"°i• •°It'

C es-•' AD-A258 456

ions will be in black•"ll
whitoe 11111'

DTIC
SELECTE !

C D
A Computational Study of a Supersonic Mixer-Ejector
Exhaust System
1
T. J. Barber
United Technologies Research Center, East Hartford, CT

IoNovember 10, 1992

I R92-957930

I,
!
.......
... .?":•• t...U k.................:-~

•-o
Ia amz
Contents

1 Abstract 1

2 Background 1

3 Ejector Plume Analysis Approach 1

3.1 Geometry and Grid Issues .............................. 3

3.2 Flow Boundary Conditions .............................. 5

4 Discussion of Computational Results 5

4.1 Parametric studies ................................... 6

5 Full Scale Nozzle Performance Asssessment 7

6 Conclusions 9

7 Acknowledgements 9

8 Appendix: Ejector Internal Flow Analysis 12

' For

U;... .e&m#ti (=
St-A per telecon, Mr. Shreeve, ........
Naval Postgraduate School/Code 67SF.
Monterey, CA 93943 V

JK 12-8-92 "" 'z.--t, Codes

I
111v oe
 List of Tables

1 Mixer-Ejector Geometry ......... .................................. 22

2 Flow Conditions for Mixer-Ejector Cases ................................ 22

3 Hot Flow Mixer Ejector Percent Mixing ....... ......................... 24

List of Figures

1 End View of Advanced Mixer Ejector . ................................ 3

2 Side View of Advanced Mixer Ejector . ............................... 4

3 Inital or Exit Plane Data ......... ................................. 11

4 Data Comparison at x = 3.3in ........ .............................. 13

5 Data Comparison at x = 10.3in ........ .............................. 14

6 Data Comparison at x = 40.3in ........ .............................. 15

17 Centerline Total Temeprature Decay................................16

8 Grid Resolution Study ......... ................................... 17

1
9 Turbulence Initialization Study ........ .............................. 18

10 Vorticity Initialization Study ........ ............................... 19

11 Full-Scale Centerline Total Temperature Decay ........................... 20

1 12 Full-Scale CrossPlane Total Temperature Distributions ..................... 21

1 13 PT Distributions for a Supersonic Slot-Nozzle Ejector: AR = 3.0 ............... 26

14 PT Distributions for a Supersonic Mixer-Ejector: AR = 3.0 ................... 27

15 PT Distributions for Supersonic Mixer-Ejector: AR = 4.5 ................... 28

S16 Measured and Calculated PT In a Supersonic AR = 3.0 Mixer-Ejector at Z = O.OOft 29

17 Measured and Calculated PT In a Supersonic AR = 3.0 Mixer-Ejector at Z = 0.15ft 30

1 18 Comparison of Cross Flow Velocity Fields in Supersonic Mixer-Ejector ........ .. 31

19 Development of Total Circulation of Secondary Flow in Supersonic Mixer-Ejector . . 32

I 20 Development of Average Vorticity of Secondary Flow in Supersonic Mixer-Ejector. . 33

21 Calculated and Measured TT Distributions for Hot Flow Supersonic Mixer-Ejectors. 34

22 Decay of TT Centerline Distributions for AR = 5.2 Supersonic Mixer-Ejectors . . . . 35

23 Measured and Calculated Exit Plane TT for Supersonic AR = 4.5 Mixer-Ejector . . 36

I 24 Measured and Calculated Exit Plane TT for Supersonic AR = 5.1 Mixer Ejector . . 37

II
 I
1 Nomenclature

I A, Ejector secondary area (in 2 )

I AA, Ejector primary area (nozzle exit area) (in 2 )

AR Ejector area ratio (A 3 /Ap)

A3 Ejector mixing duct area (in 2 )

B Slot-nozzle width (long dimension) (in)

CP Specific heat (ft 2 Isec2 IoR)

g Gravitational constant

1 HE Ejector shroud exit height (in)

HT Total enthalpy (ft 2/seC 2 )

h Slot nozzle height (short dimension) (in)

hl, h2 , h3 Metric or grid scale factors

jT Enthaply flux (ft lb f/sec)

L Length of ejector shroud (in)

Ld Length of ejector mixing duct (in)

M Local flow Mach number

M. Forward flight Mach number

Mi Jet exit Mach number

rI8 Ejector secondary weight flow (ibm/sec)

IriP Ejector primary weight flow (Ibm/sec)

mh Total integrated mass flow (ibm/sec)

NPR Nozzle pressure ratio (PTJ/Psq)

IV
 P Static pressure (psia)

PT Flow field total pressire (psia)

R Gas constant (ft 2 /sec 2 /°R)

S Entropy (ft 2/sec2/°R)

T Static temperature (°R)

TT Flow field total temperature (°R)

t Stream thrust (ibf)

U1, U2 , U3 Velocity components (ft/sec)

U Fully mixed flow velocity (ft/sec)

WE Ejector shroud exit width (in)

XP Axial penetration of nozzle exit plane into ejector shroud (in)

X or Y1 Axial distance (in)

Y or Y2 Transverse distance (in)

Z or Y3 Vertical distance (in)

I Ratio of specific heats

A Heat conductivity (ft Ibf/ft/8ec/0 R)

Q Streamwise component of vorticity

A Viscosity (slug/ftlsec)

p Density (slugs/ft3 )

V
Subscripts:

.1, P, 1 Primary, nozzle exit or jet flow state

0o Wind tunnel freestream or ambient flow state

Ssstate
Secondary flow

• Sonic flow state

Superscripts:

- Average condition

VI
1 Abstract

An analytical model of the mixer ejector based on a PNS solver approach has been applied to
analyze the flowfield downstrean of a supersonic mixer-ejector exhaust system. The method has
been used previously to analyze the plume flowfield of unshrouded mixer-type nozzles as well as the
flowfield within the mixing duct of a mixer-ejector exhaust system. Calculations are presented for
a model-scale exhaust system operating at takeoff flight conditions. Favorable comparisons with
total temperature and pressure experimental data are presented in the text. In addition paramet-
ric studies are presented examining the importance of turbulence level, streamwise vorticity, grid
resolution, etc. on the downstream plume mixing characteristics. Finally, a series of calculations
examining the effect of engine power setting (operating condition) and exhaust system size (full
scale) are analyzed.

2 Background

A three-dimensional viscous flow analysis has been developed and applied by Anderson and Barber
in a joint analytical / experimental program [1], [2], [3] to analyze the exhaust jets produced by
complex supersonic nozzles. The parabolic fully viscous method presented is an extension of a two
dimensional method presented by Anderson [4-] which was later extended to three dimensions by
Anderson and Hankins [5] but restricted to orthogonal coordinates. In this method, the primary
flow equations, which include the streamwise momentum, transverse pressure, and energy equa-
tions, are solved first for the streamwise velocity, static pressure, and total enthalpy assuming that
the crossflow velocities are known. With the primary flow variables known, the secondary flow
equations, which include the continuity, vorticity, and vorticity transport equation, are solved for
the secondary flow variables which are the streamwise component of vorticity, and the two cross
flow velocity components. The method has also been modified and applied to analyze the flowfield
within the mixing duct of the ejector [6]. The ejector analysis procedure combines empirical data
obtained from previous analytical/experimental studies with the PNS technique. Computational
studies have been reported for a flat plate shroud, flight-inlet ejector, containing either a mixer or
a slot nozzle operating at an exit Mach number of 1.5.

3 Ejector Plume Analysis Approach

In the previously conducted ejector analysis studies, the flow field within the mixing duct or internal
flow portion of the exhaust system is controlled by the mixed-out exit static pressure and the nozzle
exit flow field. Therefore, the PNS-based analysis, which required input parameters to initialize the
calculation, resorted to a combination of experimental and empirically determined parameters at the
starting plane of the calculation. For example, internal flow calculations have been performed (see
Appendix) where the experimentally determined pumping rate was specified and the equilibrium jet
approximation imposed (P. = Pj) in lieu of specifiying the ejector exit static pressure to drive the
pumping. In those calculations, the nozzle exit or jet (J) conditions have been assumed to exhibit
the same distributional characteristics as measured in the unshrouded mixer nozzle program [2].
The specific level of induced exit plane vorticity ( however, has to be empirically determined
from the experimental levels of pumping. Experience in setting this parameter and validation of the
internal flow model has been obtained by performing calculations on mixer-ejector configurations
previously studied experimentally at UTRC.

In the current analytical effort, the experimental PTJ, TTj, and Q (determined from U 2 , U3 LV data)
are combined with the freestream or coflowing conditions to generate the starting plane profile. The
PT and TT data are then normalized using the following relation,

-_F - F,(
Fj -F.

where the subscript (J) refers to the jet (primary) conditions, and the subscript (00) refers to the
free stream (secondary) conditions. In this format, the flow conditions can be rescaled to the test
conditions.

The initial cross flow velocity components U2 , U3 are also very important because they determine
the magnitude of the enhanced mixing. These cross plane velocities were determined in an indirect
manner since the analysis requires as input the streamwise component of the vorticity f)(Y2 ,Y 3 )
and then solves the Cauchy-Riemann problem given by,

-U2 + &U3 - (2)

OU2 0 Y3
3 ( -U (3)
01"2 aY3

The external or plume portion of the calculation is performed over a domain of sufficient lateral
and axial extent so as to capture the mixing of the jet to ambient conditions. This calculation
can be initialized using the exit flow field from the internal calculation and the ambient coflowing
stream. A shear layer is constructed based on a nominal flat plate boundary layer growth along

2
 I'-J WE

1
Z
z_
j~~Vl _ - HE

Figure 1: End View of Advanced Mixer Ejector

both surfaces of the ejector wall. In situations where experimental data was obtained at the ejector
exit plane, PT, TT profiles are used in conjunction with the analytically determined cross-flow field.

3.1 Geomnetry and Grid Issues

A schematic of the mixer ejector exhaust system is shown on Figs. 1 and 2 showing the end view
and side view respectively. The end view, shown in Fig. 1, has an outline of the advanced mixer
exhaust nozzle surrounded by a rectangular shroud of width WE and height HE. The side view
shown on Fig. 2 shows the advanced mixer exhaust nozzle exit plane at X/L = -1 surrounded by
a shrouded duct of length L. The current computational simulations were started at the ejector
I exit plane.

The exhaust plume calculations were conducted using a Cartesian grid. All calculations performed

3
 X/L = -1.0 -0.5 0.0

IL

Figure 2: Side View of Advanced Mixer Ejector

in this study assumed a 1/4 plane symmetry. The external calculations used a (99 by 99) cross-
planar grid composed of uniform inner grid (80 by 80) in conjunction with a outer grid stretched
geometrically to the freestream or coflowing boundary so as to avoid influencing the plume mix-
ing rate. In this inner region the grid spacing, referenced to the mixer nozzle effective hydraulic
diameter, is approximately 0.027, or about 22 points per mixer lobe. The axial mesh was also geo-
metrically stretched to allow for a more graduad evolution of the turbulent jet flow. Approximately
175 axial planes were used over the first 10 feet of the jet. Parametric mesh studies involving cross-
planar meshes ranging from (50 by 50) to (120 by 120), have demonstrated that (99 by 99) point
meshes accurately model the ejectors mixing characteristics. As a side note on computational effi-
ciency, a comparison was made for a typical internal calculation, involving approximately 300.000
grid points (50 by 50 cross-planar mesh) which required about 1.5 hours on a CRAY XMP. It should
be noted that an equivalent calculation [8], using a pressure-correction based Navier-Stokes scheme.
required 26 CRAY XMP CPU hours. An approximately 20 to 1 speed up definitely qualifies the

4
PNS approach as efficient; the calculation/ experiment comparisons below will demonstrate the
effectiveness of the method.

3.2 Flow Boundary Conditions

In the current analytical study, experimental flow conditions are interpolated onto the computa-
tional grid at the ejector exit plane, with additional flow field definition obtained by extrapolating
from the edge of the experimental data domain to coflowing or far field conditions. Traverse
mesh densities were typically (28 by 18) over the 1/4 plane for (PT, TT) and (14 by 10) over the
1/4 plane for the velocity field. The PATH code requires streamwise vorticity input, r-ther than
U2, U3 , therefore £2, was determined using second order accurate differencin- of the experimental
data. A maximum value of f., - 6000 per second was measured for the AR = 4.0 ejector case konly
case where LV data obtained). All vector components normal to planes of symmetry were zeroed
out for consistency with 1/4 plane of symmetry calculation. Color plots of the normalized (TT, PT)
and flu) data are shown in Fig. 3. Note the strong residual effect of the mixer nozzle central core,
cited in Ref. [6], and the modest alignment of Ihe vorticity "cells" with the superimposed upstream
mixer nozzle lobe side walls.

4 Discussion of Computational Results

Calculations for the AR = 4.0 ejector configuration were obtained and compared with experimental
(PT, TT) data at three planes located approximately 3.3, 10.3 and 40.3 inches downstream of the
ejector exit plane. When referenced to the nozzle effective hydraulic radius, these planes are 2.04,
6.4 and 24.9 units from the ejector exit. A more meaningful measure, however, is from the mixer
nozzle exit plane. Since the normalized ejector length is 6.5, the traverse planes are located 8.5,
12.9 and 31.4 units from the nozzle exit. The comparisons with experimental data are shown in
terms of color contours in Figs. 4, 5, and 6. The color bar for the contour plots has been doubly
expanded to highlight the comparisons as the exhaust flow approaches free stream or coflowing
conditions. Note both analysis and experiment predict the rapid decay of exhaust flow from the
lobe region of the nozzle, as well as the slow decay rate of the central core. In fact, the central
core decays in a manner similar to a slot nozzle. Additional calculations obtained beyond the last
experimental traverse plane show that the slot eventually decays to the expected round Jet in the
far field. Computational studies presented in the Appendix and in Ref. [6] have demonstrated that
the central core arises from inability of the induced secondary flow to penetrate the four cross-flow
nozzle centerline stagnation regions. A crossflow stagnation point arises when the nozzle centerline
intersects each mixer lobe bisector. The four lobe nozzle configuration therefore prod-ices the four

5
"bulls-eye" regions seen in Figs 4, 5, and 6. Figure 7 also shows a comparison of the centerline
decay of the normalized total temperature for both the experimental and computational results.
Good agreement is noted for the limited data available. Closer study notes that the rapid initial
decay reaches a plateau region and is eventually passed by the faster farfield decay rate of the more
conventional slot ejector configuration. The presence of the plateau region will be explored in more
detail in the assessment of a full-scale configuration.

4.1 Parametric studies

The comparisons of the PATH PNS analysis with experimental data have been used to validate the
computational approach. The analysis can therefore be now used to explore the dependence of a
number of key physical and numerical parameters on the rate of plume mixing.

Grid density study: One aim of the present study is to perform mixing effectiveness studies on
the exit plume. This required a complete flow field definition, specified sufficiently far downstream
so that the warmest region of the flow was only 10% from the freestream total temperature level
(TT = .05). This level occured approximately 320 diameters downstream of the exit plane. In
an attempt to minimize the PATH code computational run time for such an extended domain,
a solution on a coarser 50 by 50 mesh case Nas calculated and compared to the previous results
(obtained on a 99 by 99 mesh). Figure 8 illustrates that at a distance of 40.3 inches (24.894), the
coarse mesh jet solution already looks like a 'far-field" round jet, rather than the actual slot-like
jet. The enhanced mixing effect arises from the artificial viscosity introduced by the coarse mesh.

Initial turbulence level study: Since no measurements were performed to define the turbu-
lence characteristics of the ejector exhaust flow, analytical models are typically used to provide an
estimate in order to initialize CFD calculations. In the PATH code one typically uses an algebraic
Prandtl mixing length model to extract the initial or ejector exit levels for the turbulence energy
(k) and the turbulent dissipation (e) dependent variables of the two-equation turbulence model. A
more realistic estimate however was obtained by performing an internal flow calculation (from the
mixer nozzle exit to the ejector exit plane), as cited in the Appendix and in Ref. [6]. While the
calculation of the general flow properties is strongly dependent on an empirical model to define
the initial plane of the calculation (nozzle exit profile, pumping level, etc.). the internal flow result
can be used to more realistically set the exit plane turbulence level. Figure 9 shows a comparison
of both types of turbulence intializations at a plane 10.3 inches (6.362) downstream of the ejector
exit. The algebraic model provides too low an estimate of the exit plane turbulence level, resulting
in almost no mixing over the 10.3 inches. In contrast, the internally driven estimate, shown here

6
and in Figure 9, produced the good comparisons with experimental data.

Streamwise vorticity study: The use of convoluted surfaces (mixer lobes) to enhance mixing
levels has been demonstrated in a number of analytical and experimental studies. The reason for
this enhanced mixing has been largely attributed to the generation of streamwise vorticity through
the use of surface convolutions. A frequently raised objection to this concept claims that the
surface convolutions simply introduce additional wetted surface area. A computational study was
therefore performed to examine the impact of streamwise vorticity on jet mixing. Calculations
were conducted for similar flows, differing only in the initial level of streamwise vorticity. One
calculation was initialized with the measured exit streamwise vorticity distribution and one with
zero streamwise vorticity. The results of these calculations are compared on Fig. 10, at a plane
10.3 inches (6.362) downstream of the ejector exit plane. Without the presence of streamwise
vorticty, the plume appears to mix out as three non-interacting slot nozzles, one corresponding
to the mixer nozzle center core and two corresponding to the residual effect of the lobe hot flow
convected within the ejector to the upper and lower ejector walls. In addition to a different mixing
history, the outer "slot nozzles" are about 10% warmer than the measured flow field. The effect of
streamwise vorticity on mixing would be even more substantial if it were eliminated from the mixer
nozzle exit plane rather than from the ejector exit plane. The calculations however do illustrate
that enhanced mixing rates arise from the generation of streamwise vorticity.

5 Full Scale Nozzle Performance Asssessment

Having demonstrated the applicability of the PATH analysis procedure for analyzing mixing pro-
cesses downstream of a model-scale mixer ejector exhaust system, one can now use the PATH code
to assess the performance of a full-scale or flight-type mixer ejector configuration. If a full-scale con-
figuration was simply a length scale L enlargement of the model-scale configuration, one must still
consider what should be the corresponding modifications to the model-scale initial plane (ejector
exit plane) data for a realistic simulation of the full-scale problem. The model-scale calculations
in the previous section were initialized with (PT,TT, and Ql) data. These calculations however
demonstrated that the turbulence parameters k and e data also had to be specified.

While the full-scale or larger geometry implies a larger Reynolds number and probably thinner
boundary layers, the total pressure and temperature data are assumed to remain unchanged in
the geometrical transformation, just redistributed over the enlarged exit plane. The vorticity data
however cannot remain unchanged and must be modified. If one observes that exit flow angle of

1 7
I
the nozzle exhaust remains the same when the mixer nozzle is enlarged by L, then U 2 and U3 also
remain unchanged. The vorticity field, defined below,

- U•. 9U 2 _ model-scale (4)

O'-full-scale - aiO Y 3 L4

is reduced by L. In a similar fashion, one can observe that since the turbulence intensity k is
j proportional to u'/Ui and that the fluctuating scales are typically unaffected by geometrical scale,
k can be considered as unchanged for a full-scale simulation. On the other hand, the turbulent
dissipation c data must be modified. Recalling that if

E k3 / (5)
L

I then ej zz-acale must be decreased by L. This adso implies that the turbulent viscosity AT, given by

AT CpP-k (6)

is correspondingly larger in the full scale problem and therefore the rate of jet mixing is increased.

A geometrical scale factor of approximately 7 was selected for resizing the model-scale nozzle to

that of a typical modern fighter engine scale. Calculations for this configuration, using the PATH
code, have been largely unsuccessful, primarily due to numerical instabilities arising at the shear
layer interface of the ejector and free stream flows. An alternative approach was developed, wherein
full Navier-Stokes (elliptic) code called NASTAR was utilized for the plume calculation. Starting
conditions defined above were used to initialize a series of Navier-Stokes (NS) calculations. The
NASTAR code, developed by Rhie [91, [10] is based on a control-volume pressure-correction scheme
(similar to the "TEACH" code). Since the plume flow field is essentially parabolic, the axial
extent was subdivided into 5 segments in order to accerelerate the numerical convergence. The
computational mesh used was 75 by 75 in the crossplane, clustered near the initial ejector/free
stream shear layer interface. Upstream flow conditions were selected to maintain PT and Po,, with
TT changes (1000', 1250', 15000 F) specified to reflect different engine power settings. Figure 11
illustrates that the normalized centerline stagnation temperature decay, TT, shows little dependence
on a 50% increase in supply temperature.

An interesting point to note about the centerline decay characteristics is the extremely long axial
extent required to reach ambient conditions. A major factor leading to this feature is the plateau
region established. Figure 12 illustrates the crossplane stagnation temperature field at three axial

8
 I locations in the plume, one in the initial decay region, one in the middle of the plateau region,
and one at the end of the plateau region. It is readily evident that the initial rapid decay of the
lobe region to the shroud walls suppresses further mixing in the vertical direction until the side
wall shear layers mix inward to "circular-like" jet pattern. At this point, the plume continues to
actively mix in all direction at an "effective" round jet rate.

6 Conclusions

Albeit that ejector exhaust system pumping levels are dependent on the downstream / exit pressure
field and therefore any analysis should be capable of modeling the upstream interaction effects of
the exit plane ambient static pressure (analysis is mathematically elliptic). The results presented
in the text demonstrate however that:

""an efficient PNS based computational procedure has been developed to analyze internal and
external mixing flows dominated by streamwise vorticity,

"* elliptic effects can be modeled through a combination of empirical flow models and experi-
mentally defined boundary conditions,

"* predicted mixing levels as a function of ejector area ratio (AR) confirm experimentally ob-
I served levels,

"* the importance of grid resolution, and initial turbulence level, i.e. (k,f) on the predicted
evolution of the plume has been identified, and

" the presence on an experimentally observed plateau region in the downstream region of the
plume has been confirmed through analytical predictions, and the causal effect explained.

7 Acknowledgements

I The author would like to thank Lou Chiappetta for his assistance on performing the calculations
cited in the text. In addition, the author appreciates the assistance T. Greg Tillman and Bruce L.

Morin for providing the experimental data used in the computational comparisons.

~ 9
References

[1] Anderson, 0. L., and Barber, T. J., "Three-Dimensional Analysis of Complex Hot Exhaust
Jets," AIAA 88-3705, AIAA/ASME/SIAM/APS 1st National Fluid Dynamics Congress, July
1988.

[2] Patrick, W.P., Barber, T.J., Tillman, T.G., Anderson, O.L., and Paterson, R.W., "Experimen-
tal and Analytical Studies of Jet Mixing from Complex Supersonic Nozzles," UTRC Report
R88-957270-1, Final Report for Contract N00014-85-C-0506, prepared for Department of the
Navy, Naval Air Systems Command, Washington, D.C., June, 1988.

[3] Tillman, T. G., Patrick, W. P., Paterson, R. W., "Enhanced Mixing of Supersonic Jets," AIAA
88-3002, AIAA 24th Joint Propulsion Conference, July, 1988.

[4] Anderson, 0. L., "Calculation of Internal Viscous Flows in Axisymmetric Ducts at Moderate
to High Reynolds Numbers," Intl. J. Computers and Fluids Vol. 8, No. 4, 1980, pp. 391-411.

[5] Anderson, 0. L. and Hankins, G. B., "Development of a Finite Difference Method for 3-D High
Reynolds Number Viscous Internal Flows," Computers in Flow Predictions and Fluid
Dynamics Experiments, Winter Annual Meeting ASME, November 1981, pp. 119.

[6] Barber, T. J., and Anderson, 0. L., "Computational Study of a Supersonic Mixer-Ejector
Exhaust System," J. of Propulsion and Power, Vol. 8, No. 5, 1992, pp. 927-934, also AIAA
91-0126, AIAA 29th Aerospace Sciences Meeting, Reno, NV, 1990.

[7] Tillman, T. G., Paterson, R. W. and Presz, W. M.. " Supersonic Nozzle Mixer Ejector," AIAA
Paper AIAA-89-2925, AIAA 25th Joint Annual Propulsion Conference, July 1989.

[8] Malecki, R., and Lord, W., "Navier-Stokes Analysis of a Lobed Mixer and Nozzle," AIAA
90-0453, AIAA 28th Aerospace Sciences Meeting, Reno, NV, 1990.

[9] Rhie, C. AIAA Journal, Vol. 27, No. 8, 1989, pp. 1017-1018.

[10] Rhie, C. M., and Stowers, S. T., "Navier-Stokes Analysis for High Speed Flows Using a Pressure
Correction Algorithm," AIAA J. of Propulsion and Power, Vol. 4, No. 6, 1988, pp. 564-570.

10
 00

I 0.-C
0

CLN

* 0

LUL

0~

I C,)0
0 0
Fi
0
 04

CO
C-ia

CL

I-)

w
ns-
uLt

ui~I

12~
 C?

CE(D
CC

Lii C-
x
w
D0
I-i

0
w
C., I -m

w C
x

13
 CN
CM

CiN

UE
CL

* ~w 0
2 0i

0-

0~

-J4

L(U
W II

I 14
 I

.80

Analysis

a Experiment

.60

I Ti

I .40

I U

.20
I .o0 .,,
5.0 10.0

x, ft

Figure 7: Centerline Total Temperature Decay

15
 II 04

I~
0

I 0;

00.

I IL

'In
CL~

II z
0

ILL0

0 016
I cl

CL xJ

Iw

IL II .

ww

-J4 0
I-J

1 ~0-

I 1z
III
Ix

0.10

UcJ

* ~CE
6

0
I w

I-I

wo

0
w CC

ww

LL 18
I
I
I
I
I
1 .60

I _ 1000 F
1250 F
- 1500 F

.40

'NN

.00

0. 1200. 2400. 3600.

x, in.

Figure 11:IFull-Scale Centerline Total Temperature Decay

19
 I Cý

CC

a,
*-d

L'U
L0.

LA.

IILU
m 0N

00

LUL
I L:2
00

02
!
8 Appendix: Ejector Internal Flow Analysis

An internal flow analysis capability has been developed using a combination of CFD (PNS) and
empirical data. The computational domain is described in the schematic of the internal portion of
the mixer ejector exhaust system showing the end and side view of the ejector respectively (Figs. 1
and 2). It consists of an exhaust nozzle surrounded by a rectangular shrouded duct. The end view,
shown in Fig. 1, has an outline of the advanced mixer exhaust nozzle surrounded by a rectangular
shroud of width WE and height HE. The side view shown on Fig. 2 shows the advanced mixer
exhaust nozzle exit plane at X/L = -1 surrounded by a shrouded duct of length L. The parameters
for the different test geometries which were calculated are shown on Table 1. The tests conducted
by Presz and Morin [1] included both the slot nozzle of AR = 3.0 and an advanced mixer nozzle
with approximately the same exit area.

The flow conditions for all three experimental programs are summarized on Table 2. The Presz
and Morin [1] experiment was conducted with a cold primary and a static coflow. The Tillman
et al. [2) experiment however was conducted with a hot primary and a coflowing stream having a
forward flight Mach number of 0.1.

Initial conditions for this analysis are required at the primary flow nozzle exit plane X = 0.0 (see
Fig. 2) in order to calculate the flow in the mixer ejector exhaust system. These conditions were
not measured in any of the tests cited above. However, experimental data is available from the tests
conducted by Patrick et al. [2] for same the slot nozzle and the advanced mixer nozzle operating as
free jets. This data consists of normalized total pressure PT, normalized total temperature TT, and
cross flow velocity components U2 , U 3 traverses at the nozzle exit plane. In the report by Patrick
et al. [2], a vorticity distribution was determined by estimating the vorticity distribution until the
velocity components matched the measured data. This distribution was then used in the present
calculations.

Cold Flow Mixer-Ejector Analysis: In our initial validation study, the cold flow mixer-ejector
configuration of Presz and Morin [1] was examined. In their experiments, measurements of total
pressure and Mach number at three stations inside a mixer-ejector and a slot or rectangular nozzle
ejector were obtained. Schematic views of the ejector system is shown in Figs. 1 and 2. measure-
ments were taken at duct/primary area ratios (AR) of 3.0, 4.5. 5.2 as shown in Table I for the flow
conditions given on Table 2.

Comparisons of the measured and calculated total pressure distributions for the slot-nozzle and the
advanced mixer nozzle are shown in Figs. 13 through 17 for the three planes marked in Fig. 2. The

21
I
1
Table 1: Mixer-Ejector Geometry
I
L = 10.5 (in)
Arect = 9.082 (in2 ) Slot Nozzle
AmixeT = 9.164 (in2 ) Mixer Nozzle

(Case Defin. M._ HE [WE AR A.

A/7Ap

Coldslot_[1) 0.28 4.64 6.04 3.1 20

Cold mixer [1] 0.64 4.64 6.04 3.0 1.9

Cold mixer [1] 0.46 6.81 6.01 4.5 3.3
Cold mixer - 7.86 6.01 5.2 4.2
Hot mixer [21] 0.80 4.64 6.04 3.0 1.9
Hot mixer [2] 0.52 6.81 6.01 4.5 3.3
Hot mixer [2] 0.37 7.86 6.01 5.1 4.0

Table 2: Flow Conditions for Mixer-Ejector Cases

Reference Presz & Morin Tillman et al.

M(..0 0.1
PTJ psia 51.60 49.61
PT0. psia 15.23 15.23
TTJ °R 500.00 1460.00
TT. OR 500.00 501.00

slot nozzle predictions shown in Fig. 17 agree remarkably well with the experimental observation.
These observations, when compared to previous isolated slot nozzle experiments and calculations,
demonstrate that the nozzle shear layer growth is largely unaffected by the ejector pumping, re-
sulting in little mixing of the jet flow with the freestream. The color contour comparisons for the
mixer-ejector combination are shown in Figs. 14 through 15. These figures show reasonable agree-
ment with experimental data, matching the mixing process both in structure and level. Note that
the central core feature is retained substantially longer than the lobe features. This is largely due
to the lack of penetration of the induced secondary flow and will be explored in more detail shortly.
A quantitative comparison of the calculated and measured total pressures is shown in Figs. 16 and

22
I
17 for the advanced mixer nozzle (AR = 3.0). The total pressure comparisons show the calculation
to be in generally good agreement with the experimental measurements. These figures show com-
parisons at the initial and exit plane for the slices at the center plane Z = 0.0 and off the center

I plane at Z = 0.15 which indicate quite good agreement as to general level and distribution.

The mixing effectiveness parameters Ema, ET, EH have been calculated for the slot nozzle and mixer
nozzle cases, showing that, for this cold flow condition, Em = ET = EH. Furthermore, the slot
nozzle has a mixing effectiveness of only 34% while the mixer nozzle has an improved mixing
effectiveness of about 60% for the equivalent ejector area ratios.

The development of the secondary flow or streamwise vorticity field (quarter plane only) is shown
in Fig. 18 for the AR = 3.0 and AR = 4.5 mixer ejectors. At station X/L = -1.0, the vorticity
field is similar to the free jet. The plots show the vorticity concentrated along the shear layer
with + or - values aligned to the mixer lobe sidewalls. Note that for the AR = 3.0 mixer ejector,
the tip of the lobes are very close to the upper wall. At X/L = -0.5, the AR = 3.0 ejector has
developed a very strong vortex compared to the AR = 4.5 ejector which is centered between the
outer lobes. By X/L = 0.0, much of this vortex has been dissipated. One should note the low level
of outward flow from the central or core nozzle flow. The presence of crossflow stagnation points,
where the centerline intersects the lobe bisector, can be seen as the mechanism for producing the
four "bulls-eye" regions seen in Figs. 14 and 15. These results differ from the free jet results
reported in Ref. [2] in that the strong vortex does not develop. A more objective measure of the
secondary flow mixing is the absolute circulation given by,

IIFI JI ds (7)

and the average value of the vorticity given by

I= IrFl
A
(8)

which are plotted in Figs. 19 and 20. At X/L = -1.0, the circulation is the same for all area ratios.
It can be seen in Figs. 19 that for the AR = 3.0 where the tip of the lobe is very close to the wall,
the absolute value of the circulation reaches peak values much higher than for the other cases. On
the other hand, from Fig. 20 it can be seen that the average vorticity decreases with increasing area
ratio. While each calculation starts with the same level of circulation, the effect of AR produces a
greater effect on the average vorticity. This behavior was graphically seen in Fig. 18.

23
 Table 3: Hot Flow Mixer Ejector Percent Mixing

* AR1 4.4815.21
Em 60. 54.
ET 55. 50.
EH 56. 51.

Hot Flow Mixer-Ejector Analysis: Additional code validation was obtained in hot flow mixer-
ejector comparisons with the data of Tillman et al. (2]. In their experimental studies, total temper-
ature profiles were measured at the ejector exit plane for duct / primary area ratios (AR) of 3.0,
4.5, 5.1. The specific flow conditions for these cases are given on Table 2. Initial plane conditions
at the primary nozzle exit plane (X/L = -1.0) were obtained from the data contained in a report
by Patrick et al. [2].

Comparisons of the measured and calculated total temperatures at the mixer ejector exit plane are
shown in Figs. 21 through 24. In Figure 21, it can be seen from the experimental data that the
mixing patterns previously seen in the cold flow cases is still present. The effect of the heated jet
has raised increased the total pressure drop and resulted in a partial mixing of the hot exhaust flow.
Again the most prevalent structure at the exit plane is the residual effect of the nozzle core flow. The
effect of increased area ratio (AR) results in greater ejector pumping. This is seen at the exit plane
in Fig. 21 and along the centerline (from nozzle exit to the farfield) in Fig. 22. The axial length has
been normalized by d, the diameter of a round nozzle having an equivalent exit area to the mixer
nozzle. At the ejector exit, the temperature ratio has drop almost 50%. Similar centerline decay
results have been previously reported for mixer-type nozzles, however in this ejector configuration,
the overall exit plane temperature level has been reduced from that observed in the isolated mixer
nozzle case. It should be noted that while one frequently assesses a jet's mixing performance in
terms of it's centerline history, such three-dimensional configurations exhibit flow nonuniformities
off of the centerline. In such cases, the integrated mixing parameters Eqns (11) - (13) or the cross
plane plots represent better measures of mixing. Detailed comparisons of measured and calculated
total temperature are shown in Figs. 23 and 24. In addition, for these cases the degree of mixing
for both area ratios are shown in Table 3. The degree of mixing appears to decrease with increasing
area ratio for the two area ratios indicated in Table 3.

24
 References

[1] Presz, W. M., and Morin, B. L., "Supersonic Mixer-Ejector Program," UTRC Report
UTRC89-19, July 1989.

[2] Tillman, T. G., Paterson, R. W. and Presz, W. M.. " Supersonic Nozzle Mixer Ejector," AIAA
Paper AIAA-89-2925, AIAA 25th Joint Annual Propulsion Conference, July 1989.

3 [3] Tillman, T. G.. Patrick, W. P., Paterson, R. W., "Enhanced Mixing of Supersonic Jets," AIAA
88-3002, AIAA 24th Joint Propulsion Conference, July, 1988.

25
 Experimental Calculated X/L

P - -1.0

1.0 M

0.5 - -0.5

0_______,_____
0 F --
0.0

Figure 13: Pi l)itribztitu~i foi it SupIvrýolii" Slot-Nozzle Ejector: AR 3.0

26
91 10 56.1
 Experimental Calculated X/L

P
T -1.0

1.0

0.5 -0.5

0.0

Figure 14: P 1 Distributions for a .iupersonic Mlixer-E-jector: AR 3.0

27
91-10-56 4
 Experimental Calculated X/L

PTw
-1.0
1.0 ,

0.5 -0.5

- 0.0

Figure 15: P Distribuiion, tbzr Supersonzic Alixer-Ejector: AR - 4.5

28
91-10-56-2
I_

2.00
PNS Exp. Z Loc.
* 0.218
* 0.000

1.20 s o N o a-n

PT

0.40-

-0.40-,
-0.300 -0.150 0.000 0.150 0.300
Y

Figure 16: Measured and Calculated PT In a Supersonic AR = 3.0 Mixer-Ejector at Z = O.OOft

29
 2.00

PNS Exp. Z Loc.

I 0.218
* 0.000

1.20-

FT

0.40-

-0.40-
-0.300 -0.150 0.000 0.150 0.300
Y

Figure 17: Measured and Calculated PT In a Superso: c AR = 3.0 Mixer-Ejector at Z = 0.15ft

30
 XlL =-1.0

- 20,000

X/L = -0.5

-20,000

X/L = 0.0

AR =3.0 AR = 4.5

Figure 18: Comparisonof Strearnwise Vorticitv Fields in Supersonic Mixer-Ejector

91-10-56-5
31
 Experimental Computational

TT

1.0

0.5

AR=5.2

01

AR=4.5
Figure 19: Calcuiaredand Mleasuired T1 Distribution, for Hot Flow Supersounic Mixer-Ejectors

32
91 1 56 -3
i 1.20
PNS Exp. X/L Loc.

*•0.0
I -" -1.0

0.80O

100"
N.
\ /

0.001 , , ,,0.

-0.300 -0.150 0.000 0 150 0.30U

Y

Figure 20: Measured and Calculated Exit Plane TT for Supersonic AR = 4.5 Mixer-Ejector

33
 The higher the AR, the more accurate the experimental and computational data is.

1.20
PNS Exp. X/L Loc.
* 0.0
* -1.0

0.80- 0

TT

* 0
g O /
0 .40

\ N l

U U
0.00, I , , I ,
-0.300 -0.150 0.000 0.150 0.300
y

Figure 21: Meaaured and CalculatedExit Plane TT for Supersonic AR = 5.1 Mixer Ejector

34