-
A Study of Quantitative Concentrations of Hydroxyl (OH) in
Laminar Flat Flames Using Planar Laser Induced Fluorescence
(PLIF)
A Thesis Presented
By
Adrienne Murphy Jalbert
To
The Department of Mechanical and Industrial Engineering
In partial fulfillment of the requirements
For the degree of
Master of Science
In
Mechanical Engineering
In the field of
Thermofluids
Northeastern University
Boston, Massachusetts
August, 2011
-
-2-
Abstract In these times of striving for cleaner, more efficient
fuels, research is needed to provide understanding that will allow
technologies to improve. Planar laser induced fluorescence (PLIF)
is a popular, non-intrusive technique for studying flame structure.
A radical species, hydroxyl (OH), is known to be found in the
region between burned and unburned gases in a flame. By tuning the
PLIF system to the proper wavelength, OH can be excited and its
fluorescence can be imaged, resulting in a map of concentrations.
Knowing where OH is located allows the flame front to be defined.
Furthermore, if the concentrations of OH can be quantified then
this numerical data can be used in theoretical models.
Methane is a well documented fuel and has data readily available
for concentrations of OH at specific distances above the burner
surface. This data was used to create a calibration curve which
could then be applied to PLIF images to quantify the concentration
of OH in fuels that do not have documented data. Hydrogen was
selected as the fuel to be studied because of its promising future
in alternatives to fossil fuels. Only by increasing our
understanding of the fuel’s burning behavior will we be able to use
it most efficiently. Data collected can be used to verify
theoretical models that predict burning efficiency, emissions and
work on how to improve them.
The works of past students were verified and improved upon. As
each step of the establishment of the PLIF system was repeated to
gain a full understanding of the system, it was well documented.
All of these documents were collected into a thorough manual of
procedures that will help the next student quickly and easily learn
the system.
-
-3-
Table of Contents List of Figures and Tables
..............................................................................................................................
4
Acknowledgements
.......................................................................................................................................
5
1. Background
...........................................................................................................................................
6
1.1. Purpose
.........................................................................................................................................
6
1.2. Literature Review
..........................................................................................................................
6
1.3. PLIF Basics
.....................................................................................................................................
7
1.3.1. Quantum Mechanics
.............................................................................................................
8
1.3.2. OH Specifics
...........................................................................................................................
9
2. Experiment
..........................................................................................................................................
10
2.1. Experimental Setup
.....................................................................................................................
10
2.2. Wavelength Selection
.................................................................................................................
11
2.3. Concentration Calibration
...........................................................................................................
14
2.3.1. Thermocouple Calibration Method
....................................................................................
15
2.3.2. LaVision Calibration Method
...............................................................................................
16
2.3.3. Calibration Method Comparison
.............................................................................................
16
2.3.4. Concentration Calibration Conclusion
....................................................................................
19
2.4. Transient System
.........................................................................................................................
20
2.4.1. System Establishment
.........................................................................................................
20
2.4.2. Current System Status
.........................................................................................................
21
3. Experimental Results
..........................................................................................................................
23
3.1. Concentration Calibration Validation
.........................................................................................
23
3.2. Fuel Flowrate Comparison
..........................................................................................................
27
3.3. Equivalence Ratio Comparison
...................................................................................................
30
4. Condensed Manual
.............................................................................................................................
35
5. Conclusions and Future Recommendations
.......................................................................................
36
References
..................................................................................................................................................
38
-
-4-
List of Figures and Tables
Figure 1.1 - Basic concept of PLIF system
.....................................................................................................
8Figure 1.2 - Hypothetical energy level diagram, adapted from [12]
.............................................................
9Figure 1.3- OH fluorescence spectra showing Q1(8) transition
..................................................................
10Figure 2.1 - PLIF system
..............................................................................................................................
11Figure 2.2 - OH spectra of (1-0) band from 282nm to 284nm
....................................................................
12Figure 2.3 - Incorrect OH spectra without FCU enabled
.............................................................................
13Figure 2.4 - Calibration image with horizontal bands of constant
concentration ...................................... 14Figure 2.5 -
Concentration calibration sample plot
....................................................................................
15Figure 2.6 - LaVision calibration data (methane/air, P=1 bar,
Φ=1) ...........................................................
16Figure 2.7 - Temperatures reported by a radiation corrected
thermocouple at varying heights above a burner [19]
..................................................................................................................................................
17Figure 2.8- Thermocouple correction comparison of Fredette [4] to
Cattolica [19] .................................. 18Figure 2.9 -
Comparison between LaVision and thermocouple calibrations
.............................................. 19Figure 2.10 -
Constant volume cylinder in place for experiments
..............................................................
20Figure 2.11 - Andrews' successful timing system final images [12]
............................................................
21Figure 2.12 - Weak transient image, methane/air, P=1atm, Φ=1
..............................................................
22Figure - - H2 air=13.8) ...... 24Figure 3.2 - Comparison of three
stoichiometric hydrogen-air flames with same calibration data
........... 24Figure 3.3 - Comparison between two independent
concentration calibrations ......................................
25Figure 3.4 - Comparing average concentration to maximum
concentration .............................................
26Figure 3.5 - Inconsistent PLIF image taken with clogged burner
................................................................
26Figure 3.6 - Stoichiometric methane calibration image
.............................................................................
27Figure 3.7 - Calibrated stoichiometric methane-air flames with
varying flowrates on common scale ...... 28Figure 3.8 - Calibrated
stoichiometric methane-air flames with varying flowrates on
independent scales
....................................................................................................................................................................
28Figure 3.9 - Image post processing inconsistency test, all images
should match ....................................... 29Figure 3.10
- Comparison of [OH] varying with total flowrates of methane-air
and hydrogen-air ............ 30Figure 3.11 - H2-air flames with
varying equivalence ratios displayed on a common concentration
scale
....................................................................................................................................................................
31Figure 3.12 - Effect of equivalence ratio on value and location
of OH concentration ............................... 32Figure 3.13 -
Flowmeter calibration curve
..................................................................................................
33Figure 3.14 - Adjusted calibration curve
.....................................................................................................
34Figure 5.1 - (L to R) Fredette's flame, clogged burner flame,
post HCl bath flame .................................... 37
-
-5-
Acknowledgements First and foremost I would like to acknowledge
my adviser, Dr. Hameed Metghalchi. As busy as you are as the chair
of mechanical and industrial engineering at Northeastern
University, you still always had time to offer me assistance and
positive encouragement. I would also like to thank you for telling
me about the BS/MS program and inviting me to join your laboratory.
I would never have reached this accomplishment if not for you.
To my lab mates, Casey Bennett and Ali Moghaddas, thank you for
always dropping what you were doing and helping me out. Whether it
was with equipment set up or a theoretical concept that I did not
quite understand, neither of you ever hesitated to assist in any
way you could. And thank you for keeping me sane when things got
stressful.
Dr. Mohammad Janbozorgi and Ghassan Nicolas, my theoretical
combustion colleagues, thank you for your support. Without the
crash courses in chemical kinetics, Fortran and mathematics I would
not have the understanding behind my experimental research which I
now do.
My BS/MS predecessors, RJ Andrews and Colin Fredette, thank you
for paving the way for me by setting up this system. I can only
hope that my work will be as helpful to future students as yours
was to me.
Many of the Mechanical and Industrial Engineering staff have
come to my aid when I have needed them most. Bridget Smyser, you
were always able to help me find chemicals that I needed in a pinch
and keep things moving with laser safety installations. Joyce
Crain, your speed and efficiency more than once got me out of a
stressful jam. John Doughty, you could always help me find a tool
or fix a part when no one else could.
Finally I would like to thank my supportive family. You have
kept me focused on the end goal and picked me up at the toughest
times when I did not think that I would ever graduate!
-
-6-
1. Background
1.1. Purpose The purpose of this research is to study the
concentration of hydroxyl (OH) in laminar flames to get a better
understanding of flame structure and behavior. OH is found in the
region between burned and unburned gases and can therefore be used
to mark the flame front. This thesis will combine the master’s
thesis works of Raymond Andrews and Colin Fredette, alumni of the
Northeastern University laminar combustion laboratory. Andrews’
thesis was to set up a planar laser induced fluorescence (PLIF)
system and design a timing system for transient combustion events
in 2008. Fredette then developed a calibration method for
quantifying concentrations of radical species using a stationary
flame in 2009.
Along with combining past efforts of this lab, this thesis will
also serve as a user’s guide to the PLIF system. Previous students
have never overlapped with each other. This has resulted in each
user of the PLIF equipment needing to teach themselves the entire
system from scratch. While this is a very thorough way to ensure
that the entire system is understood, it is also very time
consuming. Learning a system of this complexity can easily take six
months when thesis work is only being done part time. A manual of
eighteen step by step procedures has been put together to accompany
this thesis to help future students learn the system quickly and
then be able to focus the bulk of their time on new developments in
their research.
1.2. Literature Review Much work has been done studying the
concentration of OH in flames using PLIF. Labs around the world
have used laser diagnostics to study OH in both methane/air flat
flames and transient flames. A thorough literature search was
conducted using the Engineering Village literature database
Compendex and the online search engine Google Scholar. The review
first targeted quantitative flat flame PLIF then moved on to the
more complex transient investigations.
A. Arnold et al. of Paul Scherrer Institut (PSI) [1] in
Switzerland used a combination of 2-D PLIF and 1-D UV absorption
spectroscopy to quantify the concentration of hydroxyl in a flat
flame. The energy profile of the laser sheet was imaged using 1-D
UV absorption spectroscopy before and after passing through the
flat flame. Looking at the difference in energies at a certain
height, the energy absorption of the flame at each height above the
burner could be determined. Using the energy absorption and the
known flame conditions, a 1-D algorithm could be used to calculate
the concentration of the species of interest at a specific height
[2]. This quantity could then be applied to the 2-D PLIF image of
the flat flame for each height, quantifying the concentration. The
data from Arnold’s work at PSI is actually reported in the user’s
manual of the calibration burner [3] used in this thesis. According
to LaVision (the company that developed much of the PLIF system
equipment), Arnold’s data is to be used as calibration data for
quantifying OH concentrations.
C. Fredette of Northeastern University in Boston developed
another quantification method for the concentration of OH in flat
flames for his master’s thesis [4].Using a thermocouple, Fredette
measured the temperature of the flame at a specific height above
the burner. This temperature, along with the known environmental
conditions and flame species, could be entered into STANJAN [5], a
chemical
-
-7-
equilibrium computing program, to calculate the concentration of
OH at the point of the thermocouple bead. See the Thermocouple
Calibration Method section for more detail.
This first part of the review confirms that quantification of OH
in methane/air flat flames has been sufficiently studied and
documented already. Studying flat flames is a stepping stone to
being able to analyze a transient combustion system. Data found by
past flat flame works should be able to be applied to the more
involved transient systems.
To decide what this thesis could do to add to knowledge in the
combustion field it first had to be determined what has already
been done and where the field currently stands in transient flame
studies. Mansour of Cairo University studied flame kernel structure
using OH laser induced predissociation fluorescence (LIPF) in 2007
[6]. The results of this study report qualitative OH concentrations
only. In 2010, Müller of the Institute of Reactive Flows and
Diagnostics in Germany used PLIF to study OH in flame kernels
within an optically accessible internal combustion engine [7].
Again only qualitative data was collected. This literature review
concluded that there is a void in current research in quantitative
concentration profiles of laminar spherically propagating
flames.
1.3. PLIF Basics Planar laser induced fluorescence (PLIF) is a
technique commonly used to optically study the structure of flames.
It can be used to visualize concentration and temperature fields.
Typically, a tunable laser system is used that can be set to a
precise wavelength (often accurate to within 0.001 nm) [8]. Once
tuned, the beam is converted into a planar sheet which passes
through the combustion event and excites a species of interest
through photon absorption. The photon will only be absorbed if its
energy is equal to the change in energy needed for a molecule to
change energy states [9]. By knowing the energy required by the
molecule of interest to change energy states, the laser can be
tuned to the best wavelength that will allow excitation of only
that species. Absorption will be followed by spontaneous emission
of the photon. When this spontaneous emission occurs, photons are
released at random in all directions. Any photons that happen to
travel into the camera lens will then be captured by the camera’s
intensifier. This means that the images taken do not collect all of
the fluorescence but only a portion of it. Since spontaneous
emission is completely random in direction, the probability of a
photon going in any direction is equal to any other direction.
Therefore, the image taken represents a relative distribution of
molecules, not the total fluorescence of an event. See Figure 1.1
for a general overview of the PLIF system:
-
-8-
Figure 1.1 - Basic concept of PLIF system
PLIF is a popular option for researchers because of its
selectivity, sensitivity and non-intrusiveness. The selectivity
comes from the tunable laser system. Being able to tune the
wavelength to the thousandth of a nanometer (picometer) allows
scientists to target exactly the species they are interested in,
even if it transitions near a different species. This method,
according to the LaVision Tunable LIF manual [8], can detect
molecules in the sub-ppm range, making it a very sensitive data
acquisition technique. Finally the fact that the procedure does not
interfere with the combustion event is very beneficial. It is not
necessary to add tracers to the air/fuel mixture or to disturb the
flame in any way. This way the data collected will be comparable to
real world situations without added experimental inaccuracies.
1.3.1. Quantum Mechanics In order to understand the data
collected by a PLIF system, one must first understand the basic
principles of quantum mechanics. According to quantum mechanics a
molecule cannot exist with an arbitrary amount of energy. There are
discrete levels of energy specific to each type of molecule known
as stationary states that they must exist within. When energy is
absorbed or emitted by a molecule it can jump to a different
stationary state. Absorption occurs when a molecule gains energy
equal to the amount of energy required to make it to jump from a
lower energy state to a higher energy state. Emission occurs when
the molecule, for some reason, releases energy and causes it to
drop from a higher energy state down to a lower energy state.
Causes of emission may include collision with another molecule or
simply that the excited state is unstable and the molecule returns
to a more stable lower energy state. When absorption and emission
occur back to back it is the phenomenon known as fluorescence
[10].
Diatomic molecules such as OH have three different types of
energy: electronic, vibrational and rotational. According to the
Born-Oppenheimer approximation, these three types of energy are all
independent of each other. There are two different spectra that can
be observed. The excitation spectrum is studied by sweeping the
laser through a range of wavelengths for varying absorption and
observing the amount of emission. This spectrum is covered in more
detail in the Wavelength Selection section where peak finding is
explained. Then there is the emission spectrum, which is studied by
tuning the laser to one specific absorption wavelength and
observing the fluorescence of different emission
-
-9-
lines [8]. The one specific wavelength is selected by studying
the excitation spectrum and finding at what wavelength the maximum
excitation of the desired molecule occurs.
Quantum numbers are used to state the energy level of a
molecule. While there are many different forms for reporting
quantum numbers, they are often given in pairs as (v’, v’’) to give
the (excited, ground) vibrational states. Vibrational separation
energies are evenly spaced and much larger than rotational
separation energies, which are not evenly spaced. This results in
clusters of rotational energy levels separated by large vibrational
energy gaps approximately 1000 times larger than rotational
separations [11] (see Figure 1.2, not to scale)
Figure 1.2 - Hypothetical energy level diagram, adapted from
[12]
1.3.2. OH Specifics Hydroxyl was selected as the radical to be
studied in this work because it is known to be found in the region
between burned and unburned gases in flames. This allows it to be
used as a flame front marker. Northeastern’s combustion lab has a
background in studying this species. The current PLIF system has
been optimized by previous students to work with OH, [4] and
[12].
The Q1(8) line within the OH A2Σ+←X2Π (1-0) transition band
targeted in this study occurs when a wavelength of approximately
283.55 nm is absorbed. This absorption raises the molecule from its
X2Π ground state to the A2Σ+
Figure 1.3
excited state from which it spontaneously emits photons with a
wavelength around 308 nm. The Q1(8) line was recommended by
LaVision because of its low temperature sensitivity [13]. The Q1(8)
transition is expected to give off a strong fluorescence signal
relative to neighboring transitions. below shows the Q1(8) emission
level relative to other peaks at nearby wavelengths measured in
arbitrary units.
-
-10-
Figure 1.3- OH fluorescence spectra showing Q1(8) transition
It can be seen that Q1(8) is one of the highest peaks in the
spectra. This fact coupled with its temperature insensitivity makes
it an ideal wavelength to work at. The stronger the emission of OH
is in experimental images, the higher the resolution of the
concentration can be. For this work, the Q1(6) line actually ended
up being used, as will be discussed in further detail in following
sections on Wavelength Selection.
2. Experiment
2.1. Experimental Setup The laminar flame lab’s PLIF system
consists of a pump laser, tunable dye laser, energy monitor, sheet
optics, combustion device and camera. The pump laser is a Nd:YAG
Spectra-Physics Quanta-Ray Lab 190. By adjusting the position of
the Nd:YAG crystals, the solid lasing medium within the laser’s
harmonic generator, the wavelength can be set to 1064nm, 532nm,
355nm or 266nm [14]. It is set for the current work to generate a
532nm wavelength beam. This beam leaves the pump laser and is
directed into the dye laser by two right angle mirrors. The Sirah
PrecisionScan SL dye laser uses a Rhodamine 6G liquid lasing medium
dye to finely tune the pumped beam to a fundamental wavelength of
566nm. Before leaving the dye laser, the 566nm beam passes through
a frequency doubling crystal which halves the wavelength to a final
output of 283nm [15]. This ultraviolet beam passes through an
energy monitor, where about 5% of each pulse’s energy is diverted,
measured and recorded for future corrections [16]. The remaining
95% of the beam then passes through a set of sheet optics where a
cylindrical lens converts the beam into a flat laser plane [17].
This planar laser then passes through the combustion event,
inducing fluorescence. Depending on what type of experiment is
being run either a flat flame burner (calibration and stationary
flames) or a spark ignition constant volume cylindrical bomb
283.4 283.45 283.5 283.55 283.6 283.65 283.7 283.75
Rela
tive
Inte
nsit
y [a
rbit
rary
uni
ts]
Wavelength [nm]
Q1(8) Q2(7)
R2(16) Q21(8) Q12(7)
P1(4)
-
-11-
(transient experiments) is used as a combustion device. Finally,
a camera positioned perpendicular to the laser sheet captures an
image of the fluorescence. A simple mock up of the PLIF system can
be seen below in Figure 2.1:
Figure 2.1 - PLIF system
The camera has a filter on it that only allows wavelengths near
308nm to pass through to be imaged. The OH transition being excited
by the 283nm beam emits photons around this wavelength so the light
particles released by fluorescence enter the camera and are
captured on the intensifier. The more photons there are, the more
OH must be present and the brighter the image is. This allows us to
locate OH throughout the plane of the flame that the laser excited.
Raw images are recorded in arbitrary units of intensity counts
where higher count areas mean higher OH concentrations.
The entire system was covered in great detail by Andrews and
Fredette in their master’s theses [4] and [12]. For an in depth
description of each piece of equipment, please refer to their
theses.
2.2. Wavelength Selection At different wavelengths, different
transitions of OH can be excited. Each of these transitions will
emit different intensities of fluorescence. By finding the
wavelength that induces the highest level of emission within a
range of interest, experiments can be run using the optimum
wavelength to capture images with the strongest fluorescence
[8].
Experiments performed by previous students on this PLIF system
were run using a UV wavelength of 282.672nm. After speaking with
experts at LaVision, it was learned that there is normally a higher
peak found around 283.55nm that is called the Q1(8) line. As
mentioned in the section on OH Specifics, this transition line is
recommended because of its low temperature dependence and fact that
it is typically the highest peak within the (1-0) transition band.
Exact wavelength locations of OH transitions can vary depending on
the room temperature and laser tuning. Knowing that the (1-0) band
occurs around 283nm, a scan was performed over a range of
wavelengths to locate the highest transition line within the band.
The current dye laser set up is optimized to study the (1-0) band.
The Rhodamine 6G dye has its peak fluorescence efficiency of 28% at
a fundamental wavelength of 566nm which, upon passing
-
-12-
through the doubling crystal, is cut in half to 283nm. The
doubling crystal in the dye laser, which doubles the frequency
while halving the wavelength and keeping velocity constant, is
compatible with wavelengths of 210 to 290 nm. This eliminates the
option of studying the other common option of the (0-0) band, which
occurs around 308nm.
Between recommendations from laser professionals, hardware
optimizations that have already been performed and limitations of
current equipment, it was decided to stay with the (1-0) band.
Knowing that this band is around 283nm, a scan was performed from
284nm to 282nm to find the maximum fluorescence intensity in that
range. A scan is performed by lighting the flat flame burner and
taking a fluorescence image of the flame at each wavelength. The
computer program averages the intensities in each image and plots
this average against the wavelength that the picture was taken at,
creating a plot similar to that shown below in Figure 2.2. Scans
are performed starting at a higher wavelength and stepping down by
0.001nm at a time. They are run in decreasing wavelength because
the stepper motor that controls the resonator of the dye laser is
more accurate when stepping in this direction, allowing a step of
only 1pm to be performed with confidence. A global maximum of the
scan range is expected to be found at the suggested Q1(8) line
around 283.55nm. The spectra obtained from scanning a
stoichiometric methane-air flat flame is shown below in Figure
2.2:
Figure 2.2 - OH spectra of (1-0) band from 282nm to 284nm
The vertical axis of Figure 2.2 above is in arbitrary units
(a.u.). Since only the wavelength of the highest peak is desired,
the heights of the peaks are all relative and the only important
information is which is the tallest. Therefore, the dependent units
can be reported as arbitrary.
A few key peaks have been labeled in the OH spectra above. One
of the first things to notice is that the Q1(8) line is not the
global maximum of the scanned range. The maximum actually occurs at
the Q1(6) line, which is around 282.92nm. Another point of interest
is that the peak occurring at 282.67nm, where previous students
conducted experiments, is shorter than the Q1(6) peak. The Q1(8)
peak is approximately 60% of the height of the Q1(6) peak. The peak
used by Andrews and Fredette at 282.67nm is about 93% of the
maximum peak.
282 282.2 282.4 282.6 282.8 283 283.2 283.4 283.6 283.8 284
Rela
tive
Inte
nsit
y [a
.u.]
Wavelength [nm]
Q1(6)
Q1(8)
Peak used by predecessors
-
-13-
Despite the fact that the Q1(8) line was recommended, it was
decided to continue with the Q1(6) line for this work. The maximum
intensity consistently occurred at this transition, giving the
strongest concentration images possible with the current setup. All
of the following work will be conducted at a wavelength of
approximately 283.9nm. Peakfinding scans were run daily to ensure
that the optimum wavelength was used for the exact room and laser
conditions that day. The maximums found by these scans only varied
by a few picometers from day to day.
The frequency conversion unit (FCU) is very sensitive to
temperature and can de-tune overnight just because of fluctuations
in room temperature [18]. To get the best results, an FCU table
should be created in the Sirah 2.6 software every day that a
peakfinding scan is being run. If the FCU table is not up to date,
then the FCU should not be enabled in the peakfinding scan. When
the FCU is enabled, that means that the doubling crystal and
compensator will rotate along with the resonator as wavelengths are
scanned. If the doubling crystal does not rotate during the scan,
you may receive a peakfinding curve similar to that shown below in
Figure 2.3:
Figure 2.3 - Incorrect OH spectra without FCU enabled
Two key points to notice about this bad peakfinding scan is that
the X axis displays the fundamental wavelength (around 566nm). This
should be a giveaway that the FCU is disabled. The second point is
that instead of looking like the correct spectra shown previously
in Figure 2.2, this spectra looks like a bell curve. While this may
seem like it would be correct to have a maximum at the ideal
wavelength surrounded by decreasing peaks, this is wrong. The
overall bell distribution of peaks tells us that the FCU’s doubling
crystal was not rotating with the resonator. This lead to the only
tall peak being at the wavelength that the doubling crystal was
tuned to. Instead of tuning the FCU to only one wavelength, an
-
-14-
FCU table must be created that generates a relationship between
wavelengths and peak crystal rotational location. The peakfinding
scan can then access the data in this table and rotate the doubling
crystal as needed during the scan to get the maximum intensity at
each wavelength so that a true peak across the scanned range can be
found.
2.3. Concentration Calibration The concentration calibration was
performed using a McKenna flat flame burner made by Holthius &
Associates (see Fredette’s thesis [4] for burner details). With a
stable flat flame, horizontal bands of constant concentration can
be seen, as shown in Figure 2.4:
Figure 2.4 - Calibration image with horizontal bands of constant
concentration
Knowing the concentration at certain heights above the burner
will allow the user to change the color scale from arbitrary counts
to absolute concentration of the species being studied. To do this
one would place a mask around a location within the flame with a
known concentration. Then the known quantitative concentration at
that location is entered. The DaVis software will then average the
intensity counts of all of the points inside of the drawn mask and
assign the given concentration to the calculated average intensity,
adding a single point to a calibration curve. This is repeated at
different locations with other known concentrations. The result
will be a concentration curve similar to that shown in Figure 2.5
that the program can then apply to experimental images, converting
the relative concentration counts to absolute concentrations.
-
-15-
Figure 2.5 - Concentration calibration sample plot
Once the program has calculated a conversion from arbitrary
intensity counts to quantitative concentration, the calibration
data can be applied to any image in the project. This allows the
user to conduct a calibration using the stationary flat flame with
known concentrations and then map the data onto a transient flame
with unknown concentrations. As long as the same environmental
conditions and laser settings are used for the experimental images
as are used for the calibration image, the concentration conversion
will be valid. As long as it is being excited under the same
conditions, a OH molecule should emit the same amount of energy
regardless of what type of flame it is produced in, be it rich or
lean, transient or stationary.
2.3.1. Thermocouple Calibration Method The calibration method
used for previous work in the laminar flame speed laboratory
involved using a thermocouple to measure the temperature of a flat
flame at different heights above the burner. With the measured
temperature and knowing that the flame is in a standard environment
(T=298K, P=1atm) the reactants and products of the flame could be
entered along with this information into a program called STANJAN
[5]. This program, developed by a professor at Stanford, uses the
given temperature, pressure and species information to calculate
the concentration of a desired species. Since the temperature
measurement from a thermocouple is used, the calculated
concentration is only valid at the thermocouple bead. In the DaVis
software a circle can be drawn around the thermocouple bead in an
image to tell the program where it is located. The program then
averages the intensity counts within that circle and can apply the
concentration calculated at that point. The thermocouple bead is
then translated to a different height above the burner and the
process is repeated. A similar concentration curve to the one above
is formed and then applied to experimental images. See Colin F
Fredette’s master’s thesis for a more detailed explanation of this
calibration method [4].
-
-16-
2.3.2. LaVision Calibration Method Another method of
quantification would be to use the data presented in the LaVision
calibration burner product manual [3]. Data used in the LaVision
manual was experimentally collected by Arnold [1] and Cattolica
[19]. The two references agree closely with each other as seen in
figure Figure 2.6.
Figure 2.6 - LaVision calibration data (methane/air, P=1 bar,
Φ=1)
The calibration data given in the manual defines known
concentrations at specific heights above the burner surface in a
stoichiometric methane/air flat flame (p=1bar, Φ=1).
Arnold collected data using a single laser pulse method as
described in the literature review. Considering error from shot
noise, temperature, transmission measurement and spectroscopic data
inaccuracies Arnold et al estimated a total accuracy of 20% for
their single shot data.
In 1982, Cattolica used two different methods, laser-absorption
spectroscopy and molecular-beam mass spectrometry (MBMS), to find
the quantitative concentration of OH in stoichiometric methane-air
flat flames at atmospheric pressure. For his laser-absorption
spectroscopy method, Cattolica used radiation corrected
thermocouple measurements and assumed a Boltzmann distribution of
the concentration of a measured rotational OH level. Unfortunately,
the thermocouple correction used could not be found to compare with
Fredette’s. The MBMS approach measured the mass to charge ratio of
molecules in the flame and could then use the detected masses to
determine concentration. Cattolica concluded that the peak OH
concentrations of the two methods agreed within their estimated
experimental error.
2.3.3. Calibration Method Comparison In order to compare the two
techniques described previously, the thermocouple approach was
performed using Fredette’s correction on a flame experimentally
similar to those used by Arnold and Cattolica (matching
stoichiometry and mixture flowrate). Then the concentrations
calculated from the thermocouple method could be directly compared
to those found in the manual.
One of Fredette’s recommendations for future users of the PLIF
system at the end of his thesis was to verify the validity of the
assumptions made for his thermocouple correction. This calibration
comparison
1.0E+02
1.0E+03
1.0E+04
0 2 4 6 8 10
OH
mol
e fr
acti
on [p
pm]
height above burner [mm]
Arnold
Cattolica
-
-17-
seemed like a good opportunity to check the thermocouple
correction as well. The original correction was designed and used
on a hydrogen-air flat flame. Since this work has dealt mostly with
methane, the correction was tested on a methane-air stoichiometric
flame. In the same Cattolica paper that LaVision specified for
concentration calibration there is also data reporting temperatures
versus the distance above the burner surface. Cattolica et al
report using “radiation corrected thermocouple measurements” but
never go into any further detail regarding what these corrections
actually were. They simply present a plot of their temperatures in
their paper, shown in Figure 2.7 below:
Figure 2.7 - Temperatures reported by a radiation corrected
thermocouple at varying heights above a burner [19]
This plot was imported into Data Thief [20], a program that can
take plots and extract numerical data. By simply marking three
reference points with known coordinates then defining the
beginning, end and color of the curve of interest, the program
prints out data that can be imported into Excel. The solid curve
labeled thermocouple, shown in Figure 2.7, was “stolen” for
comparison with experimental data.
Cattolica reports having used a mixture flow rate of 0.316
liters/second with a stoichiometric methane-air flame (Φ=1). This
flame was replicated experimentally to compare temperatures, being
sure to take into account the correction factors for the Omega flow
meters being used. Temperature measurements were taken every 0.5mm
starting on the burner surface and working up to Cattolica’s
reported maximum of 7mm. The experimental points are plotted along
with Cattolica’s curve from Figure 2.7 above in Figure 2.8
below:
-
-18-
Figure 2.8- Thermocouple correction comparison of Fredette [4]
to Cattolica [19]
The experimental data set using Fredette’s correction reported
temperatures on average 50 K lower than Cattolica. This tells us
that either Cattolica was more conservative with corrections, or
Fredette missed a contributor to energy loss from the thermocouple.
Even still, the percent difference between experimental and
expected measurements was at most 3.2% over the range of heights
studied. While this seems like an acceptable difference for the
temperatures, it creates an unacceptable difference between
concentrations calculated by STANJAN (see Table 1), which was
Fredette’s technique for calibrating. A worst case example was
checked at the experimental peak temperature 1mm off the surface of
the burner. The elements C, H, O, and N were entered into STANJAN
along with species CH4 (1 mole), O2 (2 moles) and N2 (7.52 moles)
as the reactant mixture components for a stoichiometric methane-air
reaction. Additional species that were included were H2, O, N, NO,
NO2, H2O, H2O2, OH, HO2, CO and CO2
Table 1 - Comparison of [OH] calculated using temperatures from
experiment and literature
and the calculation constraints used were constant pressure and
temperature. The results are shown below:
Cattolica Current Work Difference % Difference
Temperature [K] 1818 1760 58 3.3% STANJAN OH Mole Fraction
2.45E-04 1.55E-04 9.00E-05 58.1% STANJAN OH Mass Fraction 1.51E-04
9.54E-05 5.56E-05 58.3%
As seen in Table 1, the seemingly slight difference in
temperature causes an unacceptably large discrepancy in OH mole
fractions as calculated by STANJAN.
Then, the concentrations calculated using STANJAN were compared
to the concentrations reported in literature. It was found that the
estimated concentrations at specific heights above the burner from
the thermocouple method were all about 2% of the concentrations
from the LaVision method. To try to get
300
500
700
900
1100
1300
1500
1700
1900
-1.0 1.0 3.0 5.0 7.0
Tem
pera
ture
[K]
Height Above Burner Surface [mm]
Current Work
Cattolica
-
-19-
a better idea of how the data compared the LaVision data was
plotted against the thermocouple data resulting in Figure 2.9
below:
Figure 2.9 - Comparison between LaVision and thermocouple
calibrations
The fact that there is a strong linear trend between the results
of the thermocouple correction method and the LaVision method
suggests that some kind of scaling factor was missed in the
thermocouple method. It is not likely that an important
contribution to energy loss was neglected when assumptions were
made and is now causing this error. This may be a contributor but
the fact that such a tiny percentage of the expected concentration
was calculated means that there is something more significant
happening here.
2.3.4. Concentration Calibration Conclusion Looking at the fact
that Arnold and Cattolica used two different experimental methods
and yet their data for methane/air flames still agreed closely, it
was decided that they were likely more correct. This, along with
the fact that this data was provided by the manufacturer, resulted
in the conclusion that their data should be used for calibration.
This project uses Arnold’s data because the experimental set up
that Arnold et al used was closer to that of the current work.
Being able to closely replicate the experiments led to greater
confidence in the accuracy and applicability of the calibration
data to the current work.
Fredette’s calibration method was not used in the current work
because of the uncertainty introduced by the thermocouple. In order
to get an accurate temperature reading of the thermocouple,
corrections must be made for the radiation loss from the bead and
the conduction loss down the lead wires [4]. The results from the
thermocouple correction could not be confirmed by other
experimental data found in literature. The added complexity of
thermocouple corrections was avoided in the current work by simply
using calibration data supplied by LaVision in the McKenna burner
manual.
R² = 0.9933
0
1000
2000
3000
4000
5000
6000
7000
0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00
Arn
old'
s M
ole
Frac
tion
[ppm
]
Thermocouple Mole Fraction [ppm]
-
-20-
2.4. Transient System A constant volume cylindrical vessel with
a central spark ignition system was used to study spherically
expanding flames. Using partial pressures the cylinder would be
filled with the desired ratio of fuel and air to bring the total
system pressure to a pre-determined value. The ignition system
would then create a spark in the center of the cylinder, initiating
a flame which would then propagate outwards spherically until it
extinguished on the internal wall of the vessel.
Figure 2.10 - Constant volume cylinder in place for
experiments
Windows at the cylinders ends and on either side allowed images
to be taken of the transient flame while a laser was pulsed
through, inducing fluorescence of OH just like was done with the
flat flame burner. Theoretically, if the same laser settings
(including wavelength and power), pressure and initial temperature
are used inside the cylinder as were used in performing the
concentration calibration with the flat flame burner, the
calibration data should be able to be applied to transient images,
quantifying the concentration of OH.
2.4.1. System Establishment The master’s thesis of RJ Andrews
[12] was to establish a timing system for transient combustion
events. Andrews successfully set up a PLIF system that could image
a spherical flame at the radius desired. His final images proving
his success are shown in Figure 2.11:
-
-21-
Figure 2.11 - Andrews' successful timing system final images
[12]
Unfortunately he did this with a now outdated version of the
software, DaVis 7.1.1. Since then the software was updated to DaVis
7.2 and many of the menus and settings used were eliminated.
Therefore, the present work includes re-learning the transient
system and trying to optimize it with the current software.
2.4.2. Current System Status Much time was spent trying to
exactly replicate Andrews’ settings in the new software. After
speaking with the developers of the recording software it was
learned that transient imaging was far simpler with the new DaVis
7.2. All of the same settings that were used for imaging the flat
flame would again be used and just one delay would be entered that
would allow the flame to expand to the desired radius before
imaging. Once a satisfactory timing between laser and camera was
established it could be applied to any flame, regardless of
stationary or transient. This realization greatly simplified
matters and allowed the bulk of time and energy to be spent on
optimizing images.
Transient images taken were weak, to the point of barely being
detectable within the background noise, as shown in the raw image
below in Figure 2.12:
-
-22-
Figure 2.12 - Weak transient image, methane/air, P=1atm, Φ=1
This weak image was thought to be the result of low power output
of the dye laser. The pump laser power was consistently around 4W
throughout this work. The dye laser was much more variable and
ranged from a minimum of 160mW up to a maximum of 240mW. Past
experiments by Fredette were run around 260mW. In an attempt to
increase the power the Rhodamine 6G dye was changed and all
components in the dye laser were tuned. Changing the dye did not
solve the problem of weak signal.
After speaking with LaVision engineers, it was found that
experiments were not being conducted at the ideal wavelength.
Having performed peakfinding scans WITHOUT the FCU enabled for
previous experiments, a wavelength of approximately 282.93nm was
being used to start with. LaVision engineers suggested using the
Q1(8) transition of OH which occurs around 283.55nm, as mentioned
previously in the Wavelength Selection section. A peakfinding scan
was performed with the flat flame burner and the FCU enabled and
the Q1(8) peak was found to be at 283.545nm. Even though this was
not the highest peak of the scan (recall Figure 2.2) the dye laser
was set to this wavelength and the transient experiments were
attempted again. This new wavelength was coupled with another
suggestion from LaVision to use the laser on a different setting.
Typically for transient images the laser lays dormant until the
record button is pressed. The laser then flashes one single time
according to the timing set in the software to induce fluorescence
in time for imaging. It was recommended to try running the laser on
adjust mode which would have the laser flash repeatedly at 10Hz.
Then, when the record button is pressed, the image acquisition
would be timed to fall on one of these laser flashes. The idea
behind this was that sometimes the first shot of the laser after
lying dormant is not as powerful as the repeated flashes just by
nature of the laser. In an attempt to improve the transient images
both of these suggestions were combined but the result was a weak
image similar to that shown in Figure 2.12.
Unfortunately, time ran out before a good transient image could
be taken with the system. The fact that there is clearly an image
of the spherical flame, weak as it may be, in Figure 2.12 tells us
that the timing set up is correct. Having been able to successfully
image a flat flame proves that the camera and laser
-
-23-
are working together properly. The partial pressure filling
method of the cylindrical vessel was validated using a gas
chromatograph as part of the master’s thesis of another student in
this lab, Casey Bennett [21]. The cylinder and fill lines were
checked for leaks to make sure that the mixture was getting to and
staying in the cylinder. No leaks were found. The remaining
possibilities could be that the laser power is not high enough or
that there is simply not much OH to image in the transient flames
being created.
3. Experimental Results
3.1. Concentration Calibration Validation To determine if the
concentration calibration technique using the LaVision data was
valid, it was first checked if the calibration could be repeated
reliably. To do this, first a stoichiometric methane-air flame with
flowrates matching the total mixture flowrate of Cattolica (0.316
liters/second) was imaged. The image was corrected by subtracting
out the background noise, zeroing out laser energy fluctuations and
correcting for laser sheet inconsistencies. The processed image was
then sent to the concentration calibration module in DaVis 7.2
where the calibration data from LaVision could be applied to the
image at specific heights above the burner. This created a
calibration curve similar to that in Figure 2.5 that was stored in
the project and could be applied to any image to convert color
counts to concentration using the developed relationship, as
discussed in the concentration calibration section.
Once a calibration curve was obtained, stoichiometric
hydrogen-air flames with the same mixture flowrate were imaged.
Three sets of hydrogen images were taken. Each set of 100 images
was averaged together, corrected in a similar method as the methane
flame and then calibrated for concentration using the calibration
curve. A note should be made here: the calibration curve created
with the methane flame should be able to be applied to any flame,
flat or transient, regardless of fuel type, so long as the flame
was imaged using the exact same dye laser wavelength and power as
the calibration image and is at the same environmental conditions.
OH will fluoresce with a consistent wavelength and emit a
consistent strength signal so long as it is excited by a consistent
wavelength and power. To ensure that you are not confusing energy
transitions of OH and applying incorrect concentrations, be sure to
create a new concentration calibration curve every day. Peak
wavelength and power of the dye laser can change over night without
even touching a thing! The FCU is highly temperature sensitive and
can de-tune rapidly. This will drastically change the color scale
counts on which the concentration calibration is based.
The concentration at each pixel along a vertical line,
representing a constant radial location, was drawn through the
maximum concentration point and data was exported for each of the
three calibrated flame images (see one of these in Figure 3.1 below
with vertical line along which data was taken):
-
-24-
Figure 3.1 - - H2 air
This data from along the vertical line in
=13.8)
Figure 3.1 could then be used to plot concentrations against
height above the burner at the same radial position in three
different images. The three different images were all
stoichiometric hydrogen-air flames with equivalent flowrates. If
the calibration and post processing method is repeatable, all three
data sets should be equal. The plot of these three tests is shown
in Figure 3.2:
Figure 3.2 - Comparison of three stoichiometric hydrogen-air
flames with same calibration data
As can be seen in Figure 3.2 all three tests are quite similar.
The vertical offset in the areas of low concentration are likely
due to inconsistencies in background imaging. So this test showed
that hydrogen flames can be imaged repeatedly and come up with
consistent data through post processing.
0
500
1000
1500
2000
2500
3000
-20 -10 0 10 20 30 40
Conc
entr
atio
n [p
pm]
Distance from burner surface [mm]
Test 1
Test 2
Test 3
-
-25-
However, these three images all shared one common set of
calibration data. The real question is, can a concentration
calibration be performed repeatedly and reliably?
The next step was to simply perform the entire process again to
see how two sets of data taken on different days might compare to
eachother. Another methane flame was imaged. A new concentration
calibration curve was developed using the same data. Three more
hydrogen-air flame images were taken and calibrated. At this point
all six data sets (three from calibration 1 and three from
calibration 2) were adjusted so that the lowest reported
concentration was set to zero. The minimum concentration of each
set was simply subtracted from each point of the data set,
translating each curve without changing its actual shape. This
added step was just to clean up the noise that the background
subtractions had missed during image processing. At his point, the
consistency of concentration calibration is being tested, not the
background subtraction, so it was deemed acceptable to cancel out
this noise. In some cases there were negative concentrations which
of course do not make sense.
Having decided that the three test curves from calibration 1
were similar, they were averaged together. The three test curves
from calibration 2 were also averaged and a plot was created that
could only compare the two calibrations in the same manner that
Figure 3.2 compared the three tests within calibration 1. This
resulted in Figure 3.3:
Figure 3.3 - Comparison between two independent concentration
calibrations
Looking at Figure 3.3, it is clear that the curves from two
completely independent calibrations agree closely with each other.
This has now shown that not only the imaging process is repeatable
but the complete concentration calibration is repeatable as
well.
One area of concern in this study was the assumption that
concentration of OH is constant across the burner at each height
from the surface. Looking back at the fluorescence image in Figure
3.1, this is clearly not true. In order to test just how inaccurate
this assumption was, concentrations along horizontal lines were
studied to get the average concentration across the image at each
height. This was
0
500
1000
1500
2000
2500
-20 -10 0 10 20 30 40
OH
Con
cent
rati
on [p
pm]
Distance from burner surface [mm]
Calibration 2
Calibration 1
-
-26-
performed for both calibration data sets and plotted against
distance from burner surface. These averaged results were then
compared to the concentrations along only the vertical line
containing the maximum concentration. As can be seen in Figure 3.4
below, averaging the concentrations at each height greatly
decreased the maximum concentration from 2400 ppm to 1500 ppm.
Figure 3.4 - Comparing average concentration to maximum
concentration
This discrepancy is due to inconsistencies of the flat flame.
These inconsistencies could be caused by build up within the bronze
sintered plug of the burner that is highly calibrated to evenly
distribute the air-fuel mixture across the surface of the burner.
When the flat flame burner was first being used for this thesis, it
had been sitting in a dusty lab untouched for over a year. The
first images taken showed signs of extreme clogging. Figure 3.5 was
taken with a clogged burner:
Figure 3.5 - Inconsistent PLIF image taken with clogged
burner
0
500
1000
1500
2000
2500
-20.00 -10.00 0.00 10.00 20.00 30.00 40.00
OH
Con
cent
rati
on [p
pm]
Distance from burner surface [mm]
Cal 1 Avg
Cal 1 Max
-
-27-
Clearly this flame is nowhere near flat. The manufacturer was
contacted and a hydrochloric acid bath was performed per their
recommendation. After completing the HCl bath, images were greatly
improved to look more like that shown in Figure 3.1 above. However,
the flame is still not perfectly flat. Had time permitted, the
burner would have been sent back to California for a complete
recalibration. Images taken after the HCl bath were sent to the
manufacturer were approved as acceptable for continuing study. A
complete recalibration would have reduced the difference between
average and maximum concentrations shown in Figure 3.4, but has
been left as a recommendation for the next student who uses the
PLIF set up.
3.2. Fuel Flowrate Comparison A study was performed to see how
the flowrate of the air-fuel mixture affects the concentration and
location of OH in stoichiometric flat flames. It is expected that
as the flowrate is increased, the maximum concentration will
increase. Since the fuels being studied are hydrocarbons, a higher
flowrate means more hydrogen which allows for more OH formation.
This study will serve as the basis for understanding some of the
differences between methane-air and hydrogen-air stoichiometric
flames needed to continue this work.
First, a concentration calibration was performed using the same
total flowrate used by Cattolica in [19]. This resulted in the
calibration image shown in Figure 3.6below:
Figure 3.6 - Stoichiometric methane calibration image
In an attempt to correct for the inconsistency of the burner
which creates disturbances in the smooth bottom of the flame, only
the portion of the flame outlined in the box in Figure 3.6 above
was used for the concentration calibration. This was done because
this portion of the flame had the smoothest bottom and contained
the maximum intensity in the image. This rectangle contains heights
above the burner from approximately zero to ten millimeters.
-
-28-
Once a concentration calibration curve was created using the
image above, it was applied to four different stoichiometric
methane-air flames. Methane flowrates of between 0.75 and 4 SLPM,
stepping 0.25 SLPM at a time, were selected and used with
calculated air flowrates to maintain an equivalence ratio of one.
100 images were taken at each flowrate. For each flowrate the
images were averaged, energy corrected, background subtracted,
sheet processed and then calibrated for concentration using the
curve created from Figure 3.6 and Cattolica’s data. The final
images from four of these flowrate tests can be seen below, all set
to the same concentration scale in Figure 3.7:
Figure 3.7 - Calibrated stoichiometric methane-air flames with
varying flowrates on common scale
It is immediately apparent that the concentration increases with
flowrate. Also, the thickness of the flame front increases with
flowrate. This detail is easier to see when each image is scaled to
show its own maximum concentration on their respective color
scales, as was done for Figure 3.8 below:
Figure 3.8 - Calibrated stoichiometric methane-air flames with
varying flowrates on independent scales
-
-29-
The minimum of each concentration color scale was set to zero
and the maximum concentration in the image was found by the
software and set to the maximum of the color scale. It can be seen
clearly now in Figure 3.8 that the red region of high OH
concentration, which marks the flame front, becomes thicker as the
flowrate of methane is increased while maintaining stoichiometry.
This lead to the understanding of how important it is to match not
only the equivalence ratio but the volumetric flowrate as well for
calibration data to be applicable to a flame image.
A detail worth noting in Figure 3.8 is the smoothness of the
bottoms of the flames. As the flowrate increases, the bottoms of
the flames become considerably less smooth. Another detail that is
perhaps the most noticeable thing when comparing the four images is
that the background of the 4 SLPM image is blue, predicting an OH
concentration of around 2000 ppm throughout, while the rest of the
backgrounds are black with concentrations around zero. This would
mean that there are 2000 ppm of OH on the stainless steel of the
burner and in the ambient air when the flowrate is 4 SLPM, which of
course is not true! This discrepancy could mean that the background
subtraction did not work properly and 2000 ppm have been added to
each pixel, giving a falsely high maximum concentration. Or perhaps
there was a fluctuation in ambient light captured during the 4 SLPM
test that made the common background image incompatible with the
experimental image. Looking at Figure 3.10 below, it can be seen
that the data point for 4 SLPM for methane does not fall as close
to the trendline as the other points, further indicating that this
may be an error. This issue was further investigated and it is
believed to be due to inconsistencies in batch processing within
the DaVis software. To test the repeatability of the software, a
stationary flame was lit and three sets of 100 images were taken
right in a row, within less than a minute of each other. These
three sets were then processed using the exact same background
image, sheet image and concentration calibration. Theoretically,
all three final images would be the same, but as can be seen in
Figure 3.9 below, they clearly are not:
Figure 3.9 - Image post processing inconsistency test, all
images should match
In the images above, all color scales were set to have a minimum
of zero and a maximum of the highest concentration in each image.
The left most image has a maximum concentration of about 3200 ppm.
The solid light blue background of that same image is predicted to
be around 400 ppm. The right most picture has a maximum
concentration of about 2600 ppm and a background of almost zero.
This strongly supports the suspicion that the entire image has
simply been shifted in intensity. After conducting this post
processing consistency test, it was decided that final images would
be rejected if their backgrounds were not close to zero. Many
processed images would end up having negative OH concentrations,
which of course does not make sense. Or, as seen in Figure 3.8, the
entire image would
-
-30-
be scaled too high. When dealing with numerical data, these data
sets could be corrected by subtracting the lowest concentration
from each point to zero out the background.
A similar flowrate comparison was repeated for stoichiometric
hydrogen-air flames, still using the methane-air flame for
concentration calibration. It was again found that concentration of
OH increased with the flowrate.
Figure 3.10 - Comparison of [OH] varying with total flowrates of
methane-air and hydrogen-air
When plotted on the same graph, it becomes obvious that methane
has much higher concentrations of OH at comparable flowrates to
hydrogen. The maximum concentrations of OH increase by almost 280
ppm per SLPM for stoichiometric methane-air flat flames but by only
about 40 ppm per SLPM for stoichiometric hydrogen-air flat flames.
This study has shown that hydrogen flames give off a much weaker
LIF signal than methane flames. This knowledge can be used in
future experiments. Knowing what to expect of experimental images
relative to the calibration image can help confirm that the system
is working properly.
3.3. Equivalence Ratio Comparison Once the calibration procedure
was validated and the effect of flowrate on [OH] in stoichiometric
flames was studied, the next topic of interest for this work was to
study the effect of equivalence ratio on quantitative [OH] in a
hydrogen-air flame. A key question that arose when starting this
equivalence ratio study was, in order to have comparable flames,
should the fuel flowrate be kept constant and the oxidizer flowrate
changed to achieve the desired equivalence ratio? Or should the
total mixture flowrate be kept constant and both fuel and oxidizer
flowrates adjusted to sum up to the same total each time? For all
of the studies perfomed in this work, flowrates and equivalence
ratios were determined using the ideal gas law. Knowing that flat
flames would occur at atmospheric pressure, room temperature and
using the universal gas constant, a direct ratio of volumetric
flowrates and moles could be used. Looking at this fact, it was
thought that in order to make flames that are truly comparable, the
number of moles
0
2000
4000
6000
8000
10000
12000
14000
0.0 10.0 20.0 30.0 40.0
Max
imum
[OH
] [pp
m]
V̇total[SLPM]
Hydrogen
Methane
-
-31-
of the premixed air-fuel mixture per minute should be kept
constant. This would mean, assuming ideal gas behavior, that the
total mixture flowrate must remain the same. After this
consideration, it was decided that tests would be run using
Cattolica’s total mixture flowrate of 0.316 liters per second to
have flames be comparable to each other and to literature.
Just as was done in the flowrate study, a methane flame was
first imaged to establish a calibration curve for the post
processing of experimental images to be taken. Then, 100 images
were taken of hydrogen-air flames at each equivalence ratio of
interest. Images at each ratio were averaged, post processed and
calibrated for quantitative concentration. Five equivalence ratios
were studied and their final images can be seen, all displayed on a
common scale, in Figure 3.11 below:
Figure 3.11 - H2-air flames with varying equivalence ratios
displayed on a common concentration scale
Thanks to the common concentration scale used for the five
images in Figure 3.11 above, it can be clearly seen that the
concentration of OH decreases as the equivalence ratio increases.
This is not at all the expected result. The peak concentration was
expected to be at stoichiometric conditions, with the concentration
decreasing as the mixture moves away from stoichiometric in either
direction. Average concentrations at each height above the burner
along the full height of each image was extracted and plotted to
get a more quantitative comparative view of these results, as shown
in Figure 3.12:
-
-32-
Figure 3.12 - Effect of equivalence ratio on value and location
of OH concentration
There are a few interesting points to note about Figure 3.12
above. First off, the visually drawn conclusion that the maximum
concentration increases as the equivalence ratio decreases is
quantitatively confirmed, which is unexpected and will be discussed
in greater detail. Also, notice that the peak concentration of each
equivalence ratio occurs around the same height off the burner
surface (approximately 1mm). One detail that was not easily noticed
in the flame images is that the higher peak concentrations seem to
more rapidly decrease in concentration as height above the burner
increases, with the exception of Φ=1.2. Compare the shallow slope
from 1 to 5 mm of the Φ=1.1 curve to the nearly vertical Φ=0.8
curve across the same range. This shows that the flame front of the
lean mixtures is much thinner than that of the rich mixtures. This
would be expected because in a lean flame, all of the fuel gets
consumed whereas in a rich flame, excess fuel will be able to
travel further away from the burner surface before being
burned.
Finally, notice that the concentration is not zero at a height
of zero, as one might expect. Looking back up at any of the
fluorescence images you will see sort of a blue glow below the
flames. This blue glow is the concentration that is plotted below
the surface of the burner. The cause of this is most likely
fluorescence reflecting off of the burner surface. Fluorescence
occurs in all directions and the camera only collects the photons
that enter its intensifier. This means that there are photons
which, when emitted, travel down towards the burner surface where
they are reflected into the camera. Since the event of fluorescence
lasts only 20ns and the gate of the camera is 200ns, some
reflections will certainly be collected.
It is clear that quantitatively, the concentration of OH is
higher at the lean equivalence ratios than at stoichiometric
conditions, which should not be the case. This raised concerns that
perhaps the flowmeters are off calibration and that the desired
stoichiometries were not actually being experimentally created. To
test this, a rotameter was connected to the Omega digital
flowmeters that are used to monitor oxidizer and fuel flowrates to
the flat flame burner. Calibration curves were created
0
500
1000
1500
2000
2500
3000
0 5 10 15 20 25 30
OH
Con
cent
rati
on [p
pm]
Height Above Burner [mm]
Φ=0.8
Φ=0.9
Φ=1
Φ=1.1
Φ=1.2
-
-33-
by plotting the flowrate measured by the Omega meter against the
value measured by the Dwyer Instruments rotameter. Data was
collected for air, methane and hydrogen. Since the Omega meters are
calibrated for N2
Figure 3.13
and the Dwyer meter is calibrated for air, correction factors
from the two manufacturers were used to convert the displayed
flowrates on the meters to the actual flowrates of the specific
gases being studied. This test resulted in the plot shown in
below:
Figure 3.13 - Flowmeter calibration curve
Since both meters are reporting standard liters per minute
(SLPM), the resulting curves are expected to be linear with a slope
of one. As can be seen by the equations of each line displayed on
the plot in Figure 3.13, this is not the case for all three gases.
The slope of the air line is almost one, so the measurement of air
flowrate taken by the Omega meter is deemed accurate. The methane
and hydrogen curves however are far from having a slope of one. The
two fuels do have r-squared values close to one though, which means
that their data has a strong linear correlation. This leads to the
belief that simply the manufacturer’s scaling factor is off. The
dwyer scaling factor is a physical conversion using the specific
gravity of the fuel so it is trusted. The Omega conversion factors
however are simply numbers taken from a table with no physical
background. The fact that the Omega meters are digital also leads
to the thought that over time they more easily fall out of
calibration than a rotameter, which simply depends on the
properties of the fluid being measured and the size, shape and
density of the channel and float, which are not likely to change
much over time.
This lead to the conclusion that the Omega conversion factors
for methane and hydrogen should be adjusted to make the calibration
curves have a slope of one. By adjusting the conversion factor for
methane from 0.72 to 0.59 and the conversion factor for hydrogen
from 1.01 to 0.95, Figure 3.13 was adjusted to Figure 3.14:
y = 0.9667x + 0.2056R² = 0.9982
y = 0.7436x + 0.1818R² = 0.9927
y = 0.2521x + 0.4391R² = 0.9985
0
1
2
3
4
5
6
0 5 10 15 20
Dw
yer
rota
met
er [S
LPM
]
Omega digital [SLPM]
AIR
CH4
H2
-
-34-
Figure 3.14 - Adjusted calibration curve
The vertical shift of the hydrogen curve giving it a large
y-intercept is of concern because this means that the hydrogen
readings on the Omega meter experience a shove error reporting
approximately 1.7 SLPM too low for each measurement. After
accounting for the new conversion factors and this shove error for
hydrogen, it was found that the stoichiometries had all been
shifted. Correcting for this shift and looking back at the data
displayed in Figure 3.12, it was found that the peak concentration
actually did occur around stoichiometric conditions. A table
showing the original and converted equivalence ratios is shown
below:
Table 2- Corrected equivalence ratios using calibration
curves
original Φ corrected Φ 0.5 0.73 0.6 0.82 0.7 0.9 0.8 0.97 0.9
1.06
1 1.13 1.1 1.21 1.2 1.3 1.3 1.4 1.4 1.49 1.5 1.58
These corrected equivalence ratios were calculated in a
spreadsheet that was created to calculate equivalence ratio by
entering the flowrates displayed on the Omega meters. A cell for
the Omega conversion factor was adjusted to the new factors found
from the calibration curves and the shove error
y = 0.9993x + 0.2443R² = 0.9927
y = 1.0058x + 1.6644R² = 0.9985
0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20
Dw
yer
rota
met
er [S
LPM
]
Omega digital [SLPM]
CH4
H2
-
-35-
was added to the flowrate after conversion. As seen in Table 2
above, the peak concentration that was originally believed to have
occurred at the lean equivalence ratio of 0.8 actually occurs at
approximately stoichiometric conditions of Φ=0.97.
Fredette had performed a flowmeter calibration curve procedure
of his own, using the stoichiometries and peak temperatures to
adjust the conversion factor given by Omega. When his corrections
were used, data did not match theory even though his results had
worked in 2009. This leads to a suspicion that the Omega flowmeters
change overtime, slowly falling out of calibration. This means that
as each student starts to work with the system they should check
the meters in the way that they see most fit to confirm that they
are truly creating the flames with the equivalence ratio
desired.
4. Condensed Manual One of the most significant contributions of
this thesis to the laminar flame lab was the creation of a
condensed manual outlining all of the key procedures needed to run
a PLIF experiment. Before this new condensed manual existed, the
user had to rely on a separate manual for each piece of equipment,
for both software packages and for the overall system making a
total of over ten manuals to flip through. When a question arose,
for example, about the camera, it was never known if the answer
would be found in the camera manual, the software manual or the
system manual. Just figuring out which manual to go to became a
time consuming process. Even when the proper manual and section was
found, detail would often be given for a specific example that was
not similar to the Northeastern PLIF set up, making the information
difficult to translate to the question at hand. Over the past two
years as these manuals were deciphered and new tips and tricks were
suggested from professionals at LaVision or Spectra Physics, the
essential procedures specific to the laminar flame lab’s PLIF
system were documented so that future students would have one go to
source for experimental procedures.
In the past, no two students working on the PLIF system have
overlapped, greatly slowing down the progress that could be made on
the system by new students needing to learn the incredibly complex
system from scratch every couple of years. The aim of this manual
is to allow a new student to get the system up and running quickly
so that the majority of their time can be spent collecting data and
furthering the understanding of flame structure in laminar flames.
This manual certainly does not replace all of the other manuals,
but acts as more of a quick start guide. When more detail is needed
than is given in the condensed manual, the original manuals from
the manufacturers should be consulted.
One example of a procedure that has been made more clear by
detailing it specifically to this lab is the peakfinding scan.
Selection of wavelength is essential in PLIF experimentation.
Someone first starting out with this system probably does not know
the best range to scan nor what step size to use. While the
manufacturer’s manual [8] gives a very thorough description of how
to go through the peakfinding process in the DaVis software, they
are doing it for a different OH transition with a different range.
They also go into quite a bit of detail on steps that are not
necessary to complete a successful scan for our specific system and
can greatly complicate and confuse the process for a new student.
The biggest problem with this manufacturer’s manual procedure for
peakfinding is that it does not once mention
-
-36-
that this scan must be coupled with data taken in a different
program in order for it to be completed correctly! A relationship
between FCU position and resonator must be established in the Sirah
Control software before a peakfinding scan in the LaVision DaVis
software can be properly conducted. Since the peakfinding procedure
is in a LaVision manual, it does not once mention the Sirah
software. This lead to peakfinding being performed incorrectly in
this lab for over a year! Finally, a LaVision engineer visited and
walked through the peakfinding process, showing this essential step
of establishing the relationship between FCU and resonator. Since
not all DaVis software is coupled with Sirah software, this
procedure specific to our lab, was left out of the manufacturers
manual. Without having spoken the the engineer at LaVision, this
step would never have been known. See the Wavelength Selection
section for sample peakfinding excitation spectra that were taken
with and without the coupled Sirah step to learn more about what
this key step really does.
Peakfinding is just one of almost twenty different procedures
that have been tailored specifically to our PLIF system here at
Northeastern. Some of the procedures included in this condensed
manual do not exist anywhere else. They are simply tips and tricks
that were picked up from visits from LaVision and Spectra Physics
engineers. One of these new procedures is a step by step
walkthrough of how to clean the flashlamp assemblies. When a
Spectra Physics engineer came to replace an electrical board in the
Sirah dye laser in the spring of 2011, he also peaked the power of
the pump and dye lasers and careful notes were taken to pick up
some professional tips that could be passed on to future students.
One of his tricks was to clean the two flashlamp assemblies inside
the pump laser with a 10% hydrochloric acid solution. The engineer
said that one of the main reasons that he has found for decrease in
pump power is a buildup of some kind of condensation inside these
assemblies, making their reflective surfaces cloudy. If the
internal surfaces of the flashlamps are cloudy, they cannot as
efficiently reflect photons to generate a strong beam. By
disassembling the flashlamps and wiping down each component, this
condensation can be removed and the reflectivity restored, greatly
increasing the pumps power which will directly increase the dye
laser power.
Other procedures covered in the manual include regular
maintenance to keep the system running smoothly, experimental image
processing, how to set up and fill the constant volume combustion
chamber and more.
5. Conclusions and Future Recommendations A concentration
calibration method developed by the company that designed and set
up the PLIF system has been tested. It greatly simplifies the
calibration process, removing the need for a thermocouple with
radiation corrections and chemical equilibrium calculations. By
simply applying the given concentrations to an experimentally
similar flame, a calibration curve can be created and applied to
different fuels and flame types so long as environmental conditions
and laser settings are kept constant.
Finally, in the past two master’s theses to come out of this lab
issues with the Omega flowmeters have been noticed. In this thesis
the conclusion that the calibration changes with time was drawn.
An
-
-37-
adjustment method using a rotameter was developed to draw a
calibration curve and correct the flowrates displayed on the
digital Omega meters.
The creation of a condensed PLIF system manual will speed up the
learning curve of future students. The theoretical and technical
knowledge placed in the manual took about two years to gather.
Hopefully, having all of this information in one place will allow a
new student to quickly get a handle on using the system and be able
to spend their time conducting experiments and analyzing data
instead of troubleshooting and learning the system.
A recommendation that will greatly improve the calibration
system would be to send the burner back to Holthius &
Associates and have it rebuilt. Over the years of sitting in the
lab unused, dust and other foreign particles may have built up
inside the porous plug of the burner. As a result, the “flat flame”
is not as flat as it could be. When Fredette was using the new
burner, flames were as flat as they should be with a freshly
calibrated burner. After lying dormant for almost two years, the
flame images appeared rough with voids in the flame where air/fuel
mixture could not travel through the bronze plug of the burner.
Upon recommendation from Holthius & Associates, a hydrochloric
acid soak was performed in an attempt to dissolve and dislodge any
foreign particles inside the burner. The progression of flat flame
photographs is shown in Figure 5.1 below:
Figure 5.1 - (L to R) Fredette's flame, clogged burner flame,
post HCl bath flame
While the hydrochloric acid bath clearly made a big difference,
the flame is still not back to its original quality and should be
professionally re-calibrated.
A second recommendation to improve the performance of the PLIF
system would be to have a laser technician come to campus and tune
the dye laser. A professional tuning should only be scheduled once
the laser system is understood fairly well and the user can take
both flat flame and transient images comfortably. This way you can
get the most out of a visit from someone by knowing what questions
to ask. Also, the laser power decreases over time so you want to be
able to run experiments the day of or the day after a professional
tuning. One thing that would be worth trying while the laser
engineer is in the lab would be to measure power both with their
power meter and this labs power meter to compare values and find
what the peak on our Newport meter is. This was not done during the
last professional tuning and it is now left unknown if the pump
actually lost a watt of power overnight or if there was just a
difference between the two power meters.
-
-38-
References [1] A. Arnold, R. Bomback, B. Käppeli, A. Schlegel,
“Quantitative measurements of OH concentration fields by
two-dimensional laser-induced fluorescence,” Applied Physics B,
vol. 64, pp.579-583, 1997.
[2] H.M. Hertz and M. Aldén, “Calibration of Imaging
Laser-Induced Fluorescence Measurements in Highly Absorbing
Flames,” Applied Physics B, vol. 42, pp. 97-102, 1987.
[3] LaVision GmbH, “Calibration Burner Product-Manual,”
Anna-Vandenhoeck-Ring 19, D-37081 Göttingen, Germany, 2009.
[4] C. Fredette, “Quantitative Hydroxyl (OH) Concentration
Calibration by Use of a Flat Flame Burner, Thermocouple and Planar
Laser Induced Fluorescence (PLIF) System,” MS thesis, Northeastern
University, Boston, MA, 2009.
[5] W.C. Reynolds, STANJAN, Chemical Equilibrium Calculation,
Colorado State University,
http://navier.engr.colostate.edu/tools/equil.html, August 2011.
[6] M. Mansour, N. Peters and L. Schrader, “Experimental study
of turbulent flame kernel propagation,” Experimental Thermal and
Fluid Science, vol. 32, issue 7, pp. 1396-1404, 2008.
[7] S.H.R. Müller, B. Böhm, M. Gleißner, S. Arndt and A.
Dreizler, “Analysis of the temporal flame kernel development in an
optically accessible IC engine using high-speed OH-PLIF,” Applied
Physics B, vol. 100, pp. 447-452, 2010.
[8] LaVision GmbH, “Tunable LIF Product-Manual,”
Anna-Vandenhoeck-Ring 19, D-37081 Göttingen, Germany, 2007.
[9] R.E. Stevens, H. Ma, C.R. Stone et al., “On planar
laser-induced fluorescence with multi-component fuel and tracer
design for quantitative determination of fuel concentration in
internal combustion engines,” Proc. IMechE, vol. 221 Part D: J.
Automobile Engineering, 2007.
[10] S.M. Blinder, “Introduction to Quantum Mechanics in
Chemistry, Materials Science and Biology.” Complementary Science
Series, Elsevier Academic Press, 2004.
[11] P. Signell, “Diatomic Molecules: Properties From
Rotation-Vibration Spectra,” Project PHYSNET, Physics Bldg.,
Michigan State University.
[12] R. Andrews, “Measurement of Hydroxyl (OH) Concentration of
Transient Premixed Methane-Air Flames by Planar Laser Induced
Fluorescence (PLIF) Method,” MS thesis, Northeastern University,
Boston, MA, 2008.
[13] LaVision GmbH, “Flamemaster Training: Introductory
Presentation,” Anna-Vandenhoeck-Ring 19, D-37081 Göttingen,
Germany, 2011.
[14] Spectra-Physics, a division of Newport Corporation,
“Quanta-Ray PRO-Series Lasers: Pulsed Nd:YAG Lasers, User’s
Manual,” Mountain View, CA. November 2002.
http://navier.engr.colostate.edu/tools/equil.html�
-
-39-
[15] Sirah Laser-und Plasmatechnik GmbH, “Pulsed Dye Laser:
Service Manual,” Germany, 2006.
[16] LaVision GmbH, “Energy Monitor: Device-Manual,”
Anna-Vandenhoeck-Ring 19, D-37081 Göttingen, Germany, 2005.
[17] LaVision GmbH, “Sheet Optics (divergent): Device-Manual,”
Anna-Vandenhoeck-Ring 19, D-37081 Göttingen, Germany, 2004.
[18] T. Horstmann, C. Gray, LaVision technical staff,
www.lavision.com, 2011.
[19] R.J. Cattolica, S. Yoon, E.L. Knuth, “OH Concentration in
an Atmospheric-Pressure Methane-Air Flame from Molecular-Beam Mass
Spectrometry and Laser-Absorption Spectroscopy,” Combustion Science
and Technology, Vol. 28, pp. 225-239, 1982.
[20] B. Tummers, DataThief III. 2006 http://datathief.org/
[21] C. Bennett, “Laminar Burning Speed and Flame Structure of
1,1-Difluoroethane (HFC-152A)/Air and Difluoromethane (HFC-32)/air
Mixtures,”MS thesis, Northeastern University, Boston, MA, 2008.
http://www.lavision.com/�http://datathief.org/�
Northeastern UniversityAugust 01, 2011A study of quantitative
concentrations of hydroxyl (OH) in laminar flat flames using planar
laser induced fluorescence (PLIF)Adrienne Murphy JalbertRecommended
Citation
List of Figures and TablesAcknowledgements1. Background1.1.
Purpose1.2. Literature Review1.3. PLIF Basics1.3.1. Quantum
Mechanics1.3.2. OH Specifics
2. Experiment2.1. Experimental Setup2.2. Wavelength
Selection2.3. Concentration Calibration