U.S. Department
of Transportation
Federal Aviation
Administration
Nitric Oxide
Measurement Study
Office of Environment
and Energy
Washington, D.C. 20591
Probe Methods
Volume II
Report Numbers:
FAA-EE-80-29
USAF ESL TR-80-13
NASA CR-159862
USN NAPC-PE-38C
EPA-460/3-80-014
MAY 1980
M.B. Colket, III
M.F. Zabielski
L.J. Chiappetta
L.G. Dodge
R.N. Guile
D.J. Seery
-------
This document is disseminated under the joint sponsorship
of the Federal Aviation Administration, U.S. Air Force,
U.S. Navy, National Aeronautics and Space Administration,
and the Environmental Protection Agency in the interest of
Information exchange. The United States Government assumes
no liability for the contents or use thereof.
-------
Technical Report Documentation Page
FAA-EE-80-29
4. Title and Subtit
2. Government Accession No.
Nitric Oxide Measurement Study: Probe Methods -
Volume II
7. Author's) M_ B. Colket, III, M. F. Zabielski,
J-. J. Chiappetta, L. G. Dodge, R. N. Guile, D. J. Seery
9. Performing Orgomiotion Nome and Address
United Technologies Research Center
Silver Lane
East Hartford, CT 06108
12. Sponsoring Agency Nome ond Address
U.S. Department of Transportation
Federal Aviation Administration
Office of Environment and Energy
.Washington. DC 20591
3. Recipient's Cotalog No.
5. Report Date
March 31, J98Q
6. Performing Organization Code
8. Performing Organization Report No.
R79-994150-2
10. Work Unit No. (TRAIS)
11. Contract or Grant No.
DOT FA77WA-4081
13. Type of Report and Period Covered
14. Sponsoring Agency Code
15. Supplementary Notes
Funding for this study was provided by an Interagency Committee.
Contributing agencies and report nos. are: DOT-FAA (FAA-EE-80-29); USAF (ESL TR-80-
13); NASA (CR-159862); USN (NAPC-PE-38C); and EPA (EPA-460/3-80-014).
16 Abstract ~
Experimental facilities used in studying the performance of probes and sampling
systems for measuring NO are described. A critical review of the literature on probe
measurements of NOX is given with emphasis on reported results indicating that probes
may perturb the total concentration of NOX in a flame. Also, sample line and
chemiluminesct-nt analyzer phenomena are reviewed. A model of probe aerodynamics
including heat transfer is presented. Kinetics of NO loss are examined and quenching
criteria for measuring nitric oxide in flames are given. Sampling probes are de-
scribed that were designed to preserve NO and are suitable for measurements on small
and large combustors. Probes were designed to cool the gases both convectively and
aerodynamically. Performance of these probes is compared with model predictions.
Concentrations of nitric oxide were measured using several probes for each of three
flame environments. The values measured with each probe are compared and related
to seed levels of NO. In addition, concentration profiles required to compare probe
measurements with optical measurements are provided.
The Nitric Oxide Measurement Study is in three volumes:
Optical Calibration - Volume I;
Probe Methods - Volume II;
Comparison of Optical and Probe Methods - Volume III.
17. Key Words
Nitric oxide, probe sampling, chemilumin
escent analysis, aerodynamic analysis.
18. Distribution Statement
Document is available to public through
the National Technical Information
Service, Springfield, VA 22161
19. Security Clossif. (of thil report
Unclassified
20. Security Classif. (of this page)
Unclassified
21. No. of Poges
112
22. Pr
Form DOT F 1700.7 (8-72)
Reproduction of completed poge outhorized
-------
ACKNOWLEDGMENTS
This contract was administered by the Federal Aviation Administration.
Funding for this work was provided by an Interagency Committee representing
the Federal Aviation Administration (FAA), Air Force, Navy, National
Aeronautics and Space Administration (NASA), and the Environmental Protection
Agency (EPA).
The assistance of Mr. D. L. Kocum, Mr. R. P- Smus, Mr. D. D. Santos and
Mr. R. L. Poitras during the experimental portions of this study is gratefully
acknowledged. The authors also would like to acknowledge the contributions of
the following UTRC staff: Mrs. B. B. Johnson and Mr. C. Foley for data reduction
and report preparation; and Messrs. P. N. Cheimets, M. E. Maziolek, W. T. Knose,
and M. Cwikla for facilities support.
-------
ABSTRACT
Experimental facilities used in studying the performance of probes
and sampling systems for measuring NO are described. A critical review of the
literature on probe measurements of NO is given with emphasis on reported
results indicating that probes may perturb the total concentration of N0y in
a flame. Also, sample line and chemiluminescent analyzer phenomena are reviewed.
A model of probe aerodynamics including heat transfer is presented. Kinetics
of NO loss are examined and quenching criteria for measuring nitric oxide in
flames are given. Sampling probes are described that were designed to preserve
NO and are suitable for measurements on small and large combustors. Probes
were designed to cool the gases both convectively and aerodynamically. Perfor-
mance of these probes is compared with model predictions. Concentrations of
nitric oxide were measured using several probes for each of three flame envi-
ronments. The values measured with each probe are compared and related to seed
levels of NO. In addition, concentration profiles required to compare probe
measurements with optical measurements are provided.
ii
-------
TABLE OF CONTENTS
ACKNOWLEDGEMENTS i
ABSTRACT ii
TABLE OF CONTENTS iii
LIST OF FIGURES v
LIST OF TABLES vi
I. INTRODUCTION 1-1
II. EXPERIMENTAL FACILITIES II-l
A. General II-l
B. Flat Flame Burner II-l
C. Test Section for Large Scale Burners II-3
1. IFRF Burner II-3
2. FT12 Burner Can II-8
3. Temperature Measurements II-8
D. Sampling Systems 11-12
1. Scott Exhaust Analyzer 11-12
a. Pumping Requirements 11-13
2. TECO Analyzer 11-13
E. Mass Flow Measurement 11-14
III. DESIGN OF GAS SAMPLING PROBE III-l
A. Losses of Nitric Oxide in Sampling System III-l
1. N0/N02 Interconversion III-2
a. The Bodenstein Reaction III-4
2. NO Reduction in Sampling Probe III-8
X
3. Losses in Sampling Line 111-13
4. Response of Chemiluminescence Analyzer 111-14
5. NO Converter 111-16
6. Summary III-17
iii
-------
TABLE OF CONTENTS (Cont'd)
B. Quenching in Gas Sampling Probe 111-17
1. Kinetics of NO Decomposition 111-20
2. Description of Computer Program for Probe Analysis. . 111-24
a. Sudden Expansion Losses 111-26
b. Aerodynamic Quench 111-27
C. Design of Probes 111-30
1. Probes for Combustor Measurements 111-30
2. Probes for the Flat Flame Burner 111-35
a. Microprobe 111-42
IV. EXPERIMENTAL RESULTS IV-1
A. Flat Flame Burner IV-1
1. Uncooled, Stainless Steel Probe IV-12
B. IFRF Burner IV-12
C. FT12 Measurements IV-20
D. Experimental Verification of Probe Model IV-25
1. Pressure Profiles for the Reference Probe IV-25
2. Mass Flow Measurements IV-28
a. Macroprobe IV-28
b. Miniprobe IV-28
c. Microprobe IV-30
3. Discussion IV-30
V. RESULTS AND DISCUSSION V-l
VI. CONCLUSIONS VI-1
REFERENCES R-l
-------
LIST OF FIGURES
Fig. No. Title
II-l. Top View of Flat Flame Burner and Assembly II-2
II-2. Atmospheric Pressure Combustion Facility II-5
II-3. Swirl Burner Assembly II-6
11-4. FT12 Assembly II-9
II-5. Pt-Pt/13% Rh Aspirated Thermocouple 11-11
II-6. Experimental Set-up for Measurements of Mass Flow . 11-15
II-7. Indicated Reading at Constant Mass Flow and
Varying Operating Pressure 11-16
II-8. Metered Mass Flow Rates vs. Flow Meter Reading. . . 11-17
III-l. Model for Calculating Sudden Expansion Loss .... 111-28
III-2. Drawings of Macroprobes 111-31
III-3. Reference Probe 111-32
III-4. Tip of Reference Probe 111-33
III-5. Calculated Temperature and Pressure Profiles
for Reference Probe 111-34
III-6. Calculated Cooling Curves for Macroprobes 111-36
III-7- Drawing of Miniprobe 111-38
III-8. Stainless Steel Tipped Miniprobe 111-39
III-9. Calculated Cooling Curves for Miniprobe 111-40
111-10. Calculated Cooling Curves for Miniprobe at
Varying Back Pressure 111-41
IV-1. Horizontal Temperature Profile over CH,/02/N2
Flat Flame IV-2
IV-2. Vertical Temperature Profile over CH4/02/N2
Flat Flame IV-3
IV-3. Normalized Nitric Oxide Profiles over CH4/02/N2/NO
Flat Flame IV-6
IV-4. Vertical Profiles of Nitric Oxide over Flat Flame
Burner IV-7
IV-5. Nitric Oxide Measured vs. Nitric Oxide Seed .... IV-10
IV-6. Temperature Profile Across IFRF Combustor IV-15
IV-7- Normalized Nitric Oxide Profiles Across IFRF
Combustor IV-18
IV-8. Temperature Profiles Downstream of FT12 Combustor . IV-21
IV-9. Normalized Nitric Oxide Profiles Across Optical
Axis for FT12 Combustor IV-24
IV-10. Profiles of Static Pressure for the Reference
Probe IV-26
IV-11. Relative Mass Flow vs. Back Pressure for Several
Probes IV-29
-------
LIST OF TABLES
Table No.
Title
II-A.
II-B.
II-C.
III-A.
III-B.
III-C.
III-D.
IV-A.
IV-B.
IV-C.
IV-D.
IV-E.
IV-F.
Operating Conditions for the Flat Flame Burner. . . II-4
Operating Conditions for the IFRF Burner II-7
Operating Conditions for the FT12 Combustor .... 11-10
Flow Conditions in Non-Ideal Sampling System. . . . III-6
Comparison of Calculated and Measured NO and NOo. . III-9
Reaction Mechanism for NO Decomposition 111-21
Estimated Fractions of NO Decomposition 111-23
Mole Percent of Stable Species for Flat Flame
Burner IV-4
Measured Concentration of NO(ppm) Using Uncooled,
Stainless Steel Probe over Flat Flame Burner. . . IV-13
Mole Percent of Stable Species for IFRF Burner. . . IV-16
Comparison of Nitric Oxide Measurements using
the Reference Probe - IFRF Burner IV-19
Mole Percent of Stable Species for FT12 Combustor . IV-22
Comparison of Nitric Oxide Measurements using
the Reference Probe - FT12 Combustor IV-25
vi
-------
I. INTRODUCTION
Since Johnston (1971) and Crutzen (1970, 1972) independently suggested
that the injection of nitric oxide (NO) into the upper atmosphere could signif-
icantly diminish the ozone (Oo) concentration, an accurate knowledge of the
amount of NO emitted by jet aircraft has been a serious concern to those
involved in environmental studies. This concern intensified when McGregor,
Seiber, and Few (1972) reported that NO concentration measured by ultraviolet
resonant spectroscopy were factors of 1.5 to 5.0 larger than those measured by
extractive probe sampling with subsequent chemiluminescent analysis. These
initial measurements were made on a YJ93-GE-3 engine as part of the Climatic
Impact Assessment Program (CIAP) which was one of four studies (CIAP, NAS,
COMASA, COVOS (see References)) commissioned to determine the possible environ-
mental consequences of high altitude aircraft operation, especially supersonic
aircraft. After those studies were initiated, economic factors strongly
favored the production and operation of subsonic aircraft. Nevertheless, since
the subsonic aircraft fleet is large and does operate as high as the lower
stratosphere, interest in the causes of the discrepancies between the two NO
measurement methods continued. Few, Bryson, McGregor, and Davis (1975, 1976,
1977) reported a second set of measurements on an experimental jet combustor
(AVCO-Lycoming) where the spectroscopically determined NO concentrations were
factors of 3.5 to 6.0 higher than those determined by the probe method. In
this set of measurements, optical data were obtained not only across the
exhaust plume but also in the sample line connecting the probe with the chemilum-
inescent analyzer. The sample line optical data seemed to agree with the
chemiluminescent analyzer data; hence, it was suggested that the discrepancies
were due to phenomena occurring in the probe. These results stimulated a third
set of measurements involving ultraviolet spectroscopy (Few et al, 1976a,
1976b), infrared gas correlation spectroscopy (D. Gryvnak, 1976a, 1976b) and
probe sampling on an Allison T-56 combustor. The measured ratios of the
ultraviolet to the probe values typically ranged between 1.5 and 1.9 depending
on the data reduction procedure. The ratios of the infrared to the probe
values varied between 1.1 to 1.5 also depending on the method of data reduction.
In addition to these engine and combustor data, evidence supportive of the
accuracy of the ultraviolet spectroscopic method, i.e., calibration data and
model predictions, was presented by McGregor, Few, and Litton (1973); Davis,
Few, McGregor and Classman (1976); and Davis, McGregor, and Few (1976).
Nevertheless, it was still not possible to make a judgment on the relative
accuracy of the spectroscopic and probe methods. The most significant reasons
for this were: the complexity of the spectroscopic theory and computer model
required to infer concentration from optical transmission; the inadequate
treatment of probe use; and the incomplete exhaust temperature and pressure
data which are necessary for a valid comparison of the methods. Recently,
Oliver et al (1977, 1978) as part of the High Altitude Pollution Program has
ranked these discrepancies as a major and a continuing source of uncertainty in
atmospheric model predictions.
1-1
-------
The purpose of this investigation was to identify and determine the
magnitude of the systematic errors associated with both the optical and probe
sampling techniques for measuring NO. To accomplish this, the study was
divided into three parts. The first was devoted to calibrating the ultraviolet
and infrared spectroscopic methods. This entailed developing procedures which
could provide known concentrations of NO over a wide range of temperatures and
pressures, and also reviewing and correcting the ultraviolet spectroscopic
theory used in the engine and combustor measurements cited above. The second
part of this study was focused on sample extraction, transfer, and analysis of
chemiluminescent instrumentation. The sampling methods were used on three
successively more complicated combustion systems starting with a flat flame
burner and culminating with a jet combustor. The results are presented in TASK
II Report: Probe Methods. In the third part of this study, optical measurements
were made on the same three combustion systems operated at the same conditions
used for the probe measurements. The results of the optical and probe measure-
ments were compared and are given in TASK III Report: Comparison of Optical
and Probe Methods.
This report, i.e., TASK II, is devoted to the important processes in the
extractive sampling and measurement of NO from combustion streams. Described
are a flat flame burner, swirl combustor, and jet combustor along with their
support facilities and operating parameters. Information on the temperature,
analytical, and mass flow instrumentation is given. Problems associated with
loss of nitric oxide in a sampling system are reviewed and the results of
previous investigations are analyzed. Several different probes were used for
sampling the flame gases, and their designs were selected using a computer
program describing the principal aerodynamic and heat transfer processes
encountered in a probe. Direct experimental measurements of the fluid mechanics
within probes is presented verifying this model. In addition, temperature and
concentration profiles necessary to compare optical and probe measurements are
provided. These results are summarized and discussed and major conclusions are
given.
1-2
-------
II. EXPERIMENTAL FACILITIES
II. A. General
A principal objective of this study is to identify the relative merits
of probe and UV optical measurements of NO in the exhaust of aircraft engines.
To reach this goal, three combustion systems of varying degrees of complexity
were examined. These systems were:
1. CH^/C>2/N2 flame over a flat flame burner:
= 0.8, 1.0, 1.2, P = 1 atm, mTOTAL s 2.75 g/sec
2. C-jHg/Air flame in a swirl burner:
4, = 0.8, 1.0, 1.2, Swirl = 0.63 and 1.25, P = 1 atm, m-p^j^ s 71g/sec
3. Jet A/Air flame in a modified FT12 combustor: Idle, Cruise,
and Maximum Continuous, P = 1 atm,
Physical details and operation of the flat flame burner have been described
previously (Dodge, et al . , 1979); consequently, only a brief overview will be
presented here. For the other two flames, each burner assembly could be
installed separately into a single combustor housing with the associated fuel
lines and flow controls modified accordingly. This facility and the burner
assemblies are described in detail in this chapter. In addition, techniques
for temperature measurements with corrections, sample gas transfer and analysis,
and mass flow measurements are reviewed. Details of sample probe construction
will be discussed in the following chapter since their designs were defined by
model predictions.
II. B. Flat Flame Burner
The flat flame burner is made of sintered copper and has two zones: the
main zone (containing the main flame seeded with nitric oxide) with dimensions
of 17.5 x 9.2 cm or 161 cm and the (unseeded) buffer zone with an area of 76
cm^. A methane flame was burned above the buffer flame to provide a hot zone
in the wings of the flame. The burner was enclosed by a stainless steel
shroud/chimney with optical ports to separate windows (quartz or salt) from the
flame. The ports were purged with nitrogen at room temperature to reduce the
local nitric oxide concentrations within these ports. A top view of the burner
is shown in Figure II-l. Temperatures were measured using a butt-welded, Ir/60%
Ir-40% Rh thermocouple coated with a mixture of Yttrium and Berylium oxides.
* The stoichiometry, $, is defined to be (f/a)/(f/a) ^ where 4, = ! when the
fuel (f) and air (a) are at the stoichiometric ratio.
II-l
-------
TOP VIEW OF FLAT FLAME BURNER AND ASSEMBLY
I
r ->
BUFFER ZONE-
S.S. SHROUD
INERT
GAS PURGE
II
1
1
MAIN GAS FLOW ^
(SEEDED W/NO)
X
X
\
y
x
/
>
QUARTZ WINDOWS-^
i|A
i
i
v.
^^ DEAD SPACE
SCALE
to
I
o
5 cm
-------
The diameter of the bead and coating was approximately 90 microns (0.0035
inches). Gases were individually tnetered using critical flow orifices.
Separate mixes of gases (N2, C>2, CH^, NO, H2, Ar) were blended for the
main and buffer flows. Details of these facilities are provided in the Task I
Report (Dodge, et al, 1979). The flames examined in this program are listed in
Table II-A.
II. C. Test Section for Large Scale Burners
The combustor test section used for the swirl burner and FT12 combustor
is shown schematically in Fig. II-2. It consists of a water-cooled, double-
walled chamber 50 cm in diameter (i.d.) and 150 cm long. Four (4) rows of
eight (8) viewing ports are provided in the combustor section at 90° intervals.
It was constructed at UTRC specifically to investigate flame phenomena with
various optical and probing techniques. The two burner systems were designed
to fit inside the burner housing. One is a swirl burner and is a scaled down
verison of the burners designed at the International Flame Research Foundation
(IFRF). The second is a modified FT12 burner and shroud. The optical axis
used for subsequent optical measurements was the center of the third window
from the far right in Fig. II-2. All probes (sample and thermocouple) were
designed to translate across this axis.
II. C. 1 IFRF Burner
The IFRF burner assembly, as shown in Fig. II-3, is a model of those burners
described by Beer and Chigier (1972) and consists of a central fuel nozzle and
an annular air supply. A movable vane block arrangement provides variable air
swirl intensity from a swirl number of 0 to 2.5; in this case, the swirl number
is defined as the ratio of the tangential to the axial momentum divided by the
radius of the exit quarl. The swirl number was calculated using the appropriate
equations in the text by Beer and Chigier (1972) and, as they demonstrate,
experimental and theoretical values agree fairly well for this type of burner.
An axially adjustable, 19 mm diameter, fuel feed tube can be equipped with
various pressure atomizing of air-assisted fuel spray nozzles.
This swirl burner has been tested previously during internal programs at
UTRC and has been used recently to study the combustion of a coal/oil slurry
(Vranos, et al, 1979). In the present program, gaseous propane was used for
fuel and a nozzle was constructed to inject the fuel radially into the swirling
air flow. Six stable operating conditions were selected for these tests and
provide three stoichiometries and two swirls. Input conditions are listed in
Table II-B. The optical axis and the probe tips were located 87.5 cm
downstream of the quarl exit.
The two swirl levels used in these tests (0.63 and 1.25) were selected by
performing a series of tests on flame stability. At lower swirl numbers, the
propane flame was relatively long and unstable and was not considered to be
suitable for this series of tests. Beer and Chigier argued that below a swirl
number of 0.6 axial pressure gradients are insufficient to cause internal
II-3
-------
TABLE 11-A
OPERATING CONDITIONS FOR THE FLAT FLAME BURNER*
Test Condition Bl B2 B3
;N2 (g/sec) 2'15 2-20 2'07
m 02 (g/sec) 0.512 0.466 0.494
m CH4 (g/sec) 0-103 0.116 0.149
T inlec (K) 285 285 285
P (psia) 14.7 14.7 14.7
* 0.8 1.0 1.2
Without Seed
II-4
-------
ATMOSPHERIC PRESSURE COMBUSTION FACILITY
TRANSITION DUCT
AIR
SWIRL BURNER
4 ROWS OF 8 QUARTZ WINDOWS
WATER-COOLED, DOUBLE-WALLED. CONSTRUCTION
-COMBUSTOR SECTION (60 IN. LONG X 20 IN. DIA)-
I
in
-------
SWIRL BURNER ASSEMBLY
I
8
SWIRL VANE ADJUSTMENT/INDICATOR
STATIC PRESSURE
SWIRL VANE
ARRANGMENT
STATIONARY VANES
MOVEABLE VANES
COOLING WATER
Cxi
-------
TABLE II-B
OPERATING CONDITIONS FOR THE IFRF BURNER
Test Condition
mair
P (psia)
Tinlet
* *fuel (g/sec)
1, 4+
66.7
14.7
290
3.40
0.8
2. 5+
66.7
14.7
290
4.26
1.0
3, 6+
66.7
14.7
290
5.11
1.2
*
Gaseous propane
+ Swirl number (see definition in text) was 0.63 for test conditions 1-3
and 1.25 for test conditions 4-6.
II-7
-------
recirculation; however, at higher swirl intensities a recirculating zone in the
central portion of the jet is required to support a strong adverse pressure
gradient along the axis. Since recirculation zones tend to stabilize the
flame and increase the intensity of reaction, the ability to achieve stable flame
conditions only above swirl numbers of 0.6 in this research program is in agreement
with Bee'r and Chigier's analysis.
II. C. 2 FT12 Burner Can
A modified FT12 combustor can used in this program was 29.5 cm in length
(11.5 cm shorter than the original can) and 13.0 cm in diameter. It was
altered to make all air addition holes symmetric. This can was welded at its
exit to a shroud and was placed in the test section with the swirl burner
removed. As shown in Fig. 11-^, a flow straightener was placed in the burner
housing upstream of the combustor can and appropriate fuel lines and cable for
spark ignition were fed through the housing. A standard fuel nozzle for the
FT12 (Pratt & Whitney Part No. 525959) was used in this series of tests.
Three flight conditions, i.e. idle, cruise, and maximum continuous, were
simulated. A simulation was necessary since the test section could only
sustain a maximum pressure of four atmospheres, while the cruise and maximum
continuous flight conditions required pressure above 6 atmospheres.
In addition, since gas sampling and accurate definition and examination
of the optical path is simplified by operating at one atmosphere, all experiments
were performed at one atmosphere. Simulated flight conditions at this lower
operating pressure, were calculated by equating Mach numbers. This is a common
test procedure and is useful in simulating equivalent fluid flow patterns and
heat transfer. In this case, the mass flow rate was reduced appropriately to
maintain the chamber pressure at one atmosphere and the inlet temperature was
identical to the flight conditions. The simulated flight conditions are listed
in Table II-C. The optical and probe axis was 78 cm downstream from the exit
of the FT12 burner can.
II. C. 3 Temperature Measurements
Exhaust temperatures from the IFRF and FT12 combustors were made using a
water-cooled, double shielded, aspirated thermocouple probe with a bead made of
Pt/Pt-13%Rh. The probe was manufactured by Aero Research Instrument Co. (Part
Number T-1006-6 (25) R) according to specifications described by Glawe, et al.
(1956) but was modified (by water cooling and material substitution) to increase
the temperature range to above 1370 K (2000°F). . ohotograph of this probe
is shown in Fig. II-5. Radiation and conduction corrections were made according
to equations supplied by the manufacturer. For the measurements made in this
study (P = 1 atm, Mach no. « 1), these equations can be written
Tgas U) = thermocouple U) * * TRC
11-8
-------
FT12 ASSEMBLY
-FLOW STRAIGHTENER
L7
o
o
o
o
o
o
L
MODIFIED FT12
BURNER CAN
/ ii
TJ1
FUEL
J IGNITER
CABLE
AIR
to
I
o
I
00
(JV
I
~1/3 ACTUAL SIZE
-------
TABLE II-C
OPERATING CONDITIONS FOR THE FT 12 COMBUSTOR
m
ar
P (psia)
1 inlet
* *fuel (i/sec)
f/a
Idle
485
14.7
335
5.15
.0106
Cruise
463
14.7
515
6.62
.0143
Max imum
Continuous
454
14.7
524
6.92
.0152
Jet A
11-10
-------
Pt-Pt/13% Rh ASPIRATED THERMOCOUPLE
(WATER OUT NOT VISIBLE)
u>
I
o
I
p
-------
where the radiation and conduction correction, ATRC, is given by
ATR(, = 0.55 C1 (K)
GI (K) - 0.00917 Tchermocouple (K) - 7.136
and M is the Mach number. For the FT12, no correction was made since
L TRC was negligible « 10 K). For the IFRF burner, corrections were
typically on the order of 50 K. Additional description on the operation of
aspirated probes is given by Land and Barber (1954).
II. D. Sampling Systems
Two instrumentation systems were used for measuring the products of
combustion. The first is the Scott Exhaust Analyzer used in the Task I of this
program and the second is a chemiluminescence analyzer made by Thermo Electron
Corporation (TECO). The Scott system was used during the initial tests with
the flat flame burner to obtain concentrations of the major species. The Scott
system was thereafter dedicated to the larger scale combustor tests and the
TECO instrument was used for subsequent measurements of NO/NOX over the flat
flame burner. The Scott system was dedicated to the combustor measurements for
two reasons. First of all, it was impractical to move this system between
facilities and secondly it was determined that the Scott package (under the
conditions of the flat flame tests) did not satisfy the Federal requirements
for total instrument response time and could not be easily modified to meet
those requirements.
II. D. 1 Scott Exhaust Analyzer
The Scott Model 119 Exhaust Analyzer provides for the simultaneous anal-
ysis of CO, C02, NO or N02, Oj and Cotal hydrocarbons (THC). The
analyzer is an integrated system, with flow controls for sample, zero and
calibration gases conveniently located on the control panel. The incoming gas
simple passes through a refrigeration condenser (
-------
analyzer was stainless steel, and was operated at a temperature of
approximately 1000 K. A Scott Model 150 Paramagnetic Analyzer is used to
measure the 02 concentration in the gas sample. Concentration ranges available
with this instrument were from 0-1% to 0-25% on several scales, with a nominal
accuracy of + 1% of full scale. A Scott Model 116 Total Hydrocarbon Analyzer is
used to measure the hydrocarbon concentration in the gas sample. This analyzer
utilizes an unheated flame ionization detection system to provide for measurement
of hydrocarbons (as methane) in concentration ranges from 0-1 ppm to 0-10%, with
a nominal accuracy of + 1% of full scale. Output signals from the various
analyzers are displayed on chart recorders and a digital display.
The sample line was teflon-lined aluminum. The typical operating temperature
was 380 K. When sampling from the large combustors, water was removed from
the sample by two traps cooled to 3° C. The first was located 6 ft. beyond
the probe exit and the second was housed in the Scott analyzer.
II. D. la Pumpmig^ Requirements
Two problems specific to this sampling/probe system were encountered.
First of all, the orifice diameter of the macroprobes (2mm) was sufficiently
large that a separate vacuum pump (17.5 cfm) was required to reduce the back
pressure of the probe to the very low values ( - l/10th of an atmosphere)
required in this program. This vacuum pump was attached directly to the probe
via a line one inch (2.54 cm) in internal diameter and three feet (90 cm) long.
The second requirement was that, at the reduced pressures, the pumping capacity
of the sampling system must be sufficient to deliver flow to the analytical
instrumentation. To accomplish this task, a MB-301 pump and two MB-118 pumps
(metal bellows) were assembled in a series/parallel arrangement. These pumps
were in addition to the two MB-118 pumps in the Scott analyzer and the vacuum
pump associated with the CLA. As discussed in Section IV. D. 1, even with this
pumping capacity the deliverable flow was marginal at the lowest of probe back
pressures.
II. D. 2 TECO Analyzer
The Thermo Electron Corporation (TECO) Model 10AR Chemiluminescent NO/NO
X
analyzer was used for the reported data for the flat flame burner. This
instrument has a stated minimum detectable concentration of 50 ppb and a
maximum limit of 10,000 ppm. Linearity within any of its eight operating
ranges is given as ± 1%. A TECO Model 300 Molybedenum NOX Converter was used
for the NOo determinations. Sample was delivered to this analyzer at atmospheric
pressure by a metal bellows pumps (Metal Bellows MB-118). The sample line was
12 ft of treated teflon line (Technical Heaters, Inc.). The back pressure of the
probe was continuously monitored using a Matheson test gauge (0-760 mm, absolute).
11-13
-------
II. E. Mass Flow Measurements
The purpose for measuring mass flow rates through the sampling probes was
to provide data for comparison with model predictions. Using the experimental
arrangement depicted in Fig. II-6, flow rates through probes of three orifice
diameters (75, 635, and 2000 microns) were measured at varying external tempera-
tures and probe back pressures. To prevent water condensations, heating tape
was used between the probe and exit of the mass flow meter. The Hasting meters
which were used for these mass flow measurements, operate by siphoning a small
but constant fraction of the gas flow and passing it over a series of heated
thermocouples. Cooling of the thermocouples due to the gas flow is measured by
the meter and is primarily a function of the mass flow and the specific heat of
the gas. According to the manufacturer, mass flow calibrations made for one
gas can be related to another gas or mixture of gases by the ratio of specific
heats.
To accommodate the wide range in mass flow due to changes in orifice size
(m >r d orifice) anc* temperature (m >r l/\/Y), three low pressure drop, Hastings
mass flow meters (ALU-100, ALU-5K, ALU-20K) were used. Full scale on these
Hastings meters were 100, 5000, and 20,000 seem (.00191, 0.0953, and 0.382
g/sec of nitrogen, respectively). The reported accuracy is ± 1% of full
scale. In addition, these units each have pressure drops of less than 0.13
torr at an operating pressure of 760 torr (1 atm) and less than 1 torr at
30 torr. The transducers were installed in the lines according to the manufac-
turer's recommendations. The assembled plumbing with meter was checked for
leaks by pressurizing to 100 psig.
To verify their operation two tests were made. First, a constant mass flow of
nitrogen was passed through the transducer while the operating pressure of the
transducer was varied by adjusting the valve between the transducer and a
vacuum pump. At each constant mass flow, the meters produced a constant
reading (typically + 0.5%) independent of the operating pressure. In agreement
with the specifications, pressure measurements up and downstream from the
transducer indicated a small pressure drop « 3 torr). Throughout the pressure
range investigated (100-760 torr), the range of mass flows were supplied using
critical flow orifices upstream of the transducers. Typical data are presented
in Fig. II-7.
The second test was a check on the linearity and the calibration of these
units. The experimental set-up was identical to the check for constancy of
indicated reading with varying pressure and constant mass flow. Experimental
data are shown in Fig. II-8. Although the linearity appears to be quite good
(other than a small deviation with the ALU-5K meter), a noticeable discrepancy
with the calibration was found for the ALU-100 (>/> 7%) and ALU-20K (^ 102)
meters. For these tests, the calibrations obtained at UTRC were used. Less
than 12 change was observed for the mass flow measurements shown in Fig. II-8
when the lines were heated to 60° C.
-------
EXPERIMENTAL SET-UP FOR MEASUREMENTS OF MASS FLOW
EXHAUST
I
VACUUM
PUMP
MFTFR .,, I nvM
»
»
^-\ r^-i i- HEATING TAPE
ix^xj "lUllUuI " ""
MASS
(j\ FLOW /
-------
INDICATED READING AT CONSTANT MASS FLOW AND VARYING OPERATING PRESSURE
8
O ALU-100
A ALU-100
D ALU-100
A ALU-20K
0 ALU-5K
1.0
0.8
LU
_j 0.6
_i
=>
U.
LL
O
z
Q 0.4
cr
0.2
O -O
A
-
D - D --- -- ---- - - -- D ^
I
I
I
I
Mw
(Ng
(C3/SEC) SCCM
0.00182 953
0.00144 75.4
0.193 10.110
0.0410 2150
0.00064 33.5
100 200 300 400 500 600
PRESSURE DOWNSTREAM OF FLOW METER
(TORR)
700
O
-------
METERED MASS FLOW RATES VS. FLOW METER READING
O ALU 20K
A ALU 5K
D ALU 100
FIG. n-8
tn
ID
_i
<
O
O
c/3
6000
5000
4000
100
O 3000
g 2000
cc
1000
60
40
20
I
0.2
0.4 0.6 0.8
FRACTION OF FULL SCALE
1.0
1.2
24000
22000
20000
18000
16000
m
IE
m
CO
14000 co
1.4
12000
10000 O
8000
6000
4000
2000
0
79-10-85-10
11-17
-------
III. DESIGN OF GAS SAMPLING PROBES
III.A. Losses of Nitric Oxide in Sampling System
Although this report focuses on probe measurements of nitric oxide and
associated probe phenomena, it is clear that a gas sample probe (and the
corresponding extraction of a gas sample from a flame environment) is only one
part of a sampling system. Since any portion of this multistep process (from
gas sampling to species analysis) could cause errors, it is worthwhile to
review this process. Typically, the individual steps include:
1. Extraction of the sample from the flame environment without
perturbations external to the probe due to local temperature changes
or catalysis (due to the presence of a probe).
2. Quenching of the flame gases inside the probe by rapid temperature
and, usually, pressure reduction without the occurrence of hetero- or
homogeneous kinetics.
3. Removal from the flame environment and transfer of the sample to the
instrumentation without condensation or reactions on walls.
4. Water removal using, for example, an ice trap to minimize condensation
and/or interference in the detectors yet without condensing or absorbing
other species of interest.
5. Filtering of the gas sample for particulates.
6. Pressure recovery using a non-interfering pump to produce gas samples
at pressures required by the detector(s).
7. Analysis of the sample for the species of interest with known or
calculable corrections for the presence of interfering species.
In specific cases, certain items such as the water trap, filter, or pump
may not be necessary due to combustor conditions, instrumentation, and required
pressures; while in other systems, additional facilities, such as a storage
capability, may be needed.
In regard to the sampling and analysis of nitric oxide (NO) or total
nitrogen oxides (NO and NOo), nearly all of the above steps have been sus-
pected and examined as a source of sample perturbation. For this report,
problems associated with sampling both NO and N02 are important due to the
known interconversion between these species. Although many authors have
discussed selected problems associated with the measurement of nitrogen oxides,
III-l
-------
perhaps the most comprehensive reviews have been written by Cernansky (1976)
and Tuttle, et al. (1973). Complications with measurements of nitric oxides
such as those mentioned in these reports and with measurements of other gaseous
emissions led the Federal government to write regulations (according to recom-
mendations by the Society of Automotive Engineers (E-31 Committee)) for the gas
sampling and measurement of aircraft emissions (Federal Register; 1973 and
1976). In spite of these efforts, many uncertainties in the measurements of
nitrogen oxides remain. Primarily, these uncertainties include:
1. Interconversion between nitric oxide and nitrogen dioxide within
the sampling probe or sampling line.
2. Chemical reactions within the probe that reduce nitrogen oxides to
molecular nitrogen or other nitrogeneous species.
3. Sampling line losses of NOY (i.e., NO and N09).
A ^
4. Improper calibration or corrections for a chemiluminescent detector
for the presence of species other than nitrogen (the usual diluent in
calibration gases).
5. Low efficiency for the NO- * NO converter or complete reduction of
nitrogen oxides to molecular nitrogen or other nitrogeneous species in
the absence of oxygen.
Experimenters can also observe apparent losses of nitric oxide due to a
variety of experimental problems, including unconditioned sample lines, small
leaks, and even under unusual operating conditions (e.g., flow rate through
probe is less than that required by the analytic instruments) reverse flow
through a bypass valve that may dilute the gas sample. The above items are
discussed in detail in the following sections.
III.A.I N0/N02 Interconversion
Prior to the early seventies, it was believed that very little nitrogen
dioxide was formed during combustion processes and that nitric oxide made up
nearly all of the emissions of nitrogen oxides. Since that time, however, many
experimenters (e.g. Anon, 1971; Schefer, et. al., 1973; Merryman and Levy,
1974; Allen, 1975; Kramlich and Malte, 1978; Amin, 1977; Cernansky and
Singh, 1979; Johnson, et. al., 1979, and Clark and Mellor, 1980) have
probed various combustion systems for nitrogen oxides and have found large
NO^/NO ratios. Throughout this decade the source(s) of this measured N07
has been questioned. Although the flame, probe, and the sampling line have
each been suspected as its source, it is apparent now that each system must be
analyzed separately. In a gas turbine combustor, for example, nitric oxide
formed in the primary zone may be converted to N02 by relatively cold air
entering from the dilution holes (Chen et. al., 1979). Alternatively, as the
II1-2
-------
gases in a probe are cooled to approximately 1000K, flame radicals are
quenched and may be converted to the hydroperoxyl radical (H02) which can
oxidize nitric oxide via the reaction
NO + H02 -» N02 + OH (III-1)
Kinetic analyses by Johnson, et. al. (1979) and Kramlich and Malte (1978) for
cooled probes indicate that this reaction is of prime importance in the con-
version of NO to N02. The reaction
NO + 0 + M -»N02 + M (III-2)
although considered, is relatively unimportant due to the short lifetime of
atomic oxygen. In fact, the model by Johnson, et. al. predicts that if all the
NOy begins as N02, some will be converted to NO. In general, they conclude
that for these flames (producing small quantities of NOj,, ^ 10 ppm) , no
relationship exists between the measured N0/N02 ratio and the actual ratio in
the flame.
Schefer, et. al., (1973) observed the unusual result that cooled probes
(quartz and stainless steel) indicated virtually no nitric oxide but several
ppm of NO-? in an opposed jet combustor (premixed, propane) and with uncooled
probes the nitrogen oxides were composed nearly entirely of NO. Since the
probes were placed in the reaction zone and sampled only partially burned
gases, the authors argued that within the uncooled probes exothermic reac-
tions continued, thereby heating the surfaces to feed the catalytic conversion
of NO- to NO (similar to NOV converters used with CLA). They concluded, there-
/ A
fore, that the very high N02/(NO + N02) ratios (nearly 1) obtained with the
cooled probes are realistic measurements. Based on the recent studies by
Johnson, et. al. (1979) and Kramlich and Malte (1978), it seems more reasonable
that Reaction (III-l) is at least partly responsible for conversion of NO to
N02 in the cooled probes, especially in light of the presence of oxygen and
unburned fragmented hydrocarbons which are known to produce the H02 radical
during decomposition. Any N02 similarly formed in the uncooled probes would
undoubtedly be reconverted to NO on the hot surfaces.
Other experimenters have considered the N02 to NO conversion (based on
the same principle as a catalytic NOX converter) to be important for uncooled
stainless-steel probes. Benson, Samuelsen, and Peck (1976) and Benson and
Samuelsen (1976, 1977), for example, have examined a similar phenomena in
simulated (heated) probes and in the presence of carbon monoxide, hydrogen, and
unburned hydrocarbons. Their work cannot be directly applied to probe behavior
since no (overall) kinetics were derived from their work and, more importantly,
the residence time (~ one second) in the simulated probes was much longer than
expected residence times in an uncooled probe. This data is much more
descriptive of the behavior of catalytic converters.
III-3
-------
In the case of either the NO to N02 oxidation via Reaction (III-l)
or the surface reduction of N02 to NO, neither mechanism can be used for
quantitative predictions at the present time due to the inability to describe
accurately and simultaneously the fluid dynamics, heat transfer, and chemical
kinetics (both hetero- and homogeneous) occurring within a probe. (Note: The
present study contributes substantially to the understanding of the fluid
mechanics and heat transfer for a certain class of sampling probes. Section
III.B.2)
Another possible mechanism for conversion of NO to N02 is the reaction
NO + NO + 02 * 2N02 (III-3)
that may occur in the sample transfer lines or in the instrument lines leading
to the reaction or measurement chamber. Although Cornelius and Wade (1970)
have concluded that this reaction was unimportant in their system, it should be
noted that each sampling system should be examined since this reaction is a
strong function of the nitric oxide concentration and the total pressure. For
systems with low concentrations of nitric oxide « 250 ppm), low oxygen concen-
tration, or low sampling line pressures and residence times, this reaction is
undoubtedly insignificant; however, for systems with large NO concentrations
due to seed NO or fuel nitrogen, with high oxygen concentrations, or with high
sampling line pressures even in only part of the system, this reaction may
convert substantial fractions of the nitric oxide to nitrogen dioxide. Although
this phenomena was observed in the first phase of this program (Dodge, et. al.,
1979), complete details of this conversion were not reported. Since this
reaction is also of interest in this part of the program, further details of
these measurements have been given in the following section.
Since the rate constant for Reaction III-3 is well known, the contribution
of this reaction can be estimated once initial NO and 02 concentrations are
identified, and pressures and residence times throughout the sampling system
are measured.
III.A.la The Bodenstein Reaction
The reaction
NO + NO + 02 + 2N02 (III-3)
under certain conditions may contribute to conversion of NO to N02 in a
sampling system. It is undoubtedly not a three-body (or termolecular) reac-
tion but rather represents a sequence which either forms the dimer,
NO + NO t (NO), (III-4)
III-4
-------
(N0)2 + 02 + 2N02
or one that forms the nitrate
NO + 02 .>. N03 (III-6)
N03 + NO + 2N02 (IH-7)
Although this has been discussed by many authors (see review by Baulch, et.
al., (1970)) the actual route is not resolved. Nevertheless, this reaction has
been examined by many experimenters and its rate constant is known better than
+ 50%, kli;[_3 = 1.2 x 109 exp (+523/T) cc2/mole2-sec (Baulch, et. al., 1970;
Hillard and Wheeler, 1977) when the reaction rate is defined as in Equation
III-8.
The rate of loss of NO with respect to time due to this reaction can be
written
d[NO] 9
dt = -2km_3 [NO]-* [02] (III-8)
where brackets, [ ], represent concentration of the molecule in moles/cc.
According to this equation, the rate of the Bodenstein reaction is dependent on
the square of nitric oxide concentration and the first power of the oxygen
concentration. Since number densities are directly proportional to pressure,
this reaction rate is also indirectly dependent on the total pressure to the
third power.
This reaction then becomes quite important in sampling systems where
either high concentrations of nitric oxide or oxygen exist, high pressures
exist, or long residence times occur. Alternatively, their opposites will tend
to disfavor Reaction III-3. Specifically, this reaction may be of importance
when NO or fuel nitrogen is added to the flame and when high sample line
pressures exist, even for short line lengths.
During the course of the Task I investigation (Dodge, et. al., 1979) high
concentrations of NO were added to a H2/02/Ar flat flame so that infrared
optical measurements could be made. Simultaneous probe measurements were also
made during these tests. For the particular sampling conditions used in these
experiments, Reaction III-3 was found to contribute significantly to the
conversion of NO to N02. The results can best be understood by examining the
complete sampling system. In Table III-A, a review of different components of
the sample system from the probe tip to the reaction chamber in the CLA is
given. Estimates of lengths of line, local pressure, residence times, and
temperatures are provided. This sampling system is only part of the complete
III-5
-------
TABLE III-A
FLOW CONDITIONS IN NON-IDEAL SAMPLING SYSTEM
Reaction Chamber
Probe and Sample
Transfer Line
Refrigerator
(2)
Metal Bel-
lows Pumps
Flowmeter and Con
nection to CLA
H20 Re-
moval
Bypass
Valve
Distance
cm (feet)
550 (18)
1070 (35)
460 (15)
470 (15.5)
Pressure
torr (atm)
350 (0.46)
350 (0.46)
785 (1.03)
760 (1.0)
Eat imated
Residence
Time (sec)
1.6
4.2
3.7
3.7
Temperature
CO
110
20
20
Conditions for
1600 K
-------
SCOTT Instrument system in which the gas can be transferred to any of five
analytical instruments. In a more compact system containing only a
CLA, it is expected that line lengths and therefore residence times will be
noticeably shorter. For the estimates provided in Table III-A, the individual
residence times were based on total residence times and estimated mass flow,
temperature, cross-sectional area, etc. in each section. The total residence
time was measured as the time betweeen the moment when NO is first visually
observed after a toggle valve is opened to add seed NO to the flame (i.e., when
the flame turns a greenish-gray color) and the moment when the indicated nitric
oxide concentration begins to rise rapidly. The pressures at various locations
in the sampling system were obtained by placing several pressure gauges along
the sample lines. The sampling system was operated to meet the Federal Require-
ments for a sampling system with the following exceptions. First of all,
the sample line was held at 110 °C rather than the required 150 °C since, in
this case, the sampled gas consisted of 1^2/^2 combustion products and
condensation of hydrocarbon fragments was not of concern. Secondly, the total
response time of the analytical system (from probe tip to 15% response at the
detector as defined by the Federal Register) was approximately 18 seconds vs.
the required time of less than nine seconds. For sampling over the flat flame
burner with the SCOTT instrument package, this time could not be greatly
reduced at the flame temperatures of these measurements (
-------
= 2kTTT »[09] t- + 1 (III-9)
'
where lNO]Q is the nitric oxide concentration before passing through
the ith section of the sampling line and [NO]^ is the concentration at the
end of this section. By summing Equation III-9 for each section (accounting
for pressure variations appropriately) and assuming that the measured NOX
estimates the original nitric oxide concentration at the probe tip, the final
value of NO just prior to the CLA may be calculated. These values were esti-
mated for several experimental seed levels of NO at various flame conditions.
Calculated and experimental values of NO and N02 are reported in Table III-B.
Good relative agreement is found between the calculated and measured values of
N02 (see ratios) although the calculated value is approximately a factor two
higher than the measured value. This factor of two difference is not due to
an error in the rate constant but is probably due to inaccurate estimates of
the flow parameters in the sampling system and/or loss of N0~ in the system.
In any case, the similarities in the trends and magnitudes of conversion
between the calculated and measured values of N02, provides good evidence that
Reaction III-3 is responsible for the conversion of NO to N02 in this sampling
system.
III. A. 2 N0x Reduction in Sampling Probe
For uncooled stainless-steel probes sampling fuel-rich flames there
exists much evidence that nitrogen oxides can be reduced (probably to molecular
nitrogen) within the probe. This phenomena is not unexpected since it is the
same as that observed when attempting to use a stainless-steel catalytic
converter to convert N02 to NO in the presence of fuel-rich gases. Reduction
by stainless-steel probes has been observed by Halstead, et. al. (1972) who
compared an uncooled quartz-lined probe with an uncooled stainless-steel lined
probe and by others (England, et. al., 1973; Cernansky and Singh, 1979; and
this work). This problem is easily eliminated or at least drastically reduced
by using cooled probes for sampling fuel-rich flames.
Under stoichiotnetric or fuel-lean conditions very little evidence exists
that indicates probes significantly alter total concentrations of nitrogen
oxides. England, et. al (1973) have observed some dependence of NO concentra-
tion with changes in probe type; total NO , however, is not reported in this
report and some of the differences may be due to interconversion between NO and
NOo. In addition, the only major differences are noted between the cooled
and uncooled probes. Few, McGregor, and coworkers (1972, 1975, 1976, and 1977)
have compared UV optical measurements of NO to probe measurements and have
concluded that the probe measurements are up to a factor of six lower than the
optical measurements. These conclusions are, however, subject to question due
III-8
-------
TABLE III-B
COMPARISON OF CALCULATED AND MEASURED NO AND NO,
Measured
Calculated
1
Calc/Meas,
Calculated
°2
%
5.0
5.0
5.0
7.9
7.9
Measured
NOX
(ppm)
4891
7266
7114
3312
4533
NO
( ppm )
4727
6833
6715
3189
4272
N022
(ppm)
164
433
399
123
261
NO
(ppm)
4550
6539
6416
3080
4110
N02
(ppm)
341
727
698
232
423
N02
2.08
1.68
1.75
1.89
1.62
1
Estimated concentration at CLA assuming the initial NO concentration at the
probe tip is equal to the measured NO
Measured NO minus measured NO
X
III-9
-------
to errors in the calibration procedures and theoretical model (Dodge, et. al.,
1979). Cernansky and Singh (1978) have observed some differences in total
NOX for a variety of probes sampling fuel-lean, stoichiometric and fuel-rich
flat flames; however the differences are typically small « 15%).
Recently, Clark and Mellor (1980) have compared NO and NOX measurements
using several different probes in a model gas turbine combustor. They report
some rather large differences between measurements of NO using blunt and
tapered tip probes (as much as a factor of three); however, measurements of
total N0x indicate relatively small scatter (~20%) which is typical of the
day to day variations in their combustor and/or analysis system. It should be
noted that this agreement is achieved despite high measurements of hydrocarbons
(>4%) that indicate the probe is sampling within the reactive flame zone.
Optical and probe measurements have been compared at other laboratories
and in general good agreement is found. Meinel and Krauss (1978) have made in
situ measurements of NO in both H2/air and C^Hg/air laminar premixed
flames using a UV resonant lamp. In lean flames, agreement between the probe
and optics is excellent, i.e., within several percent. For rich flames, the
optics produce values which are not greater than the probe values but are
approximately 20 to 25 percent lower. Falcone, et al. (1979) have made optical
measurement using an infrared, tunable diode laser. For a lean flat flame,
0 = 0.67, the optical measurements are about 20 percent higher than the probe
values; however, they conclude that the primary uncertainties are associated
with the laser system since the probe measurements agree well with the seed
values of NO.
Bilger and Beck (1975) have made probe measurements on a turbulent diffu-
sion, hydrogen/air flame and compared these measurements to those of previous
work on the same or similar system. The measurements of the major species,
i.e., Hj, H^O, O^, in general agreed quite well with only a small axial
shift observable. The NO measurements, however, differed quite noticeably with
the more recent results (using a probe with a slender nose) suggesting peak
NO concentrations as much as 30 to 35 percent higher than earlier measurements
with a probe having a blunt nose profile. In addition, they compare data where
substantially different NO profiles between the small and large probes are
observed with the small probe producing the highest NO. A substantial shift was
also observed in the major species. For the large probe, the NO profiles were
dependent on the flow rate, but no difference in the major species were observed,
In these tests, it is not too surprising that different probes produce different
results since the length of the turbulent flame is on the same order as the
probe diameter. For example, the flame length is approximately 5 to 6 mm and
the blunt-nosed probe has a diameter of 6 mm very near to its orifice. This
large probe certainly must be considered a poor design for probing the reac-
tion zone of such a flame. The slender nosed probe is a better design with an
initial diameter greater than 1.2 mm but even this probe quickly tapers back to
U mm diameter. The differences in the profiles (both NO and major species) may
be due to the presence of a large heat sink, i.e., the probe, or possible
111-10
-------
stagnation zones in front of these probes. The difference between the NO
profiles obtained using the large probe at various flow rates may be due to the
phenomena stated above or to NO/NOj interconversion within the probe. NOj
was measured only for the smaller probe and in that case was found to be
negligible. Bilger and Beck also conclude that the slender nosed probe provides
a more realistic measurement of nitric oxide.
Bryson and Few (1978) at Arnold Research Organization (ARO) have observed
relatively large differences (^ 50%) in total nitrogen oxides between a tubular
probe and either a 'quick-quench' or dilution probe (the latter two agree).
Since this discrepancy is rather large, it is worthwhile to investigate the ARO
study in detail to see if this report identifies areas that require further
research.
Quite clearly, there are several significant conditions of the Bryson and
Few study that are different from other reports. First of all, in their study
supersonic exhaust from an AVCO-Lycoming engine is sampled. In all other
papers that report measurements using different probes, subsonic flows are
examined. Secondly, three distinct, water-cooled probes were compared: a
tubular inlet probe, a 'quick-quench1 probe, and a dilution probe. In addition,
the stainless-steel tubular probe was constructed to accept inserts of copper
or fused silica. Alternatively, other studies primarily examined effects due
only to changes in surface material. (Some papers, e.g., England, et al. (1973),
do test some modifications to probe design but in general these changes are
minor relative to the design variations found in the ARO study). In regard to
the design variations, it should be noted that the tubular probe which produced
the relatively high NOy measurements also had a very large opening (0.77 cm
i.d. vs 0.12 and 0.127 for the other two probes) and was operated with a sample
line pressure much larger than the latter two probes. The ratios of sample
line to combustor pressure were typically 0.9 for the tubular probe and less
than 0.5 for the other probes. Undoubtedly the pumps used for the 'quick
quench1 and dilution probe were insufficient to choke the tubular probe. Under
choked conditions (not well defined for a constant area tube where the flow
should friction choke at the exit rather than at the entrance), the probe would draw
approximately 40 times the flow of the smaller probes. Based on the above
analysis, it seems likely that the different NO^ measurements between a
tubular probe and either a "quick-quench1 or dilution probe may be associated
with differences in sampling pressure and flow rate through the probe and/or
the existence of a stagnation zone in front of the probes.
A stagnation zone, for example, in front of these fairly blunt probes in
a supersonic stream could perturb the gas samples. For air at a Mach number
of 1.15, for example, a stagnation of the flow results in a 25% rise in temper-
ature. For the tubular inlet probe, the problem may be the most severe. First
of all, the gas decelerates to a very low Mach number at the entrance (estima-
ted to be less than 0.1) and consequently stagnation temperatures must be
111-11
-------
approached even in the absence of an external stagnation zone. Secondly, the
rate of cooling is undoubtedly much slower in this probe than in the other two
since the sample tube is quite large (producing a low surface to volume ratio
for heat transfer) and since the pressure and mass flow through this tube are
significantly larger. The problem should be less important when sampling ex-
haust gas at lower temperatures because the magnitude of the temperature rise
is less and since kinetics are slower at lower temperatures. In Bryson and
Few's study, essentially no difference between the probes was found for the NO
measurements at the lowest stoichiometries which produce exhaust temperatures
approximately 600 K less than the other tests.
Sample line leaks from either the atmosphere or from a purge system may
also affect relative readings when a probe and/or pumping system create large
differences in sample line pressures and flow rates. Consider a case with a
small leak across a purge valve. If the sample line pressure is only slightly
less than the purge pressure and the mass flow through the sample line is large,
the leak would be unchoked and, if small, would dilute the sample minimally.
However, when the sample line pressure reduces to below 50% of the purge pres-
sure and simultaneously the sample flow rate decreases (such as by reducing
the probe orifice diameter) then the importance of the leak may increase sub-
stantially. Based on the data presented by Bryson and Few, the above phenomena
could explain some of the differences between the tubular inlet and the other
two probes. This effect, however, is not believed to be the cause of the dis-
crepancies since at low power levels NO measurements did not vary with probe
type. It is more likely that the increase in temperature (and pressure) due to
shock recovery and subsonic diffusion created the sampling problems encountered
by Bryson and Few.
Another aspect of the study by Bryson and Few is that, for the dilution
probes, wide scatter was observed. The authors comment that part of the scatter
may be due to estimation of the dilution ratio which was calculated via two
techniques. A factor not considered in their study is that a change in the
diluent or carrier will change the response of a chemiluminescent detector.
The response is not only a function of the quenching efficiency of a third body
as pointed out by Matthews, et al. (1977) but also dependent on the viscosity
of the sampled gas for standard commercial units (Folsom and Courtney, 1979 and
Dodge, et al. , 1979a). Although this phenomena is discussed in greater detail
in Section III.A.4, it can be estimated based on data from Folsom and Courtney
that the NO., measurements made when diluting with argon should be increased
by approximately 10 percent depending on the diluent ratio. No data were found
for the case of a helium carrier, but since the viscosity of helium and nitrogen
are similar and if helium is less efficient than nitrogen for the reaction
N02* + M * N02 + M (111-10)
which competes with
N02* + N02 + hv (III-ll)
111-12
-------
then one would expect that the data obtained with a diluent of helium should be
reduced by several percent. We estimate that these corrections should reduce
the uncertainty for data obtained using the dilution probe.
Additional research that may indicate that probes perturb measurements of
total nitrogen oxides has been performed by England, et al. (1973); however, in
this work, only nitric oxide data are presented. Consequently, definitive
statements on the loss of total NOX cannot be made. For uncooled, stainless-
steel probes in rich flames, the NO measurements are significantly less than
for cooled probes. As described previously this behavior is not at all un-
expected for 0 > 1.0 (rich flames). In spite of the fact that the paper
indicates a fall-off in NO for these uncooled probes even for 0.6 < 0 < 1.0,
this may be an artifact of the curve-fitting technique (no actual data points
are published). In any case, no detailed mechanism has been presented to
explain these data. For two cooled probes and even an uncooled quartz tube, NO
profiles are nearly identical for the lean flames and are similar for fuel-rich
flames. Slight discrepancies on the fuel-rich side may be due to NO/N02
interconversion rather than loss of total NOy.
In summary, there is only one known report that suggests that changing probes
may change the total concentration of nitrogen oxides by a significant amount
(other than uncooled probes in rich flames). This study by Bryson and Few
shows that a tubular inlet probe produces higher values (by 50%) than two other
probes. Although insufficient data are available, it is possible that pheno-
mena associated with the higher operating pressure, an external stagnation
zone, or different flow rates may contribute to the observed results. Moreover,
it should also be noted that apparent differences in NOy may be observed
between different probes if NO/N02 interconversion varies between the probes
and N©2 exiting from one of the probes is lost in water traps or an ineffic-
ient converter.
III.A.3 Losses in Sampling Line
As Tuttle, et al. (1973) have discussed, NOo can be lost in sampling
lines for any of various reasons. These include reduction of N0£ on parti-
culate filters, loss in water traps, or loss or reduction of N02 in uncon-
ditioned stainless-steel sampling lines. Dimitriades (1967) claimed to have
observed NOo losses in both water traps and Drierite columns, but these
losses are not quantified and it consequently is difficult to estimate the
impact on a measurement where water removal prior to analysis is necessary.
Dimitriades does, however^ comment that this loss is specific to N©2 and that
he did not observe similar losses of NO. The same comment cannot be made for
the other items listed above. For example, it has been observed in our labora-
tory that unconditioned stainless-steel lines will also produce an apparent
loss of nitric oxide. This loss mechanism can usually be eliminated by flowing
an NO calibration gas through new lines for 15 minutes to 1/2 hour. Once
conditioned, no evidence was found that indicated the necessity to repeat this
process. In regards to the loss of N02 on filter paper containing carbon or
soot particles, Tuttle, et al. observed losses of N02 using an NDUV detector
111-13
-------
(for N02) but did not look for the possible conversion to NO or for similar
losses of NO due to the presence of soot. Gas-phase, NO-hydrocarbon reactions
are typically as well known and as fast as N02-hydrocarbon reactions; by
analogy, one might expect similar reactions between NO and deposits of soot.
Each of the above problems represents a real concern and should be con-
sidered in the construction of any sampling or analysis system. Presumably,
the line conditioning problem can be easily rectified by using teflon coated
tubing and fittings throughout or using clean stainless-steel tubing which has
been conditioned for several (> 15) minutes using an NO or N02 calibration
gas. Nitric acid should not be used as a cleaning agent since it leaves
nitrates and/or nitrites as residues.
Water condensation is a more difficult problem to solve. Bryson and Few
(1978) ran without a water trap and then between measurements flushed out the
lines with dry nitrogen in the analyzer to remove any condensed water. (This
procedure complicates data reduction since, as discussed Section III.A.4,
water can have a strong effect on the response of the CLA and was not accounted
for by Bryson and Few.) Although this technique eliminates the retention of
large quantitites of water within the analyzer some condensed water along the
inner walls must remain and presumably could contribute to NOo absorption.
The best methods would include the modification of the analysis equipment so
that (1) the entire system (both sampling and analyzer) are heated substantially
above the dew point (based on the maximum pressure in the sampling system and
the initial water concentration) or (2) sample line and analysis instruments
all operate at reduced pressure (low enough to prevent condensation). For
example, for a liquid-fueled gas turbine engine operating at an overall stoi-
chiometry less than 0.33, it can be estimated that all lines should be heated
above 35°C if the maximum sample pressure is only 800 torr (15.5 psia).
Although these modifications represent changes over the present specifications
in the Federal Register (1973, 1976), it is not clear whether such steps are
necessary since the loss of N02 in a trap is not understood quantitatively
and may be dependent on trap geometry and capacity; among other things.
The loss of NO,, due to the presence of soot on particulate filters (or
even coated on sample lines) is a more difficult problem to address. Presumably,
no easy solution can be found for a combustion system that produces large
quantities of soot. Fortunately, Federal requirements for smoke emissions help
to reduce potential complications due to soot in sampling lines. As in the
above case for water absorption, losses cannot be estimated quantitatively.
III.A.4 Response of Chemiluminescence Analyzer
As mentioned previously in Section III.A.2, the response of a commercial
chemiluminescent analyzer (CLA) is dependent not only on the concentration of
nitric oxide but also on the fluid mechanical properties and quenching effi-
ciency of the carrier gas. Response changes due to changes in quenching
efficiency have been reviewed by Matthews, et al. (1977). From the well-known
chemiluminescent reaction sequence
111-14
-------
NO + 03 -» N02* + 02 (111-12)
-»N02 + 02 (111-13)
*
N02 + M -> N02 + M (111-10)
*
N02 -*N02 + hv (III-ll)
it is clear that Reactions 111-10 and 11 are in competition. The presence of
species such as water and carbon dioxide which are more efficient than nitrogen
at quenching the excited N02 molecule (N02 ) will produce less of a
response than when nitrogen is the carrier for equivalent concentrations of
nitric oxide. The presence of species that are less efficient at quenching
(e.g., argon) correspondingly will produce a greater response.
These comments are true only in the case of a CLA for which the pressure
in the reaction chamber is maintained constant (such as the original design by
Fontijn). Unfortunately, the technique used to maintain constant pressure in
the reaction chamber of low pressure commercial units is via flow restriction
using capillary tubes. Since the flow rate through these tubes is dependent
upon on the fluid mechanical properties of the carrier, the pressure and, of
course, the concentrations of NO and the quenching species (M) in the reaction
chamber are dependent on these same properties. This phenomena was observed
several years ago by one of these authors (M. F. Zabielski) when comparing mass
spect rometr ic and chemiluminescent data when argon was used as the carrier. In
this case, the chemiluminescent detector was found to have a reduced response
relative to measurements with nitrogen carrier; this direction is opposite to
that predicted according to an analysis of quenching phenomena. Subsequently,
Dodge, et al. (1979a) modelled this problem by assuming frictional choking in
the capillary tubes. Stimulated by the original UTRC work, Folsom and Courtney
(1979) performed a detailed empirical study on the effect of carrier gas on the
responses of commercial Beckman and Thermo Electron instruments. Although
Folsom and Courtney qualitatively explained their data in terms of relative
viscosity and quenching data, the model developed at UTRC (Dodge, et al . 1979a)
appears to be in good quantitative agreement with the data obtained both at UTRC
and by Folsom and Courtney on the Thermo Electron instrument. (The UTRC model
would have to be modified for the atmospheric pressure instruments (e.g. Beckman
or McMillan) due to a different design.
In general, corrections due to differences in a carrier of pure nitrogen
and that of combustion gases are not large (typically < 10%). Exceptions to
this rule include flames using an unusual carrier (argon, for example) or no
carrier at all; samping systems where the water is not removed (especially for
stoichiome tries near $ = 1.0); or samping systems were the sampled gases are
diluted with a species other than nitrogen. In these cases, the combination of
viscous and quenching effects ought to be considered, since the response change
of a CLA can amount to as much as 15% or -possibly more.
111-15
-------
III.A.5 NOX Converter
The purpose of an NOX converter is to reduce nitrogen dioxide to nitric
oxide via the overall reaction
2N02 2ND + 02 (111-14)
in the presence of a heated metallic surface (usually stainless-steel). The
NO thus formed, along with the initial NO,is detected in a chemiluminescent
analyzer. The N02 concentration is obtained by subtracting from this response
the instrument response when the converter is bypassed.
Problems associated with the operation of the converter have been discussed
in detail by Tuttle, et al. (1973). The problems can be placed into the three
classifications: anomalies, inefficiencies, and NOy reduction. Tuttle, et
al., for example, report an anomaly observed by A. Nelson of Pratt and Whitney
Aircraft. In this case, a span gas of 91 ppm NO in nitrogen but void of N02
(verified by the Saltzmann technique) indicated a concentration of 95 ppm when
passed through the detector. As opposed to a decrease or apparent loss of NO,
this increase appears very unusual since it suggests a generation of nitric
oxide. Although Tuttle reports this as an anomoly, it is likely that the
calibration of the instrument shifted (if the operator did not adjust for
changing flow rates due to the pressure drop across the converter) or perhaps
nitric acid was used as a cleaner which may outgas nitrogeneous species. It is
more typical that a slight but noticeable reduction in indicated NO (several
percent) will be observed when a span gas of NO is directed through the converter,
Presumably this loss is due to a reduction of nitric oxide (to nitrogen) in the
converter.
Practically, the best efficiency for conversion of N02 to NO is about 97
to 98 percent, and efficiencies of 90 to 97 percent are typically achieved for
fuel-lean gases. These efficiencies, however, are dependent on NO,, concen-
trations and type, condition, and temperature of the converter. Most converters,
for example, will perform, at least momentarily, when fuel-rich gases with
NOX are passed through, however, stainless-steel converters at high tempera-
tures will last only a matter of seconds before total NOX is destroyed.
Molybdenum converters are more useful since they efficiently convert N0« to
NO without loss of total NO,, for up to a minute or so (depending on flame
stoichiometry and converter temperature). The increased activity of molybdenum
for fuel-rich gases is due to its lower operating temperature ( - 450°C for
molybdenum vs. - 700°C for stainless-steel). The activities of the converters
are easily recovered by flowing an oxygen rich mixture through the heated tubes
for a few minutes. In fact, alternately flowing air through the converter when
sampling fuel-rich flames is a common procedure. This technique should be used
with caution, however, since resultant surface oxidation deteriorates the
converter and may reduce its conversion efficiency.
111-16
-------
III.A.6 Summa ry
In the above discussion, it is shown that many phenomena may contribute to
erroneous measurements of nitrogen oxides. These include interconversion
between NO and NCU, loss of total NOy, and misinterpretation of data
(assuming use of CLA). Interconversion and loss can occur in the probe, sample
line (water trap and filter included) or converter. Perhaps the biggest errors
may be caused by the use of uncooled probes sampling fuel-rich gas. Under the
same flame conditions, NC^ measurements (using a converter/CLA) can only be
made through careful use of the converter. In any case, efficiency checks on
the converter operation should always be made. In the case of lean or stoichio-
metric flames, only one report (Bryson and Few) was found that indicated
substantial discrepancies (~50%) in NOX when measured using different water-cooled
probes. These discrepancies may be due (at least in part) to differences in
sample line pressure and/or flow rate. The results of another report which
indicates discrepancies (Bilger and Beck) are unreliable since at least one of
the probes was poorly matched to the combustion system. The only other signifi-
cant probe effect is the conversion from NO to N02 due to oxidation by the
H02 radical. This reaction, however, has been found to be of significance
only when relatively low concentrations of NO are present (-50 ppm). It would
be expected that at larger concentrations when the NO/HC^ ratio is large both
in the flame and probe, oxidation of NO via the HC^ radical is relatively
unimportant.
III.B Quenching in Gas Sampling Probes
Gas samples extracted from a flame environment are quenched by a rapid
reduction in either pressure or temperature, and typically both. For a bi-
molecular reaction, a reduction in pressure by a factor of ten will reduce reaction
rates by a factor of 100. Decreasing the gas temperature from 2000 to 1200 K for
a reaction with an activation energy of 40 kcal/mole (typical for overall
hydrocarbon oxidation) will decrease the reaction rate by a factor of 750. The
quenching of flame gases is required not only to stop ongoing reactions (non-
equilibrium conditions) but also to prevent a shift from equilibrium (or
quasi-equilibrium) conditions such as may be present in a post flame zone.
Ideally, a probe should quench or "freeze" the flame gases exactly at the
concentrations present where the probe tip is located.
Techniques used to quench flame gases include quenching by dilution,
convection, or expansion and each of these have been discussed in some detail
by Tine (1961). Briefly, quenching by dilution is accomplished by adding a
low temperature diluent to the flame gases which acts to absorb heat from the
extracted sample. Quenching by convection is performed by heat transfer to
cooled walls within the probe and quenching by expansion is accomplished by
111-17
-------
accelerating the gases supersonically through a nozzle to drop the static
temperature and pressure. Although expansion cooling is undoubtedly the
fastest (-108 K/sec vs. - 106 K/sec for cooling by convection), the
associated phenomena are not well understood and two major drawbacks accompany
this technique. First of all, convection cooling to cooled walls must also
take place while the gas is flowing supersonically; otherwise, upon return to
subsonic flow conditions, the static gas temperature will return near to flame
temperatures. Secondly, substantial pressure losses, both friction and normal
shock losses, are suffered. Details of these phenomena will be discussed later
in this chapter.
In spite of the many years that probes have been in use, there exist many
uncertainties in regard to their quenching behavior. Beal and Grey (1953) have
argued that any of the above three techniques provide sufficiently high quench-
ing rates to freeze the concentrations of stable species at or near concentra-
tions present in the flame zone. Alternatively, Halpern and Ruegg (1958) have
found that changes in quenching rates due to changes in probe design, internal
diameter; and sample flow rate may vary measured ratios of CO/CC>2 and ^7
HoO. These latter conclusions may be subject to error since the authors
apparently used probes that were comparable to the size of the burner. The
presence of a large, cooled probe will significantly perturb the flame environ-
ment. Other authors have reported discrepancies in CO and NO measurements for
various flame conditions using different probes (Bryson and Few, 1978; Bilger
and Beck, 1975; and England, et al., 1973) but it is not clear whether these
discrepancies are due to differences in quenching rates rather than to flame
perturbations, stagnation of the flow, or sampling line/analysis phenomena. It
should be noted that species existing in low concentrations (several hundred
ppm or less) are especially susceptible to any of the above effects. Examples
include NO, N02 and in lean flames, CO. Small absolute changes (-25-50 ppm)
can reflect a large relative change for these molecules. In fact, increased
rates of quenching may not necessarily solve problems associated with measuring
these species since radical termination on the walls or in the gas phase may
perturb concentrations of these species. Indeed, a model developed by Kramlich
and Malte (1978) makes the unusual prediction that higher quench rates increase
the conversion (in the probe) from NO to N02.
Not only are there uncertainties in the effect of quenchir? rates, but also
actual knowledge of quenching rates is unknown. Typically, approximate fluid
mechanic models are used to estimate cooling rates and pressure reductions;
however, virtually no experimental evidence exists to verify these calculations.
For example, the operation of quartz microprobes has been misunderstood for
years. Fristrom and Westenberg (1965) argued by analogy that since a large
diverging nozzle sustains a supersonic flow, a smf '1 probe nozzle should
perform similarly. Although this extrapolation is Somewhat suspect, many
experimenters have operated on this premise with no experimental verification
and very little, if any, theoretical analysis (e.g., Friedman and Cyphers,
1955; Lyon, et al., 1975; Kramlich and Malte, 1978; Lengelle and Verdier,
1973). More recently, questions regarding these conclusions have been raised
111-18
-------
in the literature (Bilge^ 1975, Amin, 1977; Seery, et al., 1977; Cernansky and
Singh, 1979) with fewer and fewer experimenters tacitly assuming the existence
of quenching by expansion. Even so, arguments used to question supersonic
expansion are primarily phenomenological with only a few studies attempting to
examine analytical details of the fluid mechanics (Seery, et al., 1977; Cohen
and Guile, 1970; Amin, 1977) and with virtually no experimental investigation
of the problem.
The lack of understanding of this problem can be explained by the fact
that the fluid mechanic and thermal status of the sample gas in a probe is
altered by a number of factors. These include: heat transfer from the uncooled
probe tip to the sample gas and from the sample gas to the probe coolant; skin
friction; flow area changes; pressure losses associated with shock systems,
sudden area expansions, and turns within the sample passage; and, chemical
reaction. For probes designed to achieve an aerodynamic quench (see III.B.2.a),
sudden-expansion losses, turn losses, and chemical reaction are neglected and
shock losses are avoided until the aerodynamic quenching region of the flow is
completed. To understand the influence of the competing mechanisms, it is
convenient to assume that the sample flow is both steady and one-dimensional
and that manufacturing techniques are sophisticated enough so that the internal
geometry of the probe (especially in the region of the probe tip) is reasonably
close to the geometry analyzed. (The wall temperature distribution within the
probe tip is difficult to calculate and, for small diameter probes, the internal
shape of the tip is difficult to control during manufacture and also difficult
to examine once constructed.)
Even though a model may be developed to describe heat transfer, skin
friction, shock losses, and the effect of area changes, each of these phenomena
produces different relative effects as the probe operating conditions are
varied. Consequently, it is necessary to examine each experimental condition
individually and a probe designed from such a study is usually a compromise.
The above discussion indicates the complexities of the operation of a
probe in terms of kinetics, fluid mechanics, and heat transfer. Due to the
discrepancies between optical and probe measurements observed by McGregor, Few
and coworkers, and the above problems, the present program focused on understanding
probe and sample line behavior and its impact on sample analysis. Although
substantial effort was placed on examining possible homogeneous mechanisms for
NO reduction, no mechanism was found that predicts loss of nitrogen oxides
within water-cooled probes. Interconversion between NO and N02 may affect NO
measurements and was observed in this study under certain conditions. This
phenomena, however, was of little importance since, in the comparisons of
optical and probe measurements, NO., was approximately equal to NO. Moreover,
recent studies on this problem (Johnson, et al., 1979; Kramlich & Malte, 1978)
suggest that total NO,, is conserved. Knowledge of the N02/NO ratios may be
important in understanding the total emissions problem (N02 can be visible in
sufficient concentrations and NOo acts as an initiator in the photochemical
smog cycle).
111-19
-------
Fluid mechanics and heat transfer mechanisms were also examined in this
study. Particular effort was focused on the phenomena of aerodynamic cooling
since this technique is not well understood and since potentially the most
significant benefits (i.e., the fastest quenching rates) can be obtained.
III. B.I Kinetics of NO Decomposition
The loss of nitric oxide within a gas sampling probe, if indeed it occurs,
should be explainable in terms of either hetero- or homogeneous phenomena. As
reviewed in the TASK I report (Dodge, et. al., 1979), very little information
is available on reaction mechanisms or rates for NO decomposition on walls.
Nevertheless, for walls that are directly water cooled, virtually no NO can be
lost on either quartz or stainless steel walls for a typical residence time in
a probe of less than one second. (This statement assumes that the walls are
properly conditioned, as in the case of sampling lines.) Near the orifice of
a probe tip, wall temperatures may be significantly higher than 300 or 400 °C
and in cases even approach the material softening point. Under these cir-
cumstances, the surfaces are certainly catalyt ically active and the potential
for altering flame concentrations exists. Although a detailed analysis of
molecular diffusion was not performed for this work, it was estimated that if
the residence time in this portion of the probe is held to less than 10 micro-
seconds, then the molecules undergo only a few collisions with the wall. Since
the efficiency of NO decomposition should be significantly less than one, then
decompositions of NO by collisions with a hot wall should be negligible.
Homogeneous mechanisms for the decomposition of NO have been examined by
many workers (e.g., Hanson, et. al., 1974; Flower, et. al . , 1975; and Koshi and
Asaba, 1979). A listing of possible reactions which lead to the gas phase
reduction of NOX are listed in Table III-C. Reactions that oxidize NO to N02
are not included here since no net loss of NO results. (Assuming, of course,
that care was taken to avoid loss of N02 in the sampling line and/or sampling
system.) If the temperature is assumed to be constant, an estimate of the con-
tribution of each of the reactions in Table III-C can then be made. For small
fractional conversions of NO, the fraction of NO consumed by Reaction x is
estimated by
fraction = = k [x] At 111-23
N0
0
(or = k [x](M] At where applicable)
111-20
-------
TABLE III-C
REACTION MECHANISM FOR NO DECOMPOSITION
Rate Constant
1
Reaction
Number
111-15 NO + NO + N20 + 0 4.9 x 10
12
E/R
33,770
111-16 NO + 0 + N + 09 2.32 x 109 1 19,445
111-17 NO + N
111-18 NO + M
111-19 NO + H
111-20 NO +
111-21 NO + H20
N-, + 0
N + OH
H + HNO
111-22 NO + H + M * HNO + M
1.63 x 10
13
N + 0 + M 1.41 x
i21
1.35 x 10
5.75 x 10
OH + HNO 2.0 x 10
1.8 x 10
14
12
14
16
-1.5 77,250
24,760
28,890
36,510
-300
Ref.
Koshi and
Asabi
(1979)
Hanson, et.
al. (1974)
Baulch, et.
al. (1973)
Baulch, et.
al. (1973)
Flower; et.
al. (1975)
Baulch et.
al. (1973)
Baulch, et.
al. (1973)
Jensen and
Jones (1978)
1 "^
k = ATnexp(-E/RT), units in cm /mole, sec
111-21
-------
where [x] represents the concentration of the collision partner and kx, the
rate constant of Reaction x. The time increment, At represents an estimated
quenching time, after which the reaction is effectively frozen due to a drop in
the rate constant, concentrations, or both. Rate constants were obtained from
literature values and are listed in Table III-C. Concentrations of stable and
radical species were equated to the maximum equilibrium value for the three
flat flames examined in this study. The concentration of NO was assumed to be
5000 ppm, a value larger than the maximum seed concentration used in this (TASK
II) or the TASK III study. These concentrations, in terms of mole fraction,
are listed in Table III-D and the fraction of NO lost due to each of these
reactions, as calculated by Equation 111-23, is also given. The calculations
are based on a (local) probe pressure of 1/2 atmosphere and have been done for
average static temperatures of 2000 and 1400 K. In addition, the quench time
was assumed to be one millisecond. Even by assuming an uncertainty of one
order of magnitude in rate constants or concentrations, it is clear from these
results that only the (overall) three body reaction could possibly contribute
to loss of nitric oxide. The conclusion that a three body reaction dominates
under conditions of elevated temperature and reduced pressure is extremely
unusual. It is more likely that the negative temperature dependence used for
this reaction is too weak and, in fact, the rate constant decreases much more
rapidly at high temperatures than indicated by the rate constants given in
Table III-C. Other (overall) three body reactions typically have significantly
higher negative activation energies. Since this rate constant has only been
measured up to temperatures of 700 K, it should be expected that errors will
result by extrapolation of the rate constant to flame temperatures. In addition,
Reaction 111-22 should be less important than indicated by Table III-D since
the hydrogen atom concentration will decrease rapidly as the gas is cooled.
Even if this reaction does cause conversion of NO to HNO, the product will
primarily reform nitric oxide after abstraction of the hydrogen atom, i.e.,
R + HNO - RH + NO (111-24)
which occurs at nearly collision frequency. In this reaction, R represents any
radical species. Some HNO could possibly be lost via
HNO + HNO + N20 + H20 (111-25)
HNO + NO + N20 + OH (111-26)
or possibly by HNO + H * NH + OH (111-27)
or HNO +0 + NH » 02 (111-28)
followed by N20 + R + RO + N2 (111-29)
and NH + NO * N2 + OH (111-30)
111-22
-------
TABLE III-D
ESTIMATED FRACTIONS OF NO DECOMPOSITION
O
Fraction of NO Decomposed
Reaction
Number
111-15
111-16
111-17
111-18
111-19
111-20
111-21
111-22
'See text
React ion
Partner
NO
0
N
M
H
H2
H20
H
xM
Estimated Mole Fraction Due to Each Reaction
of Reaction Partner 2000 K
5 x 10~3 3.5 x 10~6
1.5 x 10~5 1.3 x 10~5
7.7 x 10~8 3.8 x 10~6
1.0 8.1 x 10~10
5.7 x 10~5 9.9 x 10~5
2.6 x 10~2 2.4 x 10~4
.15 1.1 x 10~3
5.7 x 10~5 1.0 x 10~2
1.0
1400 K
3.6 x 10~9
2.0 x 10~7
5.5 x 10~6
1.3 x 10"16
7.0 x 10~7
7.0 x 10~7
6.3 x 10~7
2.2 x 10~2
2From Equation 111-23, assuming that pressure = 1/2 atm, At - 1 millisecond
111-23
-------
but none of these reaction mechanisms is considered very likely. Reaction
111-25 is sterically improbable and undoubtedly represents a multistep process.
In any case, its reaction rate is dependent on the square of the HNO concen-
tration which would be quite low. Reaction 111-26 is slow since it has a
significant activation energy of 26 kcal/mole (Wilde, 1969). Reaction 111-27
is similarly slow due to its high endothermicity (^24 kcal/mole). If the
oxygen atom concentration is high enough to allow Reaction 111-28 to proceed at
a high rate, then it is highly likely that the radical NH will be oxidized to
form NO rather than reduced to form No.
Other reaction mechanisms (such as to produce HN02) were also examined,
but conclusions similar to those above were obtained. No mechanism could be
found which indicated loss of NO in a water-cooled sampling probe. Based on
the analysis, the following quenching criteria were selected.
1. In the first portion of the probe tip where high wall temperatures (>
600 K) may exist, the residence time of the gas must be less than 10
microseconds.
2. The gas sample must be "quenched" within 1 millisecond, and
3. The gas is considered to be "quenched" when the total temperature
falls below 1000 K (with an initial temperature approximately 1800 K).
This last criteria is based on the fact that after 1 millisecond radicals will
have undergone hundreds of collisions with the walls. Recombination of these
radicals, therefore, must occur within 1 millisecond and equilibrium radical
concentrations at 1000 K are too low to allow any further reactions with NO
before the gas is further cooled to water temperature (typically less than 100
mi 1liseconds).
III.B.2 Description of Computer Program for Probe Analysis
The UTRC Probe Design computer program is based upon an equation which
describes the change in local Mach number as a function of heat transfer, skin
friction, area variations and thermal property changes for a steady, one-
dimensional flow. Using the influence coefficient approach of Shapiro (1953),
this equation is
dM2 9 7 - 1 o dA 9 dQ , Y ~ 1 9 dx
(1-M*) = -2(1 + M2) + (1 + YM2) + YCTU + M2) 4f
M2 2 A cpTs 2 D
(111-31)
-(1 + YM2) *M- (1 - M2) H
M Y
111-24
-------
where M is the Mach number; YJ the ratio of specific heats; A, the cross-
sectional area; Q, the heat transfer rate; c , the heat capacity at constant
pressure; Tg, the static temperature; f, the skin friction coefficient; x,
the axial distance from the probe orifice; D, the tube diameter; and fl/f> tne
molecular weight. The heat transfer rate is calculated from
dQ = h(Tr - Tw) TtDdx (111-32)
where h is the heat transfer coefficient, T is the recovery temperature, ,
and TW is the wall temperature. To determine the wall temperature distribution
within the probe tip, a computer program called TCAL was used (see end of this
section). For both the portion of the probe tip and the constant area section
that are cooled directly by the water in the coolant passage, the wall temperat-
ure is determined by the overall heat transfer rate from the sample gas to the
coolant. To provide a conservative (i.e., low) estimate of the heat transfer
from the sample to the water, a higher than anticipated coolant temperature
(150 °C) was used in the calculations. The heat transfer coefficients reported
by Kays (1955) for laminar flow and Rohsenow and Choi (1961) for turbulent flow
were used. For the skin friction term (4f), correlations reported by Eckert
and Drake (1959) were used; these correlations are applicable to fully developed
laminar and turbulent flow in smooth tubes.
To calculate the profiles of temperature, pressure, etc., the terms in
Equation 111-31 are evaluated at a given axial location and the new Mach number
at the next axial location is calculated by a simple, forward marching proce-
dure. The gas properties at the new location may then be calculated using
standard relationships, such as the ideal gas law. When the flow is supersonic
initially, conditions downstream of a normal shock system are calculated using
the Rankine-Hugoniot relationships. The position of the shock is determined
using one of two procedures. The position may be specified by the user to
occur at a physically realistic location in the probe (e.g., at a bend or sud-
den area increase). Alternatively, the shock may be positioned by the program
so that the stagnation pressure behind the shock is just equal to the desired
back pressure.
The program performs various checks on the calculations to verify the
stability of the computational procedure. For example, the program compares
*In continuum flow, the flow velocity is necessarily zero at the wall due to
frictional forces. This deceleration of the flow results in an increase in
the flow temperature near the wall. The flow temperature at the wall is termed
the recovery temperature and is always somewhat less than the temperature that
would result from decelerating the flow isentropically to zero velocity (i.e.,
the stagnation temperature) (Schlichting, 1960).
111-25
-------
the computed results with results obtained using differential equations for
selected flow parameters (Shapiro, 1953). In addition, the step size is
selected small enough so that results are essentially independent of the step
size. Typically, the initial step size is between one and two orders of
magnitude smaller than the orifice diameter.
In order to estimate the internal wall temperature distribution of the
probe tip, it is necessary to analyze the heat transfer between the tip and
the external environment, sample gas, and cooling water. Heat transfer calcu-
lations were performed using TCAL, a computer program formulated at Pratt &
Whitney Aircraft to evaluate the temperature distribution and heat fluxes in
combustors, turbines, and other structural members in gas turbine engines. In
the program, a finite difference representation of the heat conduction equa-
tion (a time-dependent version of Laplace's equation) is solved by a relaxa-
tion technique. The inputs to the program include a description of both the
geometry and material properties for the hardware being analyzed, the proper-
ties of fluids around the hardware, and the heat generated on the surface or
within the device. Optional boundary conditions include time dependent surface
and internal heat generation. Material properties are permitted to vary with
temperature and boundary temperatures are allowed to vary with time.
III.B.2a. Sudden Expansion Losses
The computer program (probe design deck) used in this study for analysis
and design of gas sampling probes assumes that the gas flow is one-dimensional.
It, furthermore, requires that the flow within the probe be attached to the
probe walls; flow separation is not modeled. In the case of probes designed to
operate with a choked orifice followed by a supersonic expansion, this require-
ment is usually satisfied. If this type of probe is operated unchoked, then it
is probable that the subsonic flow within the expanding probe tip will separate
from the probe walls due to the high divergence angle of these walls. Detailed
analysis of the boundary layer accompanying subsonic flow in the tip of one the
large probes used in this study revealed that the subsonic flow is indeed
separated. Furthermore, it was demonstrated experimentally that for probe
operation at high back pressure (>l/2 external pressure) the probe orifice was
unchoked. Consequently, the following procedure was included to account for
the pressure loss associated with separated internal flow.
During the process of diffusing a subsonic flow, frictional forces and the
adverse pressure gradient may be sufficient to dissipate the forward momentum
of the flow, i.e., the flow will separate. Practical subsonic diffusers limit
the wall angle of the device to a few degrees. Wall angles greater than this
limit (such as encountered with flows over rearward facing steps or with flows
in probes with short, high area ratio tips) tend to cause flow separation. The
loss in stagnation pressure due to flow separation is termed the sudden expan-
sion loss .
111-26
-------
The sudden expansion loss may be calculated by assuming that the static
pressure in the separated region is equal to the (unknown) static pressure at
station 2 (see Fig. III-l). Thus, for one-dimensional flow, the continuity and
momentum equations may be applied to the control volume shown in Fig. III-l
(dashed line). After some manipulation, the system of equations becomes:
PT \PT/1 \PT/1 \A1 / (111-33)
PT
Ll
PT2 (A/A*)9
_ f _ _ * (111-34)
PT1 A2 (A/A*)L
<
where f = P(l + "VM ) and represents the stream thrust per unit area. A is
the flow area, M, the Mach number, P, the static pressure, P, the stagnation
pressure, and *y, the ratio of specific heats. A is the area at the throat
and the subscripts 1 and 2 refer to the upstream and downstream conditions,
respectively. Writing Equations 111-33 and 34 in terms of the Mach numbers,
Mi and ^2, and the area ratio, ^l f^i , and solving the equations simultaneously
yields an equation which may be solved explicitly for M2 . Then, the pressure
ratio, Pj2/Pxi' may be calculated from either Equation 111-33 or 34. The sudden
expansion loss is 1 - Pf2/PTl ^t *s> °^ course, assumed that the initial con-
ditions (M, , Pj) and the points of flow separation and reattachment (which
defines A2/Ai) are known to sufficient accuracy.
III.B.2b. Aerodynamic Quench
From extensive use of the computer program, some qualitative features on
the operation and design of 'aeroquench' probes have been obtained. These
facts are reviewed here along with a more detailed description of the phenomena
of aerodynamic cooling.
111-27
-------
MODEL FOR CALCULATING SUDDEN EXPANSION LOSS
i
r-o
00
01
Ol
I
-------
An aerodynamic quench in a sampling probe is accomplished by reducing the
static temperature to some level determined from the kinetic rates and maintain-
ing it at this level until the stagnation temperature of the gas is also reduced
to near this level. The reduction of the static temperature is achieved by
accelerating the flow to a high supersonic Mach number. During this process,
the random kinetic energy in the gas is converted to directed translational
energy and, hence, a net cooling of gas occurs. The reduction of the stagnation
temperature is achieved via heat transfer to cooled walls of the probe. (The
stagnation temperature is the temperature the gas will reach if decelerated
adiabatically to zero velocity.) This heat transfer to the walls is of prime
importance since without it the gas would eventually shock heat during its
recovery to subsonic flow conditions and return to temperatures approaching the
original flame conditions. Since an aerodynamic quench is obtained by accelerat-
ing the flow, a significant stagnation pressure loss results due to friction.
Therefore, whether an aerodynamic quench can be achieved is highly dependent
upon the pumping system used with the probe. The pumping system sets the back
pressure which is the maximum pressure behind the shock system that terminates
the supersonic portion of the flow.
In a probe designed to quench aerodynamically, the flow is accelerated
from a Mach number of unity at the orifice to a high supersonic Mach number via
the area expansion in the probe tip. This expansion must be large enough to
achieve the reduction in static temperature required yet not so large as to
cause friction to reduce the stagnation pressure below that achievable by the
pumping system. Larger area ratios would increase the Mach number and con-
sequently increase frictional losses. The flow then enters the constant area
section where the stagnation temperature is reduced by heat transfer. Some
heat transfer also occurs in the tip but is generally only a small part of the
total heat transfer because the residence time of the flow is short in this
region. In the constant area portion, both the stagnation pressure and the
supersonic Mach number of the flow are reduced due to friction. If calcula-
tions indicate that the Mach number is reduced to unity in the constant area
section, either a shock must instead occur at some upstream location where the
shock may be stabilized or the flow in the probe is subsonic throughout. In
either case, no aerodynamic quench is possible. The constant area section must
terminate prior to the occurrence of choking but only after the desired reduction
in stagnation temperature is achieved. This section is generally terminated by
a sudden expansion to stabilize a shock system that reduces the Mach number to
a subsonic value. The expansion must be large enough to result in a relatively
low Mach number so that additional stagnation pressure losses are small; other-
wise, the flow will choke in the subsonic portion of the probe. It is impor-
tant to note that the probe must not contain a bend prior to the desired
shock system location.
111-29
-------
III.C Design of Probes
Using the above assumptions for the probe model and experimental conditions,
two sets of probes were designed. One set was designed for sampling over the
flat flame burner, while the other was designed for sampling exhaust from the
IFRF burner or the FT12 combustor. In these design studies, the primary design
Criteria was to reduce the gas temperature below 1000 K within one millisecond
Csee Section III.B.I.a). Furthermore, it was assumed that the back pressure
was 0.35 atmospheres (266 torr), a pressure that should be easily achievable
for most sampling systems. A special effort was made to develop a probe design
for which true aerodynamic quenching would be expected.
III.C.I Probes for Combustor Measurements
For sampling the exhaust of the IFRF burner and the FT12 combustor, two
water-cooled probes were designed: one for cooling by expansion and the other
quenched by convection. These probes are denoted the reference probe and EPA
probe respectively and drawings of these probes are depicted in Fig. III-2.
As apparent from this figure, pressure taps and thermocouples have been located
in these probes to facilitate comparison of experimental and computed data.
Externally, both probes are identical and photographs of the reference probe
with its water-cooled mount are shown in Figs. III-3 and III-4. Due to their
size, they have been denoted macroprobes. They were each designed to enter
from a side window of the combustor but with a right angle bend to face the
oncoming flow. The probes are sufficiently long so that they could traverse
across the internal dimension of the combustor along the optical path.
The geometry of the reference probe with two sections of constant area
follows the guidelines described earlier (Section III.B.2b) for a probe that
quenches by expansion. As will be apparent, the second expansion has been
located far enough downstream so that the recovery shock returns the gas to
temperatures near 1000 K when the initial gas temperature is 1800 K. Also
observe that the tip has been carefully contoured to minimize the possibility
of coalescence of compression waves (i.e., shock waves) by providing a smooth
transition at the junction of the tip and the first constant area section. It
is very important to note that even with this ideal design, aerodynamic quench-
ing could not be achieved (based on computer program prediction) unless the
back pressure was reduced to below 7 percent («^50 torr) of the ambient flame
pressure which in this case was atmospheric pressure. This pressure was much
lower than the desired operating pressure of 0.35 atmospheres and, as described
in the experimental section, creates some special sampling problems.
A calculated profile showing the static temperature and pressure and the
total temperature vs. time is shown in Fig. III-5 for assumed flame tempera-
tures of 1800 and 1400 K. In the case of the higher temperature, internal
positions from the probe tip are also provided. Although these curves indi-
cate that the static temperature rapidly decreases to below 1000 K (in less
111-30
-------
DRAWINGS OF MACROPROBES
-------
REFERENCE PROBE
-------
TIP OF REFERENCE PROBE
FIG. 111-4
79-10-85-?
-------
CALCULATED TEMPERATURE AND PRESSURE PROFILES FOR REFERENCE PROBE
1800
I
UJ
QC
OC
LJJ
CL
5
LLJ
1600
1400
1200
1000
800
600
23
---- Tex, = 1400K
POSITION FROM TIP (CM) AT 1800 K
7 10 15 2022 25 30
TT T
SUDDEN
EXPANSION
0.10
0.08
0.06 <
LU
QC
0.04
0.02
<0
I
01
I
400
J I I I I I I I
10'5
10'4
RESIDENCE TIME (SEC)
10'3
10-2
O
-------
than 10 microseconds), the total temperature is still high and does not fall
below this level until about 1/2 millisecond for 1800 K inlet temperatures.
The two primary features in the static temperature curves are the minimum at
about 25 microseconds indicating the termination of the divergence in the tip
and the rapid rise due to shock recovery at about 200 microseconds.
Calculated cooling curves for the EPA probe which cools primarily by
convection are shown in Fig. III-6 for the two assumed inlet temperatures, 1400
and 1800 K. For these profiles a back pressure in the probe was assumed to be
0.35 atmospheres (266 torr). According to these calculated curves, the static
temperature (nearly equivalent to total temperature for subsonic flow) reduces
to below 1000 K in approximately 0.3 milliseconds. For comparison, calculated
data from the reference probe when operated at the same back pressure is pre-
sented. In this case, the reference probe cools by convection. This probe
which was designed to achieve an aerodynamic quench is clearly relatively poor
at cooling by convection since the static temperature does not reach 1000 K
until 2.3 milliseconds, a difference of nearly one order of magnitude from the
EPA probe. The inflection near this location is due to the area change at 22
centimeters from the tip.
For both probes, when cooling by convection, it may be noted that the
program indicates the existence of a normal shock within a centimeter of the
tip orifice. In each case, temperature recovery occurs within several micro-
seconds and returns the gas from less than 1000 K to within approximately 100 K
of the original gas temperature. It is this operating mode that is probably
typical of most probes previously believed to be aerodynamically quenched. The
probe is indeed choked, the gas flows supersonically for a period, and the
initial quenching rates are rapid with temperatures falling below 1000 K;
however, a shock system quickly forces the gas to the elevated temperatures and
the initial quenching, although real, is only temporary. Further examination
of the EPA probe design with the computer program indicated that no back
pressure existed for which the shock system could be pulled into the constant
area section and produce an aerodynamic quench. The difficulty was not due to
the small discontinuity in the slope at the end of the tip but rather due to
excessive friction and pressure losses in the probe.
III.C.2 Probes for the Flat Flame Burner
Several probes were selected for sampling over an atmospheric CH^/02/N2
flat flame. The orifice diameter for these probes (miniprobes) was selected to
be 0.025 inches (635 microns) which was large relative to the quartz microprobes
used in the first phase of this program but was necessary to supply the mass
flow required by the analytical instrumentation. Using this orifice diameter
and assumed flame conditions of 2000 K and one atmosphere, no practical geometry
could be found over an experimentally realizable back pressure range for which
an aerodynamic quench, as defined above, could be achieved. Consequently, all
design efforts for these probes were focused at convection cooling below 1000 K
and within one millisecond.
111-35
-------
CALCULATED COOLING CURVES FOR MACROPROBES
0.91
i
u;
o
I
O
I
IX
I
1800
1600 -
5 1400
01
tr
I
1200 -
O
C/3
1000
800
600
1.0
POSITION (cm) FOR TT = 1800K AND REFERENCE PROBE
END OF TIP AREA CHANGE
2.0 3.0 i 4.0 5.0 10 15 201 25 30
BACK PRESSURE = 0 35 aim
EPA PROBE
REFERENCE PROBE
T = 1400K \ TT =
TT = 1800K \
TO
10'4 ID'3
RESIDENCE TIME (sec)
T]
O
-------
Designs of two water-cooled probes were selected for cooling by convection.
One probe was entirely stainless-steel. The computer calculations, however,
indicated that the wall temperature of this probe was reduced to only about 750
K at the location for which the gas residence time was approximately 10 micro-
seconds. This temperature was not considered low enough to minimize wall
reactions in the probe tip. In order to satisfy this quenching criterion, a
copper tipped, water-cooled probe but which is otherwise identical to the
stainless steel probe was also constructed. Calculations for the copper-tipped
probe indicated not only that the wall temperature fell to around 500 K within
a gas residence time of 10 psec but also that the initial cooling rate of the gas
was significantly faster. A relatively blunt tip does offer an alternative to
the change in material, but this was not considered feasible since such a probe
could significantly perturb the flame and since flow separation at the entrance
of the tip (due to the large angle) was considered to be very likely. A
catalytic surface effect due to the copper was considered possible but believed
to be negligible because of the short residence time in the tip. A drawing and
photograph of the stainless-steel probe are shown in Figs. III-7 and III-8
respectively. The third, water-cooled probe was constructed completely of
quartz. Its design was similar to the quartz microprobe used in Task I of this
study (Dodge, et al. , 1979), however, the orifice was enlarged to 635 microns
by shortening the tip (see, for example, the description by Fristrom and
Westenberg (1965)). For comparison, an uncooled stainless-steel probe was
constructed which, except for a smaller outside diameter (0.635 cm) and no
cooling passages, was geometrically identical to the other metallic probes.
Model predictions for the copper tipped and stainless-steel tipped probes
are compared in Fig. III-9. The positions apply to the stainless-steel probe
although they also approximate the positions for the copper tipped probe.
Clearly, there is little difference between the calculated cooling curves for
these probes after approximately one millisecond of residence time; however,
the initial cooling rates are sufficiently faster with the copper tip so that
at residence times less than 200 microseconds, gas temperatures are nearly 150 K
cooler. Note that for both of these probes, the gas flow chokes at the tip,
accelerates supersonically and then shock heats within one centimeter of the
probe orifice.
The effect of back pressure on temperature profiles and cooling rates
was also examined. Model predictions for the copper-tipped probe at two back
pressures, 0.35 and 0.167 atm (266 and 127 torr, respectively), are shown in
Figure 111-10. It is clear that at the lower back pressure the recovery shock
occurs later in time and at a cooler static temperature, the static temperature
returns to temperatures 100 K lower than those at the higher back pressure, and
the total temperature (nearly the same as the static temperature under subsonic
flow) reduces to below 1000 K in about 140 microseconds, approximately half the
time required at the higher pressure. In addition, a 50% drop in the back
pressure provides another advantage to quenching since the rates of bimolecular
reactions are dropped by a factor of four, and even unimolecular reaction rates
slow by a factor of two.
111-37
-------
DRAWING OF MINIPROBE
TIP
I
LJ
CD
XX XX X X
0.95cm
I
r
TJ
-------
STAINLESS STEEL TIPPED MINIPROBE
GAS SAMPLE OUT
WATER IN AND OUT
f .^ -..__, ... ' _^g,__^ -jy- ' ;
-J . I - -J ,
cm 1 2 3
ORIFICE. 635 MICRONS -
ID
I
s
I
U
-------
CALCULATED COOLING CURVES FOR MINIPROBES
2000
i
i~
o
cr
ID
cr
LU
Q.
O
o
I
CO
en
I
1600
1200
800
400
INITIAL STAGNATION TEMPERATURE = 2000K
A',Si IMI D RA( K PRESSURE = 0 35 aim
WATER-COOLED
6,15 MICRON ORIFICE. ML TALI 1C PROBES
0 70
POSITION (CM) FOR S.S TIPPED PROBE
2 34
7 8 9 10 12 15 20
T
IIIITT
T
T
END OF
PROBE TIP
Cu TIPPED PROBE
SS TIPPED PROBL
10'
,-5
,-4
10'
RESIDENCE TIME (SEC)
10'
,-3
10'
-2
O
CD
-------
CALCULATED COOLING CURVES FOR MINIPROBES AT VARYING BACK PRESSURE
1400 -
Cu-TIPPED PROBE
TTOTAL = 1800K
10'
CO
o
00
U1
6 8 10'4
RESIDENCE TIME (SEC)
-------
Detailed calculations were not performed for the quartz probe or the
uncooled probe. The primary differences for these probes will of course be
surface material for the former and, for the latter, surface material at elevated
temperatures and lack of rapid cooling (some cooling is expected since the
probe itself will radiatively and conductively cool and this cooling in turn is
passed onto the gas via conduction).
III.C.2a Microprobe
In the course of analyzing the behavior of probes, quartz microprobes were
also examined since they had been used in the first part of this program.
These calculations using the current version of the UTRC Probe Analysis computer
program indicate that supersonic flow cannot be sustained within the tip region
of so-called "microprobes". Since supersonic flow can be maintained in larger
probes (if the back pressure is set at a prescribed level), it is desirable to
explain the differences in the flow characteristics within each type of probe
in terms of the physical dimensions of the probes.
For a constant property flow, it can be shown that the local Mach number
varies according to:
dM2 1 - M2 dA 0 dT 0 dx
.
= -2 + (1 + YM2) ° + YM* 4f (111-35)
M2 j + Y - 1 M2 A TQ
where M is the local Mach number, A, the area, Y, the ratio of specific heats;
T , the stagnation temperature; f, the skin friction coefficient, x, the distance
from the probe orifice, and D, the internal diameter.
Since this relationship holds for a flow passage of any size, the difference
between flows within small and large probes must lie in the relative contribution
of each term (area change, heat transfer, friction). For a supersonic flow (M
> 1) within a passage of increasing area (dA > 0) and with cooling (dT < 0),
the Mach number will increase if the absolute value of the area change and heat
transfer terms in Eq. (111-35) exceed the frictional contribution; the Mach
number will decrease if the opposite is true.
For example, consider the following case for both a micro-and macroprobe.
Assume that the probes are used to sample a flow whose stagnation pressure and
temperature are one atmosphere and 2000 K, respectively. Assume that the flow
is choked at the entrance of a probe whose tip is a conical section with a
half-angle of 7 degrees.
For a microprobe, assume that the initial probe sample passage diameter
is 0.01 cm (4 mils). It can be shown that the flow accelerates initially (even
in the absence of heat transfer). The Mach number reaches a peak value of 1.71
111-42
-------
at 0.015 cm from the tip and thereafter decelerates to unity (chokes) at 0.167
cm. This second choke point is unrealistic and indicates that instead a shock
must occur near the tip or that subsonic flow must exist throughout. The
reversal of the Mach number change is explained by noting that the flow in this
case is laminar. It can be shown for such a flow that as the distance from the
orifice increases, both the area change and heat transfer terms decrease and
the frictional term remains approximately constant. Thus, the friction term
eventually dominates the change in the Mach number equation.
For a macroprobe, assume that the initial diameter is 0.2 cm (0.079 in.)
which is identical to the orifice of the macroprobes in this study. In this
case, the flow within the probe is turbulent. It can be shown for fully
developed turbulent flow that all contributions to the change in Mach number
decreases as the diameter of the passage increases. Calculations indicate that
friction effects eventually dominate (as they do in the laminar case) and can
choke the flow under the right circumstances; however, the distance to the
choke point may be long enough to provide sufficient length to achieve an
aerodynamic quench. Choking is avoided by encouraging the formation of a
normal shock (such as by a sudden expansion) and slowing to a low subsonic Mach
number. Consequently, the rate of frictional losses is drastically reduced.
Thus, in either case, the friction contribution eventually dominates the
variation in Mach number. If supersonic flow exists within a microprobe, it
cannot be sustained for any useful length; rather a normal shock must exist
very near the probe tip and a low subsonic Mach number must exist thereafter.
This unrealistic shock location (within a few orifice diameters of the tip)
together with the fact that boundary layer growth has been ignored, lead to the
conclusion that supersonic flow is not likely within a microprobe. Boundary
layer growth is less important in a macroprobe, and the required shock location
is more realistic. Thus, supersonic flow can exist within a macroprobe.
111-43
-------
IV. EXPERIMENTAL RESULTS
The objectives of the experimental measurements on the three combustors
were two fold. First of all, probe measurements of seeded nitric oxide would
be made, compared with expected concentrations, and causes for any existing
discrepancies between the various probe measurements and/or the expected
concentrations analyzed. Secondly, sufficient probe data (specifically,
concentration and temperature profiles) must be obtained so that subsequent
optical measurements can be properly interpreted. These measurements along
with experimental data on probe behavior are presented in this chapter.
IV.A. Flat Flame Burner
Three flame stoichiometries, 4> = 0.8, 1.0, and 1.2, were examined and the
run conditions are listed in Table II~A. With the nitrogen purge passing
through the optical ports, vertical and horizontal temperature profiles were
obtained. A typical horizontal profile (for 4> = 0.8) is shown in Fig. IV-1 .
These measurements are corrected for radiation losses as outlined in Task I
Report by Dodge, et al. (1979) and were obtained along the optical axis. The
burner surface was located 2 centimeters below this axis while the visible
flame sheet was a few millimeters above the burner surface. Estimates of
uncertainties in the radiation corrections and individual and repeated measure-
ments are indicated in the error bars. Vertical profiles of uncorrected
thermocouple temperatures for the three flames are shown in Fig. IV-2. Since
the optical beam is less than a centimeter in diameter and centered at 2
centimeters above the burner, the beam encompasses a region for which devia-
tions (due to height) are less than + 15K. As may be observed by comparing
these two figures, radiation corrections are on the order of 140 K for these
f1ame s.
Measurements of stable species using the Scott Instrument package and
the quartz, water-cooled probe (all water-cooled probes produced the same
results within 10 percent) are reproduced in Table IV-A. Equilibrium data at
the adiabatic flame temperature (Gordon and McBride, 1971) for the flat flames
are also presented in this table for comparison. It may be noted that reasonable
agreement is obtained between the measured and equilibrium values. Noticeable
differences are observed for the CO and CC^ concentrations. Although these
differences may be due to interconversion within the probe, most of these
differences may be explained by differences in actual and adiabatic flame
temperatures and to incomplete combustion of carbon monoxide. Only in the case
of the stoichiometric case ( 4> = 1.0) is the measured total of the CO and C02
concentration significantly different from the equilibrium calculations (^8%).
(Data on unburned hydrocarbons are not reported since their concentrations even
in the fuel-rich flame were less than 1000 ppm.)
IV-1
-------
HORIZONTAL TEMPERATURE PROFILE OVER CH4/O2/N2 FLAT FLAME
= 08
O THERMOCOUPLE TEMPERATURE
CORRECTED FOR RADIATION
-2000-
I
K)
-11
-10
-9 -8
o
-5
-4
-3-2-101 23
POSITION FROM CENTERLINE (CM)
10 11
31
-------
FIG. ET-2
VERTICAL TEMPERATURE PROFILE OVER CH4/02/N2 FLAT FLAME
O 0 = 0.8
A = 1.0
D = 1.2
1700
1600
cc
LU
a.
1500
1400
UNCORRECTED FOR RADIATION
I
1 2
HEIGHT ABOVE BURNER (CM)
79-10-85-14
IV-3
-------
TABLE IV - A
MOLE PERCENT OF STABLE SPECIES FOR
THE FLAT FLAME BURNER
(Wet Basis)
Experimental
CO1 CO,1 H~02 N03 Temp. (K)
0.8
1.0
1.2
3.2
0.24
.01374
.064
4.1
6.6
6.55
4.8
12.6
14.1
17.8
77.6
78.7
73.3
1740
1815
1800
Equilibrium
*
0.8
1.0
1.2
o2
3.15
0.13
0 . 7ppm
CO
0.0051
0.109
3.60
co2
6.43
7.09
5.50
H20
12.9
14.3
15.7
N2
77.4
78.1
72.6
Temp. (K)
1765
1905
1904
1. Measured values but corrected for the presence of water vapor.
2. Water estimated from known input conditions.
3. Nitrogen calculated by difference.
4. Error - 40% of value.
5. Based on equilibrium flame temperature.
Except where noted, the uncertainity in the experimental concentrations
is approximately ± 5% of reported (experimental) value.
IV-4
-------
Measured concentrations of nitric oxide were on the order of 5, 30, and 30
ppm for the three flames, respectively. These values at flame temperatures and
for the given optical path length are much smaller than that necessary to
produce a reasonable signal-to-noise ratio for the optical measurements.
Typically, concentrations of 700-1500 ppm are required for the UV resonant
lamp measurements and even higher NO densities are required for the infrared
gas correlation technique. Consequently, all subsequent studies were made
using nitric oxide premixed with the inlet gases.
Using the stainless steel tipped, water-cooled probe and the TECO CLA,
nitric oxide horizontal profiles were obtained for a given seed level. Profiles,
normalized to the input seed level for the flames 4> = 0.8 and 41 = 1.2 are shown
in Fig. IV-3. Data for the 4> = 1 .0 flame is nearly identical to the data for
the $ = 0.8 flame except the centerline fraction is slightly higher (^0.88)
for an NO seed level of 840 and 980 ppm calculated on a wet and dry basis,
respectively. For the lean flame, NOX values were typically 5 to 7 percent
higher than the NO and, for the stoichiometric and rich flames, they were about
3 percent higher than measured NO. The excellent repeatability of the burner
and sampling conditions is indicated by the double set of points on the right-
hand side of this figure.
Vertical profiles for the three flames and the three water-cooled probes
are reproduced in Fig. IV-4. (Note, the data in this figure are not normalized.)
Agreement between these profiles is generally quite good (within 6%) except for
relatively low values obtained by the stainless steel tipped probe when sampling
the rich flame. Since the front part of this tip (J" 1 cm) becomes very hot
(with a red-orange glow), it is believed that catalytic reactions take place
similar to those occurring in an N02 to NO converter. This conclusion is
consistent with the computed results from TCAL indicating that the stainless
steel tip is insufficiently cooled. Although residence times in this portion
of the probe are very short (« 1 msec), the wall temperatures very near the
orifice are significantly hotter than in a stainless steel converter (1100 to
1200 K vs. 1000 K). The profile may be associated with the presence of hydrogen
that would be expected to decrease with height above the flame. Although this
mechanism was not verified, it seems highly likely considering the strong NOX
reducing effect that hydrogen has in hot stainless steel tubes (Benson and
Samuelsen, 1976, 1977). The relatively low values obtained in the = 0.8
flame for the copper-tipped probe are unexplained. Although this difference of
7 percent may be due to a catalytic effect of a copper surface, it is unclear
why good agreement is found for the other flames. In any case, this difference
is considered to be small.
From the above results, it is seen that the recovery of the nitric oxide
seeded into the flame is not quantitative. At the seed concentrations, approxi-
mately 15, 12, and 38 percent of the initial NO was lost for the = 0.8, 1.0,
and 1.2 flames, respectively. Attention was directed to determining whether
the loss occurred in the inlet gas lines (for the unburned/premixed gas), in
the flame itself, or the probe/sampling system. Losses in the post-flame zone
were believed to be unlikely since essentially flat vertical profiles were
obtained (Fig. IV-4).
IV-5
-------
NORMALIZED NITRIC OXIDE PROFILES OVER CH4/O2/N2/NO FLAT FLAME
c6 = oe AND 0=12
O
NO SEED LEVEL
'WET' 'DRV
08 RSO 971 PPM
12 H,'H 1011 PPM
-1.0
I
o>
-A
0.8
*
O
8 8 8
"O --- O
-- 06$-
6 Q -
I
I
I
Q
LU
UU
CO
O
z
Q
LU
QC
CO
LU
I
0.4
0.2
I
1
1
-11 -10
-9
o
I
CO
01
I
-7 -6 -5-4-3-2-101 234
POSITION FROM CENTER OF BURNER (cm)
10 11
co
-------
FIG. ET-4
VERTICAL PROFILES OF NITRIC OXIDE OVER FLAT FLAME BURNER
O s.s. TIPPED)
A Cu TIPPED > WATER COOLED PROBES
D QUARTZ 1
900
1
LU
2
O.
| 800 j
0
S
DC
Z
LU
Z 700
0
O
O
D
LU
DC
| 600
LU
1
= 0.8
SEED 971 PPM
j- g B D B
A
A
~
_
I I
2 3
j
.# = 1 .0
SEED 980 PPM
. A A A A
8 § e 8 8
r
j
I I
2 3
HEIGHT ABOVE BURNER (CM)
#=1.2
900
P
LU
a.
CL
w 800
0
f
DC
1
Z
LU
O
8 ?0°
o
Z
a
1 1 I
LLJ
DC
g 600.
LU t
SEED 1011 PPM
) s g 8 0
e e 8
_ 8 o ° °
i i
1 2 3
HEIGHT ABOVE BURNER (CM)
IV-7
79-10-85-11
-------
These loss figures can be slightly adjusted to account for phenomena
associated with the flame and experimental apparatus. First of all, some NC^
was present as indicated by NO readings, and, consequently, the percent
losses can be decreased by 3 to 5 percent. Although it is uncertain whether
the presence of NC>2 is due to the flame or probe, the fraction of NC^ was not
large enough relative to experimental uncertainties to warrant a detailed
investigation. Alternatively, if one assumes that the NO formed in the flame
without seed NO is also formed when NO is added, then the losses in the
stoichiometric and rich cases can be increased by approximately 3 percent
(about 30 ppm in 1000). The estimated losses at centerline of NO may be
decreased by about 3 to 5 percent due to dilution from the nitrogen purge in
the optical ports (see Dodge, et. al., 1979). Uncertainties also include the
inaccuracies in blending from the mixing apparatus and in the calibration and
analysis. It is estimated that the sum of these uncertainties is on the order
of 3 percent since measured NO concentrations generally agreed to within 3
percent of the calculated values when NO was blended only with nitrogen, and
the gas sample was extracted within 1 mm of the burner surface with no flame
present and with the nitrogen purgp off. This procedure provides an extremely
powerful and important experimental tool to check the operation of the complete
apparatus. It was used to verify not only the accuracy of the blending methods
but also the integrity and behavior of the sampling system. For example, this
procedure was used several times to identify the existence, although generally
not the location, of sample line leaks and indicated the necessity for condi-
tioning new stainless steel lines used in the sampling line or detector system
(see Section III.A.3). This technique is particularly advantageous over other
alternative tests such as a vacuum check for leaks, since it can easily be
performed prior to and after any given experiment and does not require the
opening and resealing of the sample line or probe connection to the sample
line. (Moreover, it also identifies the existence or lack thereof of other
loss mechanisms of NO such as line conditioning or other line losses.)
With all of the adjustments and uncertainties mentioned above, the percent
NOX lost for these flames becomes 9, 8 and 34 percent respectively with an
uncertainty of about ± 5 percent in each number, i.e., for the lean flame, the
estimated real loss can range from 4 to 14 percent. If the natural NO formed
in the flame is not included, then these numbers become 9, 5, and 31 percent
for the 4> = 0.8, 1.0, and 1.2 flames, respectively.
To check whether NO was conserved prior to the exit from the surface of
the burner, the following test was made. The probe tip was placed near the
surface (within 0.4 cm) but still above the flame sheet and NO and NO mea-
surements were made. Then the mass flow of the carrier gas (N^) was increased
to push the visible flame sheet well above the probe tip ( ^ 2 cm) and NO
and NOX measurements were repeated. This test was performed using the copper
tipped probe on the lean and the rich flames with approximately 1000 ppm of
IV-8
-------
seed NO and no nitrogen purge. With the flame below the probe tip for the lean
flame, the probe measurement indicated a 9 percent loss of NO . With the flame
pushed above the tip and correcting for increased diluent in the flame, measure-
ments of NOX were 2 percent higher than the calculated value, with 10 percent
of this NOX measured as N02. In the case of the NOX value, exact correc-
tions due to changes in the viscosity and quenching efficiency of the carrier
gas (Dodge, et. al., 1979a) were not made since the constituency of the carrier
was uncertain after it passed through the converter. Estimates for both NO and
N0x, indicated that less than a 3 percent correction would be required. For
the rich flame, N0x measurements indicated a loss of 31 percent of the seed
NO with the probe at 0.4 cm above the burner; but with the flame pushed above
the probe tip, the loss was only 6 percent of the original NO and at least half
this amount could be due to viscosity and quenching effects. The remaining
differences are considered to be negligible. Again, 10 percent of the measured
NOX was N02 . (in the rich case, the NO converter (molybdenum) could be
operated for a short period of time, approximately one minute or so, before
reconditioning was necessary).
These results indicate that NO is conserved in the gas handling apparatus
A.
and up to the burner. One may argue that by lifting the flame off the burner,
the temperature in the sintered copper surface is reduced and consequently
catalytic action which would depend on stoichiometry is reduced. It is not
believed that this is the case, however, due to other evidence obtained at
UTRC, (discussed later in this section).
These measurements also verify the operation of the sampling line and
analysis system. First of all, the status of the sampling system is the same as it
would be during sampling of the flame since these were sequential measurements
and since other gases were sampled along with nitrogen and nitric oxide. The
other gases presumably were CH/ and 02 although some CO, C02, H2, and
HoO could be formed in the hot tip. The presence of N02 is probably due to
Reaction III-3, i.e., NO + NO + 02 -* 2NOo. Using an analysis similar to
that described in Section III.A.la, this reaction can convert approximately 5
percent of the NO to N02 in the sampling system when sampling the unburned
gases.
The only remaining possibilities are NO losses in the flame zone or in the
probe. To shed further light on this problem, NO measurements were made at
various levels of seed for the three flames. These data are reported in Fig.
IV-5. Examination of these data indicate that the loss mechanism is concentra-
tion dependent. In the rich flame, for example, 63 percent of the NO is re-
covered (as NO) at 1000 ppm whereas at 2900 ppm only 52 percent is recovered.
At least in the case of the rich flame, it is unlikely that any NO will be
converted to N0? in the sampling line via Reaction III-3 since virtually no
oxygen exists in the combustion products. These data are in fact not unique
since very similar results have been obtained by Falcone (1979) using a quartz
probe. For reference, equilibrium concentrations for NO (Gordon and McBride, 1971)
IV-9
-------
FIG. ET-5
NITRIC OXIDE MEASURED VS. NITRIC OXIDE SEED
NO. 4 = 0.8
NO. 0 = 1.0
NO. * = 1.2
NOX. ^=0.8
NO CALCULATED ON 'DRY1 BASIS
S.S. TIPPED. WATER-COOLED PROBE
3000
NOMEASURED = NOSEED
2000
D_
D.
Q
LU
DC
C/5
LU
O
1000
1000
2000
3000
N°SEED -
79-10-86-IS
IV-10
-------
at the measured flame temperatures and for the three stoichiometries 0.8, 1.0,
and 1.2 are approximately 1400, 300, and 3 ppm, respectively. Highest NO
equilibrium concentrations are obtained in the lean case since flame temper-
atures are not that much different and since significantly more oxygen is
present in this flame (see Table IV-A).
The data presented in Fig. IV-5 are not sufficient to identify
which of the two, i.e., flame front or probe or possibly both are responsible
for the loss of NO. These data do, however, indicate two facts. First of all,
the phenomena is concentration dependent. (This fact is important when analyz-
ing data for subsequent optical measurements for which a wide range of NO
concentration was supplied.) Secondly, these results are not only consistent
with possible losses in a water-cooled probe (for which a mechanism does not
exist) but also are consistent with a move towards equilibrium conditions in
the flame. Reactions would only be rapid in the reaction zone of the flame
rather than the post flame zone since concentrations of reactive species are
the highest in these regions, e.g., superequilibrium radical concentrations.
Consequently, losses in the post flame zone would be minimal which is in
agreement with experimental measurements. Although reaction mechanisms may be
similar to that suggested by McCullough, et. al., (1977), a detailed analysis
was not considered feasible here due to the existence of very high gradients
(temperature and concentration) and the resultant difficulty of incorporating
uncertain transport properties of reactive species at elevated temperatures.
Although it could be argued that in the lean case, the equilibrium concentration
of 1400 ppm is higher than the lowest concentrations where losses were still
observed, it should be remembered that at least a portion of the loss would be
occurring in the early part of the reaction zone where temperatures and there-
fore equilibrium concentrations are much lower.
Although the above data are inconclusive, it is considered most likely
that the loss of NO is primarily due to flame kinetics rather than to probe
phenomena. This conclusion has been reached since the differences between NO
measured using the different probes is small relative to the magnitude of loss
of NO, especially and quite noticeably in the case of the rich flame. Similari-
ties between the different probes are observed in spite of noticeable changes
in surface material and quenching rates. Furthermore, several experiments were
performed where the back pressure was varied from about 100 to 600 torr.
Again, these data reproduced the above results well in spite of changes in
quenching rate (see, for example, the discussion in Section III.C.2) and even
unchoking the sampling orifice. In addition, experiments have been performed
in an adjacent laboratory at UTRC (Seery and Zabielski, 1979) using a molecular
beam/mass spectrometer system to sample a low pressure (l/10th atmosphere)
H?/07/Ar flame seeded with various concentrations of NO. In the lean
flames, NO is quantitatively recovered in the post flame zone, but in the
reaction zone a profile of nitric oxide has been observed, decreasing at first
(by as much as 35-40 percent) and then recovering back to the seed value. In
IV-11
-------
the rich cases, initial rates of NO decay are similarly fast but little or no
,ecovery of NO is observed. These results, which are analogous to the obser-
vations in this program on a CH./ 02/N2 seeded flame, strongly indicate
that the observed descrepancies are due to flame rather than probe phenomena.
Although the data when analyzed in relation to kinetic theory suggests
that the destruction of NO occurs in the flame and not in the probe, these
data are not sufficient to draw a conclusion. The ultraviolet absorption data,
however, which will be reported in TASK III Report are in agreement with the
probe data. This agreement between data obtained by two separate methods is
sufficient to conclude that destruction does take place in the flame. Since
the reaction zone where NO loss occurs is extremely small due to the high
pressure (1 atm), the present measurements cannot be used to elucidate the
kinetic details of the loss mechanism. Measurements made at subatmospheric
pressure on the other hand, could prove useful.
IV.A.I. Uncooled, Stainless Steel Probe
As described earlier, the uncooled probe is geometrically similar to the
other metallic probes. This probe, as expected, glowed red when placed in the
exhaust of the flame. Using this probe, NO measurements were similar to
measurements obtained when using the cooled probes for the lean flames although
the results were somewhat dependent on the residence time of the uncooled probe
in the flame. Data are reported in Table IV-B and times between scans are
typically 5-10 minutes. For the stoichiometric flame, the observed NO was
approximately 25-30 percent less, and for the rich flame values ranged from the
same as that measured using cooled probes to only 20 percent of that value
depending on probe history and probe back pressure, i.e., residence time. For
example, at 220 torr back pressure and when the flame is quickly changed from
the lean to rich flame (^ 15-30 seconds), cooled and uncooled probes behave
similarly but in less than a minute the indicated NO begins to fall and after
10 to 15 minutes a stable, but lower value (by a factor 0.66) is obtained.
Then by increasing the back pressure to 430 torr which correspondingly increases
the residence time substantially, the NO drops further to only 20 percent of
the initial value.
The behavior of the uncooled probe, is not unexpected since similar results
have been obtained by England, et. al., (1973) and since similar effects are
common for a stainless steel NOo-NO converter when sampling rich flame gases.
IV.B. IFRF Burner
Initially, temperature profiles were measured and their dependency on
burner operating conditions (i.e., swirl number, position of fuel nozzle, and
design of fuel nozzle) and location within the combustor was examined. The
primary objectives of these tests were to (1) find stable and repeatable
IV-12
-------
TABLE IV - B
MEASURED CONCENTRATION OF NO (PPM) USING
UNCOOLED, STAINLESS STEEL PROBE OVER FLAT FLAME BURNER5
height
above
burner
(cm)
1.5
2.0
2.5
3.0
0.81
772
762
755
752
0.82
717
712
710
702
l.O1
655
630
545
500
1.21
242
237
232
230
1.22
227
222
220
215
1.23
225
217
215
215
1.24
147
142
137
132
Back pressure 225 213 219 218 218 218 435
(torr)
Direction of down down down down down up down
Scan
Seed level 971 971 980 1011 1011 1011 1011
(ppm)
1. First scan
2. Second scan
3. Third scan
4. Fourth scan
5. These data may be compared directly to data for cooled probes presented in
Figure IV-4.
IV-13
-------
operating conditions and (2) obtain reasonably flat temperature profiles in
order to simplify the reduction of the optical data. The selected burner
conditions are described in Section II-C. Probe locations as far downstream as
practical were selected to insure that only combustion products and not un-
burned or partially burned gases were sampled and that the temperature profiles
were relatively flat. Six operating conditions were chosen and these are
listed in Table II-B. Two swirl levels were examined and at each swirl number,
three stoichiometries were tested. Although flames at lower swirl numbers were
tested, these flame conditions were relatively unstable and therefore unsuitable
for these experiments.
A typical temperature profile (corrected for radiation and conduction) for
the run condition 41 = 0.8, swirl = 1.25, are shown in Fig. IV-6. These data
were taken using the aspirated thermocouple described in Section II.C.3. The
dotted lines represent estimates based on measurements in the wings. The
change in swirl produced no measurable difference for any of the stoichiometries.
The shape of the temperature profiles for the rich and stoichiometric flames
are similar to the lean flame with the centerline temperature varying according
to stoichiometry (see Table IV-C)- In the wings, temperature measurements
(uncorrected for radiation or conduction) were made using a chrome1-alumel
thermocouple (0.010" wire diameter) inserted through the open optical ports.
The measured temperatures are much lower than are expected for adiabatic
temperatures of a C-jHg/air flame. The low temperatures observed are due to
cooling from the water-cooled walls of the expansion diameter.
Stable species were measured using the SCOTT Instrument package and both
the "EPA" and reference probes. No differences were observed between these
probes and measured concentrations were independent of back pressure. Experi-
mental data are listed in Table IV-C and equilibrium calculations based upon
the measured (not the adiabatic) temperature are also given. Data for only
one swirl level is given here since the data for the other swirl numbers is
essentially identical. In general, agreement between equilibrium and experi-
mental values are reasonable except for the C02 (and to some extent CO)
values for which the measured values are about 9% low. It is believed that
this difference is due to uncertainties in the fuel flow rate and/or the COj
calibration curve. The high experimental water value (estimated by mass
balance) for the rich flame is due to the presence of about 3.5% molecular
hydrogen (equilibrium value) and the equilibrium value of water is realistic.
The presence of ^ was not accounted for when estimating the concentration of
water.
Concentrations of nitric oxide were measured to be approximately 48, 40
and 25 ppm for the 4> = 0.8, 1.0 and 1.2 flames respectively. NO was typi-
cally, 4 to 5% higher than these numbers for the $ = 0.8 and 1.0 flames and was
not measured for the rich flame. Since these levels were too low (even with the
relatively long path length) to provide adequate signal-to-noise ratios in the
optical measurements, nitric oxide was blended with the inlet air.
IV-14
-------
TEMPERATURE PROFILE ACROSS IFRF COMBUSTOR
-------
TABLE IV-C
Mole Percent of Stable Species
for IFRF Burner1
(Wet Basis)
Experimental
* o22
0.8 4.2
1.0
1.2
CO2
9ppm5
0.063-3
3.0
C022
8.8
10.6
9.0
H203
11.7
15.6
17.3
4
N2 Temp. (K)
75.3
73.8
70.7
1200
1280
1220
Equilibrium
02 CO C02 H20 N2 Temp. (K)
0.8
1.0
1.2
3.92
<1 ppm
vl ppm
<1 ppm
0.021
3.65
9.45
11.6
9.56
12.6
15.5
13.9
73.1
71.9
68.3
1200
1280
1220
1. Although two flames were examined (two different swirls) at each
stoichioraetry, measured values of these stable species were essentially
the same.
2. Measured values but corrected for the presence of water vapor.
3. Water estimated from known input conditions.
4. Nitrogen calculated by difference.
5. Error + 40% of value.
6. Based on measured temperatures
Except where noted, the uncertainity in the experimental concentrations is
approximately ± 5% of the reported (experimental) value.
IV-16
-------
Typical profiles of nitric oxide across the optical axis, normalized
to the seed concentration of NO, are shown in Fig. IV-7. The profile data
were obtained using the "EPA1 probe and did not vary over the back pressure
range examined (100-350 torr). The dotted lines are estimated extrapolations
based on other similar flame measurements. Data for the other flames are quite
similar. Also included on these profiles are experimental data obtained using
the reference probe at two back pressures. At these conditions both the theo-
retical model and experimental pressure profiles inside the probe (presented
in Section IV.D) indicate that the flow in the probe is subsonic except for
possibly a small region in the tip. Thus, the flow is convectively cooled.
No differences between NO measurements using the reference probe and the EPA
probe are observed when operating both in the convectively cooled mode. For
all presssures examined and for both probes NO measurements were typically
within 2 or 3 percent of the NO measurement. For Fig. IV-7, the NO seed values
for the * = 0.8, 1.0 and 1.2 flames were 184, 189 and 182 ppm dry and 162, 160,
and 150 ppm wet, respectively. In these calculations the "dry1 concentrations
were estimated assuming the vapor pressure of water at 3 torr remained in the
sample gas with a total pressure of approximately 500 torr.
Measurements were also made when supersonic flow (verified by pressure pro-
files) extended into the first constant area section of the reference probe.
The data are reproduced in Table IV-D and are compared with measurements made
using the same probe but at higher back pressures, i.e., when the gases were
convectively cooled. Although these data are within about 10%, there appears
to be some difference between the NO measured at a low back pressure (90 torr)
versus that measured at higher back pressure. In addition, the N0meas/N0seecj
ratios are smaller than those obtained at lower seed concentrations and reported
in Fig. IV-7. This latter result, in fact, is not surprising since the results
from the flat flame burner also show a concentration dependence. The former
results apparently indicate a small difference between a probe that convec-
tively cools and one that aerodynamically cools; however, it is more likely
that the observed differences are associated with the very low operating pres-
sure of the sampling system. For example, due to the very low pressures, the
pumps could deliver only half (1 cfh) the normal flow (2 cfh) to the CLA. In
either case, most of the gas was extracted with a 17.5 cfm vacuum pump immed-
iately at the exit of the probe. Although the CLA was recalibrated to the
lower flow rate, a small leak of only 2 to 3% would be difficult to detect
under normal flow conditions yet would amount to a 4-6% dilution when only half
the flow passed through the sample line. In addition, less water would be
extracted at the refrigerator since the total pressure is lower. The resultant
increase in water concentration will not only act to dilute the sample on a
relative basis, but also will provide more efficient quenching of the chemilumi-
nescent reaction (see Section III.A.4) and consequently decrease the response
of the CLA. It is estimated that an increase in the water concentration from
1% to 3% will decrease the CLA response (due to both chemiluminescence quenching
and sample dilution effects) by 4 to 5%. Consequently, it is believed that the
differences observed in Table IV-D are not due to differences in quenching
rates of the gas sample but rather due to a decrease in sample line pressure
and associated phenomena.
IV-17
-------
NORMALIZED NITRIC OXIDE PROFILES ACROSS IFRF COMBUSTOR
O 'EPA PROBE
D 1M6IN OO SS TUBE. UNCOOLED
REFERENCE PROBE AT 180 TORR
A REFERENCE PROBE AT 400 TORR
i o
« a
O
1 4
1 2
1 0
°8
06
0.4
02
00
-
-40
ffl /->/~>n no o O
j o *o n o o
;
.' ° 2 o o
t
//
»
1 1 1
-30 -20 -10 C
(j> = 08. SWIRL = 1 25
O ° ^
Q (^=10. SWIRL = 1 25 \
0 o M
00° .'
lj
(^=12. SWIRL = 1 25 II
O 1
o Oo »|
l\l
tf
V
l i l >V
) 10 20 30 4C
POSITION FROM CENTERLINE (CM)
to
o
-------
TABLE IV-D
Comparison of Nitric Oxide
Measurements Using the Reference Probe
IFRF Burner
NO (ppm)
NOV (ppm)
A.
NO (ppm)
NO (ppm)
NOX (ppm)
4> = 0.8
Swirl = 1.25
Seed = 960 ppm
895 ± 40
880
830
820
4> = 1.2
Swirl = 0.63
Seed = 470 ppm
240 ± 10
246
220
Back Pressure
= 190 torr
(0.25 atm)
Back Pressure
= 380 torr
(0.5 atm)
Back Pressure
= 90 torr
(0.12 atm)1
1. Supersonic flow extends into first constant area section of reference
probe.
IV-19
-------
Additional information on the performance of the reference probe is
provided in Section IV.D.I.
IV.C FT12 Measurements
Three flight conditions, idle, cruise, and maximum continuous were simu-
lated for this series of tests. The corresponding operating conditions are
given in Table II-C. Temperature profiles across the optical axis (same as for
the IFRF measurements) are shown in Fig. IV-8 for idle and cruise conditions.
These data were obtained using the Pt/Pt-13% Rh, aspirated thermocouple. The
profile for maximum continuous is very similar in shape and magnitude to that
for cruise. Also shown in the figure are data from a vertical profile which
indicate no difference between the different quadrants. In addition to these
measurements, centerline temperature measurements using coherent anti-stokes
Raman spectroscopy (CARS) were made (Eckbreth, et. al, 1979). For the CARS
measurements, the temperatures were 580K, 875K and 875K for the idle, cruise,
and maximum continuous conditions respectively. These data agree quite well
with the centerline thermocouple data obtained at the same position i.e., 590,
900 and 920K.
Measurements of CO, C02, and 02 using the SCOTT instrument package
and the EPA probe are listed in Table IV-E. For comparison, equilibrium data
based on the input conditions and measured gas temperature are also presented.
Good agreement is observed between the experimental and equilibrium data except
for carbon monoxide and an unexplained, high experimental value for carbon
dioxide at maximum continuous. The high concentrations of CO measured behind
the FT12 combustors are due to a quenching of the reaction from air dilution
within the combustor. With the reference probe, the COo and 02 concentra-
tions were not measured. Carbon monoxide concentrations were the same as with
the EPA probe when the reference probe was operated both in the convective
cooling mode and with supersonic flow extending into the first constant area
section of the probe.
Without seed, concentrations of nitric oxide (total nitrogen oxides) on
centerline were on the order of 3(15), 5(28), and 6(30) ppm for the simulated
flight condition idle, cruise, and maximum continuous, respectively. These
values varied as much as 20-30% from day to day for any given probe but specific
variations due to probe type or back pressure were not observed. The cause of
the uncertainty may be variations in input conditions for the combustor or
calibration of the CLA at these low NO levels. Careful attention to obtaining
accurate base line values was not given since this program focused on NO seed
levels much higher than 25 ppm. In any case, it is interesting to note that
the N02/NO ratios were typically quite high, on the order of five. The
source of the N02 is not due to the reaction in the sample line.
NO + NO + 02 -* 2 N02
IV-20
-------
TEMPERATURE PROFILE DOWNSTREAM OF FT12 COMBUSTOR
CLOSl.D SYM IDLE
OI'IN SYM CRUISE
O HORIZONTAL PROFILE
D VERTICAL PROFILE
(TOP QUADRANT)
0 A HORIZONTAL PROFILE
(Cr/AI THERMOCOUPLE)
1000
40
30
20
1C
I
O
I
CD
O!
10 0 10
POSITION FROM CENTERLINE (CM)
20
30
40
00
-------
TABLE IV-E
Mole Percent of Stable Species
for FT12 Combustor
(Wet Basis)
Experimental
COJ
H2(r
Temp. (K)
Idle
Cruise
Max. Cont.
0.14
0.19
0.20
19.0
16.8
16.5
0.25
0.20
0.17
1.7
2.8
3.3
1.9
2.6
2.7
77.15
77.60
77.33
580
870
900
Equilibrium
CO CO-
H20
Temp. (K)
Idle
Cruise
Max. Cont.
0.14
0.19
0.20
17.9
16.8
16.6m
<1 ppb
<1 ppb
<1 ppb
1.94
2.60
2.77
1.94
2.58
2.74
77.3
77.0
77.0
580
870
900
1. Measured values but corrected for the presence of water vapor.
2. Water estimated from known input conditions.
3. Nitrogen calculated by difference.
4. Based on measured temperatures
The uncertainity of the experimental concentrations is approximately ± 5%
of the reported (experimental) value.
IV-22
-------
since the rate of this reaction is strongly dependent on the NO concentration
and sample line pressure. Experiments with high seed values of NO ( * 800 ppm)
in air at much higher sample line pressure (750 vs 180 torr) indicated only 3%
conversion to N02. Instead, it is more likely that N02 is formed from NO in the
combustor during the addition of relatively cold air, in the post flame region
downstream of the combustori or in the probe from the reaction with H0~ (see
discussion in Section III.A.I). Insufficient information is available to
determine conclusively which is the primary mechanism, however, it appears
unlikely that probe reactions are responsible due to the relatively low
temperature of the gas and the necessarily low radical concentrations.
Profiles of nitric oxide were obtained with seed levels of NO at 327, 326,
and 326 ppm for the idle, cruise, and maximum continuous flames. Normalized
profiles of nitric oxide for the idle and cruise conditions are plotted in Fig.
IV-9. These data were obtained using the EPA probe. Also shown are profiles
of total nitrogen oxides. For maximum continuous, the centerline fraction of
NO (NOX) recovered relative to the seed value was 0.89 (1.08). In these
gas samples, it is clear that there are relatively large fractions of N02.
For example, idle conditions convert more than 40% of the total NO to N02. For
the same reasons discussed in the previous paragraph, it is believed that the N02
is probably formed in the combustor or post flame zone rather than the probe or
sampling line. Greater losses of NO are observed at idle in spite of the rela-
tively lean stoichiometry possibly because at this level mixing is less intense
and local variations in stoichiometry may be larger and may last longer than
those at the other power levels. In the very fuel rich eddies, losses similar
to those in the flat frame burner undoubtedly take place.
Also shown in Fig. IV-9 are experimental measurements of nitric oxide
using the reference probe. For most of these data the back pressure was
approximately 300 torr. Although good agreement is obtained between measurements
with this probe and the EPA probe, some differences are noted for one set of
NO measurements at cruise and idle. The FT12 assembly had been removed from
the test assembly and reinstalled before this second set of NO and NOX
measurements were made. Small shifts in alignment of the fuel nozzle or
variations in input conditions may be responsible for the changes in NOX
recovery although no corresponding change in the gas temperature was observed.
The reference probe was also used to sample flame gases at reduced pressure
( «/> 95 torr) where supersonic flow conditions extended into the first constant
area section of this probe (see discussion in Section IV-D). In Table IV-F
these values are compared to measurements taken at high back pressure with the
same probe. The agreement is excellent.
IV-23
-------
NORMALIZED NITRIC OXIDE PROFILES ACROSS OPTICAL AXIS FOR FT12 COMBUSTOR
o
A,
r
o
LLJ
LU
-------
TABLE IV-F
Comparison of Nitric Oxide Measurements Using
the Reference Probe - FT12 Combustor
Idle Maximum Continuous
Back Pressure NO 395 740
>f 205 torr
(0.27 atm) N0x 480 800
Back Pressure* NO 385 740
^ 95 torr
(0.13 atm) NOX 478 790
* Supersonic flow extends into 1st constant area section of reference probe
IV.D Experimental Verification of Probe Model
To verify the predicted aerodynamic behavior of gas sampling probes, two
types of measurements were made. First, internal pressure measurements were made
for the reference probe to identify if and when it operated in the aerodynamic
quench mode. The smaller probes were not instrumented due to physical limitation.
Secondly., mass flow measurements were made for probes of three sizes, macro-,
mini- and microprobes, to compare with theoretical values. These data include
not only measurements of the maximum possible flow for a given probe but also
the variation of mass flow as a function of probe back pressure.
IV. D.I Pressure Profiles for the Reference Probe
To verify the operation of the reference probe, three static pressure taps
were positioned along the constant area section and one was placed after the
bend. A detailed design of the probe including the location of these pressure
taps is given in Fig. III-l. The last pressure tap was used to approximate the
stagnation pressure at the exit of the probe and this piece of data is of prime
importance in estimating the location of shock recovery. Typical static pres-
sure data obtained when sampling the exhaust from the FT12 combustor are shown
in Fig. IV-10. The solid and dotted lines in the figure are calculated profiles
when the stagnation pressure at the exit of the probe (back pressure) was
assumed to be equal to the static pressure after the bend. For these calcula-
tions, this assumption can be shown to provide a reasonable estimation. As can
be seen, the agreement between the theoretical and calculated profiles is
excellent. The mechanism for reducing the static pressure to the measured
value was a normal shock whose position and, therefore, strength varied with
back pressure.
IV-25
-------
PROFILES OF STATIC PRESSURE FOR THE REFERENCE PROBE
TEXTERNAL = 90° K
PEXTERNAL" 1 8tm
SOLID LINES CALCULATED PROFILES
SYMBOLS EXPERIMENTAL MEASUREMENTS
I.U
0.8
s
i-
3 ^ 0.6
1 UJ
KJ (E
_,
* i
<£ 1 BACK PRESSURE 0 46 aim
£ 1 >x* "" "^
o 0.4 [- /^
< 1
^
.2
0
1 022 aim ^ ^
\ °
\ 013 aim
N - ,,^ CHOKED ^
X_ , n 1 J 1 IT) * ' 1 1 " 1
0 2 4 6 8 10 12 14 16 18 20 22
POSITION FROM TIP (cm) Z!
0
M
-------
The calculations indicate that at the lowest back pressure the flow
chokes prior to the exit of the constant area section. Since the flow in a
constant area tube can choke only at the exit of the tube, the predicted choke
location is in error due apparently to uncertainties in the back pressure and
the skin friction correlation used in the analysis. In any case, the experimen-
tal data presented in Fig. IV-10 confirm the analytical result that an extremely
low back pressure is needed to extend supersonic flow for a substantial distance
into the probe.
Experimental verification of calculated temperature profiles was considered
impractical since the presence of a thermocouple in the supersonic region will
trigger a shock wave. Even if placed downstream of the recovery shock, thermo-
couple temperatures would probably be ambiguous since the thermocouple will
also be cooled by its water-cooled jacket. Nevertheless, some temperature
measurements were made using the chromel-alumel thermocouple depicted in Fig.
III-2. Although the measurements typically were within 50K of the theoretical
calculations, this agreement may be fortuitous.
NO measurements taken with the reference probe operating at low back pres-
sure are given earlier in this chapter. Typically, for these measurements, the
back pressure was 0.12 atmospheres. Under these conditions, the very low
pressure measurements at the first pressure tap verified that supersonic flow
exists at least past this position. For the IFRF burner at 4> = 1.0, a computer
analysis indicates that the shock occurred 5.2 centimeters from the probe tip
(45 microseconds). The static gas temperature recovered to 1080K (compared to
the external flame temperature of 1280K) and cooled below 1000K in a total time
of 170 microseconds. Alternatively, at a back pressure of 1/3 atmosphere, the
computer calculations indicate that the shock occurs 0.9 cm from the tip (8.6
microseconds). After this shock, the static temperature increased to 1205K but
did not fall below 1000 K until 750 microseconds. Clearly; there is substan-
tial difference between these two modes of operation. When the gas flow was
supersonic past the first pressure tap, the time above 1000K was nearly 6 times
less than that when the flow was primarily subsonic (0.125 vs 0.74 milliseconds).
It was desirable to extend the aerodynamic quench made by reducing the back
pressure further and simultaneously measuring the nitric oxide concentration;
however, this was not possible because of the low pumping speed of sampling
pumps operated at low pressures. At a back pressure of 90 torr (0.12 atm), the
maximum flow rate through the sampling lines was only one cubic foot/hour
( s 7.9 sec/sec) which was barely sufficient to operate the chemilurainescence
analyzer. Only this amount of flow could be obtained in spite of the use of
many metal bellows pumps (see Section II.D.la) in the gas sample line and the
vacuum pump associated with the CLA. To extend the supersonic flow all the way
through the constant area section, the pressure taps indicated that the back
pressure had to be reduced to ^ 50 torr. At this low sampling line pressure,
the collection of sampling pumps would not supply sufficient flow to the CLA.
IV-2 7
-------
IV. D.2 Mass Flow Measurements
Measurements of mass flow through the micro-, mini-, and macroprobes were
made using low pressure drop, Hastings mass flow meters (ALU-100, -5K, and -20K,
respectively). Description of the experimental method used for these measure-
ments is provided in Section II-E. Hot flow data were obtained for each class
of probes and cold flow measurements were made only for micro- and miniprobes.
For the macroprobe at room temperature, the mass flow was too large for accurate
readings on the Hasting meter (mass flow rate through probe
-------
FIG. IV-11
RELATIVE MASS FLOW VS. BACK PRESSURE FOR SEVERAL PROBES
MICROPROBE
(75
-------
mass flow profiles exhibit some differences between the cold and hot flows;
however, these differences cannot be explained using our current understanding
of the fluid mechanics and heat transfer within probe tips. Additional research
is required to identify the controlling phenomena.
IV.D.2c Microprobe
Measurements using a quartz, water-cooled microprobe with an orifice diam-
eter of 75 microns (0.003 inches) were also made over the flat flame burner
( $ = 0.8) and for room temperature air. A photograph of a similar probe is
shown in the Task I report (Dodge, et. al, 1979). Normalized mass flow profiles
are provided in Fig. IV-11. For the cold flow, the data indicate that the flow
chokes for all back pressures below (approximately) one-half atmospheres. This
choking point is consistent with not only the experimental data from the macro-
and miniprobes, but also with equations developed by Shapiro (1953) for the
required pressure drop for choking at a minimum area. Air, for example, will
choke with a pressure ratio (static pressure at minimum area to stagnation
pressure) of 0.53. This ratio is relatively independent of changes in gas
properties. For the hot flow data, a substantial difference is observed
when compared to the other results. The flow does not choke until at least
a back pressure below one-fifth of an atmosphere. This result is in qualitative
agreement with the computer calculations (for probe design) that indicated that
this microprobe cannot be choked (see Section III.C.2a). Based on both an
analytical study and experimental measurements, one must conclude that under
described conditions the flow does not choke (except possibly at very low back
pressure) and consequently will not quench the gas aerodynamically. This
conclusion is in direct conflict with assumptions regarding probe behavior in
other investigations and suggests that some earlier conclusions based on
(assumed) very high rates of quenching may have to be reanalyzed.
The experimental and analytical results for the microprobe do not agree
as well as for the larger probes. In addition to the uncertainties mentioned
previously, it should be noted that the dump loss calculation procedure used
here becomes singular at a high area ratio. The curve in Fig. IV-11 was
calculated for the highest area ratio prior to the breakdown of the calculation
procedure (^550), while the actual area ratio was approximately 4500. The
results for the cold flow condition indicate that a choked orifice may exist
over a range of back pressure if the appropriate sudden expansion loss is
applied. On the other hand, the sudden expansion model for this loss is not
adequate to explain the variation in the hot flow data which indicates that the
probe orifice is not choked.
IV.D. 3 Discussion
The experimental mass flow vs. back pressure profiles indicate that, except
for the microprobe operated at high gas temperatures the probe orifices become choked
IV-30
-------
at the proper back pressure. The back pressure to ambient pressure ratio at
which choking occurs is approximately 0.5 and is consistent with calculations
for choked flow through an orifice plate (Shapiro, 1955). The probe design
model assumes that the flow within the probe is not separated. For the probes
examined in this study, however, separated flow occurs when the probe is
operated unchoked. Except in the case of the microprobe operated in a high
temperature gas stream, calculations accounting for sudden expansion losses
yield mass flows in agreement with the measured data.
It should be noted that the probe design model is generally used to design
probes for which a sudden expansion loss is not important. Either the flow is
supersonic throughout the probe tip (as in designs exploiting an aerodynamic
quench) or the tip area ratio and flow angle are small enough that the subsonic
flow in the tip does not separate. Microprobes are designed to minimize the
disturbance that the probe causes to the flame front. Unfortunately, the
geometry of a microprobe requires a better understanding of its internal flow
characteristics than heretofore appreciated or included in the analysis.
IV-31
-------
V. RESULTS AND DISCUSSIONS
In order to study the processes involved in the probe sampling of NO
and, ultimately, to compare probe and optical measurement methods properly,
three combustion devices were designed and characterized under a variety of
operating conditions. The most controlled of these devices was a flat flame
burner. Temperature and NO concentration profiles for methane/nitrogen/oxygen
flames were obtained at various heights above this burner. The stoichiometries
characterized were 0.8, 1.0, and 1.2. Centerline temperatures in the post-flame
regions varied from 1740 K to 1815 K depending on stoichiometry and height.
Thermal NO produced in these flames ranged from 5 ppm to 30 ppm. Since these
concentrations were considered too low for making accurate optical absorption
measurements, NO was seeded into the flame at concentrations of 850 ppm to 4500
ppm. The amount of seeded NO recovered in the post flame region was dependent
on stoichiometry and will be discussed below.
The second device employed was a swirl burner designed after that of the
International Flame Research Foundation (IFRF). Temperature and NO concentration
profiles were obtained for unseeded and NO-seeded propane-air flames. The
stoichiometries used were 0.8, 1.0, and 1.2. In addition, tangential or swirl
flows were introduced onto the air flow. Swirls (S) of 0.63 and 1.25 were used
where swirl is defined as the ratio of tangential momentum to axial momentum
divided by the radius of the exit. Peak temperatures measured by suction
pyrometry and corrected for radiation and convection ranged from 1200 to 1280 K.
Thermal NO concentrations ranged from 25 to 48 ppm and were principally dependent
on stoichiometry. Seeded concentrations were varied from 130 ppm to 890 ppm.
The amount of NO recovered was also dependent on stoichiometry. Temperature
and concentration profiles were obtained 87.5 cm downstream of the quarl. Data
obtained in four quadrants indicated axial symmetry. The bulk of exhaust was
contained in an expansion chamber of 50.0 cm diameter. To simplify optical
access, windowless ports with purging across the exit of the optical ports were
used. Temperature and concentration profiles through and external to the ports
were obt ained .
The third combustor, which was also installed into the above expansion
chamber, was a modified Pratt & Whitney FT12 combustor. The modifications
consisted of a reduction in length from 41.1 cm to 29.5 cm and the closing of
several air holes to provide axial symmetry. The distance from the combustor
exit to the probe (optical axis) was 78.0 cm. The combustor was operated at
three conditions: idle, cruise, and maximum continuous. The fuel/air ratios
were 0.0106, 0.0143, 0.0152, respectively. Operating air pressure was Mach
number scaled from 6.0 atm to 1.0 atm. The scaling was performed because of
facility limitations and to facilitate accurate definition of the optical path.
Centerline temperatures for the three operating conditions were 580 K, 870 K, 900 K,
respectively. Thermal NO produced spanning the range from 3 ppm to 6 ppm, while
V-l
-------
total N0x values ranged from 12 to 35 ppm. In order to obtain adequate
signal-to-noise ratios in the optical measurements, the air flow was seeded
with NO in concentrations from 270 ppm to 790 ppm. As with the flat flame
burner and IFRF combustor, the amount of NO recovered was dependent on fuel/air
ratio ( stoichioraetry).
The measurement of NO by extractive sampling is a procedure consisting of
three elements each of which must be considered in detail. The first of these
is the removal of the sample from the flame without seriously perturbing the
flame while rapidly reducing sample temperature and pressure. The second element
involves the transfer of the gas from the probe to the analyzer. This transfer
includes, usually, the removal of water and particulates. The third element is
the actual analysis of NO with a chemiluminescent analyzer. This study has
been concentrated on the first two elements. However, information obtained
outside this particular study has been applied to the third element.
Two separate sets of probes were designed and evaluated and set for the flat
flame burner measurements and the other set for the combustor tests. For each
set of probes, a special effort was made (using a computer code for probe
design) to select a probe geometry that was capable of cooling the gas aero-
dynamically. For the flat flame burner, the probes used were: (1) a water-
cooled quartz miniprobe; (2) a water-cooled, stainless steel miniprobe; (3) a
water-cooled, stainless steel miniprobe with a copper tip; and (4) an uncooled
stainless steel miniprobe. Each of these probes had an orifice diameter of 635
V , and the metallic probes had an area ratio of 16. The area ratio for the
quartz probe was 62. In addition, a water-cooled quartz probe with an orifice
of 75 V and an area ratio of 4500 was also examined, however, only mass flow
measurements were made with this probe. For the large combustor, two water-
cooled stainless steel macroprobes were designed. The first of these, which
was defined as the EPA probe (in compliance with the Federal Register require-
ments), had an orifice diameter equal to 0.080 in (2 mm) and an area ratio of
2.4. The second macroprobe, which was defined as the reference probe, had an
orifice diameter of 0.080 in (2 mm) but with dual expansion areas. The area
ratio of the first relative to the tip was 10.6 while the second ratio relative
to the t ip was 29.7.
In addition to properly sizing the probe to minimize temperature and flame
perturbation, these probe designs were analyzed in relation to the available
kinetics of NO decomposition and the aerodynamics necessary to achieve quenching
of NO react ions.
A consideration of gas phase reactions througi, 'hich NO could be destroyed
revealed that if the gas sample was rapidly cooled ( /* 1 msec) to below 1000 K,
NO loss would not occur. In addition, it was estimated that catalytic effects
would be minimized if exposure to walls whose temperature was greater than 600 K
was kept at 10 usec or less. This kinetic analysis is consistent with the
results of this study obtained with probes satisfying the above conditions. The
V-2
-------
aerodynamic behavior of these probes were studied with a computer model based
on the original work of Cohen and Guile (1970). This program predicts the
changes in static temperature and pressure as a function of time and position
in the probe. Parameters considered were area changes, heat transfer rates,
skin friction, and probe back pressures. In addition, mass flow is calculated
for specified combustion stream temperatures and pressures, geometries, and
back pressures. The most important prediction of this model was that micro-
probes and miniprobes (as defined above) cannot achieve an aerodynamic quench.
In such a quench, the gas is supersonically expanded to reduce the static
temperature and pressure rapidly but does not shock heat until the stagnation
temperature is significantly reduced below source conditions via convective
cooling to the walls. Cooling sufficiently rapid relative to kinetic rates,
however, can still be achieved only by convective cooling to the walls under
subsonic flow conditions.
For the macroprobes (large scale), a study of various geometries for
sampling from many temperatures, revealed that an aerodynamic quench is pos-
sible. The reference probe was constructed from the model predictions. The
design for the EPA probe was selected for rapid cooling by convection. The
significant results obtained will be reviewed below.
The second element of NO extractive sampling, i.e., sample transfer, can
be subject to two related problems. The first is the conversion of NO to
N02. The second is the loss of NOo once converted from NO. The principal
mechanism for the conversion is
NO + NO + 02 -» 2N02
This conversion is related directly to the partial pressure of 02 and to
the square of the partial pressure of NO present. Since the reaction is
relatively slow, conversion can be minimized by minimizing the sample transfer
time. For laboratory flames, this reaction is only important for lean flames
and, more critically, for high seed levels of NO. For large scale combustors
which typically have an overall lean stoichiometry. the above reaction is
important. Federal Register (1976) regulations for sample transfer time (2
seconds) seem adequate. A vital part of the sample transfer process, however^
is the time spent in the analytical instrumentation. For some instruments,
this time can be considerable. An example of this problem was considered and
analyzed experimentally. The rate coefficient for the above reaction was
obtained with good agreement ( ^ 50%). The Federal Register (1976) requires 9
seconds to 15% response for N0x- For low values of NO (-M00-200 ppm) this
seems adequate; however, for high concentrations (> 1000 ppm) careful considera-
tion of this reaction is important to the measurement of NO. If conversion of
NO to N02 occurs, then NO loss (NOX loss) is possible in water traps and
particulate filters as has been previously documented (Tuttle, et. al., 1973).
V-3
-------
The final element to be addressed is sample analysis with a chemiluminescent
detector. For laboratory flames with unusual carriers such as He and Ar, sig-
nificant errors in calibration are possible due to quenching differences and
viscosity effects associated with sampling handling. These problems are
avoided by having calibration gases gravimetrically prepared in the desired
diluent and, if possible, analyzed by an alternative method. For laboratory
and cotnbustor flames in air, measurements made on wet samples can have errors
in excess of 10%.
The probe measurements of seeded-NO over the flat-flame burner indicated
the following. First, within the random errors of the measurements, no signi-
ficant differences were observed between the water-cooled quartz, water-cooled
all stainless steel, and water-cooled copper tipped stainless steel probes even
though there was significant visible radiation from the all stainless steel
probe and none from the copper tip. A range of back pressures (100 torr to 700
torr) were employed but the results were independent of this parameter. The
uncooled stainless steel probe produced results similar to the water-cooled
probes in the lean flame. For the stoichiometric and rich flames, NO results
were significantly lower. In addition, for all stoichiometries a systematic
difference between the NO-seed value and the NO measured by the water-cooled
probes was observed. These differences were the greatest for the rich flame
while for the lean and stoichiometric flames the differences were approximately
the same. For all stoichiometries, the amount of NO destruction was dependent
on seed concentration. This fact by itself was not sufficient to conclude that
the NO was being destroyed in the flame and not in the probe. However, the
optical measurements obtained in TASK III of this study indicate that the loss
is occurring in the flame front. The specific reaction for this loss was not
analyzed. It is possible, nevertheless, to suggest that the NO concentration
is slowly driving towards equilibrium. Similar results were observed on the
IFRF diffusion flame. Here, as with flat flame, the amount of NO- observed
cannot explain the loss. For the FT12, a loss was observed forx all operating
conditions. Although these conditions are lean overall the locally rich
regions in the combustor are responsible for the loss.
The macroprobes used in making the NO measurements on the large scale
combustors revealed no significant differences in detected NO. The reference
probe was operated both in the aerodynamic quench mode and in the convective
quench mode with no significant difference in detected NO. It is important to
note that the low back pressure required to operate a probe in the aeroquench
mode is difficult to achieve. The results do not warrant the effort necessary
to make these measurements.
Finally, the validity of the computer model for the probe design model was
investigated. Pressure measurements made internal to the reference probe
verified the predictions of this model and demonstrated that a probe could be
operated in an aerodynamic quench mode albeit at extremely low back pressures.
V-4
-------
Mass flow measurements for varying back pressures were in agreement with the
predictions for the mini- and macroprobes when sudden expansion losses were
considered. Mass flow measurements on the microprobe indicated that when
sampling at atmospheric pressure and high temperatures (^ 1800 K) the flow in
the probe is not choked until at least below 0.2 atm. This is significantly
lower than that predicted by cursory analysis (^ 0.5 atm). The nonchoking of
the probe was accurately predicted with the detailed fluid flow analysis
contained in the probe design model.
V-5
-------
VI. CONCLUSIONS
Based on the results of this study, the following major conclusions can be
drawn for well designed and properly operated probes sampling the exhaust gases of
gaseous and liquid fueled combustors.
1. Water-cooled quartz, stainless steel, and copper-tipped miniprobes
yield the same NO concentrations when sampling products from methane/
oxygen/nitrogen flames with seeded NO at temperatures up to 2000 K and
irrespective of stoichiometry. This similarity in behavior occurs
over a wide range of probe back pressures; hence, no advantage is
gained from back pressures less than 0.5 atm when sampling atmospheric
pressure flames.
2. Uncooled stainless steel probes give NO concentrations slightly smaller
(10-15%) than the cooled probes mentioned above for lean methane/oxygen/
nitrogen flames (1>1800K). For stoichiometrie and rich flames, the NO
concentrations are significantly less and, at least for the rich
flames, the amount of loss is dependent on the probe back pressure.
The destruction of NO in this uncooled probe is similar to that
encountered in NO/NOj converters operated in the absence of oxygen.
Hence, uncooled stainless steel probes are only suitable for sampling
NO in the presence of oxygen.
3. In general, for miniprobes and microprobes, aerodynamic quenching is
not possible because of fluid mechanical and geometric constraints.
However, rapid-cooling of the sample gases (within a few milliseconds)
can be achieved by convective heat transfer to the probe walls.
4. For microprobes, mass flow measurements indicate that, in the sampling
of high temperature gases, choked flow does not occur at the classical
pressure ratio. This result was predicted by the probe model and
suggests that quenching processes may be less efficient than those
estimated in previous studies. A kinetic analysis indicates that
quenching rates are still sufficiently fast for sampling NO in exhaust
gases; the impact of slower quenching rates on gases sampled from
reactive flame zones may have to be examined.
5. Unlike the small scale probes, it is possible to construct a large
scale water-cooled probe that can produce aerodynamic quenching of the
sample; however, measurements made on both gaseous and liquid fueled
combustion systems yielded essentially the same results regardless of
the quenching mode, i.e., aerodynamic or convective. Given this fact
plus the complexity of probe construction and the difficulties in
achieving the low probe back pressures required of the aerodynamic
mode, there is no advantage to aerodynamic quenching in the measure-
ment of NO.
VI-1
-------
The model predictions of pressure distribution and mass flow within
the large scale probe agreed well with the experimental data. In
addition, the model accurately predicted that the microprobe would not
choke at the classical pressure ratio when sampling at high tempera-
tures. Based on these results, it can be stated that the fluid dy-
namic and heat transfer processes have been adequately described in
the model for the case of fluid mechanical choking at the probe
ori fice.
A kinetics analysis of gas phase reactions known to destroy NO indi-
cated that no significant loss of NO would occur during the sampling
process if the sample temperature was reduced to 1000 K in approxi-
mately 1-2 milliseconds. The results of this study are consistent
with this analysis.
A review of the literature indicated no definitive study where large
differences (> 20%) between measurements of total nitrogen oxides from
different probes were observed when properly designed probes and
sampling lines were used and correct calibration of the chemilumi-
nescent analyzer was performed. A properly designed probe is one that
does not perturb the flame environment, does not stagnate the flow, is
water-cooled, operates at a back pressure to external static pressure
ratio low enough to aid in quenching of reactions, and finally has hot
walls (> 600 K) which are limited only to the front portion of the tip
such that the local gas residence time is on the order of 10 micro-
seconds or less. Proper sampling lines are those in which the resi-
dence times are short relative to the time required for the conversion
of NO to N0« and the loss of NO in water traps and particulate
filters. Correct calibration of the chemiluminescent analyzer con-
sists of accounting for the influence of gases other than N? on the
introduction of NO into the chemiluminescent reaction chamber and the
collisional deactivation of excited N02-
VI-2
-------
REFERENCES
Allen, J. D.: Combustion and Flame, 24, 133 (1975).
Amin, H.: Combustion Science and Technology, 15, 31 (1977).
Baulch, D. L. , D. D. Drysdale, and D. G. Home: "Critical Evaluation of
Rate Data for Homogeneous, Gas Phase Reactions of Interest in High Temperature
Systems". The University, Leeds, England, 1970.
Baulch, D. L. , D. D. Drysdale, D. G. Home, and A. C. Lloyd: Evaluated Kinetic
Data for High Temperature Reactions, Vol 2, Butterworth, London, 1973.
i i
Beer, J. M. and N. A. Chigier: Combustion Aerodynamics, Ed. by J. M. Beer,
John Wiley and Sons, Inc., New York, 1972.
Benson R. and G. S. Samuelsen: Presentations at the Western State Section of
the Combustion Institute, Fall Meeting, Paper No. 76-39, October 18-20, 1976
and Spring Meeting, Paper No. 77-7, April 18-19, 1977.
Benson, R., G. S. Samuelsen and R. E. Peck: Presentation at the Spring
Meeting Western State Section of the Combustion Institute, Paper No. 76-11,
April 19-20, 1976.
Beal, J. L. and Grey, J. T.: Journal of the American Rocket Society 23,
p. 174 (1953).
Bilger, R. W.: Probe Measurements in Turbulent Combustion. PURDU-CL-75-02,
Purdue University, School of Mechanical Engineering, May 1975.
Bilger, R. W. and R. E. Beck: Fifteenth Symposium (International) on
Combustion, The Combustion Institute, Pittsburgh, 541 (1975).
Bryson, R. J. and J. D. Few: "Comparisons of Turbine Engine Combustor
Exhaust Emissions Measurements Using Three Gas-Sampling Probe Designs",
AEDC-TR-78-7, November (1978).
Cernansky, N. P.: AIAA Aerospace Sciences Meeting, Paper No. 76-139,
Washington, D. C., January, 1976.
Cernansky, N. P. and S. Singh: "Sampling and Measuring NO and N02 in
Combustion Systems". Final Report for NSF Grant No. 76-22112, Drexel
University, April 1979.
Chen, J. Y., W. J. McLean and F. C. Gouldin: Paper No. 79-17 presented
at the Western States Section of the Combustion Institute, Spring Meeting,
April 1979.
R-l
-------
REFERENCES (Cont'd)
CIAP (Climatic Impact Assessment Program) DOT-TST-75-51, 52 (1975).
Clark, J. A. and A. M. Mellor, Paper presented at the Gas Turbine
Conference and Products Show, ASME, Paper No. 80-GT-71, March, 1980.
Cohen, L. S. and R. N. Guile: AIAA Journal, 8^, 1053 (1970).
COMESA (Committee on Meterological Effects of Stratospheric Aircraft)
U.K. Meterological Office, Bracknell (1975).
Cornelius, W. and W. R. Wade: Transactions of Society of Automotive
Engineering, Paper No. 700708, 1970.
COVOS (Comite d'Etudes sur les Consequences des Vols Stratospheriques)
Societe Meterologique de France, Boulogne (1976).
Crutzen, P. J.: Quart. J. Royal Met. Soc., 96, p. 320 (1970).
Crutzen, P. J.: Ambio, 1_, p. 41 (1972).
Davis, M. G., W. K. McGregor and J. D. Few: Arnold Engineering Development
Center Report AEDC-TR-74-124 (AD-A004105), (1976a).
Davis, M. G., J. D. Few and W. K. McGregor: Arnold Engineering Development
Center Report AEDC-TR-76-12 (ADA021061), (1976b).
Davis, M. G., W. K. McGregor and J. D. Few: J. Quant. Spectrosc. Radiat.
Transfer, 16, p. 1109 (1976).
Dimitriades, B.: Journal of Air Pollution Control Association, 17, p. 238
(1967).
Dodge, L. G., M. B. Colket, M. F. Zabielski, J. Dusek and D. J. Seery: "Nitric
Oxide & Measurement Study: Optical Calibration". Report prepared for DOT/FAA
Contract No. FA77WA-4081 (1979).
Dodge, L. G., M. F. Zabielski, L. M. Chiappetta and D. J. Seery: United
Technologies Research Center, Unpublished report (1979a).
Eckbreth, A. C., P. A. Bonczyk and J. F. Verdieck: "Investigations of CARS and
Laser-Induced Saturated Flourescence for Practical Combustion Diagnosis".
Final ki-port under EPn contract 68-02-3105, United Technologies Research Center
K79-954403-13, September, 1979.
LcKtrt, L. K. b. and R. M. Drake: Heat and Mass Transfer (New York; McGraw-Hill,
1^5^), pp. 15--158.
R-2
-------
REFERENCES (Cont'd)
England, C., J. Houseman and D. P. Teixeira: Combustion and Flame
2,0, p. 439, (1973).
Falcone, P. K., R. K. Hanson, and C. H. Kruger: Paper No. 79-53 presented
at the Western States Section of the Combustion Institute, October 1979.
Falcone, P.: Mechanical Engineering Department, Stanford University, Private
Communication, 1979.
Federal Register, Vol. 38, No. 136, July 17, 1973, Vol. 41, No. 181,
September 16, 1976.
Few, J. D., R. J. Bryson, W. K. McGregor and M. G. Davis: in Proc. Intl.
Conf. Environmental Sensing and Assessment, Las Vegas, Nev. (Sept. 1975).
Few, J. D., R. J. Bryson, and W. K. McGregor: Arnold Engineering Development
Center Report AEDC-TR-76-180 (1977).
Flower; W., R. K. Hanson, C. H. Kruger: Fifteenth Symposium (Int'l) on
Combustion. The Combustion Institute, p. 823 (1975).
Folsom, B. A. and C. W. Courtney: Proceedings of the Third Stationary
Source Combustion Symposium Volume II Advanced Processes and Special Topics.
EPA 60017-79-050b, Feb., 1979.
Friedman, R. and J. A. Cyphers: Journal of Chemical Physics, 23, p. 1875,
(1955).
Fristrom, R. M. and A. A. Westenberg: Flame Structure, McGraw-Hill, New
York, 1965.
Glawe, G. E., F. S. Simmons, and T. M. Stickney: NACA TN 3766, 1956.
Gordon, S. and B. J. McBride: "Computer Program for Calculation of Complex
Chemical Equilibrium Compositions, Rocket Performance, Incident and Reflected
Shocks, and Chapman-Jouquet Detonations". NASA SP-273 (1971).
Gryvnak, D. A., D. E. Burch: AFAPL-TR-75-101, (1976a).
Gryvnak, D. A., D. E. Burch: AIAA Fourteenth Aerospace Meeting, AIAA No.
76-110, Washington, D. C., (I976b).
Halpern, C. and F. W. Ruegg: Journal of Research of the National Bureau
of Standards, 60, p. 29 (1958).
R-3
-------
REFERENCES (Cont'd)
Halstead, C. J., G. H. Nation, and L. Turner: Analyst 9_7, p. 55, (1972),
Hanson, R. K., W. F. Flower and C. H. Kruger: Combustion Science and
Technology, £, p. 79, (1974).
Milliard, J. C. and R. W. Wheeler: Combustion and Flame 29^, p. 15.
(1977).
Jensen, D. E. and G. A. Jones: Combustion and Flame 32, p.l, (1978).
Johnson, G. M. , M. Y. Smith and M. F. R. Mulcahy: Seventeenth Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh,
p. 647, 1979.
Johnston, H. S.: Science, 173, p. 517.
Kays, W. M.: Numerical Solutions for Laminar-Flow Heat Transfer in
Circular tubes, Transactions ASME, p. 1265, November 1955.
Koshi, M. and T. Asaba: International Journal of Chemical Kinetics,
9^, p. 305. (1979).
Kramlich, J. C. and P. C. Malte: Combustion Science and Technology,
J^, p. 91, (1978).
Land, T. and R. Barber: Transactions of the Society of Instrument
Technology £, p. 112, (1954).
Lengelle G. and C. Verdier: Gas Sampling and Analysis in Combustion
Phenomena. AGARD-AG-168, July, 1973.
Lyon, T. F., W. C. Colley, M. J. Kenworthy and D. W. Bahr: Development
of Emissions, Measurement Techniques for Afterburning Turbine Engines.
AFAPL-TR-75-52 , October, 1975.
Matthews, R. D., R. F. Sawyer, and R. W. Shefer: Environmental Science,
and Technology, 11, p. 1092, (1977).
McCullough, R. W., R. H. Kruger, and R. K. Hanson: Combustion Science
and Technology J_5_, p. 213, (1977).
McGregor, W. K., J. D. Few and C. D. Litton: Arnold Engineering
Development Center Report AEDC-TR-73-182, AD-771 642 (1973).
R-4
-------
REFERENCES (Cont'd)
McGregor, W. K., B. L. Seiber and J. D. Few: in Proc. Second Conf.
Climatic Impact Assessment Program, Cambridge, Mass. (Nov., 1972).
Meinel, H. and L. Krauss: Combustion and Flame _33, P- 69 (1978).
NAS (National Academy of Science), "Environmental Impact of Stratospheric
Flight", Climatic Impact Committee (1975).
Oliver, R. C., E. Baver and Wasylkiwskyj: DOT-FAA-AEE-78-24.
Oliver, R. C., E. Baver, H. Hidalgo, K. A. Gardner and Wasylkiwskyj:
DOT-FAA Rpt. FAA-EQ-77-3 (1977).
Rohsenow, W. M. and H. Y. Choi: Heat, Mass and Momentum Transfer;
Englewood Cliffs, New Jersey; Prentice-Hall, p. 192, Eq. 8-34, 1961.
Schefer, R. W., R. D. Matthews, N. P. Cernansky, and R. F. Sawyer,
Paper presented at the Western States Section of the Combustion Institute.
Paper No. WSCI 73-31, 1973 Fall Meeting.
Schlichting, H.: Boundary Layer Theory, Fourth Edition, McGraw-Hill,
New York, 1960.
Seery, D. J., R. N. Guile, L. J. Chiappetta: Paper No. 14, Presented
at Eastern States Section of the Combustion Institute, Nov. 10-11, 1977.
Seery, D. J. and M. F. Zabielski: Private communication, 1979.
Shapiro, A. H.: The Dynamics and Thermodynamics of Compressible Fluid
Flow, Vol. I, (New York; Ronald Press, 1953), p. 228.
Tine, G.: Gas Sampling and Chemical Analysis in Combustion Processes
Pergamon Press, 1961.
Tuttle, J. H., R. A. Shisler, A. M. Mellor: "Nitrogen Dioxide Formation
in Gas Turbine Engines: Measurements and Measurement Methods". Report
No. PURDU-CL-73-06, The Combustion Laboratory, Purdue University, 1973.
Wilde, K. A.: Combustion and Flame j_3, p. 173, (1969).
Vranos, A., B. A. Knight, J. J. Sangiovanni, L. R. Boedeker, and D. J. Seery:
Feasibility Testing of Micronized Coal-Oil (MICO) Fuel in a Model Gas Turbine
Combustor, UTRC R79-954451-1, May 1979.
»D.S. GOVERNMENT FEINTING OFFICE:1981 0-725-402/1111
R-5
------- |