A CFD study of jet mixing in reduced flow areas for lower combustor emissions

22 Pages·0.88 MB·English

by  Smith, C. E.

/,,,u,--7o
NASATechnical Memorandum_!044ii ::: /_,,2 _ -
AIAA-911=2_460 .... _..................... _.................... _...... ............. .. ,. ......
. , ±.... - :.... 7 .... :2..... ± : =_: -,:_-_:-:2-7:Y _-::-= i ,_--=_-.... _ ° LT: 2_7- L _ ,2_-=i-=_-: i
, A CFD__Study of Jet Mixing in
Reduced Flow ___e-asfor Lower
Combustor Emissions =-
C2E. Smith and M.V. Tal_Ilikar
CFD Re_search Corporation
Huntsville, Alabama -
--k: ± -_ L - _L_ :2 - L :L !L
and
J.D_Holdeman ...........................
Lewis Research Center
............. Cleveland, Ohio ' _ _ _ -_ :: _ : .... :_::_::_-i _:: _ _ -_,_
Prepared f6r-ih_- ..... :/-::- ::-7-7 i-:_ ----i-: _7_J , : . " " "
27th Joint Propul_0n Corffe_rence_ .................
cosponsored by the AIAA, SAE, ASME,-and ASEE ""
.......... Sacramento, Ca!ifomg j_e.24-27, i99:!-_=]:_:_ ,=. :_ =_2_--___,,=__:=_. : _--
It
- z -:
= :=
_ __ .=
N91-23185
(NASA-TM-I0441I) A CFD aTUDY OF JET MIXING
IN R_DUCED FLnW APEAS FOR LOWER COMBUSTnR
. <.g
EMISSIONS (NASA) 17 p CSCE 21E
Unclas
=: : 63/0? 0014928

A CFD STUDY OF JET MIXING IN REDUCED FLOW
AREAS FOR LOWER COMBUSTOR EMISSIONS
C. E. Smith*
M. V. Talpallikar**
CFD Research Corporation
Huntsville, AL 35805
J. D. Holdeman***
NASA Lewis Research Center
Cleveland, OH 44135
Abstract
1. Introduction
The Rich-burn/Quick-mlx/Lean-burn (RQL) The design of low NOx combustors isa subject
combustor has the potential of significantly of ongoing research at NASA Lewis Research
reducing NOx emissions in combustion chambers Center as applied to High Speed Civil Transport
of High Speed Civil Transport (HSCT) aircraft. (HSCT) aircraft. One combustor design presently
Previous work on RQL combustors for industrial under studyis the Rich-burn/Quick-mix/Lean-burn
applications suggested the benefit of "necking (RQL) combustor. Originally conceived and
down" the mixing section. In this study, a 3D developed for industrial combustors I-2, the RQL
numerical investigation was performed to study the concept utilizes staged burning, as shown in
effects of neckdown on NOx emissions and to Figure 1. Combustion isinitiatedinafuel richzone
develop a correlation for optimum mixing designs atequivalence ratiosbetween 1.2 and 1.8, thereby
interms of neckdown area ratio. The results of the reducing NOx formation by depleting the available
study showed that jet mixing inreduced flow areas oxygen. Bypass air is introduced in a quick-mix
does not enhance mixing, but does decrease section and lean combustion occurs downstream
residence time at high flame temperatures, thus at equivalence ratios between 0.5 and 0.7. A key
reducing NOx formation. By necking down the design technology required for the RQL
mixing flow area by four, apotential NOx reduction combustor isamethod of rapidly mixingbypass air
of sixteen-to-one is possible for annular withrich-bum gases. Rapid and uniform mixingis
combustors. However, there is a penalty that required for producing low amounts of NOx while
accompanies the mixing neckdown: reduced oxidizingCO produced inthe rich-burnsection.
pressure drop across the combustor swirler. At
conventional combustor loading parameters, the Generic research on dilution jet mixing ingas
pressure drop penalty does not appear to be turbine combustors has been performed in the
excessive. past3,and is applicable to RQL combustors. Good
* Manager, ApplicationProjects
Member AIAA
** Project Engineer
Member AIAA
*** Senior Research Engineer
Member AIAA
engineeaingcorrelationswere developedfor
inches) was performed to understand the cause of
optimummixingofdilutionjetsincan,rectangular
NOx reductioninreduced flow area. And finally, a
andannulargeometries4.Insearchofimproved
1-D computer code was used to calculate the
mixingschemesr,ecentworkhasbeenperformed
overall pressure lossofacombustor and toassess
on staggered dilution jets in rectangular
the penalty ofmixinginaneckdown section. Each
geometriess,asymmetrijcetsincangeometrie6s,
part ofthe studywill be discussed inthe following
andslotsincangeometrie7s. sections.
An important aspect of jet mixing that
warranted further investigation was the effect of
"necking down" the mixing flow area. The mixing Geometry
section has been typically necked down in RQL
combustors to promote better mixing and prevent A "no neckdown" case was selected as the
backflow8-11. In Reference 2 itwas experimentally baseline. The baseline configuration (see Figure
shown that neckdown of the mixing section 2) consisted of three components: inlet pipe,
produced lower NOx emissions. The experiments converging section, and mixing section. The inlet
did not provide the data base to identify why pipe was 6.0 inches (0.152 m) indiameter and 3.0
neckdown produced lower NOx emissions or how inches (0.0762 m) in length. The convergence
to optimize NOx reduction. Hence, this study was section connected the inlet pipe to the mixing
undertaken to investigate the effects of area section and was 0.866 inches (0.022 m) in length.
reduction on NOx formation inthe mixing section, The mixing section had a diameter of six inches
and to develop design correlations to optimize (Le. no neckdown_ a_ad twelve equally-spaced
mixing inreduced areas. slots located on its perimeter. The slots'
centerlines were located one mixing section
diameter downstream of the exit plane of the
converging section. The aspect ratio of each slot
Parametric numerical calculations were was 4-to-1, with the largest dimension of 1.31
performed to quantify potential improvement from inches (0.033 m) positioned in the direction of the
neckdown and to understand the physical mainstream. The mixing section extended two
mechanisms causing low NOx. Both 3-D CFD mixing section diameters downstream of the jet
numerical analysis and 1-D analysis were centerline.
employed. The 3-D numerical calculations were
made using the CFD code named REFLEQS. Grid
REFLEQS has been developed to analyze
turbulent reacting flows12, and has undergone a The baseline grid had 20,160 cells (56x20x1 8
considerable amount of systematic quantitative cells in x,r, 0directions). Figure 3 shows two views
validation for both incompressible and of the baseline grid. The grid distribution is
compressibleflows. Over 30validationcases have
non-uniform with greater grid density inthe vicinity
been performed to date, and good to excellent of the slot as well as the combustor wall. The
agreement between data and predictions has domain in the 0-direction extends from the jet
been shown13-14. Further, ithas been shownthat centerline to between the jets. Only a pie section
REFLEQS is a viable tool in modelling complex with a central angle of 15" was analyzed to
geometries and intricate flow patterns involved in conserve grid points. The grid distribution in each
mixingconcepts of lowemissioncombustors5-7. direction is described below.
The studywas divided into four parts. First, a
baseline mixing configuration was analyzed and
assessed for grid independence. Second, the X0 < x < X1 inlet pipe 4 cells uniform
baseline configuration was optimized interms of
numberofslots. Third,aparametricvariationofthe Xl <x<X2 converging 2cells uniform
mixing diameter (from six inches down to four section
X2<x<X3 pre-slot 10 cells matched last cell Numerical Details
to the 1st cell in
the slot The numerical details of the baseline
calculation (as well as all calculations in this paper)
X3<x <X4 slot 8cells uniform included:
X4<x <)(5 post-slot 22 cells matched 1stcell . Wholefield solution of u momentum, v
to 1-D to last cell in the momentum, w momentum, pressure
slot correction, turbulent kinetic energy (k),
turbulence dissipation (¢), total enthalpy,
Xs<x <XL 1-Dto exit 10 cells matched 1st cell and mixture fraction.
to last cell of
previous domain 2. Second order central differencing of
convective and diffusive fluxes;
Radial Direction
3. Variable fluid properties;
Ro< r<RL 20 cells grid refined at the
combustor wall 4. Adiabatic walls;
with algebraic
packing factor of 5. Standard k-_modelwith wallfunctions;
1.4
6. Turbulent Prandtl number of 0.9;
Angular Direction
. Instantaneous heat-release model and
eo<0<el slot 6cells uniform one-step NOx model (details of the
reaction models are discussed in
01<O<OL 12 cells matched 1st cell reference 7); and
with last cell inthe
slot 8. Six chemical species.
The values of the grid variables in the different
zones discussed above are given inTable 1. Boundary_ Conditions
The baseline case had a jet-to-mainstream
Table 1. Grid Data
momentum flux ratio (J) of 36 and a
jet-to-mainstream mass flow ratio of 1.94. Specific
boundary conditions are stated below.
-0.1744 -0.1490 .0.1236
Mainstream Flow
.0.1524 -0.1270 -0.1016
-0.0167 -0.0139 -0,0111 AxialVelocity = 35.4 m/s (116.2 ft/s)
unit=
Temperature = 2221 "K(3538 "F)
0.0167 0.013g 0.0111 x~m
R~rn Density = 1.864 kg/m3 (0.1163 Ibm/ft3)
0.1524 0.1300 0.1016
0 - deg Composition = 0.134 CO, 0.068 CO2,
0.3048 0.2540 0.2032 (massfraction) 0.006 H2, 0.096 H20, 0.696
0.0000 0.0000 0.0000 N2
Turbulent kinetic = 300.0 m2/s2 (3.2x103
0.0762 0.0782 0.0762
energy, k ft2/s 2)
0.0000 0.0000 0.0000 Dissipation of = 5.5x105 m2/s3
3.1400 3.1400 3.1400 turbulent kinetic (5.92xl 06 ft2/s3)
energy,
15.0000 15.0000 15.0000
The mass fractions of the species were
equilibrium concentrations for propane and air at
an equivalence ratio (¢) of 1.6. The turbulent The combustor wall was treated as a no-slip
kinetic energy corresponded to a high turbulence adiabatic wall (zero enthalpy gradient). Wall
intensity (40%) typically encountered at the exit of functions were used for the calculation of wall
combustor primary zones15. However, the solution shear stress and near wall turbulent quantities (k
has been shown to be relatively insensitive to inlet and _).
turbulent kinetic energy.
Centerline
JetElow (Slot)
The computational boundary at the centerline
Radial Velocity = 120.3 m/s (394.6 ft/s) was assumed to be asymmetryplane.
Temperature = 811 "K (1000 "F)
Density = 5.82 kg/m3 (0.36 Convergence
tbrn/ft3)
Composition = 0.232 02, 0.768 N2 The summations of all error residuals were
(mass fraction) reduced five orders of magnitude, and continuity
Turbulent kinetic = 219.0 m2/s2 (2.3x103 was conserved in each axial plane. Typically,
energy, k ft2/s2) convergence required approximately 300
Dissipationof iterations. Approximately 6 CPU hours were
turbulent kinetic = 1.2x105m2/s3 (1.3x106 required on an Alliant FX/8 mini-supercomputer
energy, c ft2/s3) (configured one computational element per job).
For comparison, the Alliant computer speeds are
The radial velocity corresponded to a liner ~20times slower than aCray X-MP.
Ap/p of 0.03. The assumed turbulent kinetic
energy gave aturbulence intensity of 10%, typical Besults for Baseline Case
of dilution jets15.
The calculated isotherm results are presented
in Figure 4. Figure 4a shows isotherms inthe x-r
plane through the center of the slot (0 = 0).
The exit boundary condition was a fixed Although a 15° pie section was numerically
pressure boundary with pressure set at 200 psia analyzed, the results in Figure 4are shown as a30°
(13.6x10 s N/m2). All other variables (velocity pie sectionfor ease of understanding. The cold jet
components, physical properties, turbulence has penetrated to about the center of the mixing
variables, species concentrations, etc.) were zero section. Reaction istaking place at the interface of
gradient. the two flowstreams as evidenced by isotherms
near stoichiometric temperature. At x/D=0.15,
Transverse Boundaries
Figure 4b shows kidney-shaped isotherms behind
the jet. Figure 4c shows the velocity vectors at
The transverse boundaries were assumed to x/D=0.5. The velocity vectors show the vortex
be symmetry planes. As a check for potential roll-up behind the jet which is atypical feature of a
asymmetric and/or periodic flow behavior, the jet incrossfiow.
transverse boundaries were moved between slots
(doubling the computational grid) and periodicity In Figure 5, NOx emissions are presented in
was enforced. No discernible difference was terms of NOx Emission Index (El) as afunction of
observed between the two solutions. Hence, to axialdistance. NOx El isderived from the sum of
conserve grid points, transverse boundaries were volumefractions of NO and NO2,and expressed as
assumed to be symmetric, and positioned on the equivalent gramsof NO2 perkilogram offuel. The
jet centerline and between jets. valueofNOx El onediameter downstream ofthe jet
centerline (x/D=1.0) is8.14.
4
4. Grid Independence Study developed for circular holed dilutionjet mixing and
jet-to-mainstream mass flow ratios (mj/moo) of
Two sizes ofgridswere employed to check for approximately 0.5. The accuracy of equation 1for
grid independence. The baseline gridwas 20,160 high aspect ratio slots (44o-1) and high mass flow
cells and the fine grid was 68,040 cells. The fine ratios (1.94) studied in this investigation is not
grid was obtained by increasing the grid density by certain.
50% in each of the three directions and
maintaining the same stretching factors. Hence, as a preliminary step to studying flow
area reductionon mixing, a parametric study was
Computational results for the two grids are performed to determine the optimum number of
presented in Figure 6. Quantitatively, they are slots for J=36. The number of slots was
nearly the same. However, the fine grid solution parametrically varied from 10to 14on the baseline
shows slightly greater jet penetration and less geometry. As the number of slots was varied, the
temperature dissipation. central angle of the pie section changed, but the
jet-to-mainstream mass flow ratio (mj/m=) was held
To estimate numerical error caused by grid constant by varying the slot open area. The slot
resolution, the Richardson extrapolation method aspect ratio was maintained at4for allcases.
was employed. The Richardson extrapolation
method utilizes a Taylor series expansion on the The same number of grid cells was used in all
baseline and fine grid solutions to obtain an cases, including the number of grid cells inthe slot.
approximate solution based on zero discretization However, since the slot width-to-transverse
error. The values of NOx El at x/D=1.0 are 8.14, dimension varied in each case, cell density in the
7.97, and 7.47 for the baseline grid, fine grid and O-directionvaried between cases. This variation is
zero error grid, respectively. Based on this finding, thought to have minimal impact on the trends
hundreds of thousands of grid cells would be discussed below.
required to obtain a grid independent solution.
Such finegridswere not practical inthis study. For Figure 7 shows the predicted isotherms in the
a comparative study such as this, it was felt the 0=0 plane for different numbers of slots. The jet
baseline grid is sufficient in accuracy, and should penetration increased as an inversefunction of the
give qualitative engineering answers. number of slots and led to backflow as the number
of slots was reduced to 10. From previous
5. Optimization on Number of Slots experience in reference 6, jet backflow on the
mixingsection centerline leads to poor mixingand
It has been shown in the past 16-19that excessive NOx formation in the combustor. So,
temperature distributions are similar when J and further decrease in number of slots below 10 was
orifice spacing are coupled. Since the number of not considered necessary for this analysis. ASthe
orifices follows from orifice spacing, optimum number of jets was increased to 14, the individual
mixing in a can occurs when the following jets did not penetrate to the mixing section
expression is satisfied4: centerline. Such underpenetration has been
shown to be poor from amixing viewpoint.
n = ___.__ (1)
C Table 2shows NOx and CO emissions at x/D of
1.0 as a function of number of slots. NOx
where emissions decrease with the increase in the
n= optimum number of holes number of slots. Going by NOx emissions alone,
C = experimentally derived Gonstant~2.5 the 14 slot case would be judged to be the
J=jet-to-mainstream momentum flux ratio optimum mixing configuration. However, for the 13
and 14slot cases, CO has gone unreacted on the
Using equation 1, the optimum number of centerline of the mixer due to underpenetration of
slots would be 10 or 11 depending on the the dilution jets. The 12 slot configuration has the
roundoff. However, this correlation was lowest NOx El while exhibiting no CO at x/D--1.0.
Basedonthisanalysis,the12slotconfiguration
same as the baseline case except the exit
was selectedas the optimummixerfor this boundary condition.
geometryandtheseflowconditions.
Table3. Jet VelocityandTurbulence Data
Table2. NOxandCOEmissionsatx/D=1
for VariableNumberofSlots
Slots _xx El CO El
120.3 173.2 270.6
iii!iii!i!ii!i!!!!ii!iiiiiii
10.07 0.00
i1_::!__:jii_i_i_ Ii 219.0 454.1 1098.0
;ii;i!i
8.73 0.00
::.+::.::+:.::..:;++:;..:;+.;:+.:;+.::++::+:-:+;:+,:+:+:.:+:.:.:.: : 1.2 xl0 s 4.3x10 s 2.0x10 s
iiiiiiiiiii!!i ! iii8ii.i1 i 4iii i' o.oo
The pressure at the exit plane for the 6 inch
+::;--;;).::.:;++::.+:.::++::+.::+.::.+:+:-::,:.:+:.::,:.::.:I.+:::::-:
..,..,.,.....,......,.....,...........,,...... diameter case was set to be 200 psia (13.6 x 105
i!iii!iiii!!!!i i i!!iliiiiiiiiiiiiii
7.37 1.22 N/m2). For the 5 inch and 4 inch neckdown
:::::.::;::+:::::::;::,:::::::+:::::::.:::::+:::::::,:::::::+:::::::+:::::::+:::::::,:::::::.:::::+:::::::.:::::::::::::.::::::::::::: diameters, the exit pressure was set at 198.8 psia
:::::,:::::+:::::::.:::::+:::::::.:::::.:::::+:::::::.:::::,:::::.:::::.:::::.:::::.:::::.:::::.:::::.:::::.:::::.:::::.:::::.:::::+:::::::.:::: (13.52 x 105 N/m2) and 195.3 psia (13.28 x 105
6.75 5.71
N/m2), respectively. The lower pressures were
!i.iiiilil);iiiiii!iiiiiii}i!ii!ii!!iiiil
determined by assuming isentropicflow expansion
fromthe fiveorfourinchdiameter mixertoa 6inch
6. Parametric Study of Area Reduction diameter exit. This precluded the necessity of
modeling a diffuser at the exit of the five or four
Using the optimized 12 slot geometry, three inchmixerinthe CFD calculations.
neckdown configurations were analyzed to assess
the effect o! flow area reduction on NOx emissions. The grid distribution inthe axial and the radial
The three mixing section diameters were 6, 5, and direction was identical except for the size of the
4 inches (0.1524, 0.127, and 0.0762 m). As the slot. The grid distributions for the three
flow area was reduced, the velocity of the configurations are given inTable 1.
mainstream flow in the mixing section increased
proportionately to the flow area reduction. The Figure 8 shows the isotherms in the plane
resulting reduction in mainstream static pressure in through the jet centerline (0=0) for allthree cases.
the mixing section increased the pressure drop Figure 9 shows isotherms at an r-e plane one
across the slots, thus increasing the jet velocity. mixing section diameter downstream of the jet inlet
For incompressible flow, the increase in (x/D=1.0). In this figure, a full 360° plane is
mainstream and jet velocities exactly displayed, although the computations were
counterbalanced, and the jet-to-mainstream performed on a 15° pie section. The identical
momentum flux ratio (J) remained constant as the nature of the flow patterns shows that the mixing
mixing flow area was reduced. characteristics were identical for each case.
The slot size was adjusted according to the There was some concern that flow separation
variation in diameter of the mixing section to was not predicted at the inlet 1o the four-inch
ensure a constant mass flow ratio (mj/m=). The diameter mixing section. To investigate this
turbulence parameters at the jet inlet had to be concern, a number of cases were run with fine grid
rescaled according to slot size and the jet velocity. in the converging section and immediately
The jet velocity and turbulence parameters at the downstream. Cases were run with and without
jet inlet for each mixing diameter are given inTable dilution jets. Without dilution jets, flow separation
3. The rest of the boundary conditions were the was predicted for laminar flow, but not for turbulent
flow (although a somewhat thick boundary layer
was calculated downstream of the contraction). A flow area reduction of 4.0 appears to be
However, with dilution jets, the mainstream flow possible in conventional combustor designs (see
"sensed" the jet blockage and started accelerating next section for details), giving a potential NOx
at the entrance of the mixing section. Hence, flow reduction of 16-to-1 in an annular combustor
separation (and a thick boundary layer) was (84o-1 inacan combustor).
avoided inthe neckdown mixing sections.
7. 1-D Pressure Loss Analysis
In Figure 10, NOx El is plotted as afunction of
axial location for the three different neckdown There is apenalty involved inreducing the flow
diameters. NOx decreased as the mixing section area of the mixing section. By necking down the
diameter decreased. For all the cases, CO was mixing section,atotal pressure drop occurs across
completely depleted by x/D of 1.0. The NOx Elfor the mixer, backpressuring the combustor. The
the six inch diameter mixing section was 8.14 at backpressure causes a reduced pressure drop
x/D of 1.0, while the NOx El for the four inch across the combustor swirler. The reduced swirler
diameter mixing section was 2.43, a 3.35-to-1 air pressure drop results in lower atomizing
reduction. velocities and worse atomization quality.
The formation of NOx is controlled by local To investigate this backpressure effect, a 1-D
temperature, local oxygen concentration, and local flow model of a combustor was developed. This
residence time. Since mixingwas identicalfor the model was similar to the 1-D model discussed in
three mixing diameters analyzed, the local reference 2 that showed good agreement with
temperatures and oxygen concentrations must be experimental pressure loss measurements. Figure
identical.This left residencetime asthe parameter 11 shows the basic elements of the model,
causing reduced NOx levels. Residence time is consisting of 1) primary zone section, 2)
reduced in neckdown mixers intwo ways: higher converging section, 3) constant-area mixing
velocities andshortermixinglengths. section, and 4) diffuser. The primary zone was six
inches in diameter, and the mixing section
An engineering correlation was developed to diameter was varied between six and three inches.
approximate NOx emissions attainable by reduced
flow areas. The correlation (based on residence The hotmainstream gases inthe primary zone
time considerations) is expressed below: section were isentropically accelerated into the
mixing section. In the mixing section, the 1-D
NOxn_.do,, = Aneck_o,,. Hned_down (2) momentum equation was used to solve for static
NOxnoned_down Ano neckdown Hno neckdown pressure at the exit of the mixing section. The jet
velocity was assumed to enter radially, and
where complete (Le. uniform) mixing and reaction was
A = flow area assumed. An iterative solution procedure was
H = height (diameter in can, duct height in used, in which the inlet pressure of the hot gases
annulus) was iterated until a combustor exit pressure of 200
pslawas attained.
A comparison of CFD results with equation 2 is
shown inTable 4. To better understand the relationship of
combustor loading parameter on backpressure
Table 4. Neckdown Effect on NOx Emissions penalty, calculations were performed with two
_P__2T fro rr_L,;cn'6"w reference velocities: 50 and 100 f/s. The
reference velocity is defined as
! i i i iIiiiii1i .i0 iiIi 1iii°i1!iii! ! i
Vrel= rh (3)
1.,3i1,,i
pA
_i_::_iii_i?:_:_i_!_i_i_i_:3:!_!3_is_i_ I 3.38]
7
where
effect is caused by increased total
pressure loss across the mixing section.
rh =total combustor airflow Analysis showed the penalty for neckdown
p =combustor inletdensity to be relatively minor for conventional
A =area ofthe inlet(6in. diameter) combustor loadingparameters.
A combustor reference velocity of 50 f/s 9. Acknowledgements
corresponds to conventional combustor design
practice. The authors wish to thank NASA Lewis
Research Center for funding this work under
Figure 12 presents the predictions of swirler NASA Contract NAS3-25967. Our thanksalso are
pressure drop versus mixing flow area. For extended to Ms. Kathy W. Rhoades for preparing
demonstration purposes, a six percent Ap/p was thistypescript.
assumed across the swirler for no mixing
neckdown. As the mixingflow area was reduced, 10. Reference_
the pressure drop acrossthe swirierwas reduced.
For a combustor reference velocity of 50 f/s, a 1. Mosier, S.A., and Pierce, R.M., "Advanced
4-to-1 flowarea reductionproduced afour percent Combustion Systems for Stationary Gas
Ap/p across the swirler. Such a swirler pressure Turbine Engines," Vol. I,EPA Contract 68-02-
drop should be acceptable to combustor 2136, 1980.
designers. However, for a combustor reference 2. Pierce, R.M., Smith, C.E., and Hinton, B.S.,
velocity of 100 f/s, it is evident that excessive "Advanced Combustion Systems for
backpressure would result, making the three-inch Stationary Gas Turbine Engines," Vol. III, EPA
diameter mixingdesignimpractical. Contract 68-02-2136, 1980.
3. Holdeman, J.D., "Summary of NASA-
To get confidence in the 1-D model, results Supported Experiments and Modeling of the
fromthe 3-D CFD calculationswere compared with Mixing of Multiple Jets with a Confined
the 1-D predictions. Figure 13 shows the Crossflow for Controlling Exit Temperature
comparison and good agreement between 1-D Profiles in Gas Turbine Combustors," AIAA
and 3-D calculations. Paper 91-2458, 1991.
4. Hoideman, J.D., Srinivasan, R., and White,
C.D., " An Empirical Model of the Effects of
Curvature and Convergence on Dilution Jet
The overallconclusionsofthisstudyare" Mixing", AIAA Paper 88-3180 (NASA TM
100896), July 1988.
. By reducing residence time at high flame 5. Smith, C.E., "Mixing Characteristics of Dilution
temperatures, mixing in a "neckdown" Jets inSmall Gas Turbine Combustors," AIAA-
mixing section significantly reduces NOx 90-2728,1990.
formation. A design correlation was 6. Talpallikar, M.V., Smith, C.E., and Lai, M.-C.,
developed for NOx reductionattainableby "Rapid Mix Concepts for Low Emission
area reduction,as shown inequation2. Combustors in Gas Turbine Engines," NASA
CR-185292, 1990.
Area reduction of 4.0 appears to be 7. Talpallikar, M.V., Smith, C.E., Lai, M.-C., and
possible in conventional combustor Holdeman, J.D., "CFD Analysis of Jet Mixing
designs, giving a potential NOx reduction in Low NOx Flametube Combustors," 36th
of 16-to-1 inan annular combustor (8-to-1 ASME International Gas Turbine and
inacan combustor). Aeroengine Conference, 1991.
8. Russell, P.L., et aL, "Evaluation of Concepts
. The penalty for neckdown manifests itself for Controlling Exhaust Emissions from
in reduced pressure drop across the Minimally Processed Petroleum and Synthetic
combustor swirler. This backpressure Fuels," ASME Paper 81-GT-157, 1981.
8