atmX=0.1
5,6,7 8 9 10This Work
Figure 4.3 Expanded view of absorption spectra for the selected H2O line pairs in the
near-infrared region based on the HITEMP database; evaluated for P=1 atm, 10%
H2O, 90% air.
Chapter 4
62
Figure 4.3 shows the calculated candidate H2O spectra (P=1 atm, 10% H2O and 90% air) based
on the HITEMP database [Rothman 1998]; spectroscopic constants are listed in Table 4.4. Figure
4.3 allows visual inspection of these candidate line pairs. First, note that typically one or both
transitions absorb strongly at room temperature, which makes these line pairs sensitive to
interference from absorption by ambient water vapor in room air. Consequently it may be
necessary to purge the optical path outside the target measurement zone.
6
5
4
3
2
1
0
Line
stre
ngth
Rat
io S
ensi
tivity
(dR
/R)/(
dT/T
)
30002500200015001000500Temperature [K]
Line pair 5 @ 1818.8nm [deltE"=2781.06cm-1] LIne pair10 @ 1800.5nm [deltE"=2331.97cm-1] Line pair 2 @ 1863.2nm [deltE"=2017.59cm-1] Desired minimum sensitivity
Figure 4.4. Calculated temperature sensitivity of line strength ratio as a function of
temperature for line pair 2, 5 and 10 based on the HITEMP database.
The line pairs are also ranked based on their temperature sensitivity in Table 4.4. Figure 4.4
shows the temperature sensitivities of the line strength ratio of line pair 2, 5 and 10 as a function
of temperature, using Equation (3.9) and data from HITEMP. Line pair 5 has the largest
temperature sensitivity because it has the largest difference of lower state energy for the two
candidate transitions. If the sensitivity is greater than 1.065, and if the integrated absorbance of
individual transitions can be determined within an accuracy of 4%, the subsequent error in the
temperature will be less than 5%.
All of the line pairs in Table 4.4 provide excellent time-resolved measurement of temperature
when the absorbance is fit with a set of Voigt lineshapes. However, real-time measurements
Temperature sensing using H2O transitions near 1.8 µm
63
require a data reduction strategy sufficiently simple for rapid temperature computation, e.g. using
ratios of the peak absorption to avoid time consuming fits to the line-shape. Nearly half of our
candidate line pairs (5, 6, 7 and 9) have enough interferences from nearby lines that the direct use
of the peak absorption ratio is insufficient to determine temperature, and a correction scheme is
needed for real time applications. For this reason, we have excluded these line pairs from further
consideration, though it is certainly possible to develop correction algorithms or look-up table
strategies that would allow use of these line pairs for real time applications.
Ideally, the peak absorption ratio should also be insensitive to pressure. Since the HITEMP
database does not provide a value of self-broadening coefficient 2 2H O H Oγ − for most of the
transitions in Table 4.4, the sensitivities of the ratio of peak absorption to pressure and mole
fraction are not calculated here. Since pressure effects could become important for real-time
temperature measurements in practical systems, it is clear that experimental data for line-
broadening are critically needed in support of accurate temperature sensing.
From the above discussion, Figure 4.3, and Table 4.4, we see that some line pairs have good
temperature sensitivity but limited line spacing, thus they are unsuitable for high pressure
applications since the pressure-broadening mechanism will make them indistinguishable, while
others have large line spacing but moderate temperature sensitivity. In addition, interference
absorption by ambient water vapor should be minimized. Considering all these factors, more than
half of the line pairs in Table 4.4, namely line pairs 1, 2, 3, 4, 8 & 10, should be suitable for use
with a single-laser sensor for combustion applications at modest elevated pressures.
Line pairs 8 and 10 are the two most promising choices because of their large temperature range,
good temperature sensitivity, and relative isolation from other neighboring transitions. We have
selected line pair 10 for detailed investigation here due to laser availability. In order to proceed
with development of this potential temperature sensor, we must first validate the
HITRAN/HITEMP data base, as past work has revealed discrepancies in these data bases,
especially at high temperature.
4.2.2 Spectroscopy Experiments, Results and Discussions
Measurement strategies based on absorption spectroscopy techniques require the accurate
determination of important spectroscopic parameters of the probed species. Hence a primary
Chapter 4
64
focus of this initial work is to present survey spectra of H2O near 1.8 µm to validate available
spectroscopic data or obtain improved values for the needed spectroscopic parameters for the
target transitions and their neighbors. These spectroscopic measurements form the basis for the
design of a single diode-laser sensor system for non-intrusive measurements of H2O
concentrations and gas temperature.
To gas manifold andvacuum pump
DFB1800 nm
InGaAsDetector
Heated Quartz Cell
TransmittedIntensity
Wavemeter
InGaAsDetector
Solid Etalon
P 1-100 Torr MKS Baratron
20 cm
Flipper mirrors
FunctionGenerator
Water
LaserController
ParabolicMirror
N - Purged Area2
Figure 4.5 Experimental schematic of the measurement system for determining
spectroscopic parameters.
Figure 4.5 illustrates the experimental arrangement of a diode laser, a heated quartz cell,
appropriate mirrors and lenses, and two InGaAsP detectors for measuring an etalon trace and the
transmitted intensities. One distributed-feedback InGaAsP laser emitting near 1.8 µm is used in
this study. The output from the laser is collimated using a parabolic mirror to reduce etalon
interference. The ILX Lightwave LDC-3900, which includes temperature and current controllers,
drives the diode laser. A function generator is used to ramp the laser injection current and thus
tune the wavelengths of the laser over the desired absorption features. The 20-cm long static cell
is made of quartz with 0.5°-wedged windows mounted at a 3° angle to minimize interference
effects in the transmission signal. The cell temperature is measured with four type-S
thermocouples that are equally spaced along the cell axis. The temperature deviations along the
cell are determined to be <2%. A pressure gauge (MKS Baratron with a full scale deflection of
Temperature sensing using H2O transitions near 1.8 µm
65
100 Torr and accuracy of <1%) is used to measure the cell pressure, and a mechanical pump
evacuates the cell. A solid etalon with a known free spectral range (FSR=2.01 GHz) is used to
convert transmitted intensity data from the time domain to the frequency domain. Distilled liquid
water contained in a flask is used as a source of water vapor to measure spectroscopic parameters.
The flask is pumped down for an hour prior to measurements to remove all gaseous impurities.
The laser, detectors, and optics are enclosed in a nitrogen-purged area to prevent interfering
absorption by room air. The experimental profiles are best fit using Voigt profiles to get
fundamental spectroscopic parameters of the probed water transitions, and the temperature
dependence of the line pair’s intensity ratio.
The needed fundamental spectroscopic parameters, including line strengths, line-center
frequencies and lower state energies, are measured using low-pressure H2O in a heated cell.
Figure 4.6 shows the reduced pair of experimental profiles corresponding to the absorption
features at 944 K. Based on the HITEMP database [Rothman 1998], there are 4 transitions, as
shown in Figure 4.7. “Line 1” and “line 4” are low temperature lines, which are included in
HITRAN and HITEMP. “Line 2” and “line 3” are high temperature lines, which are included
only in HITEMP. The line strengths for these four transitions are plotted versus temperature in
Figure 4.8, as calculated using HITEMP and measured in our heated cell. The agreement between
the inferred E” and the HITEMP value confirms the spectroscopic assignment. For “lines 1, 2, 4”,
the general agreement between measured spectroscopic parameters and those published in the
HITEMP database is good, so we recommend adopting the values listed in the HITEMP database
for further use. Note “line 3” is a weak transition and ignored for this work.
Chapter 4
66
0.20
0.15
0.10
0.05
0.00
Abs
orba
nce
1.21.00.80.60.40.20.0Relative Frequency [cm-1]
-1.0
0.0
1.0R
esid
ual
(%)
Exp DataVoigt fit
T=944KPH2O=17.44 TorrL=40.8 cm
Figure 4.6 Reduced H2O line-shape (line pair #10) recorded in a static cell at T=944
K, PH2O=17.44 Torr.
10x10-3
9
8
7
6
5
4
3
2
1
0
Line
str
engt
h [c
m-2
/atm
]
5554.85554.65554.45554.25554.05553.85553.65553.4Frequency [cm
-1]
Line 1
Line 2
Line 3 Line 4
T=1000K
Line position
Line 2=5553.80 cm-1
(HITEMP)
Line 2=5553.86 cm-1
(Experiment)
Line 3=5553.99 cm-1
(HITEMP)
Line 3=5554.04 cm-1
(Experiment)
Figure 4.7 Line strength of the transitions contributing to line pair #10 near 1.8 µm at
1000 K based on HITEMP parameters.
Temperature sensing using H2O transitions near 1.8 µm
67
24x10-3
22
20
18
16
14
12
10
8
6
4
2
0
Line
stre
ngth
S [c
m-2
/atm
]
30002800260024002200200018001600140012001000800600400
Temperature [K]
3.0x10-3
2.5
2.0
1.5
1.0
S [c
m-2
/atm
]
1000950900850800750Temperature [K]
4.0x10-3
3.5
3.0
2.5
2.0
1.5
1.0
S [c
m-2
/atm
]
800700600500Temperature [K]
Line 1 (HITEMP) Line 2 (HITEMP) Line 3 (HITEMP) Line 4 (HITEMP) Line 1 (Experiment) Line 2 (Experiment) Line 4 (Experiment)
Figure 4.8 Calculated and measured line strengths for the components of line pair
#10 as a function of temperature. “Line 2” is the high temperature transition at
5553.86 cm-1; “Line 1” is the low temperature transition at 5554.18 cm-1.
The line positions are also measured using a wavelength meter. The measured “line 1” and “line
4” position agree with HITEMP within the precision of the IR wavelength meter (0.01 cm-1), but
the measured positions agree less well for “line 2” and “line 3” near 5553.86 cm-1 and 5554.04
cm-1, respectively. We suggest using our measured positions in future work. The spectroscopic
parameters of the selected H2O transitions are listed in Table 4.5.
Table 4.5 Spectroscopic data for the selected H2O line pair.
Line # HITRAN Frequency
[cm-1]
Measured Frequency
[cm-1]
Line strength @ 296 K
[cm-2/atm]
Lower state energy [cm-1]
2 5553.80 5553.86 7.298E-7 3314.883 3 5553.99 5554.04 3.628E-7 3139.505 1 5554.18 5554.18 7.662E-3 982.912 4 5554.21 5554.21 9.200E-3 173.365
Chapter 4
68
At atmospheric pressure “line 1” and “line 4” are blended. The line strength of “line 4” is about
7% of “line 1” at 1000 K and 5% at 1500 K. The effect of “line 4” is included in the data analysis
to improve accuracy; therefore the peak ratio in this case is defined as “line 2”/(“line 1”+“line 4”).
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
Peak
Rat
io R
320028002400200016001200800400Temperature [K]
X=8% X=10% X=12%
Figure 4.9 The ratio of peak absorbance coefficients, Rpeak(line pair #10), calculated
as a function of temperature for various values of water mole fraction at 1 atm.
(based on HITEMP database [Rothman 1998])
The ratio of peak absorption is often used to infer temperature for real-time measurements.
Unlike the line strength ratio, the peak absorption ratio is also dependent on water mole fraction
and pressure via the line-shape function. To illustrate the relative insensitivity of these factors on
the peak absorption ratio of the selected line pair, Figure 4.9 shows the peak absorption ratio as a
function of temperature for values of water mole fraction in the range 8%~12% at a constant
pressure of 1atm. The result show that a 20% change in water mole fraction only leads to a 1%
change in the measured gas temperature. Hence, the peak absorption ratio of this line pair is
relatively independent of water mole fraction, which enables simplified data reduction for real-
time temperature measurements.
Temperature sensing using H2O transitions near 1.8 µm
69
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
Ratio
320028002400200016001200800400Temperature [K]
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
Tem
pera
ture
sen
sitiv
ity (d
R/R
)/(dT
/T)
Peak Ratio Linestrength Ratio
Peak Ratio Sensitivity Linestrength Ratio Sensitivity
Figure 4.10 The ratio of line strength and peak absorbance coefficients and their
sensitivity to temperature versus temperature for the line pair #10.
Figure 4.10 presents the line strength ratio and peak absorption ratio for line pair #10 and their
corresponding temperature sensitivities as a function of temperature. At temperatures below 960
K, the line strength ratio is less than 0.2, which makes accurate measurements of the line strength
ratio difficult. At temperatures above 3300 K, although the sensitivity is still good, the absorption
coefficients of the two transitions becomes quite small. Thus, the selected H2O line pair for
temperature measurement is suitable for use in the temperature range 960~3300 K.
4.3 Combustion Demonstration
4.3.1 Temperature and Concentration Measurements
Figure 4.11 illustrates the arrangement employed for a demonstration measurement in a
laboratory burner. Light from a distributed-feedback InGaAsP diode laser emitting near 1.8 µm is
directed across a flat diffusion flame stabilized on a Hencken burner. The diode laser is
temperature and current controlled (ILX Lightwave LDC-3900), and injection current tuned (SRS
DS345) across the two absorption transitions. The beam path is purged to avoid interference from
Chapter 4
70
ambient water vapor. The flow rates of fuel (C2H4) and air are measured using calibrated flow
meters. Water vapor absorption is measured 1 cm above the 5 cm × 5 cm square flame.
DFB1800 nm
FunctionGenerator
LaserControllerMirror
InGaAsDetector
TransmittedIntensity
Computer
Sine wave
N - Purged Area2 N - Purged Area2
View looking down
Air
C H2 4
Figure 4.11 Schematic diagram of the measurement system applied to the Hencken
burner.
The laser is scanned at 500 Hz across the H2O line pair to record spectrally resolved absorption
line-shapes. The transmitted signal is sampled at 1 MHz, which corresponds to 2000 points in
each laser scan. The incident laser intensity I0 is determined by fitting the regions outside the
absorption lines to a low-order polynomial. Gas temperatures are inferred from the measured line
strength ratio, using Equation (2.26), for each scan (2 ms).
Temperature sensing using H2O transitions near 1.8 µm
71
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
Tem
pera
ture
[K]
5.04.03.02.01.00.0Position [cm]
Thermocouple Temperature distribution (trapezoid) Tpath-averaged (uniform)=1620 K Tpath-averaged (trapezoid)=1760 K
LcenterLedge Ledge
Figure 4.12 Measured temperatures in the burned-gas region above a C2H4-air flame
in a 5 cm ×5 cm Hencken burner.
Results are shown in Figure 4.12 for an air flow rate of 64 liter/min and fuel flow rate of 2.9
liter/min (overall equivalence ratio ~ 0.65). The thermocouple measurements (type S
thermocouples, 5-mil wires) are corrected for radiation loss. [Shaddix 1999] (The typical
correction is 50 K.) The thermocouple is traversed forward and backward to confirm stability of
the flame temperature. Since the boundary layer is not negligible in this case, it is clear that an
assumption of uniform temperature, implicit in the absorption ratio method, will lead to
systematic error. The temperature of primary interest is typically the temperature in the core
region, which is measured to be 1740 K using thermocouple. Under the assumption of uniform
temperature and mole fraction, the temperature inferred from laser absorption data is determined
to be 1620 ± 30 K along this path, i.e. 120 K (7.0%) below the true core temperature, and the
measured water mole fraction is determined to be 7.9 ± 0.3 %, i.e. 0.8% (9.2%) under the
corresponding calculated equilibrium mole fraction (8.7%).
It is of course not necessary to assume uniform conditions along the absorption path as long as
some spatial characteristics of the temperature and absorber concentration are prescribed. For
example, in the present case it is reasonable to assume a trapezoidal-shaped temperature
distribution with a boundary layer thickness estimated either from the observed thermocouple
data or simple mixing-layer analyses. A linear mixing model may be assumed for the water mole
Chapter 4
72
fraction between the combustion products value in the core region and room air humidity at the
edge of the flame so that the water mole fraction profile along the line-of-sight is also a
“trapezoid”.
Using such a simple model, the remaining unknowns, to be inferred from the peak ratio
absorption data are the core temperature (Tcore) and core water mole fraction (Xcore). We may
solve for Tcore and Xcore interactively, recognizing that the integrated absorbance is now given by:
dxTSXPAL
OHiOHi ∫=0 2,2 )(
(4.5)
We first assume an approximate core temperature from the uniform temperature assumption and
solve the integrated absorbance of one transition for the core water mole fraction Xcore. Using this
Xcore value, we solve the integrated absorbance of the other transition for Tcore; this Tcore value is
further used to solve for a new value of the core water mole fraction, and so on. For the flat flame
diffusion burner used here, we converge in 5 iterations to Tcore = 1760 K, only 20 K (1.2%) from
the radiation-corrected thermocouple value, and the water mole fraction in the core region is
about 9.0%, merely 0.3% (3.4%) from the theoretical calculated mole fraction.
This demonstration experiment confirms the sensitivity and potential accuracy of absorption-
based temperature sensing, while also illustrating the potential problems associated with
nonuniform properties along the line-of-sight. For combustion flows of the type studied here, it
may be sufficient to assume an approximate temperature distribution in reducing the data, or to
select line pairs immune to the effect of cold edges [Ouyang 1990]. Under other conditions where
temperature changes significantly along the absorption path and the relative temperature profile is
unknown, the 2-line absorption temperature technique may not yield useful results. In such cases,
it may be attractive to consider use of a larger number of absorption lines, as has been reported by
Sanders et al. [Sanders 2001] for oxygen, where multiple absorption lines were used to determine
the extent of the hot (or cold) regions in the optical path.
Temperature sensing using H2O transitions near 1.8 µm
73
4.3.2 Identification of Acoustic Instabilities
2600
2400
2200
2000
1800
1600
1400
1200
Tem
pera
ture
[K]
3.02.52.01.51.00.50.0Time [s]
2400
2000
1600
1200
800
Tem
pera
ture
[K]
0.800.700.600.50
Time [s]
Hencken BurnerC2H4/AirSpeaker on
200
150
100
50
0
T rm
s [K
]
250200150100500Frequency [Hz]
Amplitude Spectrum
Figure 4.13 Measured temperatures and its power spectrum in the burned region
above the C2H4-air flame.
The laser absorption sensor offers fast time response for line-of-sight measurements making it
well-suited for the detection of combustion instabilities. To illustrate the use of the sensor to
identify acoustic combustion instabilities, a disturbance was introduced in the flame by
modulating the fuel flow with a speaker attached to the bottom of the Hencken burner (see Figure
4.11). The speaker is driven with a 50 Hz sine wave, thus producing an oscillating gas
temperature. Figure 4.13 shows a time series of gas temperature (top panel) and the Fourier
transform (lower panel). The dominant and harmonic modes of the temperature fluctuations are
clearly shown in the power spectrum. The width of each bar is around 0.33 Hz, which
corresponds to the frequency resolution in the determination of the discrete Fourier transform
Chapter 4
74
over a 3 sec sampling interval. Note the prominent acoustic fluctuation frequency of 50 Hz. These
results demonstrate the utility of this sensor for quantitative, accurate identification of acoustic
disturbances. The ability of this H2O absorption sensor to measure concentration and gas
temperature and track fluctuations illustrates the potential of this sensor for real-time monitoring
of combustion.
4.3.3 Closed-loop Control of Mean Temperature
DFB1800 nm
FunctionGenerator
LaserController
ParabolicMirror
InGaAsDetector
TransmittedIntensity
Computer
Fuel flow Controller
Control Signal
N - Purged Area2 N - Purged Area2
View looking down duct
Air
C H2 4
FilterFocusing mir ror
Figure 4.14 Experimental schematic of the measurement system applied to the Hencken burner.
The diode laser temperature sensor was used to control the mean temperature of a C2H4/Air
laboratory flame. Figure 4.14 shows the experimental arrangement: light from a distributed-
feedback InGaAsP diode laser emitting near 1.8 µm is directed across a Hencken burner, and the
transmitted light intensity is detected with an InGaAsP detector. The diode laser is driven with an
ILX Lightwave LDC-3900 module and appropriate temperature and current controllers; the
wavelength is injection current tuned with a ramp from an SRS DS345 function generator.
Reflective optics are used throughout, including the laser collimation, to minimize back
reflections into the laser and reduce etalon interference.
Temperature sensing using H2O transitions near 1.8 µm
75
A uniform, 5-cm square flat-flame is produced with ethylene/air mixing at the surface of the
burner. The entire optical path except for the burner is purged with nitrogen to avoid absorption
from ambient water vapor. Apertures in the 7-cm diameter duct limit the laser beam to a 4 mm
diameter, which passes 2 cm above the burner surface on the diagonal of the square burner
resulting in an optical path length over the burner of 7 cm.
Figure 4.15 shows the strategy for closed-loop control of mean temperature in Hencken burner. A
proportional voltage-controlled solenoid valve is used to adjust the fuel flow rate in the current
experiment. With a constant air stream, adjusting the equivalence ratio varies the flame
temperature. The error signal fed to the control algorithm is obtained from the difference between
the measured temperature value and the desired set point value, Tdesired. The feedback control
signal Vcontrol is thus calculated from the product of the error signal (T-Tdesired) and an adjustable
gain factor G [Furlong 1996]. The gas temperature can be effectively adjusted to the desired
temperature by this closed-loop control system.
Tdesired HenckenBurner
Airm•
Fuelm•
Fuel Flow Controller
1.8 µm Single-laser Sensor
Vcontrol
Vbias
Control System
+ Vadjusted
Tmeasured
T∆
+ _
+
Figure 4.15 Block diagram showing the strategy used for closed-loop control of the mean
temperature.
Chapter 4
76
1800
1700
1600
1500
1400
Tem
pera
ture
[K]
6005004003002001000
Time [s]
Tdesired
Laser Scan=500HzSampling rate=1MHz300-scan (0.6 sec) average
Control On Control Off Control On
Figure 4.16 The temperature response to a desired set-point temperature.
1800
1700
1600
1500
1400
1300
Tem
pera
ture
[K]
1.00.80.60.40.20.0-0.2Time [s]
Tdesired 20-scan(40ms) averaged T 60-scan(120ms) averaged T 100-scan(200ms) averaged T
Laser scan rate=500HzSampling rate=1MHz
Figure 4.17 The response time of the closed-loop control system.
Temperature sensing using H2O transitions near 1.8 µm
77
Figure 4.16 illustrates the measured temperature response to a desired set-point temperature. Note
when the control is turned off, the temperature fluctuation significantly increases, and when the
control is re-applied the temperature recovers its relatively stable state. The standard deviation in
the percentage temperature variation is less than 1% when the control is on (0~200 sec &
400~600 sec ), and nearly doubles when the control is off due to natural combustion instabilities.
Figure 4.17 demonstrates the system response to a step-change in the desired temperature using
different levels of scan averaging. The (1/e) response time of the closed-loop control system is
about 0.15 sec, owing primarily to the finite response time of the fuel flow control.
4.3.4 Time-resolved measurements in a swirl-spray combustor
The 1.8µm, wavelength-scanned, direct-absorption, one-laser temperature sensor was used for
time-resolved measurements in a swirl-spray combustor in the laboratory of Professor Ephraim
Gutmark at the University of Cincinnati. A detailed description of the swirl-stabilized spray
combustor utilized in the present experiment is given elsewhere [Li 2004]. The combustion
chamber has a 100 mm square cross section and is 450 mm long, with flat quartz windows to
provide optical access. The laser beam is angled ~10o horizontally to avoid the etalons from
multiple reflections within the quartz windows. The total path length is 102 mm. The laser is
scanned at 2 kHz across the H2O line pair to record spectrally resolved absorption line shapes.
The transmitted signal is sampled at 1MHz. A laboratory-code (written in LABVIEW) was used
for post-processing data. A 4-scan average is used to improve SNR, which reduces the
temperature data rate to 500 Hz. Temperature is measured for a single axial location at 50 mm
downstream of the nozzle for gaseous fuel (propane) and liquid fuel (ethanol).
Chapter 4
78
Diode laser controller
Functiongenerator
Computer
FilterN2 purge N2 purge
SpeakerSpeaker
Mirror
Mirror
HeNelaser
FlipMirror
1.8 µmDFB laser
Collimator
Detector
Diode laser controller
Functiongenerator
Computer
FilterN2 purge N2 purge
SpeakerSpeaker
Mirror
Mirror
HeNelaser
FlipMirror
1.8 µmDFB laser
Collimator
Detector
Figure 4.18 Experimental schematic of the measurement system applied to the swirl spray
combustor.
A schematic diagram of the experimental setup is shown in Figure 4.18. Light from a distributed-
feedback InGaAsP diode laser emitting near 1.8 µm is directed across a swirl spray combustor,
and the transmitted light intensity is detected with an InGaAsP detector. Although infrared-
sensitive cards are commercially available near 1.8 µm, the laser beam is not easily observed due
to the low power of the laser. The laser beam is coaligned with a “red” HeNe visible laser beam
for alignment purposes, via a flat flip mirror. The diode laser is driven with an ILX Lightwave
LDC-3900 module and appropriate temperature and current controllers; the wavelength is
injection current tuned with a ramp from an SRS DS345 function generator. In the experiment,
most of the laser beam is enclosed by a pipe with N2 flowing to eliminate ambient absorption in
the optical path. The transmitted signals are recorded by NI-6115 DAQ card installed in a PC for
post-processing of data. A bandpass optical filter is added so that the emission noise from flame
is attenuated. Reflective optics are used throughout, including the laser collimation, to minimize
back reflections into the laser and reduce etalon interference.
Temperature sensing using H2O transitions near 1.8 µm
79
Gas Fuel (Propane) Liquid Fuel (Ethanol)
Figure 4.19 Reduced line-shapes for gas and liquid fuel.
Figure 4.19 shows the reduced pair of wavelength-scanned absorption measurements using gas
and liquid fuel in the combustor. Figure 4.19 (left panel) is for an air flow rate of 10 SCFM and
propane flow rate of 10 SLM. Figure 4.19 (right panel) is for an air flow rate of 57.5 SCFM and
ethanol flow rate of 0.1 kg/min. As expected, SNR is reduced in liquid fuel experiments due to
beamsteering effects and unburned liquid droplet interference. Although 4-scan averaging
provides sufficient precision for the current condition, additional averaging is required to improve
SNR under more noisy conditions for more practical flames.
200
150
100
50
0
T2 rm
s [K
2 ]
25020015010050
Frequency [Hz]
2400
2000
1600
1200
Tem
pera
ture
[K]
0.50.40.30.20.1Time [s]
Propane Speaker Off
Laser Scan Rate = 2000Hz4-scan average
(a) Unforced flow
40x10-3
30
20
10
0
Abs
orba
nce
(kL)
2.22.01.81.61.41.2Relative Frequency [cm-1]
Raw Data Voigt fit
40x10-3
30
20
10
0
Abs
orba
nce
(kL)
2.22.01.81.61.41.2Relative Frequency [cm-1]
Raw Data Voigt fit
Chapter 4
80
200
150
100
50
0
T2 rm
s [K
2 ]
250200150100500
Frequency [Hz]
2400
2000
1600
1200
Tem
pera
ture
[K]
0.50.40.30.20.1Time [s]
Speaker On: 100Hz
Laser Scan Rate = 2000Hz4-scan average
Propane
(b) Forced flow
Figure 4.20 Measured temperatures and its power spectrum in the burned region above the Propane-air flame (unforced (a) and forced flow (b)).
To illustrate the use of the sensor to identify acoustic combustion instabilities, a fluctuation was
introduced in the flame by modulating the air flow with two speakers attached to the fuel flow
line, as shown in figure 4.18. Measurements were made for an air flow rate of 10 SCFM and
propane flow rate of 10 SLM. With laser scan rates of 2000Hz and 4-scan averages, a time
resolution of 2ms is achievable. Figure 4.20(a) shows the measured temperature and its power
spectrum when the speaker is off. This unforced flame shows no dominant instability at this
condition. The speakers are driven with a 100 Hz sine wave, thus producing an oscillating gas
temperature. The dominant mode of the temperature fluctuations is clearly shown in figure
4.20(b), which demonstrates the utility of this sensor for quantitative characterization of acoustic
disturbances. It is concluded that fast-response of TDL T sensor allows direct measurement of
power spectrum which is not feasible with thermocouples. The time response of this sensor
suggests the potential for real-time combustion control.
Temperature sensing using H2O transitions near 1.8 µm
81
Figure 4.21 Four sensor positions investigated: 1. Top of flame 2. Under flame 3. Above flame 4. Diagonal
Figure 4.22 Power spectrum at four sensor positions investigated (Propane).
1 2
3 4
1200800400
0
T2 rm
s [K
2 ]
250200150100500
Frequency [Hz]
80604020
0
600400
2000
80604020
0250200150100500
21
3 4
Chapter 4
82
It should be noted that the position of laser beam is quite critical for measurements. To
investigate the best location to observe the temperature fluctuations, four locations at 50 mm, 100
mm, 150 mm downstream of the nozzle and diagnostic direction are examined. As shown in
figure 4.21, location 1 is near the flame tip; location 2 is in the center of the flame; location 3 is
above the flame; and location 4 is diagonal through the flame and its tip. As indicated by the
power spectra in figure 4.22, position 1 (top of flame) is the most sensitive place to observe
temperature fluctuations from fuel modulation. The ability of the 1.8 µm sensor to measure gas
temperature and track fluctuations illustrates the potential of TDL temperature measurements for
combustion sensing and control.
4.4 Summary
A single-diode-laser sensor based on wavelength-scanned absorption was shown to provide rapid
and accurate temperature measurements in a combustion environment. The strategies and criteria
to select optimum water features in the 1-2 micron wavelength range have been detailed. The ten
best NIR water transitions for temperature measurements with a single DFB laser in atmospheric-
pressure flames were determined by systematically analyzing the water spectra in this spectral
region. These optimum line pairs are all in the 1.8~1.9 µm region. The greatest advantage of
these water line pairs is the potential to measure both with a single scan for one diode laser. Even
though different laser specifications may enable other (more widely spaced) line pairs, the line
selection criteria described here may be applied for a quantitative evaluation of potential
transitions. Discrepancies between the experimentally determined spectroscopic parameters and
HITRAN/HITEMP database are also found in this region. Thus, it is absolutely necessary to
verify or experimentally determine the fundamental spectroscopic parameters in the development
of a practical sensor.
A specific line pair (#10) was investigated experimentally, and the pertinent spectroscopic
parameters determined from cell experiments, yielding improvements in the spectroscopic
database. This line pair should be applicable for temperature measurements in the range from 960
to 3300 K. Demonstration experiments were conducted in a steady and a forced Hencken burner.
The presence of cold boundary layers was shown to impact the temperature inferred assuming
uniform conditions, but a simple assumption of a trapezoidal temperature distribution was shown
to recover very accurate values for the core temperature of the flow. Experiments with forced
flames confirmed the utility of the sensor to monitor temperature fluctuations. In addition, the
Temperature sensing using H2O transitions near 1.8 µm
83
sensor is used for closed loop set-point temperature adjustment. Qualitative sensing of
temperature fluctuations and frequencies are also demonstrated in swirl spray combustor. The
results offer clear evidence that this sensor system has the flexibility, speed and accuracy to be a
useful tool for fundamental and applied combustion monitoring and control.
Chapter 4
84
85
Chapter 5: Temperature sensing using H2O transitions near 1.4 µm In this chapter, the development of a diode-laser sensor system is described for non-intrusive
measurements of gas temperature in combustion systems combining scanned-wavelength with
wavelength modulation and 2f detection. The sensor is based on a single diode laser (distributed-
feedback), operating near 1.4 µm and scanned over a spectral range targeting a pair of H2O
absorption transitions (7154.354cm-1 & 7153.748 cm-1) at a 2 kHz repetition rate. The wavelength
is modulated at a frequency f = 500 kHz with modulation amplitude a = 0.056 cm-1. Gas
temperature is inferred from the ratio of the second harmonic signals of the two selected H2O
transitions. The fiber-coupled-single-laser design makes the system compact, rugged, low cost
and simple to assemble. As part of the sensor development effort, fundamental spectroscopic
parameters of the probed transitions including the line strength, self-broadening coefficients, air-
broadening coefficients, and their temperature dependence were determined via laboratory
measurements. The sensor design includes considerations of hardware and software to enable fast
data acquisition and analysis; a temperature readout rate of 2 kHz has been demonstrated for
measurements in a laboratory flame at atmospheric pressure. The combination of scanned-
wavelength and wavelength-modulation minimizes interference from emission and provides a
robust temperature measurement that is useful for combustion control applications.
5.1 Motivation The previous chapter reported a 1.8 µm single-laser temperature sensor based on wavelength-
scanned direct absorption of two adjacent H2O lines. That sensor system has the desired
flexibility, sensitivity, speed and accuracy to be a useful tool for fundamental and applied
combustion monitoring. However, the specific sensor design was not without disadvantages. A
major limitation is that the strong absorption at room temperature of one of the lines employed
makes it sensitive to ambient air interference and subject to interference from cold boundary
layers in the combustor. Although this previous sensor could acquire temperature data at kHz
rates, the analysis of the direct absorption data required post-processing to extract accurate and
precise temperature, which significantly reduces the update rate for control applications.
These limitations are substantially mitigated in the approach reported here utilizing a wavelength-
scanned single-laser sensor architecture combined with wavelength modulation spectroscopy
(WMS) and 2f detection. By using wavelength modulation with 2f detection, the measurement
Chapter 5
86
sensitivity is improved by shifting the detection to higher frequencies where laser excess noise
and detector thermal noise are both much smaller; in addition flow-generated noise outside the
detection bandwidth is suppressed using phase-sensitive detection. [Liu 2004; Fernholz 2002;
Bullock 1997] Furthermore, the data analysis update rate is significantly increased because 2f
detection simplifies the computational analysis needed. The increased sensitivity enables the use
of relatively weak absorption transitions near 1.4 µm where fiber-coupled lasers and fiber
components are readily available from the mature telecommunication laser technology. We select
transitions that originate on energy levels with significant internal energy which have weak
absorption at room temperature, minimizing the interference from ambient air and cold boundary
layer. This new sensor system offers significant advantages for real-time, in situ measurements of
temperature for combustion control.
Water is a primary combustion product and its prevailing and relatively strong absorption spectra
in the near-infrared region make it an ideal species for temperature measurement. By suitable
choice of laser wavelength it is possible to measure temperature using a single diode laser [Zhou
2003]. The use of a single diode laser can greatly simplify the sensor system and reduce cost
compared with wavelength-multiplexing techniques.
Previous combustion control work [Furlong 1996; Furlong 1999] in our laboratory showed that
temperature is a good control variable for complete combustion and reduced emissions in a forced
vortex incinerator. Furlong et al. [Furlong 1999] multiplexed two diode lasers to infer
temperature at a 2 kHz rate from the ratio of two peak absorbances. Although this ratio was a
good control variable, it was contaminated by optical emission and other interference; thus the
absolute temperature determined by this sensor design can be subject to large uncertainty. This
problem generally limits fixed-wavelength sensors to flames without soot, fuel spray, or other
scattering interference. As shown in the previous chapter, a scanned-wavelength approach
utilizing ratios of integrated absorbance offers significant mitigation of these problems at the cost
of more complex data analysis. The laser wavelength is scanned across an absorption feature and
the zero absorption transmission baseline must be inferred. In this chapter we illustrate that this
data acquisition and processing can be significantly simplified by using a wavelength modulation
approach with 2f detection. A continuous temperature measurement readout rate of 2 kHz is
demonstrated for this scanned wavelength sensor.
Temperature Sensing using H2O transitions near 1.4 µm
87
5.2 Development of single-laser T sensor (2f) 5.2.1 Line selection
10-5
10-4
10-3
10-2
10-1
100
Line
Stre
ngth
[cm
-2/a
tm]
2.01.91.81.71.61.51.41.31.21.11.0Wavelength [µm]
2ν1, ν1+ν3, 2ν3
ν1+ν2, ν2+ν3
T=1000K HITRAN 2004
2ν1+ν2, ν1+ν2+ν3, ν2+2ν3
Figure 5.1 Linestrength of H2O in the 1 to 2 µm spectral region at 1000 K (from
HITRAN 2004 database)
Just as seen earlier for the direct absorption sensor, selection of water vapor absorption features is
an important part of the sensor design. Figure 5.1 graphically depicts the near-infrared (NIR) line
strengths of water over a range of wavelengths from 1 to 2 µm at a temperature of 1000K using
the HITRAN 2004 database [Rothman 2003], and the red bar illustrates the region where
telecommunication lasers are available. Transitions are chosen using the following requirements:
Criterion 1: Both lines need sufficient absorption over the selected temperature
range.
We assume a minimum detectable absorbance of 10-4, and a desired SNR of 10, which requires a
peak absorption be greater than 10-3. In addition, the peak absorption should preferably be less
Chapter 5
88
than ~ 0.05 to meet the “weak transition” assumption in Eqn (2.35) associated with wavelength
modulation measurements.
For our laboratory experiment, with a pathlength of 5 cm and a combustion product water vapor
mole fraction between 0.05 and 0.2 at a pressure of 1 atm, we thus require
2
3, , ,( ) ( ) 1 5% 5 10peak i H o peak i peakS T P x L S T atm cmν ν να φ φ −= ⋅ ⋅ ⋅ ⋅ = ⋅ ⋅ ⋅ ⋅ ≥ (5.1)
2, , ,( ) ( ) 1 20% 5 0.05peak i H o peak i peakS T P x L S T atm cmν ν να φ φ= ⋅ ⋅ ⋅ ⋅ = ⋅ ⋅ ⋅ ⋅ ≤ (5.2)
Hence the constraint on the product of line strength and lineshape function becomes 1 1 1 1
,0.05 ( ) 0.004i peakcm atm S T cm atmνφ− − − −⋅ ≥ ⋅ ≥ ⋅ (5.3)
which we apply in the temperature range of current interest, 1000 - 2500 K.
A total of 963 transitions in the HITRAN 2004 database meet the absorption strength criterion in
the 1.0-2.0 µm NIR region.
Criterion 2: Transitions should be relatively free of ambient H2O interference
10-2
10-1
100
101
102
103
104
105
S(T
)/S
(296
)
4000300020001000Temperature [K]
E" = 500 cm-1
E" = 1500 cm-1
E" = 1700 cm-1
E" = 2500 cm-1
Figure 5.2 Linestrength scaled by values at room temperature as a function of
temperature for H2O lines with various lower state energies.
Temperature Sensing using H2O transitions near 1.4 µm
89
As previously mentioned, it is usually necessary to purge outside the target measurement zone
with nitrogen or dry air to eliminate interference from ambient water. The amount of purging
necessary to attain a desired low-humidity set-point depends not only on the ambient humidity
level, but also on the absorption strength of the transition. If a transition has a strong absorption
coefficient at room temperature, greater care has to be taken in the purging process. Even a short,
occasionally unpurged path may reduce accuracy and lead to greater measurement uncertainty.
This difficulty can be easily mitigated by choosing transitions with desirable lower state energies.
The temperature-dependent linestrength is given by Eqn. (2.23). The ratio of line strength at
temperature and line strength at room temperature is given by:
1"
0 0 0
0 0
0
0
( )( ) 1 1exp 1 exp 1 exp( ) ( )
Q T T hc hcS T hcEkTS T Q T T k T T kT
ν ν−
− − = − − − −
(5.4)
The curves in figure 5.2 show line strength scaled by values at room temperature vs. temperature
for H2O lines at various lower state energies. As the lower state energy is raised, the line strength
ratio at elevated temperature is increased. If we require the strength in the temperature range of
1000 - 2500 K to be at least 3 times stronger than its strength at room temperature, the constraint
on minimum lower state energy becomes
1" 1700E cm−≥ (5.5)
Criterion 2 reduces the number of potential candidates to 558 lines. For measurements with cold
thermal boundary layers, this criterion helps make the sensor immune to the effect of cold edges
and improve sensor accuracy. [Ouyang 1990]
Criterion 3: The absorption lines must lie within a single laser scan and not overlap
significantly at atmosphere pressure.
Currently the typical rapid-tuning range of a single-mode DFB diode laser near 1.4 µm is ~2 cm-1.
Hence we require that the spectral separation of the line pairs must lie between 0.3 and 1 cm-1. If
line spacing is larger than ~1 cm-1, modulation amplitude will be limited by the safe current limit
and threshold of the laser at the both ends of the laser scan, this limitation could decrease the
signal-to-noise ratio (SNR) in the measurements. If line spacing is smaller than 0.3 cm-1,
Chapter 5
90
interference on the 2f peak heights will be caused by the overlap of the two absorption features,
making analysis more complicated. Criterion 3 reduces the number of potential H2O line
candidates to 188 line pairs.
Criterion 4: The two lines should have sufficiently different lower state energy E″ to
yield a 2f peak height ratio that is sensitive to the probed temperature.
It was previously mentioned that a line pair with a large difference in lower state energy is
desired to provide high temperature sensitivity [Zhou 2003]. A constraint on minimum lower
state energy difference of 700 cm-1 is proposed for the line selection " " " 1
1 2 700iE E E cm−∆ = − ≥ (5.6)
There are a total of 33 line pairs that satisfy criteria 1-4, and they all have good temperature
sensitivity in the temperature range 1000 ~ 2500 K.
Criterion 5: The two lines should be free of significant interference from nearby
transitions.
The 33 potential line pairs are examined for burnt gas conditions at 296K, 1000K and 2000K to
investigate the potential interference from neighbor transitions. In all, 21 of the promising
transition pairs are rejected because of interference from adjacent absorption features. The
remaining 12 line pairs are regarded as the most promising water vapor features for temperature
measurement in combustion environments using the selection criteria noted above. Table 5.1
summarizes the 12 line pair candidates.
Criterion 6: The wavelength should be in the 1.25-1.65 µm range.
In this work, we chose to limit the wavelength of the transitions to the 1.25-1.65 µm range, where
fiber-coupled telecommunication lasers are commercially available. Sensors built in this
wavelength range can also take advantage of extended fiber-optic component technology for
signal transmission and multiplexing. There are 4 line pairs (pairs 1-4 in Table 1) which satisfy
criteria 1-6. Table 5.2 summarizes the line selection result.
Temperature Sensing using H2O transitions near 1.4 µm
91
Table 5.1 Candidate H2O line intensity pairs for measurements of temperature and water concentration in the 1-2 µm region based on HITTRAN2004.
Line pair λ [nm] ν [cm-1]
103 S @1000K
[cm-2atm-1]
E” [cm-1]
∆E” [cm-1]
Max P [atm]
Line Spacing [cm-1]
Notes
1442.67 6931.592 4.604E-4 1813.22 1 1442.51 6932.352 4.604E-7 3072.73 1259.73 2.2 0.760 B
1397.87 7153.748 5.504E-6 2552.86 2 1397.75 7154.354 3.852E-4 1789.04 763.82 3.7 0.606 B, E
1337.44 7476.949 1.421E-4 1899.01 3 1337.37 7477.366 5.403E-7 3319.45 1420.44 5.0 0.417 A, C, D
1337.44 7476.949 1.421E-4 1899.01 4 1337.30 7477.743 3.954E-6 2746.02 847.01 5.0 0.794 A, C, D
1982.87 5043.193 9.927E-7 3058.40 5 1982.66 5043.738 9.267E-5 2246.88 811.52 2.8 0.545 B
1981.83 5045.831 2.242E-6 2915.89 6 1981.54 5046.576 5.554E-5 2105.87 810.02 8.0 0.745 B, C
1975.41 5062.250 6.234E-6 2904.43 7 1975.11 5063.017 1.141E-4 1960.21 944.22 5.9 0.767 B, C
1967.46 5082.697 5.504E-6 2690.59 8 1967.32 5083.058 1.611E-4 1843.03 847.56 3.7 0.361 B
1852.38 5398.473 3.041E-4 1908.02 9 1852.06 5399.396 1.531E-6 3032.69 1124.67 6.3 0.923 A, C
1764.78 5666.434 7.866E-7 3211.21 10 1764.57 5667.119 9.486E-6 2321.81 889.40 1.2 0.685 D
1764.78 5666.434 7.866E-7 3211.21 11 1764.51 5667.305 2.843E-5 2321.91 889.30 3.4 0.871 D
1764.57 5667.119 9.486E-6 2321.81 12 1764.45 5667.489 2.731E-7 3211.06 889.25 1.5 0.370 D
Notes: A: some interfering absorption by room air from nearby transitions, need purge
B: isolated from nearby interference C: good for high pressure experiments D: close to other promising line pair, measurement accuracy could be improved by averaging
results with nearby line pair E: verified experimentally
Table 5.2 Line selection result using the selection criteria in the near-infrared region based on HITRAN2004.
Transitions between 1.0µm and 2.0 µm 15907
Transitions, satisfying 1 963 Transitions, satisfying 1,2 558
Line pairs, satisfying 1, 2, 3 188 Line pairs, satisfying 1,2, 3, 4 33
Line pairs, satisfying 1,2, 3, 4, 5 12 Line pairs, satisfying 1,2, 3, 4, 5, 6 4
Chapter 5
92
0.010
0.008
0.006
0.004
0.002
0.000
Spe
ctra
l abs
orpt
ion
coef
ficie
nt [c
m-1
]
748074797478747774767475Frequency [cm
-1]
0.010
0.008
0.006
0.004
0.002
0.0006933693269316930 7156715571547153
T = 296 K T = 1000 K T = 2000 K
1 2
3, 4
P = 1 atmX = 10%
Figure 5.3 Expanded view of absorption spectra for the four selected H2O line pairs
in the 1.4 µm region based on the HITRAN2004 database; evaluated for P=1 atm,
10% H2O, 90% air.
Figure 5.3 shows segments of the calculated H2O (10%) absorption spectra for the four line pairs
based on HITRAN2004 [Rothman 2003] parameters; spectroscopic constants are listed in Table 2.
The pair of features at 7154.35cm-1 & 7153.75 cm-1 , labeled pair 2, is selected for several reasons,
as follows. Both features are well resolved at one atmosphere pressure, avoiding interference by
neighboring transitions. Both have similar absorption coefficients and thus will have similar
measurement uncertainty. Additionally, they have sufficiently different lower state energy E” to
yield a high temperature sensitivity while the large E” values of 1789 cm-1 and 2553 cm-1 insure
that the transitions will be strongest at temperatures much larger than room temperature, thereby
minimizing any interference from room temperature water in the measurement path. Finally, the
wavelengths of these two transitions are close enough to be covered in a single laser scan but
separated enough to be isolated from each other. These two transitions thus offer an attractive
opportunity to examine the potential for a practical single laser-based combustion temperature
sensor with potential for control applications.
Temperature Sensing using H2O transitions near 1.4 µm
93
0.010
0.008
0.006
0.004
0.002
0.000
Spe
ctra
l abs
orpt
ion
coef
ficie
nt [c
m-1
]
50445043Frequency [cm
-1]
50465045 50635062
P = 1 atmX = 10% T = 296 K
T = 1000 K T = 2000 K
5 6 7
0.010
0.008
0.006
0.004
0.002
0.000
Spe
ctra
l abs
orpt
ion
coef
ficie
nt [c
m-1
]
50835082Frequency [cm
-1]
53995398 56685667
P = 1 atmX = 10%
T = 296 K T = 1000 K T = 2000 K
8 9 10,11,12
Figure 5.4. Expanded view of absorption spectra for the selected H2O line pairs in
the 1.8µm region based on the HITRAN2004 database; evaluated for P=1 atm, 10%
H2O, 90% air.
It is noteworthy that 70% of the promising transitions (see Table 5.1) were rejected because of the
wavelength limitation in criteria 6. Our spectral simulations show that the water spectrum in the
1.8µm spectral region is stronger and has more isolated features than the 1.4µm spectral region,
and hence holds promise for gas sensing. In the future, the wavelength limitation in criteria 6 can
Chapter 5
94
be avoided when fiber-coupled laser and fiber-optic component technology becomes available in
the 1.8 µm region. Figure 5.4 shows the simulated spectra (P=1 atm, 10% H2O and 90% air) for
the 8 best candidate H2O pairs (pairs 5-12 in Table 5.1) in the 1.8 µm region based on the
HITRAN2004 database.
Table 5.1 includes an estimate of the maximum suitable pressure for a combustion temperature
1500K and a limit of 10% interference from overlap or nearby transitions. Among the 8
promising candidate H2O pairs (pairs 5-12 in Table 5.1) in the 1.8 µm region, line pair 9 has the
best temperature sensitivity, but it suffers some room air water interference from nearby
transitions. Line pair 5 and line pair 8 are suitable for relatively low pressure (< 3 atm)
application due to their relatively small line spacing. Line pair 6 and line pair 7 are well isolated
and good for high pressure application. Line pairs 10, 11 and 12 are so closely spaced that they
can be covered by a single laser scan, and although they are only useful for applications at
atmospheric pressure and below, they provide a unique sensor which can improve measurement
accuracy by averaging results from three line pairs in one laser scan.
5.2.2 Spectroscopic verification
Measurement strategies based on absorption spectroscopy techniques require accurate values of
important spectroscopic parameters of the probed species. Experimental verification of pure water
vapor spectra at low pressure in a heated cell is used to generate (or validate) spectroscopic
databases for line assignment (E”) and line strength S(T). Additional controlled experiments are
performed to provide accurate pressure broadening data.
Figure 5.5 illustrates the experimental arrangement of a diode laser, a heated quartz cell,
appropriate mirrors and lenses, and two InGaAsP detectors that measure the transmitted intensity
and a solid etalon (FSR = 2.01 GHz) to calibrate laser wavelength. One fiber-coupled DFB laser
emitting near 1.4 µm is used in this study. The output from the laser is split into two beams by a
50/50 beam splitter. One beam passes through the cell and a double-pass configuration is used to
improve SNR. An ILX Lightwave LDC-3900 is used to control laser temperature and current. A
function generator is used to vary the laser injection current and thus tune the wavelength of the
laser over the desired absorption features. The 35.6 cm long static cell is made of quartz with
0.5°-wedged windows mounted at a 3° angle to minimize interference effects in the transmission
signal. The cell temperature is measured with three type-K thermocouples that are equally spaced
Temperature Sensing using H2O transitions near 1.4 µm
95
along the cell axis. The temperature deviations along the cell are determined to be <0.5%. Two
pressure gauges (MKS Baratron with a full scale deflection of 100 Torr and 1000 Torr, accuracy
of ±1%) are used to measure the cell pressure, and a mechanical pump evacuates the cell.
Distilled liquid water contained in a flask is used as a source of water vapor to measure
spectroscopic parameters. The flask is pumped down for an hour prior to measurements to
remove all gaseous impurities. The laser, detectors, and optics are enclosed in a nitrogen-purged
area to prevent interfering absorption by room air. The experimental profiles are best-fit using
Voigt profiles to get fundamental spectroscopic parameters of the probed water transitions.
Fiber Coupled DFB 1400nm 14” Sample Path
3-Zone Tube Furnace
Transmission Detector
DAQ Computer
EtalonDetector
Purged with N2
Quartz CellPurged with N2
Mixing Tank
Air Cylinder
Low P Baratron High P Baratron
Mullite Tube
Etalon
Mirror
LaserController
FunctionGenerator
Fiber Coupled DFB 1400nm 14” Sample Path
3-Zone Tube Furnace
Transmission Detector
DAQ Computer
EtalonDetector
Purged with N2
Quartz CellPurged with N2
Mixing Tank
Air Cylinder
Low P Baratron High P Baratron
Mullite Tube
Etalon
Mirror
LaserController
FunctionGenerator
Figure 5.5 Experimental schematic of the measurement system for determining
spectroscopic parameters.
Sample raw data (50-scan average) are shown in Figure 5.6. The laser is scanned at 200Hz with a
sampling rate of 1MHz, and the transmission (I) is normalized by the unattenuated signal (I0).
Since the total scan includes the far wings on both sides of the probed features, the unattenuated
signal could be determined accurately by mathematically fitting the part of the trace with no
absorption to a simple polynomial.
Chapter 5
96
1.4
1.2
1.0
0.8
0.6
Nor
mal
ized
Raw
Sig
nal I
/I 0
4321Time [ms]
Baseline Normalized Raw Signal
T = 951 KP = 15.47 TorrL = 71.12 cm
Laser Scan Rate = 200 HzSampling Rate = 1MHz50-scan average
Figure 5.6 Sample data (50-scan average) obtained from cell experiment at T = 951
K, PH2O = 15.47 Torr.
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Abso
rban
ce
3.02.52.01.5Relative Frequency [cm-1]
-1.0
0.0
1.0
Res
idua
l (%
)
Raw Data Voigt fit
T = 951 KP = 15.47 TorrL = 71.12 cm
High T Line
Low T Line
Figure 5.7 Reduced H2O lineshape recorded in the cell at T = 951 K, PH2O = 15.47
Torr. The low T line is the line with the smaller value of lower-state E”.
Temperature Sensing using H2O transitions near 1.4 µm
97
10x10-3
8
6
4
2
0
Line
stre
ngth
[cm
-2/a
tm]
7155.07154.57154.07153.57153.0Frequency [cm
-1]
T =951 KLow T Line
High T Line 2High T Line 1
Figure 5.8 Line strength of the selected transitions at 1000 K based on HITRAN2004
parameters revealing that the high T line is composed of two lines.
Figure 5.7 shows the experimental data at 951 K. Based on the HITRAN2004 database (figure
5.8), the high T line is composed of two transitions which are labeled as “High T Line 1” and
“High T Line 2” in the figure; these lines have the same lower state energy and very similar
broadening coefficients. The experimental profiles are best-fit using Voigt profiles. The fitting
procedure minimizes the integrated squared difference between the experimental profile and the
Voigt profile.
As mentioned before, the integrated absorbance area of an individual transition is proportional to
partial pressure,
0
ln( )
I dI AS T
P L X P L X
ν
− = =⋅ ⋅ ⋅ ⋅
∫
(5.7)
where P[atm] is the total pressure, L[cm] is the path length, S(T) [cm-2/atm] is the line strength,
X is the mole fraction of absorbing species, and A[cm-1] is the integrated area. Thus, line strength
can be obtained by performing a linear fit on multiple area measurements at various pressures, as
shown in figure 5.9, and using the slope to measure the line strength. This procedure exploits the
improved pressure gage accuracy for P∆ measurements and avoids a “zero” pressure calibration
influence on the line strength uncertainty.
Chapter 5
98
14x10-3
12
10
8
6
4
2
0
Inte
grat
ed A
rea
[cm
-1]
181614121086Pressure [Torr]
Low T Line High T Line 1 High T Line 2
Linestrength inferred from the slope
T = 951 KL = 71.12 cm
Figure 5.9 Measured integrated absorbance area vs. H2O pressure at T=951K for the
“Low T Line”, “High T Line 1” and “High T Line 2”. The line strength can be
calculated from the slope.
10x10-3
8
6
4
2
0
Col
lisio
nal w
idth
∆ν c [
cm-1
]
161412108Pressure [Torr]
T = 951 KL = 71.12 cm
Low T Line High T Line
Self-broadening coefficient inferred from the slope
Figure 5.10 Measured collision width vs. H2O pressure at T=951K for the “Low T
Line”, “High T Line 1” and “High T Line 2”. The self-broadening coefficient can be
calculated from the slope.
Temperature Sensing using H2O transitions near 1.4 µm
99
The self-broadening coefficient, γself, is measured in a manner analogous to the line strength,
since the collision width is linear with pressure.
2 ) (2 )(C selfjjj
P PXν γγ∆ = = ⋅∑ (5.8)
where cν∆ [cm-1] is the collision width (FWHM)
By holding the Doppler width fixed at the appropriate value for the measurement temperature
during Voigt fits, the collision width is extracted from the overall width of the absorption profile
using the inferred (best-fit) Voigt a parameter. The broadening coefficient is determined by
performing a linear fit of the measured collision widths at various pressures, as shown in figure
5.10, and using the slope to calculate the broadening coefficient.
For measurements of air-broadening coefficient, an air-water vapor mixture is made of pure water
vapor and dry air. The air-broadening coefficients can be determined using
(2 )2
(1 )C self
air
X PX P
ν γγ
∆ − ⋅ ⋅=
− ⋅ (5.9)
where cν∆ [cm-1] is the collision width (FWHM), X is the mole fraction of absorbing
species and P[atm] is the total pressure. The uncertainty for the individual line strength,
self-broadening and air-broadening coefficient measurements is estimated to be <3% due to
measurement uncertainties of 1% in the total pressure, and primarily 2% in the area under
each Voigt profile.
Chapter 5
100
8x10-3
6
4
2
0
Line
stre
ngth
[cm
-2/a
tm]
30002500200015001000500Temperature [K]
Low T Line (HITRAN 2004 ) Low T Line (Exp Fit) Low T Line (Experiment)
E"=1789cm-1
Figure 5.11 Calculated and measured line strengths for the “Low T Line” as a
function of temperature.
3.0x10-3
2.5
2.0
1.5
1.0
0.5
0.0
Line
stre
ngth
[cm
-2/a
tm]
30002500200015001000500Temperature [K]
High T Line 1 (HITRAN 2004) High T Line 2 (HITRAN 2004) High T Line 2 (Experiment) Experiment
E"=2553cm-1
Figure 5.12 Calculated and measured line strengths for the “High T Line 1” and
“High T Line 2” as a function of temperature.
Temperature Sensing using H2O transitions near 1.4 µm
101
0.1
2
3
4
5
6
7
89
1
2γse
lf [c
m-1
/atm
]
3 4 5 6 7 8 91000
Temperature [K]
E"=1789cm-1 Low T Line (HITRAN 2004)
Low T Line (Experiment) Experiment
Experiment
2γself=0.302(296/T)0.65
HITRAN 2004
2γself=0.354(296/T)0.50
Figure 5.13 Calculated and measured self-broadening coefficients for the “Low T
Line” as a function of temperature.
0.1
2
3
4
5
6
7
89
1
2γse
lf [c
m-1
/atm
]
3 4 5 6 7 8 91000
Temperature [K]
E"=2553cm-1 High T Line (HITRAN 2004)
High T Line (Experiment) Experiment
Experiment
2γself=0.616(296/T)0.82
HITRAN 2004
2γself=0.598(296/T)0.50
Figure 5.14 Calculated and measured self-broadening coefficients for the “High T
Line” as a function of temperature.
Chapter 5
102
3x10-2
4
5
6
2γai
r [c
m-1
/atm
]
3 4 5 6 7 8 91000
Temperature [K]
Experiment
2γair=0.06212(296/T)0.51803
Exp Data Exp fit HITRAN 2004
HITRAN 2004
2γair=0.0642(296/T)0.53
E"=1789cm-1
Figure 5.15 Calculated and measured air-broadening coefficients for the “Low T
Line” as a function of temperature.
5
6
7
8
9
0.1
2γai
r [c
m-1
/atm
]
3 4 5 6 7 8 91000
Temperature [K]
2γair=0.08424(296/T)0.3439
Exp Data Exp fit HITRAN 2004
HITRAN 2004
2γair=0.1106(296/T)0.64
HighT Line E"=2553cm-1
Figure 5.16 Calculated and measured air-broadening coefficients for the “High T
Line” as a function of temperature.
The line strengths, self-broadening coefficients, and air-broadening coefficients for these three
transitions are measured over a range of temperature. They are plotted versus temperature in
Temperature Sensing using H2O transitions near 1.4 µm
103
figure 5.11 – 5.16, as calculated using HITRAN2004 parameters and measured in our cell. The
temperature dependence of the measured line strength is first fit to Eqn (5.4) with both E” and
S(296K) as free parameters. The agreement between the inferred E” and the HITRAN value
confirms the spectroscopic assignment. Once this assignment is confirmed the HITRAN value of
E” is used to determine S(296K). The experimentally determined spectroscopic parameters of the
selected H2O line pair are listed in table 5.3. Discrepancies are found between measured and
HITRAN data and we suggest using our measured results in future work.
5.2.3 2f temperature sensor validation
5
4
3
2
1
0
Det
ecto
r si
gnal
(V
olts
)
1.00.80.60.40.20.0Sampling time (ms)
Laser Scan with Modulation
Laser intensity I1
Laser intensity I2
0.06
0.04
0.02
0.00
-0.02
-0.04
2f S
igna
l (ar
b. u
nits
)
1.00.80.60.40.20.0Sampling time (ms)
2f Line Shape
2f Peak Height 1
2f peak height 2
Figure 5.17 Schematic of single-laser scanned-wavelength method.
As pointed out in chapter 2, there are two key issues for the WMS technique: intensity
modulation effects and calibration procedures. The problems associated with these two concerns
Table 5.3 Spectroscopic data for the selected H2O line pair
γair γself Line ν0
[cm-1]
S [cm-2/atm]
@296K
E” [cm-1] [cm-1/atm] @ 296K
nair nself
High T 1 7153.72 1.90E-6
High T 2 7153.75 6.15E-6 2552.9 0.0421 0.308 0.34 0.82
This
wor
k
Low T 7154.35 3.67E-4 1789.0 0.0311 0.151 0.52 0.65
High T 1 7153.72 1.90E-6 2552.9 0.0553 0.299 0.64 0.50
High T 2 7153.75 5.50E-6 2552.9 0.0536 0.299 0.64 0.50
HIT
RA
N
Low T 7154.35 3.85E-4 1789.0 0.0321 0.177 0.53 0.50
Chapter 5
104
are greatly simplified here. The laser wavelength is modulated by driving the laser with a
sinusoidally modulated injection current, which also modulates the laser intensity. Since we
detect the 2f signal at line center (there is little effect on the 2f signal at line center from this
intensity modulation [Liu 2004; Philippe 1993]), the intensity modulation is neglected here.
Because the 2f signal is proportional to laser intensity, many hardware-related parameters
including detector sensitivity, signal amplification, lock-in gain and bandwidth, etc. are required
exclusive of measuring the 2f signal itself. The usual approach is to calibrate the WMS sensor at a
reference condition (with the exact same settings) to eliminate the dependence of hardware-
related parameters. Performing such calibration is usually difficult and adds additional
uncertainty to the measurement results. The dependence of these instrumental parameters are
removed in this work by using a single-laser scanned-wavelength method (as shown in figure
5.17), since two transitions lie in the same laser scan and they share the same lock-in amplifier
and detector, and thus utilize identical hardware settings. Thus the 2f peak ratio simplifies:
1 1 2 12
2 2 2 2
( )( )
peakf
peak
Height I HRatioHeight I H
νν
= = ⋅ (5.10)
Where I1 and I2 are the laser intensity at line center of peak 1 and peak 2, respectively, and
2 1( )H ν and 2 2( )H ν are the second harmonic Fourier coefficients which are given by Eqn.
(2.40). The first term in Eqn. (5.10) I1/I2 is constant for a given laser temperature and laser scan
range. The second term is calculated using Eqn. (2.44). The calibration procedure becomes a
measurement of I1/I2. This is done here in two ways: (1) calibrate from measurements at a single
point in the low-pressure heated-cell experiment or (2) fit baseline to the low-pass filtered raw
data and determine I1/I2 directly. Once I1/I2 is determined using either of the methods above, this
sensor does not require further calibration even when the hardware-related parameters change
later. This is a unique advantage of using single-laser scanned-wavelength 2f WMS method.
Heated-cell experiments were performed to validate the sensor accuracy and reliability for the
temperature inferred from the ratio of the two transitions.
The experimental setup is illustrated in figure 5.18. The DFB diode laser operating near 1.4 µm is
driven by a 2 kHz sawtooth ramp to tune the wavelength combined with a faster 500 kHz sine
wave to provide the wavelength modulation. The output from the laser passes through the cell to
monitor the transmitted intensity. A flat mirror, which provides a double-pass configuration to
improve SNR, is used to reflect the laser beam to pass through the cell again. The laser, detectors,
and optics are enclosed in a nitrogen-purged area to prevent interfering absorption by room air. A
Temperature Sensing using H2O transitions near 1.4 µm
105
lock-in amplifier (Perkin-Elmer Model 7280) is used to measure the second-harmonic component
of the transmitted laser signal.
14” Sample Path
3-Zone Tube Furnace
Transmission Detector
DAQ Computer
Quartz CellPurged with N2
Mixing Tank
Air Cylinder
Low P Baratron High P Baratron
Mullite Tube
Mirror
2kHz ramp
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Reference signal Harmonic Signal
14” Sample Path
3-Zone Tube Furnace
Transmission Detector
DAQ Computer
Quartz CellPurged with N2
Mixing Tank
Air Cylinder
Low P Baratron High P Baratron
Mullite Tube
Mirror
2kHz ramp
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Reference signal Harmonic Signal
Figure 5.18 Arrangement for the 2f sensor validation experiments.
Two sets of static heated cell experiments were carried out to validate the 2f sensor thermometry.
The first experiment was performed with 3 Torr pure water vapor. Two different modulation
depths (0.037 cm-1 and 0.056 cm-1) are used in this experiment. The top graph in figure 5.19
shows the comparison between the measured 2f peak ratios and the theoretical simulations. A
single point at 803K (a=0.056 cm-1) is used to calibrate the 2f sensor for both values of a. Good
agreement between measurements and simulations confirm the accuracy of the measured
fundamental spectroscopic parameters over the full range of conditions studied. The
corresponding temperatures are shown in the bottom graph in figure 5.19. The temperatures
inferred from the 2f sensor are seen to agree extremely well with the thermocouple measurements
(+/- 10K).
Chapter 5
106
1400
1200
1000
800
600
Tem
pera
ture
[K]
14001300120011001000900800700600Temperature [K]
0.4
0.3
0.2
0.1
0.0
2f p
eak
ratio
P=3 TorrXH2O=100%L=71.12 cm
Simulation
a = 0.037 cm-1
a = 0.056 cm-1
Experiment
a = 0.037 cm-1
a = 0.056 cm-1Single Point Calibration
Figure 5.19 Comparison of measured 2f peak ratio with simulated 2f peak ratio (top);
Comparison of measured temperature with thermocouple temperature. (bottom)
To demonstrate the applicability of the 2f sensor at atmospheric conditions, the heated cell
experiment was repeated with ambient air at atmospheric pressure. The water concentration in the
ambient air was determined to be 1.1% using a direct absorption method. Two different
modulation depths (0.037 cm-1 and 0.056 cm-1) are also used in this experiment. The experimental
results are plotted in figure 5.20. The top graph in figure 5.20 shows the comparison between the
measured 2f peak ratios and the theoretical simulations. The signal data point at 877K (a=0.037
cm-1) is used to calibrate the 2f sensor. The inferred temperatures from the 2f sensor are plotted in
the bottom graph in figure 5.20. Comparison of the measured temperatures and thermocouple
data shows good agreement (+/- 20 K).
Temperature Sensing using H2O transitions near 1.4 µm
107
1200
1000
800
600
Tem
pera
ture
[K]
1300120011001000900800700600Temperature [K]
0.30
0.25
0.20
0.15
0.10
0.05
0.00
2f P
eak
Rat
io
Simulation
a = 0.037 cm-1
a = 0.056 cm-1
Experiment
a = 0.037 cm-1
a = 0.056 cm-1
P =1 atmXH2O=1.1%L=71.12 cm
Single Point Calibration
Figure 5.20 Comparison of measured 2f peak ratio with simulated 2f peak ratio (top);
Comparison of measured temperature with thermocouple temperature. (bottom)
Good agreement is found for both sets of experiments with different modulation depths. The
upper curves in figure 5.19 and 5.20 confirm the accuracy of the measured fundamental
spectroscopic parameters over the full range of conditions studied. The lower curves in figures
5.19 and 5.20 demonstrate the efficiency and accuracy of the 2f sensor thermometry, which
shows good potential for combustion sensing and control.
5.2.4 Real-time capabilities
Many measurement and control applications require high performance in real-time. Real-time
data acquisition and signal processing are the core of real-time performance. The hardware-
software architecture of the 1.4 µm temperature sensor is described in this section.
Chapter 5
108
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
Harmonic Signal
Transmission
Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Perkin Elmer 7280 Lock-in amplifier
Reference signal
Harmonic Signal
DFB Laser
Bias TEE
DG535 DelayGenerator
Trigger 1
Measurement Path length
Trigger 2
Real Time T
2kHz
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
Harmonic Signal
Transmission
Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Perkin Elmer 7280 Lock-in amplifier
Reference signal
Harmonic Signal
DFB Laser
Bias TEE
DG535 DelayGenerator
Trigger 1
Measurement Path length
Trigger 2
Real Time T
2kHz
Figure 5.21 Complete hardware-software framework
+
+
+ _
_
_
5.6 kΩ 5.6 kΩ 5.6 kΩ 5.6 kΩ
11 kΩ 11 kΩ
470 kΩ 470 kΩ
470kΩ 470 kΩ
11 kΩ 11 kΩ
2N3904 2N3904
2N3904 2N3904
22 μF 22 μF
0.1 μF 0.1 μF
~ 500 kHz sine wave
~ 2 kHz sawtooth wave
Output
+9V
+9V +9V
+9V +9V
+9V +9V
Figure 5.22 Diagram of the bias-tee
Figure 5.21 presents the complete schematic of hardware-software implementation for the 1.4 µm
single-laser sensor. A 2 kHz sawtooth ramp is combined with a faster 500 kHz sinusoidal signal
through a bias-TEE. The high quality bias-TEE is designed using a differential circuit which is
Temperature Sensing using H2O transitions near 1.4 µm
109
shown in figure 5.22. The values of resistors and capacitors are chosen based on the transistor’s
(2N3904) parameters and the frequencies of ramp and modulation.
3.0
2.0
1.0
0.0
Sig
nal [
V]
10008006004002000Time [us]
5
4
3
2
1
0
Sig
nal [
V]
Transmission
Trigger1Laser Scan Rate = 2000 HzModulation Frequency = 500 kHz
Figure 5.23 The trigger signal (top) and the laser transmission signal (bottom).
The detected signal, which is shown in the bottom graph of figure 5.23, is fed into a Perkin-Elmer
lock-in amplifier (Model 7280). The reference signal is provided to the lock-in by the function
generator that generates the 500 kHz sinusoidal signal. The trigger signal (Trigger 1, top graph of
figure 5.23) of the other function generator (which provides 2 kHz ramp) is connected to a digital
delay generator (Stanford Research System DG535) to generate a timing trigger signal (Trigger
2). Trigger 2 and the resulting second-harmonic components of the transmitted laser signal from
lock-in are transferred to PC for data analysis. There are two DAQ cards installed in the DAQ
computer: one is Gage CompuScope 1250 which is used for data acquisition, and the other one is
National Instruments PCI 6120 which is used to output the 2f ratio (voltage).
Chapter 5
110
Sign
al [V
]
10008006004002000Time [us]
Sig
nal [
V]
Acqu
ired
Sign
al [V
]
Data Acquisition Data Acquisition
Trigger 2
2f signal
Acquired 2f signal
Transfer & Calculation
Output Output
Rising Edge
Falling Edge
Figure 5.24 Timing diagram: timing trigger signal (Top), the second-harmonic signal
(Middle) and the acquired signal (Bottom)
Figure 5.24 presents the detailed timing diagram of the 1.4 µm single-laser sensor. The timing
trigger signal, detected second-harmonic signal and acquired signal are shown in the top, middle
and bottom graphs of figure 5.24, respectively. A laboratory code is written in C/C++ for data
acquisition and analysis. At the instant when the falling edge of trigger 2 is detected, a selected
data record (usually 1000 points) is captured into the on-board memory of CompuScope 1250 and
the acquisition is stopped. The falling edge of trigger 2 and the number of data points are adjusted
so that all the useful information (2f peaks in the current study) is captured. Once the acquisition
is complete, the signal processing program transfers the captured data from on-board memory to
PC memory. This program takes advantage of the high speed PCI bus-mastering data transfer
technique, whose transfer rate is up to 100 MB/s. Data analysis, including peak finding and ratio
calculation, is then performed on the accumulated data. When the rising edge of trigger 2 is
detected, the program outputs the calculated 2f ratio through National Instruments PCI 6120. The
system is then ready for the next acquisition. 2 kHz real-time performance is achieved for the 1.4
µm single-laser sensor, which is sufficient for many combustion sensing and control applications.
Temperature Sensing using H2O transitions near 1.4 µm
111
5.3 Combustion Demonstration 5.3.1 Identification of Acoustic Instabilities
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Harmonic Signal
Transmission Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Reference signal
Harmonic Signal
Forced Fuel (100Hz)
DFB Laser
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Harmonic Signal
Transmission Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Lock-in amplifier
Reference signal
Harmonic Signal
Forced Fuel (100Hz)
DFB Laser
Figure 5.25 Schematic diagram of the measurement system applied to a Hencken
burner.
To illustrate the potential of the sensor for monitoring fluctuations of acoustic instabilities in
combustion gases from a time history of temperature data, measurements were made on a flat-
flame Hencken burner, as shown in figure 5.25. The Hencken burner can produce a uniform,
constant-pressure flow field, as described in chapter 4 and ref. [Furlong 1998]. The burner
consists of a 5-cm square array of inner tubes (0.5-mm internal diameter), which supply the fuel
(C2H4). Coannual air streams mix with fuel stream at the top of the burner to produce an array of
shot diffusion flames. The 1.4 µm single-laser sensor is driven by an external modulation, which
consists of a 2 kHz sawtooth ramp combined with a faster 500 kHz sinusoidal signal. The second-
harmonic components of the transmitted laser signal are obtained by a Perkin-Elmer lock-in
amplifier (Model 7280) with a time constant of 1 µs. The temperature is inferred from the ratio of
2f peak heights. 2 kHz real-time data processing and reduction is achieved by a fast PC
combined with a laboratory code written in C++ as described above.
Chapter 5
112
3000
2500
2000
1500
1000
Tem
pera
ture
[K]
1.00.80.60.40.20.0Time [s]
Hencken BurnerC2H4/AirSpeaker On (100Hz)
Time [s]
Tem
pera
ture
[K]
5x104
4
3
2
1
0
Trm
s2 [K2 ]
10008006004002000Frequency [Hz]
Real-time 2000Hz100 Hz
Figure 5.26 Measured temperature and its power spectrum in the burned region
above the C2H4-air flame.
The H2O sensor was aligned to probe the burned gases 2 cm above the burner surface. A
disturbance was introduced in the flame by modulating the fuel flow with a speaker attached to
the bottom of the Hencken burner (see figure 5.25). The speaker was driven with a 100 Hz sine
wave, thus producing an oscillating gas temperature (top panel in figure 5.26) and associated
Fourier transform (lower panel in figure 5.26). The dominant and harmonic modes of the
temperature fluctuations are clearly shown in the 2 kHz real-time power spectrum. These results
are similar to those shown in chapter 4 for the 1.8 µm direct absorption temperature sensor.
However, the 1.8 µm sensor requires post data processing while this 1.4 µm sensor is a real-time
sensor. These results demonstrate the utility of this fast temperature sensor for accurate
Temperature Sensing using H2O transitions near 1.4 µm
113
characterization of acoustic disturbances, and suggest good potential for applications to real-time
combustion sensing and control.
5.3.2 Real-time measurements in a Swirl Spray Combustor The real-time 1.4 µm WMS sensor is used for measurements in a liquid-fuel swirl-stabilized
spray combustor at University of Cincinnati. A detailed description of the swirl-stabilized spray
combustor can be found in ref. [Li 2004]
Computer
Lock-in amplifier
Real-time T@ 2000Hz
Round duct (47cm) generate natural flame instability
Collimator
2kHz ramp
Diode laser controller
Functiongenerator
Modulate at f=500 kHz
Functiongenerator
+
Reference signal
Filter
N2 purge
N2 purge
Microphone
Computer
Lock-in amplifier
Real-time T@ 2000Hz
Round duct (47cm) generate natural flame instability
Collimator
2kHz ramp
Diode laser controller
Functiongenerator
Modulate at f=500 kHz
Functiongenerator
+
Reference signal
Filter
N2 purge
N2 purge
Microphone
Figure 5.27 Schematic diagram of the measurement system applied to the swirl-
stabilized spray combustor.
The experimental setup for the 1.4 µm, 2f temperature sensor is illustrated in figure 5.27. The
DFB diode laser operating near 1.4 µm is driven by an external modulation, which consists of a 2
kHz saw tooth ramp combined with a faster 500 kHz sinusoidal modulation signal. The laser
beam exits the fiber and is collimated with a lens across the flame, and filtered and detected. The
fiber optics provides set-up flexibility and ease of alignment, which is advantageous for practical
applications under industrial conditions.
A round quartz duct is utilized to generate natural flame instability. The natural flame instability
is driven by the strong coupling between oscillations in heat release and pressure oscillations of
the combustion chamber. The driving mechanisms for thermo-acoustic combustion dynamics are
reviewed in ref. [McManus, 1992]. Acoustic signals are detected by a Brüel & Kjær microphone
(Model 4939-A-011) which is located 0.5 m away from the combustor chamber. The laser beam
Chapter 5
114
is intentionally kept away from the centerline of the round quartz duct to minimize etalon
interference. The second-harmonic components of the transmitted laser signal are obtained by a
Perkin-Elmer lock-in amplifier (Model 7280) with a time constant of 1 µs. The temperature is
inferred from the simple ratio of 2f peak height. 2 kHz real-time data processing and reduction is
achieved by a fast industrial PC combined with a laboratory code written in C++.
3
2
1
0
-1
-2
2f S
igna
l [ar
b. u
nits
]
500400300200100
Time [us]
500400300200100
Propane Ethanol
Peak 1Peak 2
Peak 1Peak 2
Figure 5.28 Reduced H2O 2f line shapes (single scan) recorded in gas fuel (propane)
and liquid fuel (ethanol), a = 0.065 cm-1.
Measurements with the 1.4 µm sensor were carried out at a radial position 15 mm from the spray
centerline and an axial position 50.8 mm downstream of the nozzle exit. The total path length is
97 mm. Figure 5.28 shows the representative 2f lineshape (single scan) corresponding to the
absorption features using gas and liquid fuel. Figure 5.28 (left) is for an air flow rate of 29.0
SCFM and propane flow rate of 40.9 SLM. Figure 5.28 (right) is for an air flow rate of 57.5
SCFM and ethanol flow rate of 0.15 kg/min. As seen from figure 5.28, SNR is about same for
liquid fuel and gas fuel. Due to the superior noise properties of the high-frequency detection
(1MHz in the present study), the 1.4 µm sensor provides precise temperature measurements even
in the liquid-fueled swirl flame.
Temperature Sensing using H2O transitions near 1.4 µm
115
0.12
0.08
0.04
0.00
A. U
. rms
10008006004002000Frequency [Hz]
-1.0
-0.5
0.0
0.5
1.0
A. U
.
0.50.40.30.20.10.0Time [s]
FFT
MicrophonePropane
Figure 5.29 Measured acoustic signal and its power spectrum in the burned region
above the propane-air flame.
200
150
100
50
0
Trm
s [K
]
10008006004002000Frequency [Hz]
4000
3000
2000
1000 Tem
pera
ture
[K]
0.50.40.30.20.10.0Time [s]
FFT
Propane 1.4 µm sensor
Figure 5.30 Measured temperature and its power spectrum in the burned region
above the propane-air flame.
Chapter 5
116
1.2
0.8
0.4
0.0
A. U
. rms
10008006004002000Frequency [Hz]
-10
-5
0
5
10A
. U.
0.50.40.30.20.10.0Time [s]
FFT
MicrophoneEthanol
Figure 5.31 Measured acoustic signal and its power spectrum in the burned region
above the ethanol-air flame.
100
80
60
40
20
0
Trm
s [K
]
10008006004002000Frequency [Hz]
4000
3000
2000
1000 Tem
pera
ture
[K]
0.50.40.30.20.10.0Time [s]
FFT
Ethanol 1.4 µm sensor
Figure 5.32 Measured temperature and its power spectrum in the burned region
above the ethanol-air flame.
Figure 5.29 shows the measured acoustic signal and its power spectrum in the burned region
above the propane-air flame with an air flow rate of 29.0 SCFM and propane flow rate of 40.9
SLM. In figure 5.30, real-time temperature and its power spectrum are depicted under the same
Temperature Sensing using H2O transitions near 1.4 µm
117
condition. With a laser scan rate of 2 kHz, a time resolution of 0.5 ms is achieved. It can be seen
that the dominant mode (232 Hz) and harmonic (464 Hz) modes of fluctuations (sound pressure
and temperature), determined by calculating the magnitude of the discrete Fourier transform of
the measured time-varying pressures and temperatures from 0 to 0.5 sec, are clearly revealed. The
measured temperature fluctuation is in good agreement with the measured pressure fluctuation
from the microphone.
Figure 5.31 shows the measured acoustic signal and its power spectrum in the burned region
above the ethanol-air flame with an air flow rate of 57.5 SCFM and ethanol flow rate of 0.15
kg/min. Figure 5.32 shows the measured temperature and its power spectrum under the same
condition. With laser scan rate of 2 kHz a time resolution of 0.5 ms is achieved. It can be seen
that the dominant mode (350 Hz) and harmonic (700 Hz) modes of fluctuations (sound pressure
and temperature), determined by calculating the magnitude of the discrete Fourier transform of
the measured time-varying temperatures from 0 to 0.5 sec, are again clearly revealed. These
results demonstrate the sensor’s ability to track and quantify the temperature fluctuations, as
needed to control combustion instability.
5.3.3 Comparison with 1.8 µm sensor
A comparison of the critical differences between the 1.4 µm scanned-WMS sensor and the
scanned-wavelength direct-absorption 1.8 µm sensor is now discussed.
0.010
0.008
0.006
0.004
0.002
0.000
Abs
orpt
ion
coef
ficie
nt [c
m-1
]
7155.07154.57154.07153.57153.0Frequency [cm
-1]
0.010
0.008
0.006
0.004
0.002
0.000
5555.05554.55554.05553.55553.0
T = 296 K T = 1000 K T = 2000 K
P = 1 atmX = 10%
1.8 µm sensor
1.4 µm sensor
Figure 5.33 Calculated spectroscopic features for water line pairs in the 1.4 µm and
1.8 µm sensors based on HITRAN; XH2O = 10%.
Chapter 5
118
The 1.4 µm scanned-WMS sensor exploits the lessons learned with the earlier 1.8 µm direct-
absorption sensor described in chapter 4. The calculated spectroscopic features for the water line
pairs of the 1.4 µm sensor and the 1.8 µm sensor are shown in figure 5.33. Both sensors are built
on the wavelength-scanned, single-laser, two-line-ratio thermometry concept, which measures the
ratio of absorption on two nearby temperature water transitions with a single laser scan. This
sensor architecture simplifies the sensor hardware and reduces its cost.
The 1.4 µm sensor includes several important improvements over the 1.8 µm sensor. First, the 1.4
µm sensor minimizes ambient air interference by using transitions with weak absorption at room
temperature, as shown in figure 5.33. The existence of ambient air interference can reduce the
measurement accuracy and lead to increased measurement uncertainty. It is usually necessary to
purge the optical path outside the target measurement zone with nitrogen or dry air to eliminate
the ambient water interference. The amount of purge gas necessary to attain a desired low
humidity set-point depends not only on the ambient humidity level, but also on the absorption
strength of the transition. Due to overlap with a cold water transition, the 1.8 µm sensor has a
very strong water absorption at room temperature, and therefore the purging requirements of the
1.8 µm sensor are quite significant. This difficulty is mitigated by the 1.4 µm sensor which
requires modest purging for long ambient pathlengths and no purging for short pathlengths.
Second, the 1.4 µm sensor uses sensitive WMS to achieve better SNR and lower detection limits.
WMS is a widely used technique for sensitive trace-species detection, and can significantly
reduce the dominating 1/f noise by shifting detection to higher frequencies [Reid 1981], which
provides a substantial sensitivity enhancement compared to the direct absorption methods used in
the 1.8 µm sensor. Therefore, the 1.4 µm sensor can be used in noisy environments like the
liquid-fueled spray flame where the 1.8 µm sensor showed significant uncertainty.
Temperature Sensing using H2O transitions near 1.4 µm
119
3x10-3
2
1
0
-1
2f s
igna
l [a.
b. u
nits
]
7155.07154.57154.07153.57153.0Frequency [cm
-1]
0.010
0.008
0.006
0.004
0.002
0.000Abs
orpt
ion
coef
ficie
nt [c
m-1
]
5555.05554.55554.05553.55553.0
T = 1000 K T = 2000 K
P = 1 atmX = 10%
1.8 µm sensor
1.4 µm sensor
Figure 5.34 Comparison of the measurement strategy of the 1.4 µm sensor and 1.8
µm sensor.
Third, the 1.4 µm WMS sensor has a simpler data reduction computation than the direct
absorption method and thus can achieve better real-time performance. The data reduction
strategies used in the 1.4 µm sensor and 1.8 µm sensor are illustrated in figure 5.34. The 1.8 µm
sensor first must subtract the baseline to infer I0, and then must fit Voigt lineshapes to extract the
integrated absorbance needed to determine temperature. The non-linear and time-consuming
Voigt fitting of the lineshape often requires post-processing data analysis, and is impractical to
rapidly determine temperature in real-time. The 1.4 µm sensor uses the ratio of the 2f peak
heights to avoid time-consuming fits to the line shape, and can reach a 2 kHz real-time
measurement rate (2 kHz scan repetition rate). Hence, the 1.4 µm sensor could be utilized in
many applications where the instability and control response is > 1 ms.
Fourth, the 1.4 µm WMS sensor takes advantage of the mature telecommunication laser
techniques at 1.4 µm where fiber-coupled lasers and fiber components are readily available. The
fiber-coupled lasers and fiber optics at 1.4 µm have many attractive features that are superior to
the free-space lasers and optics at 1.8 µm. These include advanced laser performance, simple
installation, easy laser beam alignment, improved ruggedness and flexibility, and reduced overall
system cost.
Chapter 5
120
5.4 Summary
A single-diode-laser temperature sensor based on wavelength modulation absorption
spectroscopy in H2O vapor is designed for rapid and accurate temperature measurements in a
combustion environment. The strategies and criteria to select optimum water features in the 1-2
micron wavelength range have been detailed. The 12 best NIR water transition line pairs for
temperature measurements with a single DFB laser in flames were determined by systematic
analysis of the HITRAN simulation of the water spectra in the 1-2 µm spectral region. A specific
line pair near 1.4 µm was identified and investigated experimentally, and the pertinent
spectroscopic parameters were determined from cell experiments. These measurements provide
useful improvements to the current spectroscopic database for H2O for the target transitions and
their neighbors. The sensor enables a real-time temperature readout rate of 2 kHz and immunity
from ambient water vapor interference (using a common PC). Demonstration experiments were
conducted in a heated cell and a forced Hencken burner. Cell experiments confirmed sensitivity
and accuracy of the sensor. The burner experiments illustrate its ability to monitor temperature
fluctuations with potential applications to combustion control. Temperature is an important
parameter for combustion control because it is a sensitive measure of heat release. Compared
with traditional pressure-based sensor like microphone, temperature-based TDL sensor has many
advantages such as good spatial resolution, immunity to vibration and suitability in noisy
environments. Chapter 6 will present the application of the 1.4 µm sensor to combustion control.
121
Chapter 6: Application of fast temperature sensor to
combustion control
In this chapter, the 1.4 µm real-time WMS temperature sensor is used for combustion sensing in a
swirl-stabilized combustor and the potential for combustion control is demonstrated. Two
problems are investigated in the swirl-stabilized flame: lean blowout (LBO) and acoustic
instabilities. The use of the 1.4 µm WMS TDL is compared with a traditional microphone sensor.
It is shown that the 1.4 µm WMS temperature sensor has advantages for lean blowout prediction,
and is especially promising for lean blowout control. A phase-delay control strategy is used to
suppress temperature fluctuation in this experiment, and the results illustrate the potential of this
sensor for real-time combustion instability control. Conclusions are summarized in the final
section.
6.1 Motivation
The reduction of NOx emissions from practical combustion is an important problem [Martin 1990;
John 1997], and many schemes to reduce NOx emissions have been investigated. Lean premixed
combustion is one of the most effective approaches to reduce NOx emissions because of its lower
flame temperatures [Martin 1990]. Unfortunately, this scheme has two major drawbacks: First, it
is susceptible to lean blowout (LBO) of the combustion, which can lead to significant potential
safety hazards, energy loss, and substantial costs from power shut-down [Thiruchengode 2003;
Muruganandam 2003; Nair 2003]. Second, this scheme is susceptible to thermoacoustic
combustion instabilities from the coupling of heat release to acoustic oscillations [Mcmanus 1993;
Docquier 2002; Lieuwen 2003], which can lead to decreased combustion efficiency, increased
noise pollution, and serious system performance degradation. Thus, practical use of lean
premixed combustion likely will require real-time control to avoid LBO and suppress acoustic
instabilities. An important part of any control strategy is a robust sensor to monitor a good control
variable. Gas temperature is a sensitive measure of heat release and is investigated here as a
control variable.
There is a large literature for flame behavior during the lean blowout process [Thiruchengode
2004; Nair 2004; Muruganandam 2005; Prakash 2005]. This previous work shows that significant
low frequency fluctuations increase near lean blowout. Acoustic noise measurement with a
Chapter 6
122
microphone and optical chemiluminescence measured with a photo detector are the most
frequently used methods for LBO experiments. Although the microphone has the desired
sensitivity and response time, it is sensitive to the signal from the entire flame, and thus is
unsuitable to detect a blowout event in a local region of the flame. While chemiluminescence can
provide a rough measure of heat release rate, it is much less quantitative. To be a useful sensor for
this application, any potential spectroscopic strategy must have fast time response, high
sensitivity, high spectral resolution, robustness, reasonable spatial resolution and be non-intrusive
in character.
The control of combustion acoustic instabilities has also been the subject of numerous theoretical
and experimental studies [for example, Kemal 1996; Candel 2002; Lieuwen 2001; Park 2002;
Fleifil 1996; Furlong 1998]. A variety of models and mechanisms [Lieuwen 2003] have been
proposed to explain the occurrence of thermoacoustic instabilities and different experimental
methods were used to suppress these combustion instabilities. A review of these studies can be
found in refs. [Mcmanus 1993; Docquier 2002]. Previous work in this laboratory [Furlong 1996;
Furlong 1998] demonstrated that temperature is a good control variable for combustion instability
control. The aim of the current work is to illustrate the potential of our 1.4 µm WMS temperature
sensor to suppress the natural combustion instabilities in a practical swirl-stabilized flame using
active control.
In the following sections, the 1.4 µm WMS temperature sensor is first applied to a swirl-
stabilized combustor to illustrate the potential of this sensor to observe a clear indication of the
approach to the lean blowout limit. The sensor is then used to provide active feedback control to
suppress natural thermoacoustic instabilities for the same combustor. The LBO experiments show
that the 1.4 µm WMS temperature sensor has good potential to predict the lean blowout limits.
The instability control experiments show that the 1.4 µm WMS temperature sensor has the
desired flexibility, speed, sensitivity and accuracy to be a useful tool for fundamental and applied
combustion monitoring and control.
6.2 Swirl-stabilized combustor
To investigate the potential to predict the approach to the lean blow out limit, the WMS
temperature sensor is applied to a swirl-stabilized combustor at Stanford. This combustor is a
Application of fast temperature sensor to combustion control
123
copy of the swirl-stabilized combustor utilized in chapters 4 and 5. Illustration of the combustor
design is presented in figure 6.1 and a full description can be found in ref. [Li 2003].
Air
Speaker Port
Air Conditioning Chamber
Triple Annular Research Swirler
Seeding Fuel
80 cm
10.2 cm
Figure 6.1 Schematic of swirl-stabilized combustor.
The combustor is assembled vertically and the flow is from bottom to top. A combination of
honeycomb and different size mesh screens is used to create a uniform flow in the air
conditioning chamber. For the purpose of investigation and thermoacoustic instability control of
the swirl-stabilized combustion, four identical acoustic ports are placed at the air conditioning
chamber. A triple-annular research swirler (TARS) [Li 2004] is used here to investigate the
effects of various swirler configurations on combustion performance. The details of the TARS are
Chapter 6
124
given in ref. [Li 2004]. The outer, intermediate, and inner swirlers can be changed independently,
and the available swirler configurations are summarized in table 6.1, where 30o means that “the
swirler angle is 30o in the clockwise direction”, and -30o means that “the swirler angle is 30o in
the counterclockwise direction”. An inner swirler (-45 o), intermediate swirler (0o) and outer
swirler (-55o) configuration is used for current work.
Table 6.1 Different inter, intermediate and outer swirler configurations
Swirler #1 #2 #3 #4 #5 #6
Inner 30o 45o -30o -45o
Intermediate 45o 30o 0o
outer 30o 45o 55o -30o -45o -55o
The fuel is delivered through a separate main and pilot system. The main fuel is injected through
8 injection holes into outer swirler, and the pilot fuel is injected through 4 injection holes into
intermediate swirler. There are also two air-assist lines (not shown in figure 6.1) to provide a
small amount of air to premix the fuel. The fuel flow and air flow rate are independently
controlled by valves and measured using calibrated flow meters.
6.3 Lean blowout (LBO) prediction
6.3.1 Experimental setup
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
-Harmonic Signal
Transmission
Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Perkin Elmer 7280 Lock-in amplifier
Reference signal
Harmonic Signal
DFB Laser
Bias TEE
GeneratorDG535 DelayTrigger 1 Trigger 2
Real Time T
2kHz
Flat mirror
N2 purge N2 purgeFocusingmirror
Band passfilter
MicrophoneFilter
Detector
DAQ Computer
2kHz ramp
Laser
FunctionGenerator
FunctionGenerator
-Harmonic Signal
Transmission
Collimator
DAQ Computer
Modulate at f=500 kHz
+
LaserController
FunctionGenerator
FunctionGenerator
Perkin Elmer 7280 Lock-in amplifier
Reference signal
Harmonic Signal
DFB Laser
Bias TEE
GeneratorDG535 DelayTrigger 1 Trigger 2
Real Time T
2kHz
Flat mirror
N2 purge N2 purgeFocusingmirror
Band passfilter
MicrophoneFilter
Figure 6.2 Scheme of the experimental setup.
Application of fast temperature sensor to combustion control
125
The experimental setup is illustrated in figure 6.2. The hardware-software architecture of the 1.4
µm temperature sensor was detailed in chapter 5, and will not be repeated here. The DFB diode
laser operating near 1.4 µm is driven by an external modulation, which consists of a 2 kHz saw
tooth ramp combined with a faster 500 kHz sinusoidal modulation signal. The laser beam is
collimated and directed across the flame. Simultaneously with the TDL sensor signal, acoustic
signals are detected by a Brüel & Kjær microphone (Model 4939-A-011) located 0.3 m away
from the combustor chamber. The combustion chamber is a 200mm long round quartz duct
whose diameter is 90mm. The laser beam is intentionally aligned off the centerline of the duct to
minimize etalon interference. A flat mirror provides a double-pass configuration to improve SNR.
The reflected laser beam is focused, then passes through a narrow band pass filter (NB-1400-030-
B) and is detected by a large area InGaAs detector (3 mm diameter active area, ELECTRO-
OPTICAL SYSTEMS). The large area detector is used here to reduce beam steering noise. The
detected signal is filtered with a 320 kHz (818H8B-5 High Pass 8-Bit Programmable 8-Pole Filter)
and 1.28 MHz (818L8B-5 Low Pass 8-Bit Programmable 8-Pole Filter) to remove unwanted
frequency components.
7
6
5
4
3
2
1
0
Sig
nal [
V]
543210Time [ms]
7
6
5
4
3
2
1
0
Sig
nal [
V]
With beam steering
Without beam steering
Figure 6.3 Raw data with/without beam steering noise.
Chapter 6
126
Figure 6.3 illustrates the typical raw data with beam steering noise (bottom) and without beam
steering (top, this experiment). The existence of beam steering noise will reduce measurement
accuracy. Proper optical design should be adopted to minimize the beam steering effects in laser
absorption experiments. Some effective anti-beamsteering approaches are summarized in the ref.
[Sanders 2001].
-2
-1
0
1
2
2f s
igna
l [ar
b. u
nits
]
0.50.40.30.20.10.0Time [ms]
2f line shape
Low T Line Peak High T Line Peak
Figure 6.4 Reduced H2O 2f line shape (single scan) recorded in gas fuel (propane).
The second-harmonic components of the transmitted laser signal are obtained by a Perkin-Elmer
lock-in amplifier (Model 7280) with a time constant of 1 µs. Figure 6.4 shows the representative
2f lineshape (single scan) corresponding to the absorption features using gas fuel (propane). A
modulation amplitude a = 0.047 cm-1 is adopted in current work. 2 kHz real-time data processing
and reduction is achieved by a fast PC combined with a laboratory code written in C++.
6.3.2 Results and discussions Four sets of experiments were conducted to evaluate the performance of the 1.4 µm WMS
temperature sensor for lean blow out prediction. All experiments have roughly the same initial
Application of fast temperature sensor to combustion control
127
propane flow rate (1.2 SCFM) but different air flow rates (a. 27.3 SCFM b. 38.7 SCFM c. 52.0
SCFM 4. 63.9 SCFM). During the lean blowout process, the propane flow rate was decreased by
slowly adjusting a needle valve while keeping the air flow constant. All experiments were carried
out with the sensor at a radial position 22 mm from the centerline to avoid etalon noise. To
examine the effects of sensing position, each set of experiment was performed at two axial
positions (20mm and 50 mm downstream of the nozzle exit). The total path length is 156 mm for
both flame heights.
(a) (b) (c) (d) (e)
Figure 6.5 The blowout process for the first set of experiments (air flow rate = 27.3 SCFM)
Figure 6.5 shows the images of flame during the blowout process for the first set of experiment
(air flow rate = 27.3 SCFM). The initial flame is slightly fuel rich, and a luminous swirling flame
is clearly observed (image a). The full flame height is about 50 cm, and composed of three
regions: 1.) A blue flame on the top (25 cm), 2.) A pinkish flame in the middle (20 cm) 3.) A very
bright flame in the bottom (5 cm). As fuel flow decreases, the top blue flame gradually vanishes
and the middle pinkish flame and bottom bright flame decreases in length. The full flame height
decreases to 25 cm (image b). The pinkish flame slowly disappears and the bright flame zone
appears indistinguishable with further decrease of fuel flow. The total flame length reduces to 8
cm (image c). The bright flame zone disappears and the whole flame becomes dark blue. The
flame near the wall increases in length (flame height increases to 15 cm) and a flame instability is
observed (image d). As the flame approaches lean blowout limit, the flame becomes unstable,
moves to center and fills the whole chamber (image e). Finally the flame is lifted and blowout
occurs.
Chapter 6
128
Figure 6.6 (a) Microphone and 2f sensor result during the lean blowout process for the first set of
experiments. (Air flow rate = 27.3 SCFM). Laser beam: 20mm height from the injector
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
0.30
0.25
0.20
0.15
0.10
0.05
0.00
P rm
s[0,
50]/
Prm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
Application of fast temperature sensor to combustion control
129
Figure 6.6 (b) Microphone and 2f sensor result during the lean blowout process for
the first set of experiments. (Air flow rate = 27.3 SCFM). Laser beam: 50mm height
from the injector
Figure 6.6 shows the measurement results from microphone and the 1.4 µm WMS temperature
sensor during the lean blowout process for the first set of experiments. Panel (a) and (b) of figure
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Prm
s[0,
50]/
Prm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
Chapter 6
130
6.6 illustrate the results measured at an axial position 20mm and 50 mm downstream of the
nozzle exit, respectively.
The microphone signal and 2f peak ratio as a function of time during the lean blowout process are
shown in the first row of figure 6.6. The insets of each figure show the microphone signal and 2f
peak ratio from the TDL sensor for steady combustion condition (10~10.5 s) and near blowout
(39~39.5 s). There are two things to note: First, the amplitude of the microphone signal gradually
decreased during the lean blowout process because it is proportional to the total combustion
energy. Near blowout, the amplitude of the microphone signal is close to the amplitude of
background acoustic signal. At the same time, a significant decrease in the 2f peak ratio is
observed as it is directly related to temperature. Second, there is no significant difference between
steady combustion and near-blowout limit for 2f signal in position 1 (20mm downstream of the
nozzle exit, figure 6.6 (a)). In contrast, the presence of low frequency disturbance (significant
drop) in 2f peak ratio is clearly shown in position 2 (50 mm downstream of the nozzle exit, figure
6.6 (b)) as the combustor approaches the lean blowout limit. It is clearly shown that position 2 is
more sensitive to temperature fluctuation than position 1. To achieve the best performance,
therefore, it is critical to find the most sensitive position for the 2f sensor. Since the microphone
is not a location-selective component, its signal is not sensitive to its positioning.
The second row of figure 6.6 illustrates the power spectrum from the microphone signal and 2f
peak ratio from the TDL sensor for steady combustion (10~10.5 s) and near-blowout (39~39.5 s)
conditions. As mentioned above, the absolute amplitude of microphone signal changes during the
blowout process. To make a meaningful comparison, the power spectra are normalized by the
sum of all frequency components. The laser scan rate of the 2f WMS temperature sensor and
microphone’s sampling rates are set to 2000 Hz, so the upper detection limit of 1000Hz is
achieved for both sensors. The width of each bar is 2 Hz, which corresponds to the frequency
resolution in the determination of the discrete Fourier transform over a 0.5 s sampling interval. It
is clearly shown that the low-frequency components increase as the flame approaches blowout.
Such low-frequency components correspond to the local flame extinction events and become
more frequent (thus flame becomes unstable) as the flame approaches the blowout limit. Once the
rate of heat release is not sufficient to heat the fuel and air to the required ignition temperature,
the flame approaches blowout. Therefore, these low-frequency disturbances can be used as a
symptom for lean blowout prediction.
Application of fast temperature sensor to combustion control
131
The third row of figure 6.6 shows a strategy for quantitative prediction of approach to the lean
blowout limit. Since the flame extinction events are not periodic, there is no characteristic
frequency in the power spectrum. Here the ratio of the sum of low-frequency (0~50 Hz in current
work) components to the sum of all-frequency (0~1000 Hz) components is plotted. This quantity
illustrates the variety of low-frequency flame extinction events during the lean blowout process.
As shown in figure 6.6, there is a significant increase in low frequency events near blowout for
both the microphone and the 1.4 µm WMS temperature sensor. From steady combustion to near
blowout conditions, the low-frequency components increase from 2% to 12% of the total noise
power for microphone and from 5% to 28% of the total noise power for the 1.4 µm WMS
temperature sensor. However, it should be noted that the low-frequency components for the
microphone are only 5% of the total noise power after the flame blowout from the background
acoustic signals (air flow etc.). Near blowout the amplitude of microphone signal is close to the
background acoustic signal, thus the microphone’s performance for LBO prediction is limited by
the influence of background acoustic signal. In addition, the microphone can only detect the
overall sound so it is unable to detect local flame extinction events.
Similar results were obtained with other air flow rates: 38.7 SCFM, 52.0 SCFM and 63.9 SCFM.
Figures 6.7~6.9 show the corresponding flame behavior during the blowout process. The
measurement results for microphone and the 1.4 µm WMS temperature sensors are presented in
figures 6.10~6.12. For all cases the approach to LBO is captured by the WMS temperature sensor.
Chapter 6
132
Figure 6.7 The blowout process for the second set of experiments (air flow rate=38.7
SCFM)
Figure 6.8 The blowout process for the third set of experiments (air flow rate= 52.0 SCFM)
Figure 6.9 The blowout process for the fourth set of experiments (air flow rate= 63.9
SCFM)
Application of fast temperature sensor to combustion control
133
Figure 6.10 (a) Microphone and 2f sensor result during the lean blowout process for the second
set of experiments. (Air flow rate = 38.7 SCFM). Laser beam: 20mm height from the injector
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
P rm
s[0,
50]/
Prm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
Chapter 6
134
Figure 6.10 (b) Microphone and 2f sensor result during the lean blowout process for
the second set of experiments. (Air flow rate = 38.7 SCFM). Laser beam: 50mm
height from the injector
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Prm
s[0,
50]/
Prm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
Application of fast temperature sensor to combustion control
135
Figure 6.11 (a) Microphone and 2f sensor result during the lean blowout process for the third set
of experiments. (Air flow rate = 52.0 SCFM). Laser beam: 20mm height from the injector
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Prm
s[0,
50]/
Prm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
Chapter 6
136
Figure 6.11 (b) Microphone and 2f sensor result during the lean blowout process for
the third set of experiments. (Air flow rate = 52.0 SCFM). Laser beam: 50mm height
from the injector
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
P rm
s[0,
50]/
P rm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
Application of fast temperature sensor to combustion control
137
Figure 6.12 (a) Microphone and 2f sensor result during the lean blowout process for the fourth set
of experiments. (Air flow rate = 63.9 SCFM). Laser beam: 20mm height from the injector
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Norm
alize
d Po
wer S
pect
rum
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
0.30
0.25
0.20
0.15
0.10
0.05
0.00
P rms[
0,50
]/P rm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
Chapter 6
138
Figure 6.12 (b) Microphone and 2f sensor result during the lean blowout process for
the fourth set of experiments. (Air flow rate = 63.9 SCFM). (b) Laser beam: 50mm
height from the injector
2.0
1.5
1.0
0.5
0.0
2f p
eak
ratio
50403020100Time [s]
0.7
0.6
0.5
0.4
0.310.410.210.0
0.5
0.4
0.3
0.2
0.139.439.239.0
Blowout
2f T Sensor
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
10008006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
Microphone
0.30
0.25
0.20
0.15
0.10
0.05
0.00
T rms[
0,50
]/T rm
s[0,
1000
]
50403020100Time [s]
Blowout2f T sensor
Low Frequency TAll frequency T
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Nor
mal
ized
Powe
r Spe
ctru
m
8006004002000Frequency [Hz]
Near Blowout (39~39.5 s) Steady combustion (10~10.5 s)
2f T sensor
0.30
0.25
0.20
0.15
0.10
0.05
0.00
P rm
s[0,
50]/
P rm
s[0,
1000
]
50403020100Time [s]
BlowoutMicrophone
Low Frequency P
All frequency P
4
3
2
1
0
-1
Pres
sure
[V]
50403020100Time [s]
-1.0
-0.5
0.0
0.5
1.0
10.410.210.0-0.10
-0.05
0.00
0.05
0.10
39.439.239.0
Blowout
Microphone
Application of fast temperature sensor to combustion control
139
In this section, the 1.4 µm WMS temperature sensor has been used to detect approach to the lean
blowout limit in a swirl-stabilized combustor. Experiments were performed under different air
flow conditions and sensing positions. These results demonstrate the potential of this sensor for
accurate identification of low-frequency temperature fluctuations near the lean blowout limit, and
suggest strategies for lean blowout control. The ability of the 1.4 µm WMS temperature sensor to
measure the low-frequency increase near the lean blowout limit shows good potential for LBO
control.
6.4 Combustion instability control
6.4.1 Experimental setup
Detector
DAQ Computer
Harmonic Signal
Transmission
Collimator
DAQ Computer
Perkin Elmer 7280 Lock-in amplifier Harmonic Signal
DFB Laser
Real Time 2f ratio
2kHz
Flat mirror
N2 purge N2 purgeFocusingmirror
Band passfilter 1
Microphone
Filter
Band passfilter 2
Time-Delay GeneratorΔt
AmplifierGain: α
Gas Fuel (Propane)
Detector
DAQ Computer
Harmonic Signal
Transmission
Collimator
DAQ Computer
Perkin Elmer 7280 Lock-in amplifier Harmonic Signal
DFB Laser
Real Time 2f ratio
2kHz
Flat mirror
N2 purge N2 purgeFocusingmirror
Band passfilter 1
Microphone
Filter
Band passfilter 2
Time-Delay GeneratorΔt
AmplifierGain: α
Gas Fuel (Propane)
Figure 6.13 Scheme of the experimental setup.
Detailed descriptions of the 1.4 µm WMS temperature sensor can be found in chapter 5 and
section 6.2; figure 6.13 illustrates the fundamentals of its use for the phase-delay control strategy.
The DFB diode laser operating near 1.4 µm is driven by an external modulation, which consists
of a 2 kHz saw tooth ramp combined with a faster 500 kHz sinusoidal modulation signal (a =
0.047 cm-1). The laser beam is collimated and directed across the flame 15 mm above the nozzle
exit. Acoustic signals are detected by a Brüel & Kjær microphone (Model 4939-A-011) which is
Chapter 6
140
located 0.3 m away from the combustor chamber. The 200 mm long round quartz duct is replaced
with a 450 mm long quartz duct to generate natural flame instability. The second-harmonic
components of the transmitted laser signal are obtained by a Perkin-Elmer lock-in amplifier
(Model 7280) with a time constant of 1 µs. 2 kHz real-time data reduction is achieved by a fast
PC combined with a laboratory code written in C++. The C++ program conditioned the 2f ratio
and output only the ac component of the signal (in voltage), is directed into a time-delay
generator. The time-delay scheme is implemented using a dSPACE 1104 real-time control board.
The analog signal (2f ratio) is first converted to digital signal through an analog-to-digital
converter (ADC), then output as an analog signal with a digital-to-analog converter (DAC) after a
specified time delay. The delayed signal is then filtered (SR640 Low Pass Filter and SR645 High
Pass Filter) and amplified by a Phast Landmark PLB-AMP8 8-channel power amplifier. The
amplified signal is used to drive four speakers (75 Watts each) mounted 51 cm upstream of the
nozzle at the air conditioning chamber.
6.4.2 Results and discussions
1.0
0.8
0.6
0.4
0.2
0.0
2f p
eak
ratio
20151050Time [s]
2f Ratio
Control off Control On
(a) Time history of 2f peak ratio (Gas fuel propane).
Application of fast temperature sensor to combustion control
141
-10
-5
0
5
10
Am
plifi
ed 2
f pea
k ra
tio
20151050Time [s]
2f Ratio (AC)
Control off Control On
(b) Amplified ac components of 2f peak ratio.
0.5
0.4
0.3
0.2
0.1
0.0
Pow
er S
pect
rum
[a.u
. rms2 ]
10008006004002000Frequency [Hz]
Control Off (9.5∼10 s)2f Sensor
0.5
0.4
0.3
0.2
0.1
0.0
Pow
er S
pect
rum
[a.u
. rms2 ]
10008006004002000Frequency [Hz]
Control On (10∼10.5 s)2f Sensor
(c) Power spectrum with and without control
Figure 6.14 2f peak ratio and its power spectra before and after control were applied
on the swirl-stabilized combustor (Propane/Air).
Chapter 6
142
-10
-5
0
5
10M
icro
phon
e [a
rb.u
nits
]
20151050Time [s]
Control Off Control On
Microphone
(a) Time history of microphone signal
5
4
3
2
1
0
Pow
er S
pect
rum
[a.u
. rms2 ]
10008006004002000Frequency [Hz]
Control Off (9.5∼10 s) Microphone
5
4
3
2
1
0
Pow
er S
pect
rum
[a.u
. rms2 ]
10008006004002000Frequency [Hz]
Control On (10∼10.5 s) Microphone
(b) Power spectrum with and without control
Figure 6.15 Microphone signal and its power spectra before and after control were
applied on the swirl-stabilized combustor (Propane/Air).
Application of fast temperature sensor to combustion control
143
Experiments were carried out with a fuel (propane) flow rate of 1.1 SCFM and air flow rate of
46.5 SCFM. Figure 6.14 and 6.15 illustrates the raw data of the 1.4 µm WMS temperature sensor
and microphone and their representative power spectrum with and without control. Figure 6.14 (a)
shows the time history of 2f peak ratio. For this application, the ac components of the temperature
(2f peak ratio) provide the control signal. Thus, the dc component is first calculated and
subtracted from 2f peak ratio, then this ac signal is amplified at constant gain (50 in this case).
The amplified ac components of 2f peak ratio as shown in figure 6.14 (b) are then output to the
dSPACE 1104 real-time control board and delayed by ∆t (0.002 s in this case). The delayed
signal is then filtered by a band pass filter (375Hz and 425 Hz) and amplified to drive four audio
speakers to provide pressure feedback to the air supply and suppress the instabilities. The power
spectrum of 2f ratio and microphone signal are shown in figure 6.14 (c) and 6.15 (c) with and
without control. In the uncontrolled case (9.5~10 s), a 388 Hz fluctuation is clearly seen in both
2f signal and microphone signal, which confirms the ability of the 1.4 µm WMS temperature
sensor for quantitative, accurate identification of flame disturbances. With the controller on, the
temperature oscillations are reduced by a factor of 10 (figure 6.14 (c)) and the pressure
oscillations are reduced by a factor of 2. It should be pointed out, however, that the signal
detected by microphone may contain components from control speakers. Therefore, extra efforts
should be taken to remove such influence when adopting the microphone approach for
acoustically driven flames. In addition, the microphone may also suffer other interference from
background acoustic signal especially in noisy environments. Such limitations are avoided by
detecting temperature fluctuations using the 1.4 µm WMS temperature sensor, and the results
clearly indicate that the temperature sensor is able to determine the correct frequency, phase and
amplitude of the flame fluctuations, and provide the appropriate feedback. The 1.4 µm WMS
temperature sensor could be a very useful tool for active control of combustion.
Chapter 6
144
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
AC
ontr
ol O
n/A
Con
trol
Off
2.5x10-3
2.01.51.00.5Delay time [s]
Baseline
Figure 6.16 Control performances versus controller time-delay
To illustrate the effect of delay-time on the control results, experiments were performed for
different time delays. The measurement time is 20 s (10 s for control off and 10 s for control on)
for each experiment. The FFT analysis was performed every 0.5s. The maximum fluctuation
amplitude between 375 Hz and 425 Hz are calculated for control-on and control-off, and the ratio
AControl On/AControl Off is used as a measure of control performance. Figure 6.16 shows the control
performance vs. delay time (phase). The best control performance is obtained between 2.0~2.25
ms.
Application of fast temperature sensor to combustion control
145
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
AC
ontr
ol O
n/A
Con
trol
Off
10080604020Gain
Baseline
Figure 6.17 Control performances versus amplifier gain
.
Similar experiments were performed to find the optimum gain to suppress the natural combustion
instabilities. Figure 6.17 shows the attenuation of the peak frequency as functions of the amplifier
gain. The maximum reduction was found at the gain of 50. These results demonstrate the
feasibility and utility of this sensor for feedback control on combustion instabilities.
6.5 Summary
The 1.4 µm WMS temperature sensor has been successfully applied to LBO prediction and
thermoacoustic instability control. Four flow conditions and two sensing positions were chosen to
study the flame characteristics near lean blowout limit. For comparison purposes, a microphone is
used to detect acoustic signal during the blowout process. Experiments show that low frequency
components in both acoustic and temperature increase significantly near lean blowout limit, and it
can be used as a precursor for lean blowout. The microphone approach, however, has two
intrinsic limitations: First, it is susceptible to background acoustic signals and thus unsuitable for
lean blowout control under noisy environments. Second, it can only detect global acoustic signals,
so it is unable to detect local flame extinction events. Unlike the microphone, the 1.4 µm WMS
temperature sensor can be applied to a wide range of environments because it does not suffer
Chapter 6
146
background interference. Furthermore, the line-of-sight measurements enable sensitive detection
of any local flame extinction events. It was demonstrated that this sensor has the desired speed
and sensitivity to precisely detect the low frequency fluctuations near lean blow out, for LBO
prediction, which has significant potential for LBO control.
Phase-delay control was implemented using the 1.4 µm WMS temperature sensor to suppress the
combustion natural instabilities in a swirl-stabilized combustor. It is shown that the amplitudes of
natural instabilities are greatly reduced by the phase-delay control at the optimum phase and
amplifier gain. These results offer the strong evidence that the 1.4 µm WMS temperature sensor
has the desired accuracy to monitor dynamics of combustion oscillations and provide effective
feedback signal for active combustion control.
147
Chapter 7: Conclusions and future work The overall objective of this thesis is to design and develop time-resolved and real-time tunable
diode laser sensors with the potential for combustion control. A crucial element in the design of a
tunable-diode-laser optical-absorption-based sensor is the selection of optimum transitions. The
strategy and spectroscopic criteria for selecting optimum wavelength regions and absorption line
combinations are developed for two-line thermometry. The development of this design-rule
approach establishes a new paradigm to optimize tunable diode laser sensors for the target
application. Two single-laser TDL sensors are developed in this thesis. Both sensors are
demonstrated in a heated cell, a forced Hencken burner and a swirl-stabilized spray combustor.
The following sections summarize the major findings of this thesis.
7.1 Summary of the use of design rules to identify the optimum transitions for IC
engine applications
The optimum selection of H2O lines in the 1.25-1.65 µm region for TDL WMS-based
temperature measurements in an internal combustion engine is investigated in chapter 3. The
strategy and spectroscopic criteria are discussed for selection of optimum wavelength regions and
absorption line combinations for the time-varying pressures and temperatures expected during the
compression portion of an engine cycle. We have identified 14 of the water transitions in this
spectral region as promising for this application. Based on these findings, 16 potential line pairs
of H2O were considered for a wavelength-modulated absorption sensor for in-cylinder gas
temperature during the compression stroke. As part of the sensor development effort, expected
performance is modeled for a variety of engine cycles. Simulations show that these line pairs
have good potential for TDL thermometry.
7.2 Design of a single laser absorption sensor for temperature measurements using
direct absorption
Design rules for the selection of optimum transitions for a robust sensor system using two-line
thermometry and a single laser are proposed and elucidated in chapter 4 of this dissertation. The
strategy and spectroscopic criteria for selecting optimum wavelength regions and absorption line
combinations are discussed for direct absorption spectroscopy.
Chapter 7
148
Following the design rules developed for combustion conditions, a short optical path, and direct
absorption, the water vapor spectrum in the 1-2 µm near-infrared (NIR) region is systematically
analyzed to find the best absorption transitions for sensitive measurement of H2O concentration
and temperature. The use of a single laser, even with relatively narrow (~1cm-1) tuning range, can
offer advantages over wavelength-multiplexing techniques. Our investigation reveals that the 1.8
µm spectral region is especially promising, and we have identified 10 of the best line pairs for
water vapor in this spectral region for temperature measurements in flames. Based on these
findings, a pair of H2O transitions near 1.8 µm was selected as an example for the design and
development of a single-laser sensor for simultaneously measuring H2O concentration and
temperature in atmospheric-pressure flames. The greatest advantage of these water line pairs is
the potential to measure with a single scan for one diode laser.
As part of the sensor development effort, fundamental spectroscopic parameters including the line
strength, line-center frequency, and lower state energies of the probed transitions were measured
experimentally to improve the HITRAN database values. Discrepancies between the
experimentally determined spectroscopic parameters and HITRAN/HITEMP database are found
in this region. Thus, it is recommended that the fundamental spectroscopic parameters be
experimentally verified or measured during the development of a practical sensor.
The line pair selected is applicable for temperature measurements in the range from 960 to 3300
K. Demonstration experiments were conducted in a steady and a forced Hencken burner. The
presence of cold boundary layers was shown to impact the temperature inferred assuming
uniform conditions, but a simple assumption of a trapezoidal temperature distribution was shown
to recover an accurate value for the core temperature of the flow. Experiments with forced flames
confirmed the utility of the sensor to monitor temperature fluctuations. In addition, the sensor was
used for closed loop set-point temperature adjustment. Qualitative sensing of temperature
fluctuations and frequencies were demonstrated in liquid fuel swirl spray combustor. The results
offer clear evidence that this sensor system has the flexibility, speed and accuracy to be a useful
tool for fundamental and applied combustion monitoring and control.
Although this sensor has the capability to measure temperature of both the gas and liquid swirling
flames, several limitations are encountered. First, special effort is required for the laser alignment
and nitrogen purge due to the low power of the laser and strong absorption of ambient water
vapor near room temperature. Second, the 1.8 µm direct absorption sensor may require averaging
Conclusions and future work
149
to obtain sufficient SNR under noisy conditions, which limits the measurement bandwidth. Third,
complex data reduction needed for wavelength-scanned direct absorption renders it impractical
for real-time measurements.
7.3 Design of a single laser absorption sensor for temperature measurements using
WMS
The experiences using the 1.8 µm direct absorption scanned-wavelength temperature sensor led to
changes in the design rules to improve the performance. Scanned-wavelength WMS offers
several advantages that improve performance. First, the increased sensitivity of WMS over direct
absorption allows selection of weaker transitions which provide three solutions to problems
encountered with the 1.8 µm sensor. (1) This enables selection of transitions with larger internal
energy E” reducing the sensitivity to cold ambient water vapor. (2) This also provides availability
of fiber-coupled telecommunications lasers, which overlap with the 2ν1, 2ν3 and ν1+ν3 bands of
water vapor. (3) WMS has a simpler data reduction computation than the direct absorption
method and makes real-time measurements possible with a laboratory PC. The design rules to
select optimum water absorption features in the 1-2 micron wavelength range for a WMS sensor
are detailed. The 12 best NIR water transition line pairs for temperature measurements with a
single DFB laser in flames are determined by systematic analysis of the HITRAN simulation of
the water spectra in this spectral region. A specific line pair near 1.4 µm was identified and
investigated experimentally, and the pertinent spectroscopic parameters were determined from
cell experiments. These measurements provide useful improvements to the current spectroscopic
database for H2O for the target transitions and their neighbors. Gas temperature is inferred from
the ratio of the second harmonic signals of two selected H2O transitions. Demonstration
experiments were conducted in a heated cell and a forced Hencken burner. These cell and
laboratory flame experiments confirmed sensitivity and accuracy of the temperature sensor.
Both 1.4 µm and 1.8 µm sensors are based on a single-laser two-line concept, and have the
advantage of a simple optical system. Their use is demonstrated in a liquid-fuel swirl spray
combustor, providing the first demonstration of TDL sensors in such flames. The 1.4 µm WMS
temperature sensor incorporates several improvements over the 1.8 µm sensor design. The 1.4 µm
sensor uses a line pair with relatively weak absorption at room temperature to minimize
interference from ambient air in the measurement path. Second, a very compact and robust setup
Chapter 7
150
was realized by fiber optics, which has advantages for practical application under industrial
conditions. Third, a wavelength modulation spectroscopy technique is employed to shift
measurements to high frequency, and significantly reduces the dominating 1/f noise. Fourth, the
ratio of integrated area is replaced by the ratio of peak-to-peak 2f signals. We have achieved a
real-time repetition rate of 2 kHz with the 1.4 µm sensor. These changes greatly improve the
sensor performance and allow us to further demonstrate the utility of in situ temperature sensing
for combustion control.
7.4 Investigation of the 1.4 µm WMS T sensor for combustion control
To investigate the ability to predict the lean blowout limit, the 1.4 µm WMS temperature sensor
was used for experiments with a swirl-stabilized combustor. Four flow conditions and two
sensing positions were chosen to study the combustion characteristics near lean blowout limit.
For comparison purposes, a microphone is used to simultaneously detect the acoustic signal
during blowout. Experiments show that low frequency components of both acoustic and
temperature measurements increase significantly near lean blowout limit, and can be used as a
precursor for lean blowout. The microphone approach, however, has two intrinsic limitations:
First, it is susceptible to background acoustic signal and thus unsuitable for lean blowout control
under noisy environments. Second, it can only detect global acoustic signal, so it is unable to
detect local flame extinction events. Unlike a microphone, the 1.4 µm WMS sensor can be
applied to a wide range of environments because it does not suffer background interference.
Furthermore, the line-of-sight measurements enable sensitive detection of any local flame
extinction events. It was demonstrated that this sensor has the desired speed and sensitivity to
precisely detect the low frequency fluctuations near lean blowout and has significant potential for
LBO control.
Phase-delay control was implemented using the 1.4 µm WMS temperature sensor to suppress the
natural combustion instabilities in a swirl-stabilized combustor. It is shown that the amplitude of
natural instabilities is greatly reduced by phase-delay control at the optimum phase and amplifier
gain. These results offer strong evidence that the 1.4 µm WMS sensor has the desired accuracy to
monitor dynamics of combustion oscillations and provide effective feedback signal for active
combustion control.
Conclusions and future work
151
7.5 Potential Plan for future work
This thesis has presented the development of systematic sensor design rules, used such rules to
select transitions for a piston engine, and developed two-generations of single-laser temperature
sensors for combustion sensing and control. There are several directions for future work.
1. With the continued improvements in tunable diode lasers and increased availability in laser
wavelengths, a larger region of the spectrum is becoming available. The design rules
developed in this dissertation can be extended to find transitions to achieve better
performance with these new laser choices. (a.) It is possible to select two transitions which
have nearly the same air-broadened halfwidth, self-broadened halfwidth and temperature-
dependence coefficients, thus the effects of the lineshape function (the dependence of species
concentration) could be removed for a 2-line WMS temperature sensor. (b.) By extending 2-
line thermometry to multi-line thermometry, expanded temperature information could be
collected and temperature distribution measurements in non-uniform conditions could be
investigated.
2. The 1.4 µm WMS sensor can precisely detect the approach to the LBO limit in a gas-fueled
swirl-stabilized combustor with significant potential for LBO control. This could be extended
to a LBO control demonstration for both gas- and liquid-fueled combustors.
3. Phase-delay combustion instability control has been implemented in a gas-fueled swirl-
stabilized combustor using the 1.4 µm WMS temperature sensor. The extension to liquid-
fueled flames and further investigations of advanced control strategies such as model-based
adaptive control and fuel modulation are recommended. In addition, further investigations of
sensing at multiple-locations are suggested since it could allow detection of thermal waves.
4. A useful continuation of this work would be the extension to other combustion gas
components (such as NOx and CO). The investigation of the correlation between combustion
instabilities and pollutant emission levels would be very useful. This could enable
combustion control strategies based on CO and/or NOx emission levels, and enable optimized
combustion emissions.
Chapter 7
152
153
Appendix: Architecture of the real-time WMS sensor
The data acquisition and analysis for the real-time 1.4 µm WMS sensor is written using the
C/C++ programming language. The DAC computer utilizes two separate DAQ cards: (1) a Gage
CompuScope 1250, which is used for data acquisition and (2) a National Instruments PCI 6115,
which is used to output 2f ratio (voltage). These are connected as shown in the simplified block
diagram in figure A.1. The input 2f signal is first captured to the on-board memory of the Gage
CompuScope 1250 card. This data is then transferred over the PCI bus into PC RAM for further
analysis. In the bus-mastering mode, the Gage CompuScope 1250 is capable of a PCI bus data
transfer rate up to 100Megabytes/sec. [Gage 2003] After the PC calculates the ratio of the 2f
signals on the target transitions, this ratio is passed to the NI PCI 6115 card where an analog
voltage proportional to the 2f ratio is produced for use by the control computer.
PCI BUS
Input 2f A/D Gage CompusScope 1250
On-board memory Gage CompusScope 1250
PC RAM Data analysis by CPU
D/A NI PCI 6115 Output signal
Output 2f ratio
2f signal 2f ratio
Figure A.1 A simplified block diagram.
Applications Programming Interface (API) routines are supported by Gage CompuScope 1250
and National Instruments PCI 6115. The hardware functionality of the Gage CompuScope 1250
and NI PCI 6115 cards are controlled by these API routines. The detailed descriptions of these
API routines are provided in ref. [Gage 2003; NI 2003], and the manual for their use is found in
ref. [Petzold 1998]. A flow-chart of the sensor logic is illustrated in figure A.2.
Appendix
154
START
Initialize Program Structures and Variables
Initialize the driver and the CompuScope and NI PCI 6115 hardware
Initialize the CompuScope and NI PCI 6115 cards with the desired settings
Triggered? (Falling edge)
Start Capture
Transfer data using DMA method
Data analysis
NI PCI 6115 card output 2f ratio
Triggered? (Rising edge)
More data acquisition?
End
Set the CompuScope and NI PCI 6115 back to initial states
N
Y
N
Y
N
Y
Figure A.2 Flow chart for the data acquisition and analysis program.
Architecture of the real-time WMS sensor
155
The complete source codes is available from the Hanson laboratory library, and only the essential
components specific to the measurements in Chapters 5 and 6 are given here, including: DAQ
card setting, data transfer, peak finding algorithms and output signal.
1. DAQ card setting
Gage CompuScope 1250 and NI PCI 6115 settings are listed below, and comments are provided
between /* and */ using Italic fonts.
Gage CompuScope 1250
/*This parameter is the current trigger source for the data acquisition, and set to external mode
*/
board.source = GAGE_EXTERNAL;
/*This parameter is the GAGE card operating mode, and set to dual channel mode */
board.opmode = GAGE_DUAL_CHAN;
/*This parameter is the sampling rate, and set to 5 MHz */
board.srindex = SRTI_5_MHZ;
/*This parameter is the total data acquisition points, and set to 1000 */
board.depth = 1000;
/*This parameter is the current trigger slope , and set to falling edge as shown in figure 5.24 */
board.slope = GAGE_NEGATIVE;
/*This parameter is the input range for channel A , and set to 500mV */
board.range_a = GAGE_PM_500_MV;
/*This parameter is the input coupling for channel A , and set to DC */
board.couple_a = GAGE_DC;
NI PCI 6115
/* Set output at 0 volts. */
iStatus = AO_VWrite(iDevice, iChan, 0.0);
/* Configure for bipolar mode, internal reference, and external updates. */
iStatus = AO_Configure(iDevice, iChan, iOutputPolarity, iIntOrExtRef, dRefVolts,
iUpdateModeEXT);
/* Setup PFI line for external updates. (PFI0 is setup with AO_Configure.), the trigger slope is
set to rising edge as shown in figure 5.24 */
iRetVal = NIDAQErrorHandler(iStatus, "AO_Configure/ExternalUpdate", iIgnoreWarning);
iStatus = Select_Signal(iDevice, ND_OUT_UPDATE, ND_PFI_0, ND_LOW_TO_HIGH);
Appendix
156
2. Data transfer
/* Adjust this number for data acquisition start position */
location = trigger[0];
/* Adjust this number for data acquisition length for two peaks, as shown in figure 5.24*/
points = 100;
/* gage_transfer_buffer_3 is used to copy “points” data from the channel A to the supplied buffer
beginning from address” location”, this is for the peak 1 data acquisition*/
offset_a1 = gage_transfer_buffer_3 (location, GAGE_CHAN_A, chan_a_12, points);
/*This converts offset_a from a byte pointer to a sample pointer for 12, 14 and 16 bit
CompuScopes, whose samples are 16 bits.*/
offset_a1 = offset_a1 /2;
/* gage_transfer_buffer_3 is used to copy “points” data from the channel A to the supplied buffer
beginning from address” location+690”, adjust the “location+690” so that the program transfer
the data for peak 2*/
offset_a2 = gage_transfer_buffer_3 (location+690, GAGE_CHAN_A, chan_a2_12, points);
/*This converts offset_a from a byte pointer to a sample pointer for 12, 14 and 16 bit
CompuScopes, whose samples are 16 bits.*/
offset_a2 = offset_a2 /2;
3. Peak finding algorithm
The obtained data for 2f peak 1 and 2f peak 2 are best fit using a second order polynomial. 2
i i iy a bx cx= + + (A.1)
Where a, b and c are the polynomial coefficients.
These coefficients can be calculated using the following equations [Bevington 1992]: 2
2 3
2 3 4
1 i i i
i i i i
i i i i
y x xa x y x x
x y x x=
∆
∑ ∑ ∑∑ ∑ ∑∑ ∑ ∑
(A.2)
2
3
2 2 4
1 i i
i i i i
i i i i
N y xb x x y x
x x y x=
∆
∑ ∑∑ ∑ ∑∑ ∑ ∑
(A.3)
Architecture of the real-time WMS sensor
157
2
2 3 2
1 i i
i i i i
i i i i
N x yc x x x y
x x x y=
∆
∑ ∑∑ ∑ ∑∑ ∑ ∑
(A.4)
2
2 3
2 3 4
i i
i i i
i i i
N x xx x xx x x
∆ =∑ ∑
∑ ∑ ∑∑ ∑ ∑
(A.5)
Thus the peak can be found by derive equation (A.1)
2
max02 4
dy b bx y adx c c
= ⇒ = − ⇒ = − (A.6)
The source codes for peak finding are listed below:
for (k = 0; k < LG; k++)
/*for 2f peak 1*/
Tf1_SX1Y1 += k*(double)(board_info->sample_offset_32 - chan_a_12[offset_a1+k]) * sign_res;
Tf1_SX2Y1 += k*k*(double)(board_info->sample_offset_32 - chan_a_12[offset_a1+k]) *
sign_res;
Tf1_SY1 += (double)(board_info->sample_offset_32 - chan_a_12[offset_a1+k]) * sign_res;
/*for 2f peak 2*/
Tf2_SX1Y1 += k*(double)((board_info->sample_offset_32 - chan_a2_12[offset_a2+k]) *
sign_res);
Tf2_SX2Y1 += k*k*(double)((board_info->sample_offset_32 - chan_a2_12[offset_a2+k]) *
sign_res);
Tf2_SY1 += (double)((board_info->sample_offset_32 - chan_a2_12[offset_a2+k]) * sign_res);
/*Calculate polynomial coefficients for 2f peak 1*/
Tf1_a_coeff=((Tf1_SY1)*(SX2)*(SX4)+(SX1)*(SX3)*(Tf1_SX2Y1)+(Tf1_SX1Y1)*(SX3)*(S
X2)-(SX2)*(SX2)*(Tf1_SX2Y1)-(SX1)*(Tf1_SX1Y1)*(SX4)-(Tf1_SY1)*(SX3)*(SX3))/Delta;
Tf1_b_coeff=(LG*(Tf1_SX1Y1)*(SX4)+(Tf1_SY1)*(SX3)*(SX2)+(SX1)*(Tf1_SX2Y1)*(SX2)
-(SX2)*(SX2)*(Tf1_SX1Y1)-(SX1)*(Tf1_SY1)*(SX4)-LG*(SX3)*(Tf1_SX2Y1))/Delta;
Appendix
158
Tf1_c_coeff=(LG*(SX2)*(Tf1_SX2Y1)+(SX1)*(Tf1_SX1Y1)*(SX2)+(SX1)*(SX3)*(Tf1_SY1)
-(SX2)*(SX2)*(Tf1_SY1)-(SX1)*(SX1)*(Tf1_SX2Y1)-LG*(SX3)*(Tf1_SX1Y1))/Delta;
/* Calculate polynomial coefficients for 2f peak 2*/
Tf2_a_coeff=((Tf2_SY1)*(SX2)*(SX4)+(SX1)*(SX3)*(Tf2_SX2Y1)+(Tf2_SX1Y1)*(SX3)*(S
X2)-(SX2)*(SX2)*(Tf2_SX2Y1)-(SX1)*(Tf2_SX1Y1)*(SX4)-(Tf2_SY1)*(SX3)*(SX3))/Delta;
Tf2_b_coeff=(LG*(Tf2_SX1Y1)*(SX4)+(Tf2_SY1)*(SX3)*(SX2)+(SX1)*(Tf2_SX2Y1)*(SX2)
-(SX2)*(SX2)*(Tf2_SX1Y1)-(SX1)*(Tf2_SY1)*(SX4)-LG*(SX3)*(Tf2_SX2Y1))/Delta;
Tf2_c_coeff=(LG*(SX2)*(Tf2_SX2Y1)+(SX1)*(Tf2_SX1Y1)*(SX2)+(SX1)*(SX3)*(Tf2_SY1)
-(SX2)*(SX2)*(Tf2_SY1)-(SX1)*(SX1)*(Tf2_SX2Y1)-LG*(SX3)*(Tf2_SX1Y1))/Delta;
/*Calculate peak position for 2f peak 1*/
index1=-(Tf1_b_coeff)/(2*Tf1_c_coeff);
/*Calculate peak position for 2f peak 2*/
index2=-(Tf2_b_coeff)/(2*Tf2_c_coeff);
/*Calculate 2f peak 1*/
maxpeak1=(Tf1_a_coeff)-(Tf1_b_coeff)*(Tf1_b_coeff)/(4*Tf1_c_coeff);
/*Calculate 2f peak 2*/
maxpeak2=(Tf2_a_coeff)-(Tf2_b_coeff)*(Tf2_b_coeff)/(4*Tf2_c_coeff);
4. Output signal
The 2f peak ratio is output through NI PCI 6115, and the source code is as follow:
/*Specify the 2f background signal */
basevalue=0;
/*Calculate the 2f peak ratio */
peakratio=(basevalue-maxpeak2)/(basevalue-maxpeak1);
/*Output the 2f peak ratio (voltage) through channel “iChan” through NI PCI 6115, as shown in
figure 5.24*/
iStatus = AO_VWrite(iDevice, iChan, peakratio);
/* Set update mode back to initial state. */
Architecture of the real-time WMS sensor
159
iStatus = AO_Configure(iDevice, iChan, iOutputPolarity, iIntOrExtRef, dRefVolts,
iUpdateModeINT);
/* Set PFI line back to initial state. */
iStatus = Select_Signal(iDevice, ND_OUT_UPDATE, ND_INTERNAL_TIMER,
ND_LOW_TO_HIGH);
Appendix
160
161
References
T. Aizawa, “Diode-Laser Wavelength-Modulation Absorption Spectroscopy for Quantitative In
Situ Measurements of Temperature and OH Radical Concentration in Combustion gases”,
Applied Optics, 40, 4894-4903 (2001).
M.G. Allen, "Diode Laser Absorption Sensors for Gas-Dynamic and Combustion Flows,"
Measurement Science and Technology, 9, 545-562 (1998).
M.G. Allen, E.R. Furlong, and R.K. Hanson, "Tunable Diode Laser Sensing and Combustion
Control," in "Applied Combustion Diagnostics," K. Kohseoeinghaus and J.B. Jeffries, ed.
Taylor and Francis, 2002, Chapter 18, pp 479-498.
B.H. Armstrong, “Spectrum line profiles: the Voigt function,” J. Quant. Spectrosc. Radiat.
Transfer 7, 61-88 (1967).
M.P.Arroyo and R.K.Hanson, "Absorption measurements of water-vapor concentration,
temperature, and line-shape parameters using a tunable InGaAs diode laser," Applied Optics,
32, 6104-6116 (1993).
M.P.Arroyo, S.Langlois, and R.K.Hanson, "Diode-laser absorption technique for simultaneous
measurements of multiple gasdynamic parameters in high-speed flows containing water
vapor,” Applied Optics, 33, 3296-3307 (1994).
A. Ateshkadi, V.G. Mcdonell and G.S. Samuelsen “Lean Blowout Model for a Spray-fired Swirl-
stabilized Combustor”, Proceedings of the Combustion Institute, Volume 28, 1281-1288,
2000.
V.G. Avetisov, P. Kauranen, “Two-tone frequency-modulation spectroscopy for quantitative
measurements of gaseous species: theoretical, numerical, and experimental investigation of
line shapes,” Applied Optics, Vol. 35 Issue 24 Page 4705 (August 1996).
D.S. Baer, M.E. Newfield, N. Gopaul, and R.K. Hanson, "Multiplexed diode-laser sensor system
for simultaneous H2O, O2, and temperature measurements," Optics Letters, 19, 1900-1902
(1994).
D.S. Baer, V. Nagali, E.R. Furlong, and R.K. Hanson, “ Scanned- and fixed-wavelength
absorption diagnostics for combustion measurements using multiplexed diode lasers,” AIAA
Journal 34, 489-493 (1996).
C.N. Banwell and E.M. Mccash, “Fundamentals of Molecular Spectroscopy”, four edition, 1994
(McGraw-Hill International (UK) Limited).
Bibliography
162
P.R. Bevington and D.K. Robinson, Data Reduction and Error Analysis for the Physical Sciences
(McGraw-Hill, NY, 1992).
M. Born and E. Wolf, Principles of Optics, 5th ed. (Pergamon, Oxford, 1975) pp. 323-332.
J. Brouwer, B. A. Ault, J. E. Bobrow, and G. S. Samuelsen, “Active control for gas turbine
combustors,” The Twenty-Third Symposium (International) on Combustion, 1087-1092
(1990).
A.M. Bullock, A.N. Dharamsi, W.P. Chu and L.R. Poole, “Measurement of Absorption Line
Wing Structure by Wavelength Modulation Spectroscopy,” Appl. Phys. Lett. 70(10), 1195-
1197(1997).
S. Candel, “Combustion Dynamics and Control: Progress and Challenges,” Proceedings of the
Combustion Institute, Vol. 29, 2002.
A.Y. Chang, M.D. DiRosa, D.F. Davidson, and R.K. Hanson, "Rapid tuning cw laser technique
for measurements of gas velocity, temperature, pressure, density, and mass flux using NO,"
Applied Optics, 30, 3011-3022(1991).
D.E. Cooper, and R. E. Warren, “Frequency modulation spectroscopy with lead-salt diode lasers:
a comparison of single-tone and two-tone techniques,” Appl. Opt. 26 (1987), 3726-3732.
A.N. Dharamsi, A.M. Bullock, “Application of wavelength-modulation spectroscopy in
resolution of pressure and modulation broadened spectra,” Appl. Phys. B 63, 283-292 (1996).
N. Docquier and S. Candel, "Sensor Requirements for Combusiton Control,” in Applied
Combustion Diagnostics, ed. K. Kohse-Höinghaus and J.B. Jeffries, Taylor and Francis, NY,
2002, pp. 534-560.
N. Docquier, and S. Candel. “Combustion control and sensors: a review,” Progress in Energy and
Combustion Science, 28: pp. 107-150 (2002).
T. Fernholz, H. Teichert and V. Ebert, “Digital, Phase-Sensitive Detection for In Situ Diode-
Laser Spectroscopy under Rapidly Changing Transmission Conditions”, Appl. Phys. B 75,
229-236 (2002).
M. Fleifil; A.M. Annaswamy, Z.A. Ghoneim, A.F. Ghoniem, “Response of a laminar premixed
flame to flow oscillations - A kinematic model and thermoacoustic instability results
“ Combustion and Flame. Vol. 106, no. 4, pp. 487-510. Sept. 1996.
E.R. Furlong, V. Nagali, D.S. Baer, and R.K. Hanson, "Combustion Control using a Multiplexed
Diode-Laser Sensor System," paper AIAA-96-0756 at the 34th Aerospace Sciences Meeting,
Reno, Jan. 1996.
References
163
E.R. Furlong, D.S. Baer, M. Webber, and R.K. Hanson, "Combustion Diagnostics and Control
using a Multiplexed Diode Laser System," paper AIAA-96-2763 at 32nd AIAA Joint Prop.
Conf., Lake Buena Vista, FL, July 1996.
E.R. Furlong, D.S. Baer, and R.K. Hanson, "Combustion Control and Monitoring using a
Multiplexed Diode-Laser Sensor System," 26th Symposium (International) on Combustion,
The Combustion Inst., Pittsburgh, 1996, pp.2851-2858.
E.R. Furlong, R.M. Mihalcea, M. Webber, D.S. Baer, and R.K. Hanson, "Combustion Sensing
and Control using Wavelength-Multiplexed Diode-Lasers," paper AIAA-97-0320 at AIAA
35th Aerospace Sciences Meeting, Reno, Jan. 6-9, 1997.
E.R. Furlong, R.M. Mihalcea, M.E. Webber, D.S. Baer, and R.K. Hanson, "Diode-Laser Sensor
System for Closed-Loop Control of a 50-kW Incinerator," Proc. of SPIE, Vol. 3172, paper 33;
presented at 42nd annual SPIE meeting, San Diego, July, 1997.
E.R. Furlong, D.S. Baer, and R.K. Hanson, "Diode-Laser Sensors for Real-Time Control of
Temperature and H2O in Pulsed Combustion Systems," paper AIAA-98-3949 at 34th
AIAA/ASME/SAE/ASEE Joint Propulsion Conference, July 13-15, 1998, Cleveland, OH.
E.R. Furlong, D.S. Baer, and R.K. Hanson, "Real-Time Adaptive Combustion Control Using
Diode-Laser Absorption Sensors," 27th Symp. (International) on Combustion, The
Combustion Institute, pp. 103-111, (1998).
GaGe (A Tektronix Technology Company) Volume 1: CompuScope C/C++ Software
Development Kit Manual, 2003.
GaGe (A Tektronix Technology Company) Volume 2: API Reference Manual, 2003
U. Gustafsson, G. Somesfalean, J. Alnis, S. Svanberg, “Frequency-Modulation Spectroscopy with
Blue Diode Lasers,” Applied Optics, Vol. 39 Issue 21 Page 3774 (July 2000).
D.C. Hovde, J.T. Hodges, G.E. Scace, and J.A. Silver, “Wavelength-Modulation Laser
Hygrometer for Ultrasensitive Detection of Water Vapor in Semiconductor Gases,” Applied
Optics, 40, 829-839 (2001).
J. Humlicek, “Optimized computation of the Voigt and complex probability functions,” J. Quant.
Spectrosc. Radiat. Transfer 27, 437-444 (1982).
D.S. John, and S. Samuelsen, “Robust Optimal Control of a Natural Gas-Fired Burner for the
Control of Oxides of Nitrogen (NOx),” Combust. Sci. Technol., vol. 128, pp. 1-21, 1997.
A. Kemal and C.T. Bowman, “Real-time adaptive feedback control of combustion instability,”
26th Symposium (International) on combustion/The combustion institute, pp. 2803-2809,
1996.
Bibliography
164
P. Kluczynski and O. Axner, “Theoretical description based on Fourier analysis of wavelength-
modulation spectrometry in terms of analytical and background signals,” Applied Optics,
5803-5815, Vol. 38, No. 27, 1999.
P. Kluczynski, A.M. Lindberg, and O. Axner, “Background signals in wavelength-modulation
spectrometry with frequency-doubled diode-laser light I. Theory,” Applied Optics, 783-793,
Vol. 40, No. 6, 2001.
P. Kluczynski, A.M. Lindberg, and O. Axner, “Background signals in wavelength-modulation
spectrometry with frequency-doubled diode-laser light II. Experiment,” Applied Optics, 794-
804, Vol. 40, No. 6, 2001.
K. Kohse-Höinghaus and J.B. Jeffries: Applied Combustion Diagnostics (Taylor and Francis, NY,
2002).
G. Li and E. Gutmark, “Experimental Study of Flow Patterns and Reaction in a Multiple Swirl
Spray Combustor” AIAA-2003-489 41st Aerospace Sciences Meeting and Exhibit, Reno,
Nevada, Jan. 6-9, 2003.
G, Li. “Emissions, Combustion Dynamics, and Control of a Multiple Swirl Combustor” Ph.D.
dissertation, 2004.
H.J. Li, G.B. Rieker, X. Liu, J.B. Jeffries, and R.K. Hanson, “Extension of Wavelength
Modulation Spectroscopy to Large Modulation Depth for Diode Laser Absorption
Measurements in High Pressure Gases” In preparation.
T. Lieuwen, H. Torres, C. Johnson, and B.T. Zinn, “A Mechanism for Combustion Instabilities in
Premixed Gas Turbine Combustors,” Journal of Engineering for Gas Turbines and Power, Vol.
123, No. 1, pp. 182-190 2001.
T. Lieuwen, “Modeling premixed combustion- Acoustic wave interactions: A review” J ournal of
propulsion and power Vol. 19, No. 5, September–October 2003.
J.T.C. Liu, J.B. Jeffries, and R.K. Hanson, "Wavelength Modulation Absorption Spectroscopy
with 2f Detection using Multiplexed Diode Lasers for Rapid Temperature Measurements in
Gaseous Flows," App. Phys. B 78, 503-511 (2004).
J.T.C. Liu, J.B. Jeffries, and R.K. Hanson, "Large-Modulation-Depth 2f Spectroscopy with Diode
Lasers for Rapid Temperature and Species Measurements in Gases with Blended and
Broadened Spectra," App. Optics 43, 6500-6509 (2004).
X. Liu, X. Zhou, J.B. Jeffries, and R.K. Hanson, "Experimental Study of H2O Spectroscopic
Parameters in the Near-IR," paper AIAA-2005-0829 at 43rd Aerospace Sciences Meeting,
Reno, NV, Jan. 2005.
References
165
X. Liu, X. Zhou, J.B. Jeffries, and R.K. Hanson, " Precision Measurements of Line strength,
Pressure Broadening, and Pressure Shift of H2O Transitions near 1.34 and 1.39 micron for
Measurement of Gas Temperature during the Compression Stroke of an IC-engine," In
preparation.
R.J. Martin and N.J. Brown “Nitrous Oxide Formation and Destruction in Lean, Premixed
Combustion” Combust. Flame 80 238–55 (1990).
F. Mayinger, O. Feldmann, “Optical Measurements---Techniques and Applications” Series: Heat
and Mass Transfer 2nd corr. ed., ISBN: 3-540-66690-7, 2001.
KR McManus, T. Poinsot, S.M. Candel, “A review of active control of combustion instabilities,”
Progress in Energy and Combustion Science 19:1 pp. 1-29, February 1993.
R.M. Mihalcea, D.S. Baer, and R.K. Hanson, “A diode-laser absorption sensor system for
combustion emission measurements,” Meas. Sci. Technol. 9, 327-338 (1998).
T.M. Muruganandam, B. Kim, R. Olsen, M. Patel, B. Romig and J.M. Seitzman,
“Chemiluminescence Based Sensors for Turbine Engines”, paper AIAA-2003-4490, at the
39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Huntsville, Alabama,
July 20-23, 2003.
T.M. Muruganandam and J.M. Seitzman, “Origin of Lean Blowout Precursors in Swirled Gas
Turbine Combustors”, paper AIAA 2005-1163 at the 43rd AIAA Aerospace Sciences
Meeting, Reno NV, Jan 10-13, 2005.
V. Nagali and R.K. Hanson, "Design of a diode-laser sensor to monitor water vapor in high-
pressure combustion gases," Applied Optics, 36, 9518-9527(1997).
S. Nair, T. Lieuwen “Acoustic detection of imminent blowout in pilot and swirl stabilized
combustors”, Proceedings of ASME/IGTI Turbo Expo 2003 GT-2003-38074, June 16-19,
Atlanta, Georgia.
S.Nair, T.M. Muruganandam, R. Olsen, A. Meyers, J. Seitzman, B.T. Zinn, T. Lieuwen, T.Held,
and H.Mongia “Lean Blowout Detection in a Single Nozzle Swirl Cup Combustor”, paper
AIAA 2004-0138 at the 42nd AIAA Aerospace Sciences Meeting, Reno, NV, Jan 5-8, 2004.
National Instruments NI-6115 manual, 2003.
J.J. Olivero, R.L. Longbothum “Empirical fits to the Voigt line width: A brief review,” JQRST
17:233-236, 1977.
X. Ouyang and P.L. Varghese, “Reliable and efficient program for fitting Galatry and Voigt
profiles to spectral data on multiple lines,” Appl. Opt. 28, 1538-1545 (1989).
X. Ouyang and P.L. Varghese, “Selection of spectral lines for combustion diagnostics,” Applied
Optics, 29, 4884-4890 (1990).
Bibliography
166
S. Park, A.M. Annaswamy, and A. Ghoniem, “Heat Release Dynamics Modeling of Kinetically
Controlled Burning,” Combusiton and Flame, Vol. 128, No. 5 pp.217-231, 2002.
C.O. Paschereit, E.Gutmark, and W. Weisenstein, “Control of Thermoacoustic Instabilites and
Emissions in an Industrial-Type Gas Turbine Combustor,” Proc. Combust. Inst., vol. 27, pp.
1817-1824, 1998.
C. Petzold “Programming Windows”, 5th Edition ISBN: 1-57231-995-X; Microsoft Press;
November 11, 1998; 1479 pages.
L.C. Philippe and R.K. Hanson, "Laser Diode Wavelength-Modulation Spectroscopy for
Simultaneous Measurement of Temperature, Pressure and Velocity in Shock-Heated Oxygen
Flows," Applied Optics 32, 6090-6103 (1993).
S. Prakash, S. Nair , T. M. Muruganandam, Y. Neumeier, T. Lieuwen , J. Seitzman and B. Zinn,
“Acoustic Sensing and Mitigation of Lean Blow Out in Premixed Flames,” paper AIAA-
2005-1420 at the 43rd Aerospace Sciences Meeting, Reno, NV, Jan 10-13, 2005.
J. Reid and D. Labrie, “Second-Harmonic Detection with Tunable Diode Lasers – Comparison of
Experiment and Theory,” Appl. Phys. B 26, 203-210 1981.
H. Riris, C.B. Carlisle, R.E. Warren, and D.E. Cooper, “Signal-to-noise enhancement in
frequency-modulation spectrometers by digital signal processing,” Opt. Lett. 19, 144-146
(1994).
L.S. Rothman, C.P. Rinsland, A. Goldman, S.T. Massie, D.P. Edwards, J.-M. Flaud, A.Perrin, C.
Camy-Peyret, V. Dana, J.-Y. Mandin, J. Schroeder, A. McCann, R.R. Gamache, R.B.
Wattson, K. Yoshino, K. Chance, K. Jucks, L.R. Brown, V. Nemtchinov, and P. Varanasi,
"The HITRAN Molecular Spectroscopic Database and HAWKS (HITRAN Atmospheric
Workstation): 1996 Edition," J. Ouant. Spectrosc. and Rad. Transfer 60, 665-710 (1998).
L.S. Rothman, A. Barbe, D.C. Benner, L.R. Brown, C. Camy-Peyret, M.R. Carleer, K. Chance, C.
Clerbaux, V. Dana, V.M. Devi, A. Fayt, J.-M. Flaud, R.R. Gamache, A. Goldman, D.
jacquemart, K.W. Jucks, W.J. Lafferty, J.-Y. Mandin, S.T. Massie, V. Nemtchinov, D.A.
Newnham, A. Perrin, C.P. Rinsland, J. Schroeder, K.M. Smith, M.A.H. Smith, K. Tang, R.A.
Toth, J. Vander Auwera, P. Varanasi, K. Yoshino: J. Quant. Spectrosc. Radiat. Transfer 82, 5
(2003).
S.T. Sanders. “ Diode-Laser Sensors for Harsh Environments with Application to Pulse
Detonation Engines” Ph.D. dissertation, 2004.
S.T. Sanders, D.S. Baer, and R.K. Hanson. "Diode laser absorption sensor for measurements in
pulse detonation engines," Paper number 2000–0358, AIAA 38th Aerospace Sciences
Conference, Reno, NV, January 2000.
References
167
S.T. Sanders, D.W. Mattison, J.B. Jeffries, and R.K. Hanson, "Rapid Temperature-Tuning of a
1.4 mm Diode Laser with Application to High Pressure H2O Absorption Spectroscopy," Opt.
Letters 26, 1568-1570 (2001).
S.T. Sanders, J.B. Jeffries, J. Wang, and R.K. Hanson, "Wavelength-Agile Diode Laser Sensors
for Monitoring Gas Properties in Harsh Environments," Laser Applications to Chemical and
Environmental Analysis, OSA Topical Meeting, Feb. 2002.
S.T. Sanders, D.W. Mattison, J.B. Jeffries, and R.K. Hanson, "Sensors for High-Pressure, Harsh
Combustion Environments Using Wavelength-Agile Diode Lasers," 29th Int. Symp. on
Combustion, Sapporo, Japan, July 2002; Proc. of the Comb. Inst. 29 2661-2667(2002).
S.T. Sanders, D.W. Mattison, L. Ma, J.B. Jeffries, and R.K. Hanson, "Wavelength-Agile Diode
Laser Sensing Strategies for Monitoring Gas Properties in Optically Harsh Flows: application
in Cesium-Seeded Pulse Detonation Engine," Optics Express 10, 505-514 (2002).
S.T. Sanders, T. Kim, J.B. Ghandhi, "Gas temperature measurements during ignition in an HCCI
engine," SAE 2003-01-744, 2003.
F. Schreier, “The Voigt and complex error function: a comparison of computational methods,” J.
Quant. Spectrosc. Radiat. Transfer 48, 743-762 (1992).
J.M. Seitzman, R. Tamma, and B.T. Scully, “Broadband infrared sensor for active control of high
pressure combustor,” AIAA 1998-0401, 36th Aerospace Sciences Meeting, Reno (1998).
S. Schilt, L. Thevenaz, and P. Robert, “Wavelength modulation spectroscopy: combined
frequency and intensity laser modulation,” Applied Optics, Vol. 42, No. 33 6728-6738 (2003).
C.R. Shaddix, "Correcting Thermocouple Measurements for Radiation Loss: A Critical Review,"
Proceedings of the 33rd National Heat Transfer Conference, August 15-17, 1999,
Albuquerque, New Mexico.
J.A. Silver, “Frequency-modulation spectroscopy for trace species detection: theory and
comparison among experimental methods,” Appl. Opt. 31, 707-717 (1992).
J.A. Silver and D.J. Kane, “Diode laser measurements of concentration and temperature in
microgravity combustion,” Meas. Sci. Technol. 10, 845-852 (1999).
M. Thiruchengode, S. Nair, S. Prakash, D. Scarborough, Y. Neumeier, J. Jagoda, T. Lieuwen, J.
Seitzman and B. Zinn, “An Active Control System for LBO Margin Reduction in Turbine
Engines,” paper AIAA-2003-1008 at the 41st AIAA Aerospace Sciences Meeting, Reno, NV,
January 6-9, 2003.
M. Thiruchengode, S. Nair, R. Olsen, Y. Neumeier, A. Meyers, J. Jagoda, T. Lieuwen, J.
Seitzman and B. Zinn, “Blowout Control in Turbine Engine Combustors,” paper AIAA-2004-
0637 at the 42nd AIAA Aerospace Sciences Meeting, Reno, NV, Jan 5-8, 2004.
Bibliography
168
N. H. Trans, R. Kachru, P. Pillet, H. B. van Linden van den Heuvell, T. F. Gallagher, J. P. Watjen,
“Frequency-modulation spectroscopy with a pulsed dye laser: experimental investigations of
sensitivity and useful features,” Applied Optics, Vol. 23 Issue 9 Page 1353 (May 1984).
J. Wang, S.T. Sanders, J.B. Jeffries, and R.K. Hanson, "Oxygen Measurements at High Pressures
using Vertical Cavity Surface-Emitting Lasers," App. Phys. B. 72, 865-872, (2001).
M.E. Webber, J. Wang, S.T. Sanders, D.S. Baer, and R.K.Hanson. "In Situ Combustion
Measurements of CO, CO2, H2O and Temperature Using Diode Laser Absorption Sensors,"
Proc. Comb. Inst., 28. 407-413(2000).
E.E. Whiting. "New Empirical Approximation to the Voigt Profile," J. Quant. Spect. Radiat.
Transfer, 16, 611-614(1976).
X. Zhou, X. Liu, J.B. Jeffries, and R.K. Hanson, "Development of a Sensor for Temperature and
Water Vapor Concentration in Combustion Gases using a Single Tunable Diode Laser," Meas.
Sci. and Technology 14, 1459-1468 (2003).
X. Zhou, J.B. Jeffries, and R.K. Hanson, “Development of a Fast Sensor for Temperature in
Combustion Gases Using a Single Tunable Diode Laser”, Submitted to Applied Physics B
(2005).