Академический Документы
Профессиональный Документы
Культура Документы
(Revised 12/07)
PURDUE UNIVERSITY
GRADUATE SCHOOL
Thesis/Dissertation Acceptance
By DEEPAK THIRUMURTHY
To the best of my knowledge and as understood by the student in the Research Integrity and
Copyright Disclaimer (Graduate School Form 20), this thesis/dissertation adheres to the provisions of
Purdue University’s “Policy on Integrity in Research” and the use of copyrighted material.
PURDUE UNIVERSITY
GRADUATE SCHOOL
Title of Thesis/Dissertation:
DESIGN AND ANALYSIS OF NOISE SUPPRESSION EXHAUST NOZZLE SYSTEMS
I certify that in the preparation of this thesis, I have observed the provisions of Purdue University
Executive Memorandum No. C-22, September 6, 1991, Policy on Integrity in Research.*
Further, I certify that this work is free of plagiarism and all materials appearing in this
thesis/dissertation have been properly quoted and attributed.
I certify that all copyrighted material incorporated into this thesis/dissertation is in compliance with
the United States’ copyright law and that I have received written permission from the copyright
owners for my use of their work, which is beyond the scope of the law. I agree to indemnify and save
harmless Purdue University from any and all claims that may be asserted or that may arise from any
copyright violation.
DEEPAK THIRUMURTHY
______________________________________
Printed Name and Signature of Candidate
02/22/2010
______________________________________
Date (month/day/year)
*Located at http://www.purdue.edu/policies/pages/teach_res_outreach/c_22.html
DESIGN AND ANALYSIS OF
A Thesis
of
Purdue University
by
Deepak Thirumurthy
of
May 2010
Purdue University
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI 1479646
Copyright 2010 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
ii
ACKNOWLEDGMENTS
TABLE OF CONTENTS
Page
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
NOMENCLATURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Supersonic Civil Transport . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Challenges Associated with Supersonic Transport . . . . . . . . . . 4
1.3 Noise Suppression Propulsion System . . . . . . . . . . . . . . . . . 5
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Noise Suppression and Design Methodology . . . . . . . . . . . . . . . . 8
2.1 Jet Noise - A Classical Problem . . . . . . . . . . . . . . . . . . . . 8
2.2 Noise Suppression Exhaust Nozzles . . . . . . . . . . . . . . . . . . 10
2.2.1 Ejector Nozzle . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Chevrons - Passive Mixers . . . . . . . . . . . . . . . . . . . 15
2.3 Computational Techniques . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.2 Turbulence Modeling for Jet Flows . . . . . . . . . . . . . . 20
2.4 Design Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3 Chevron Nozzles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Three-Stream Separate-Flow Axisymmetric Plug Nozzle (3BB) . . . 29
3.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.2 Geometry and Mesh Generation . . . . . . . . . . . . . . . . 30
3.3.3 Boundary Conditions and CFD Methodology . . . . . . . . . 31
3.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Three-Stream Separate-Flow Chevron Nozzle (3A12B) . . . . . . . 42
3.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.2 Geometry and Mesh Generation . . . . . . . . . . . . . . . . 42
3.4.3 Boundary Conditions and CFD Methodology . . . . . . . . . 45
3.4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
v
Page
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Ejector Nozzles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3 2-D Ejector Nozzle Test Case . . . . . . . . . . . . . . . . . . . . . 60
4.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3.2 Geometry and Mesh Generation . . . . . . . . . . . . . . . . 60
4.3.3 Boundary Conditions and Numerical Computation . . . . . 62
4.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 3-D Ejector Nozzle with Clamshell Doors . . . . . . . . . . . . . . . 68
4.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.4.2 Experimental Investigation . . . . . . . . . . . . . . . . . . . 68
4.4.3 Nozzle Design and CAD Geometry . . . . . . . . . . . . . . 69
4.4.4 Grid Generation . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.4.5 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . 74
4.4.6 Numerical Computation . . . . . . . . . . . . . . . . . . . . 76
4.4.7 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5 3-D Ejector Nozzles
with Clamshell Doors and Chevrons . . . . . . . . . . . . . . . . . . . . . 108
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.3 Ejector Flow with Chevrons . . . . . . . . . . . . . . . . . . . . . . 109
5.4 Nozzle Design and CAD Geometry . . . . . . . . . . . . . . . . . . 110
5.5 Computational Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.6 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.7 Numerical Computation . . . . . . . . . . . . . . . . . . . . . . . . 117
5.8 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.8.1 Design I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.8.2 Design II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.8.3 Discussion on the centerline statistics . . . . . . . . . . . . . 120
5.8.4 Effect on the ejector mass flow . . . . . . . . . . . . . . . . . 120
5.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.10 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . 129
LIST OF REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
vi
LIST OF TABLES
Table Page
2.1 Thies and Tam’s k- turbulence model constants [39]. . . . . . . . . . . 24
3.1 Boundary conditions for the CFD simulation of the three-stream separate-
flow axisymmetric plug nozzle (3BB) [20]. . . . . . . . . . . . . . . . . 32
3.2 Boundary conditions for the CFD simulation of the three-stream separate-
flow chevron nozzle (3A12B) [24]. . . . . . . . . . . . . . . . . . . . . . 45
4.1 2-D ejector nozzle boundary conditions [46]. . . . . . . . . . . . . . . . 62
4.2 Calculation of the corrected inlet axial velocity magnitude for the CFD
simulations using the minimization of the RMS difference. . . . . . . . 77
5.1 Dimensions of the chevron on the 3-D ejector nozzle with clamshell doors
for Design I and Design II. . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.2 Boundary conditions for the CFD simulation of the ejector nozzle with
clamshell doors and chevrons. . . . . . . . . . . . . . . . . . . . . . . . 116
5.3 The effect of chevrons on the ejector mass flow. . . . . . . . . . . . . . 121
vii
LIST OF FIGURES
Figure Page
1.1 History of the commercial and military supersonic transport aircraft and
its progress. (Reproduced courtesy of P. Henne [3].) . . . . . . . . . . . 2
1.2 The noise distribution from the individual components of the airbreathing
jet engine propulsion system [6]. . . . . . . . . . . . . . . . . . . . . . . 4
1.3 A schematic representation of the 3-D ejector nozzle with clamshell doors [7]. 5
2.1 Jet noise as a result of the shear layer mixing phenomenon. . . . . . . . 9
2.2 Requirements for the pressure ratio and the area ratio as Mach number
increases [16]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Operational modes of the ejector nozzle with clamshell doors (1) Subsonic
take-off, (2) Supersonic cruise and (3) Subsonic approach. . . . . . . . 14
2.4 LS /∆ = Optimum attached free mixing layer. (Reproduced courtesy of J.
Der Jr. [17].) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Three-stream separate-flow nozzle with chevrons on the core and fan noz-
zle [18]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6 Mixing of two streams of the chevron nozzle and streamwise vortex for-
mation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.7 Methodology for the jet engine exhaust nozzle design and analysis. . . 27
3.1 The CAD geometry of the three-stream separate-flow axisymmetric plug
nozzle [20]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 The computational mesh for the three-stream separate-flow axisymmetric
plug nozzle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 3BB axial velocity magnitude contour plot corresponding to PIV experi-
ments on the Z=0 plane [20]. . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 3BB axial velocity magnitude contour plot corresponding to the standard
k- turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 37
3.5 3BB axial velocity magnitude contour plot corresponding to the realizable
k- turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 37
viii
Figure Page
3.6 3BB axial velocity magnitude contour plot corresponding to the standard
k-ω turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 38
3.7 3BB axial velocity magnitude contour plot corresponding to the k-ω shear
stress transport turbulence model on the Z=0 plane. . . . . . . . . . . 38
3.8 3BB axial velocity magnitude contour plot corresponding to the Reynolds
stress turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . 38
3.9 3BB turbulent kinetic energy contour plot corresponding to PIV experi-
ments on the Z=0 plane [20]. . . . . . . . . . . . . . . . . . . . . . . . 39
3.10 3BB turbulent kinetic energy contour plot corresponding to the standard
k- turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 39
3.11 3BB turbulent kinetic energy contour plot corresponding to the realizable
k- turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 39
3.12 3BB turbulent kinetic energy contour plot corresponding to the standard
k-ω turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . . 40
3.13 3BB turbulent kinetic energy contour plot corresponding to the k-ω shear
stress transport turbulence model on the Z=0 plane. . . . . . . . . . . 40
3.14 3BB turbulent kinetic energy contour plot corresponding to the Reynolds
stress turbulence model on the Z=0 plane. . . . . . . . . . . . . . . . . 40
3.15 Centerline axial velocity profiles for different turbulence models and com-
parison with the experimental result. . . . . . . . . . . . . . . . . . . . 41
3.16 Centerline total temperature profiles for different turbulence models and
comparison with the experimental result. . . . . . . . . . . . . . . . . . 41
3.17 Dimensions for the design of alternating chevrons [21]. . . . . . . . . . 43
3.18 The CAD geometry of the three-stream separate-flow chevron nozzle [24]. 44
3.19 The computational mesh for the three-stream separate-flow chevron noz-
zle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.20 3A12B axial velocity magnitude contour plot corresponding to PIV exper-
iments on the Z=0 and inward-facing chevron mid-plane [24]. . . . . . 49
3.21 3A12B axial velocity magnitude contour plot corresponding to WIND-
CFD results on the Z=0 and inward-facing chevron mid-plane [24]. . . 49
3.22 3A12B axial velocity magnitude contour plot corresponding to the k-ω
SST turbulence model on the Z=0 and inward-facing chevron mid-plane. 49
ix
Figure Page
3.23 3A12B axial velocity magnitude contour plot corresponding to the stan-
dard k- turbulence model on the Z=0 and inward-facing chevron mid-
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.24 3A12B axial velocity magnitude contour plot corresponding to the real-
izable k- turbulence model on the Z=0 and inward-facing chevron mid-
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.25 3A12B axial velocity magnitude contour plot corresponding to Thies and
Tam’s k- turbulence model on the Z=0 and inward-facing chevron mid-
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.26 3A12B axial velocity magnitude contour plot corresponding to PIV exper-
iments on the outward-facing chevron mid-plane [24]. . . . . . . . . . . 51
3.27 3A12B axial velocity magnitude contour plot corresponding to WIND-
CFD results on the outward-facing chevron mid-plane [24]. . . . . . . . 51
3.28 3A12B axial velocity magnitude contour plot corresponding to the k-ω
SST turbulence model on the outward-facing chevron mid-plane. . . . . 51
3.29 3A12B axial velocity magnitude contour plot corresponding to the stan-
dard k- turbulence model on the outward-facing chevron mid-plane. . 52
3.30 3A12B axial velocity magnitude contour plot corresponding to the realiz-
able k- turbulence model on the outward-facing chevron mid-plane. . . 52
3.31 3A12B axial velocity magnitude contour plot corresponding to Thies and
Tam’s k- turbulence model on the outward-facing chevron mid plane. . 52
3.32 3A12B turbulent kinetic energy contour plot corresponding to PIV exper-
iments on the Z=0 and inward-facing chevron mid-plane [24]. . . . . . 53
3.33 3A12B turbulent kinetic energy contour plot corresponding to WIND-CFD
results on the Z=0 and inward-facing chevron mid-plane [24]. . . . . . 53
3.34 3A12B turbulent kinetic energy contour plot corresponding to the k-ω SST
turbulence model on the Z=0 and inward-facing chevron mid-plane. . . 53
3.35 3A12B turbulent kinetic energy contour plot corresponding to the standard
k- turbulence model on the Z=0 and inward-facing chevron mid-plane. 54
3.36 3A12B turbulent kinetic energy contour plot corresponding to the real-
izable k- turbulence model on the Z=0 and inward-facing chevron mid-
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.37 3A12B turbulent kinetic energy contour plot corresponding to Thies and
Tam’s k- turbulence model on the Z=0 and inward-facing chevron mid-
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
x
Figure Page
3.38 3A12B turbulent kinetic energy contour plot corresponding to PIV exper-
iments on the outward-facing chevron mid-plane [24]. . . . . . . . . . . 55
3.39 3A12B turbulent kinetic energy contour plot corresponding to WIND-CFD
results on the outward-facing chevron mid-plane [24]. . . . . . . . . . . 55
3.40 3A12B turbulent kinetic energy contour plot corresponding to the k-ω SST
turbulence model on the outward-facing chevron mid-plane. . . . . . . 55
3.41 3A12B turbulent kinetic energy contour plot corresponding to the standard
k- turbulence model on the outward-facing chevron mid-plane. . . . . 56
3.42 3A12B turbulent kinetic energy contour plot corresponding to the realiz-
able k- turbulence model on the outward-facing chevron mid-plane. . . 56
3.43 3A12B turbulent kinetic energy contour plot corresponding to Thies and
Tam’s k- turbulence model on the outward-facing chevron mid-plane. . 56
3.44 Centerline axial velocity profiles for different turbulence models and com-
parison with the experimental result. . . . . . . . . . . . . . . . . . . . 57
3.45 Centerline total temperature profiles for different turbulence models and
comparison with the experimental result. . . . . . . . . . . . . . . . . . 57
4.1 Experimental setup for the 2-D ejector nozzle. (Reproduced courtesy
of [46].) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Computational mesh for the 2-D ejector nozzle. . . . . . . . . . . . . . 62
4.3 Mach number contour plot of the 2-D ejector nozzle corresponding to the
k-ω SST turbulence model. . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4 Mach number contour plot of the 2-D ejector nozzle corresponding to the
Spalart-Allmaras turbulence model. . . . . . . . . . . . . . . . . . . . . 63
4.5 2-D ejector nozzle axial velocity profile at X=3.0 in. . . . . . . . . . . 65
4.6 2-D ejector nozzle axial velocity profile at X=5.0 in. . . . . . . . . . . 65
4.7 2-D ejector nozzle axial velocity profile at X=7.0 in. . . . . . . . . . . 66
4.8 2-D ejector nozzle axial velocity profile at X=10.5 in. . . . . . . . . . . 66
4.9 2-D ejector nozzle total temperature profile at X=3.0 in. . . . . . . . . 67
4.10 2-D ejector nozzle total temperature profile at X=10.5 in. . . . . . . . 67
4.11 CAD model of the 3-D ejector nozzle without clamshell doors [44]. . . . 69
4.12 CAD model of the 3-D ejector nozzle with clamshell doors [44]. . . . . 69
xi
Figure Page
4.13 Computational mesh for nozzle walls (a) 3-D ejector nozzle without clamshell
doors (Grid I), and (b) 3-D ejector nozzle with clamshell doors (Grid II). 70
4.14 Computational mesh (Grid II) for the entire flow domain of the 3-D ejec-
tor nozzle with clamshell doors for the CFD simulation at experimental
conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.15 Computational mesh (Grid III) for the entire flow domain of the 3-D ejec-
tor nozzle with clamshell doors for the CFD simulation at take-off condi-
tions (higher NPR). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.16 Extent of the computational domain for the 3-D ejector nozzle with clamshell
doors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.17 Schematic representation of experimental boundary conditions (Simula-
tion I) and their numerical values. . . . . . . . . . . . . . . . . . . . . . 74
4.18 Schematic representation of take-off boundary conditions (Simulation II)
and their numerical values. . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.19 Experimental survey of the plenum chamber in the absence of the noz-
zle showing the nonuniformity involved in the axial velocity magnitude
distribution [44]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.20 RMS difference distribution in the axial velocity magnitude at X/DEQ =1.0
downstream of the nozzle throat. . . . . . . . . . . . . . . . . . . . . . 77
4.21 Contour plot of the normalized axial velocity magnitude on the Z=0 plane
for the 3-D ejector nozzle with clamshell doors corresponding to the k-ω
shear stress transport turbulence model. . . . . . . . . . . . . . . . . . 80
4.22 Contour plot of the normalized axial velocity magnitude on the Z=0 plane
for the 3-D ejector nozzle with clamshell doors corresponding to the real-
izable k- turbulence model with Thies and Tam’s model constants for jet
flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.23 Experimental U/UP L contour plot at X/DEQ =1.0 from the throat [44]. 82
4.24 Computational U/UP L contour plot at X/DEQ =1.0 from the throat. . 82
4.25 Experimental U/UP L contour plot at X/DEQ =1.5 from the throat [44]. 83
4.26 Computational U/UP L contour plot at X/DEQ =1.5 from the throat. . 83
4.27 Experimental U/UP L contour plot at X/DEQ =2.0 from the throat [44]. 84
4.28 Computational U/UP L contour plot at X/DEQ =2.0 from the throat. . 84
4.29 Experimental U/UP L contour plot at X/DEQ =3.0 from the throat [44]. 85
xii
Figure Page
4.30 Computational U/UP L contour plot at X/DEQ =3.0 from the throat. . 85
4.31 Normalized axial velocity profile at X/DEQ =1.0 and on the Z=0 plane. 86
4.32 Normalized axial velocity profile at X/DEQ =1.0 and on the Y =0 plane. 86
4.33 Normalized axial velocity profile at X/DEQ =1.5 and on the Z=0 plane. 87
4.34 Normalized axial velocity profile at X/DEQ =1.5 and on the Y =0 plane. 87
4.35 Normalized axial velocity profile at X/DEQ =2.0 and on the Z=0 plane. 88
4.36 Normalized axial velocity profile at X/DEQ =2.0 and on the Y =0 plane. 88
4.37 Normalized axial velocity profile at X/DEQ =3.0 and on the Z=0 plane. 89
4.38 Normalized axial velocity profile at X/DEQ =3.0 and on the Y =0 plane. 89
4.39 Computational U/UP L contour plot at X/DEQ =3.0 downstream of the
nozzle throat corresponding to the realizable k- turbulence model. . . 90
4.40 Computational U/UP L contour plot at X/DEQ =3.0 downstream of the
nozzle throat corresponding to the standard k- turbulence model. . . . 90
4.41 Comparison of centerline axial velocity profiles among experiments, the
k-ω SST, the realizable k- and the standard k- turbulence models. . . 91
4.42 Comparison of axial velocity profiles at X/DEQ =3.0 and on the Z=0 plane
between the k-ω SST and the realizable k- turbulence model. . . . . . 92
4.43 Comparison of axial velocity profiles at X/DEQ =3.0 and on the Y =0 plane
between the k-ω SST and the realizable k- turbulence model. . . . . . 92
4.44 Contour plot of the normalized axial velocity magnitude of the 3-D ejector
nozzle with clamshell doors on the Z=0 plane with streamlines showing
the flow separation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.45 Experimental U/UP L contour plot at X/DEQ =0.42 from the throat [44]. 95
4.46 Computational U/UP L contour plot at X/DEQ =0.42 from the throat. . 95
4.47 Experimental U/UP L contour plot at X/DEQ =1.0 from the throat [44]. 96
4.48 Computational U/UP L contour plot at X/DEQ =1.0 from the throat. . 96
4.49 Experimental U/UP L contour plot at X/DEQ =1.5 from the throat [44]. 97
4.50 Computational U/UP L contour plot at X/DEQ =1.5 from the throat. . 97
4.51 Experimental U/UP L contour plot at X/DEQ =2.0 from the throat [44]. 98
4.52 Computational U/UP L contour plot at X/DEQ =2.0 from the throat. . 98
xiii
Figure Page
4.53 Experimental U/UP L contour plot at X/DEQ =3.0 from the throat [44]. 99
4.54 Computational U/UP L contour plot at X/DEQ =3.0 from the throat. . 99
4.55 Normalized axial velocity profile at X/DEQ =0.42 and on the Z=0 plane. 100
4.56 Normalized axial velocity profile at X/DEQ =0.42 and on the Y =0 plane. 100
4.57 Normalized axial velocity profile at X/DEQ =1.0 and on the Z=0 plane. 101
4.58 Normalized axial velocity profile at X/DEQ =1.0 and on the Y =0 plane. 101
4.59 Normalized axial velocity profile at X/DEQ =1.5 and on the Z=0 plane. 102
4.60 Normalized axial velocity profile at X/DEQ =1.5 and on the Y =0 plane. 102
4.61 Normalized axial velocity profile at X/DEQ =2.0 and on the Z=0 plane. 103
4.62 Normalized axial velocity profile at X/DEQ =2.0 and on the Y =0 plane. 103
4.63 Normalized axial velocity profile at X/DEQ =3.0 and on the Z=0 plane. 104
4.64 Normalized axial velocity profile at X/DEQ =3.0 and on the Y =0 plane. 104
4.65 Mach number contour plot on the Z=0 symmetry plane of the 3-D ejector
nozzle with clamshell doors at take-off conditions. . . . . . . . . . . . . 105
4.66 Mach number contour plot at X/DEQ =0.5 plane downstream of the 3-D
ejector nozzle throat at take-off conditions. . . . . . . . . . . . . . . . . 106
5.1 The phenomenon of the ejector flow with chevrons. . . . . . . . . . . . 109
5.2 CAD geometry of the ejector nozzle with clamshells and chevrons, Design
I - X-section at the throat. . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.3 CAD geometry of the ejector nozzle with clamshells and chevrons, Design
II - X-section at the throat. . . . . . . . . . . . . . . . . . . . . . . . . 112
5.4 CAD geometry of the ejector nozzle with clamshells and chevrons, Design
I - Isometric view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.5 CAD geometry of the ejector nozzle with clamshells and chevrons, Design
II - Isometric view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.6 Computational mesh for ejector nozzle walls and chevrons - Design I. . 114
5.7 Computational mesh for ejector nozzle walls and chevrons - Design II. . 115
5.8 Contours of Mach number and turbulent kinetic energy corresponding to
the CFD simulation of the baseline nozzle. . . . . . . . . . . . . . . . . 122
5.9 Contours of Mach number and turbulent kinetic energy corresponding to
the CFD simulation of the chevron nozzle (Design II). . . . . . . . . . 122
xiv
Figure Page
5.10 Mach number contour plot Z=0 symmetry plane for the ejector nozzle
without chevrons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.11 Mach number contours on Y Z-plane at X/DEQ = 0.5 for the ejector nozzle
without chevrons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.12 Mach number contours on the plane in between two chevrons for the ejector
nozzle with chevrons - Design I. . . . . . . . . . . . . . . . . . . . . . . 124
5.13 Mach number contours on Z=0 symmetry plane for the ejector nozzle with
chevrons - Design I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.14 Mach number contours on Y Z-plane at X/DEQ = 0.5 for the ejector nozzle
with chevrons - Design I. . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.15 Mach number contours on Z=0 symmetry plane for the ejector nozzle with
chevrons - Design II. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.16 Mach number contours on the plane in between two chevrons for the ejector
nozzle with chevrons - Design II. . . . . . . . . . . . . . . . . . . . . . 126
5.17 Mach number contours on Y Z-plane at X/DEQ = 0.5 for the ejector nozzle
with chevrons - Design II. . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.18 Centerline axial velocity profiles corresponding to the ejector nozzle with
and without chevrons. . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.19 Centerline total temperature profiles corresponding to the ejector nozzle
with and without chevrons. . . . . . . . . . . . . . . . . . . . . . . . . 127
xv
ABBREVIATIONS
NOMENCLATURE
A Area m2
A∗ Nozzle throat area m2
Ae Nozzle exit area m2
D Diameter m
DC Control diameter of the nozzle m
DEQ Equivalent diameter of the nozzle throat m
Df an Diameter of the fan nozzle m
H Semi-height of the 2-D ejector nozzle m
I Turbulent intensity
LS Length of the mixing duct in ejectors m
M Mach number
Me Exit Mach number
Mthroat Throat Mach number
P◦ Total pressure P a or atm
Ps Static pressure P a or atm
ReD Reynolds number based on the nozzle diameter
T◦ Total temperature K or R
Ts Static temperature K or R
U Axial velocity m/s or f t/sec
UP L Axial velocity inside the plenum chamber m/s or f t/sec
V Velocity magnitude m/s or f t/sec
k Turbulent kinetic energy m2 /s2
ṁej Secondary mass flow entrained through the ejector slot kg/m3
ṁin Primary nozzle mass flow kg/m3
y Wall normal distance m
xviii
ABSTRACT
Thirumurthy, Deepak M.S.A.A., Purdue University, May 2010. Design and Analysis
of Noise Suppression Exhaust Nozzle Systems. Major Professors: Anastasios S.
Lyrintzis and Gregory A. Blaisdell.
The exhaust nozzle is an integral part of a jet engine and critical to its overall
system performance. Challenges associated with the design and manufacturing of
an exhaust nozzle become greater as the cruise speed of the aircraft increases. The
exhaust nozzle of a supersonic cruise aircraft requires additional capabilities such
as variable throat and exit area, noise suppression, and reverse thrust. The present
work is an effort to study the design and analysis of jet engine exhaust nozzle systems
such as the axisymmetric plug nozzle, the chevron nozzle and the ejector nozzle with
clamshells.
High-bypass-ratio jet engines with two or more flow streams have superior noise
suppressing and thrust characteristics. Much research has been done in the past to
study and understand the flow physics of these engines. In the present work a com-
putational fluid dynamics-based approach was used to study the jet engine exhaust
nozzle systems. First, a computer-aided-design model of a three-stream separate-flow
axisymmetric plug nozzle was created and axisymmetric flow simulations were per-
formed to study the flow field. The mean flow and turbulent kinetic energy fields
were compared with the particle image velocimetry results available in the literature.
Next, computational fluid dynamics was used to study the performance of passive
chevron mixers in enhancing the turbulent mixing. Three-dimensional calculations
were carried out to study the effect of enhanced mixing on the mean velocity and
turbulent kinetic energy flow fields. Different turbulence models were used to study
their performance in predicting chevron-based jet flows.
xx
Gas turbine engine manufacturer Rolls-Royce, and business class aircraft man-
ufacturer Gulfstream Aerospace Corporation, are collaborating on the development
of technologies for a supersonic jet. As part of this collaborative research and de-
velopment program, an ejector nozzle with clamshell doors, similar to that on an
Olympus-593 engine, which powered the Concorde aircraft, was designed and tested.
The ejector nozzle offers additional advantages such as thrust augmentation and noise
suppression.
Numerical simulations of this ejector nozzle with clamshell doors at 11.5◦ clamshell
angle and without clamshell doors were performed as part of the validation task. Mean
flow fields were predicted for low subsonic experimental conditions and compared with
the experimental data. Flow separation and recirculation zones were encountered near
the inner surface of clamshell doors. Simulations at higher nozzle pressure ratios were
also performed to simulate actual flight conditions. Flow separation prevailed at this
condition as well.
The existing new supersonic noise suppression exhaust nozzle design was improved
by the addition of chevrons and its flow field was analyzed using computational fluid
dynamics. The jet engine exhaust nozzle consisted of three-dimensional ejectors in
the form of clamshell doors and chevrons as passive mixers. Chevrons were placed
in the ejector slot to introduce streamwise vorticity and enhance mixing. It was
observed that the flow separation zone was almost removed and an improvement in
the ejector performance was obtained. Computational simulations corresponded to
take-off conditions with a nozzle pressure ratio of 1.7 and freestream Mach number
of 0.3.
1
1. Introduction
Mankind has witnessed a remarkable change in the speed of transporting goods and
people. During the 19th century the transportation method changed from horse-
powered carts traveling at 10 kmph to high speed trains transporting passengers and
cargo at 100 kmph. Speed has no limits as evidenced by the advent of subsonic
airplanes of the 20th century capable of flying at 1000 kmph [1]. Mankind was
skeptical of flying at a speed greater than the speed of sound until October 1947,
when United States Air Force Capt. Charles Yeager crossed the sound barrier and
reached Mach 1.02 in his XS-1 experimental aircraft [2].
This fascinating and challenging supersonic flight motivated many aerospace or-
ganizations to start programs related to the design of supersonic cruise aircraft and
develop related technologies. On November 29, 1962, the Concorde project, an Anglo-
French partnership, was launched and remains one of two supersonic cruise passenger
aircraft that traveled at speeds exceeding 2000 kmph, more than twice the speed of
sound. As airport regulations became more stringent, the Concorde failed to meet
requirements for performance, operating economics, development cost and environ-
mental acceptance. British Airways and Air France ended their Concorde service in
2003.
The Tupolev Tu-144 supersonic transport aircraft was a Soviet Union (now Russia)
effort to make supersonic civil transport a viable option. The project started two
years later than the Concorde. Although the Tu-144 was technically comparable
to the Concorde with a cruise Mach number of 2.5, the Tu-144 lacked a passenger
market within the Soviet Union and service was halted after only about 100 scheduled
flights. Initial plane crashes and high maintenance cost led the Soviet Union to cease
the program in 1983.
2
Figure 1.1 shows various supersonic aircraft, both military and civil, that were
designed with the motivation of supersonic cruise. No supersonic civil transport
aircraft has been produced since the end of the Concorde program. However, speed
continues to hold attraction for the civil aviation world, particularly in the executive
and corporate markets. Hence, the supersonic business jet (SSBJ) started gaining
traction despite the rising environmental concerns [2].
3.5
Military
SR−71 Commercial
3 X−B70
B−2707 US−Never Built
2.5 F−104
Cruise Mach Number
Tupolov−144 Russia
1.5 F−100
30 Years with no New
Supersonic Civil Transport
X−1
1
55 Years of Subsonic Civil Jet Transports
0.5
0
1940 1950 1960 1970 1980 1990 2000 2010 2020
Year
The technical challenges associated with supersonic flight such as sonic boom,
airport community noise, engine emissions and developmental cost are reduced with
a smaller business jet compared to a 100-seater passenger aircraft, such as the Con-
corde. Several programs were initiated to develop a viable supersonic business jet
3
Aircraft noise, unwanted sound from the aircraft, is generated when the airflow
over the aircraft structure or its propulsion system causes fluctuating pressure distur-
bances that propagate to an observer either sitting in the aircraft or on the ground.
For a subsonic cruise aircraft, these pressure fluctuations are significant during take-
off and approach, when the landing gear and high-lift devices (slats and flaps) are
deployed. In addition, if the aircraft is capable of supersonic cruise, additional noise
sources in the form of sonic boom and shock-generated jet noise add to the overall
pressure fluctuations causing severe annoyance to the community beneath the aircraft.
As previously mentioned, the effect of aircraft noise is more significant during
take-off and approach phases. Several sources can combine to increase the overall
sound pressure level. Of all the aircraft noise sources, the major noise sources during
take-off and approach are noise from the powerplant and airframe noise. Powerplant
noise is much more complex than airframe noise.
Compressor Fan
Turbine & Core Compressor
Shock Jet Turbine & Core Jet
Most of the components of a typical gas turbine engine contribute to the overall
powerplant noise, as shown in Figure 1.2. Pressure fluctuations in the airflow through
5
the propulsion system due to the fan, compressors, turbines, mixing of hot and cold
flows, nozzle and exhaust jet are the sources of powerplant noise. Based on the type
of spectral distribution, these noises can be classified as broadband noise and discrete
tones. Detailed discussion on various powerplant noise sources, except jet noise, is
outside the scope of the present work.
The performance of the exhaust nozzle is critical to the overall system performance
in the sense that it produces the required thrust efficiently during different phases
of the flight, such as subsonic take-off, transonic acceleration, supersonic cruise and
subsonic approach. In addition to the performance, noise associated with the high
speed exhaust jet is a concern in modern supersonic cruise nozzles. The aircraft
jet engine must comply with the federal aviation regulation (FAR) Stage IV noise
regulations during all the phases of flight. This poses additional requirements in the
design and performance of jet engine exhaust systems. Figure 1.3 shows a schematic
representation of a variable cycle engine with an ejector nozzle in the form of clamshell
doors.
During the past 50 years, several high speed propulsion systems have been designed
to address these challenges. Research and development programs such as SST, SCAR,
HSR and FA were initiatives to address problems related to supersonic civil trans-
port. During these programs, special emphasis was given to the jet engine exhaust
system design. Many noise suppression approaches were studied and combinations
of technologies were formulated into acoustically effective, optimum configurations
based on quantitative analyses [8]. At the supersonic cruise point, the lift to drag
ratio (L/D) of the aircraft is low and the specific fuel consumption is high relative
to subsonic jetliners. The aircraft payload weight thus becomes highly sensitive to
the nozzle efficiency. For example, the Concorde, at the cruise speed of Mach 2.2, a
1% decrease in nozzle performance was estimated to be equivalent to an 8% loss in
weight [9].
1.4 Objectives
The aim of this thesis was to study the design and analysis of noise suppression ex-
haust nozzle systems for a business class supersonic transport. A Reynolds-averaged
Navier-Stokes (RANS) equations-based computational fluid dynamics (CFD) method-
ology was followed for the analysis of the design. First, the design of a three-stream
separate-flow axisymmetric plug nozzle was studied and its CFD simulation was per-
formed. The mean flow and turbulent kinetic energy flow fields were compared to the
experimental data.
Second, a detailed computational study of the design and analysis of the chevron
nozzle was carried out. A CAD geometry of the nozzle was created and its flow field
was studied using a RANS solver. As a validation task, CFD simulations of the 3-D
ejector nozzle with and without clamshell doors were performed and the results were
compared to the experimental data. A zone of flow separation was found at the inner
surface of clamshell doors. Hence, the design of the ejector nozzle with clamshell
doors was improved by the introduction of chevrons. Chevrons introduce streamwise
7
vortices, which enhance the mixing between the nozzle flow and the atmospheric air.
A detailed CAD design and CFD analysis of this new nozzle design was performed
and documented in this thesis. It was observed that the extent of flow separation was
greatly reduced by the application of chevrons.
The organization of this thesis is as follows. A detailed discussion on the literature
survey of noise suppression exhaust nozzles, jet noise sources and the turbulence
models available in the flow solver are presented in Chapter 2. Chapter 3 covers the
CFD simulation of a baseline three-stream separate-flow axisymmetric plug nozzle
and a chevron nozzle. A comparison between the computations and experiments was
carried out to study the performance of various turbulence models. CFD simulations
of the 3-D ejector nozzle with and without clamshell doors are presented in Chapter 4.
The design and computational analysis of the 3-D ejector nozzle with clamshell doors
and chevron mixers are presented in Chapter 5. Chapter 6 summarizes the key
findings of this work and recommendations for future study.
8
In airbreathing propulsion, the main purpose of the exhaust nozzle is to increase the
kinetic energy of the exhaust gases by increasing the exit jet velocity while main-
taining the back pressure for the operation of compressors and turbines. Since the
pressure ratio across the nozzle is the driving force for the operation of any nozzle,
an ideal nozzle should closely match the exit and atmospheric pressures, and thereby
maximize the thrust for a given jet engine. The net thrust is more sensitive to noz-
zle performance than any other engine component. For this reason, it is extremely
important to obtain the highest possible nozzle performance with consideration of
nozzle cost, weight, complexity, reliability and maintainability.
The propulsive efficiency of any jet engine is directly related to the performance
of the exhaust system. Complexities involved in the nozzle design increase for high
speed transport and military aircraft. For supersonic transport, additional consider-
ations such as jet noise, shock-associated noise and better mixing of core and bypass
flow govern the exhaust nozzle design. Afterburner operation, thrust reversal, thrust
vector control and minimization of infrared signature are some of the exhaust noz-
zle design challenges associated with military applications. Design motivations other
than jet noise and better mixing of exhaust streams are beyond the scope of the
present study and will not be discussed.
One of the critical challenges associated with the performance of the supersonic
cruise propulsion system during take-off conditions is the aerodynamic noise associ-
ated with its jet exhaust. This is referred to as Jet Noise. The three primary sources
of jet noise are as follows [10]:
9
3. Radiation from the turbulent jet convected at supersonic velocity relative to the
freestream.
Figure 2.1. Jet noise as a result of the shear layer mixing phenomenon.
Figure 2.1 shows the distribution of the noise sources associated with the mixing
of primary and secondary jet exhaust streams. In the case of supersonic exhaust
streams, the jet noise is comprised of jet-mixing noise (broadband noise) and shock-
associated noise. Early theoretical work of Lighthill [11], shows that fluctuations
in the shear stress, as a result of the mixing process behind the nozzle, generate
broadband noise. This theory postulates that the acoustic power varies according to
the eighth power of the exhaust velocity (V ). In practice, most experiments show
a significant departure from the V 8 relationship both at high and low speeds. In
high speed flows, the deviation can be explained in terms of the convection speed
of the turbulent eddies which are comparable to the speed at which acoustic waves
propagate to the far field [12]. As the jet velocity increases and becomes greater than
the speed of sound, source power is dominated by kinematic effects and tends towards
V 3.
10
such as the nozzle, the jet pipe, active or passive mixers, the thrust reverser, ejectors
and thrust vectoring mechanisms depending upon the specific requirements of the
propulsion system and overall aircraft performance requirements.
The exhaust nozzle serves two primary functions. First, it controls the back pres-
sure for the overall operation of the jet engine and second, it converts the thermal
energy from the combustion of fuel and potential energy in the form of pressure
into kinetic energy in the form of exhaust jet velocity, thereby accelerating the ex-
haust gases to produce thrust. The performance of the exhaust nozzle is significantly
different when installed with the aircraft as part of the entire propulsion system.
Additional design challenges are introduced by the requirements for features such as
reverse thrust, thrust vectoring and variable area throat to facilitate the application
of afterburners in military aircraft.
A supersonic cruise propulsion system requires a noise suppressing, economically
viable, light weight and simple-to-maintain exhaust nozzle system. One of the critical
challenges associated with the design of a supersonic exhaust system during take-off
and landing is its noise suppression. The nozzle should comply with the FAR noise reg-
ulations and produce minimum community noise. During take-off, the aerodynamic
performance of the nozzle is less critical compared to the acoustic characteristics [7].
In the past forty years, research has been done in the field of supersonic noz-
zle design. The primary motivation for the design of low noise jet engine nozzle is
the stringent noise regulations for the community noise. In the past, Rolls-Royce
conducted studies to understand the powerplant noise sources associated with the
supersonic transport and reduce the noise levels associated with the Olympus-593
powerplant. The nozzle design (ejector nozzle with clamshell doors) corresponding
to the Olympus-593 powerplant of the Concorde serves as the primary motivation for
the present work. At this point, it is necessary to discuss the operation of the ejector
nozzle before going into detailed analysis. Since the present work also involves the
design of chevrons, the flow physics of chevrons is discussed.
12
Variable nozzle exit area is a requirement for the operation of the nozzle at off-
design nozzle pressure ratios when the aircraft speed increases beyond the transonic
regime. During supersonic cruise, as the Mach number increases, the nozzle pressure
ratio and the exhaust nozzle area ratio varies such that
γ
Pt γ − 1 2 (γ−1)
= 1+ Me and (2.1)
pe 2
γ+1
1 1 + γ−1 2 2(γ−1)
Ae 2
Me
= γ+1 . (2.2)
A∗ Me 2
Beyond Mach 1.2 ∼ 1.5, nozzle pressure ratios increase rapidly beyond 3.0, as
shown in Figure 2.2, requiring further expansion of the exhaust flow to keep the nozzle
perfectly expanded (pe =patm ); increase its velocity and improve nozzle performance.
When exhaust gases are not adequately expanded to near ambient pressures, they
suffer from overexpansion or underexpansion losses. These losses can be avoided
by selecting an area ratio corresponding to the nozzle pressure ratio. The required
area ratio increases with the nozzle pressure ratio as a function of flight altitude and
Mach number. In order to provide optimum thrust, the exhaust nozzle area ratio
should be varied as a function of the nozzle pressure ratio. However, weight, cost
and performance goals must be balanced to provide the best nozzle to meet aircraft
requirements.
Modern supersonic cruise aircraft use different techniques to achieve the variable
exit area capability to operate at off-design conditions. Some of the techniques involve
geometrically scheduled (F-14 and F-18 fighter aircrafts), passively controlled (F-
15 and F-16 fighter aircrafts) and fully variable area ratios (F-22 fifth generation
fighter aircraft) [16]. These techniques involve addition of divergent flaps and an
actuation mechanism to schedule the area variation of the throat (A∗ ) and the exit
(Ae ) either independently or with respect to each other. These techniques suffer from
disadvantages such as increased weight, complexity, cost, and difficulty to maintain.
13
3 3
10 10
2 2
10 10
0 0
10 10
−1 −1
10 10
−1 0 1
10 10 10
Nozzle Exit Mach Number, Me
Figure 2.2. Requirements for the pressure ratio and the area ratio as
Mach number increases [16].
As studied by Der [17] in detail and shown in Figure 2.4, the ejector performance
can be deduced from the behavior of the associated shear layer and its relation to
the ejector shroud walls. For instance, if the mixing duct length (LS ) is too small,
the shear layer will not attach to the shroud wall, leaving the secondary flow open
to a stronger influence from external conditions and susceptible to separation and
recirculation. On the other hand, if LS is long and extends beyond the attachment
point of the shear layer with the shroud walls, frictional losses will result.
Nozzle Flow
The term Chevron literally means a V-shaped pattern. In the context of jet en-
gines, a chevron nozzle features triangular serrated trailing edge as shown in Figure
2.5. When compared with a baseline axisymmetric nozzle, this additional feature
promotes streamwise vortices, which, along with the naturally occurring toroidal vor-
tices, accelerate the mixing between the jet exhaust and the surrounding atmospheric
air. Enhanced mixing results in a reduction of the exhaust velocity and therefore the
jet noise, in accordance with the Lighthill’s eighth power law (U 8 ) [11].
Other passive mixers include corrugated mixers and tabs. Corrugated mixers suf-
fer from additional weight, drag and increased specific fuel consumption. Tabs result
in a reduction of the low frequency noise but suffer from considerable high frequency
penalties. When compared with other passive mixers, chevrons offer simplicity in de-
sign, manufacturing and maintenance, with much smaller weight and high frequency
penalties [19]. Because of these advantages, chevrons are finding applications in recent
16
commercial jet engines such as the Rolls-Royce Trent-1000 and the General Electric
GE-NX.
Kenzakowski et al. (2000) [20] studied the effect of passive noise reduction devices
such as chevrons and tabs in plume mixing enhancement. They compared the mean
and turbulent flow fields, obtained using the inhouse Combustion Research and Flow
Technology (CRAFT) flow solver, with experiments conducted at NASA Glenn Re-
search Center (GRC). The main concentration in these simulations was to study the
performance of available turbulence models in predicting the chevron flow, which will
be discussed in more detail in Section 2.3.2. Janardan et al. (2000) [21] documented
the noise reducing performance of various passive mixing enhancing devices with a
three-stream separate-flow axisymmetric plug nozzle. Most of the acoustic measure-
ments presented in the above references were based on microphones alone and hence
a more detailed experimental study of the flow physics involved with chevron nozzles
was required to understand the distribution and nature of the noise sources.
The Particle Image Velocimetry (PIV) emerged as a promising candidate for this
application and is popular in characterizing jet plumes because of its capability of
capturing small turbulence scales. Bridges and Wernet (2002) [18] carried out a com-
prehensive experimental study of the turbulent flow characteristics using various mix-
ing enhancement devices on a three-stream separate-flow axisymmetric plug nozzle
using PIV. The passive mixing device alters the turbulence components significantly
and reduces the turbulent kinetic energy (TKE) magnitude in the jet mixing region
(six to nine fan diameters from the nozzle exit) while increasing it in the fan/core
shear layer near the nozzle exit (around first two fan diameters from the nozzle exit).
Bridges and Brown (2004) [22] conducted a parametric experimental study of
chevron nozzles. They performed flow field measurements using PIV and acoustic
field measurements on single-flow hot and cold jets with chevrons. The effect of
chevron count, penetration length and symmetry on the exhaust jet flow field was
studied in detail. They concluded that the chevron penetration decreases the noise
at low frequency and increases the noise at high frequency. Also, low frequency noise
17
reduction can be achieved by increasing the chevron count. Rask et al. (2007) [23]
recently conducted a detailed investigation into the primary chevron mechanisms
using PIV.
Koch and Bridges (2004) [24] studied a three-stream separate-flow plug nozzle with
chevrons arranged in an alternating fashion on the core nozzle. Chevrons arranged in
alternating fashion consisted of a chevron facing towards the nozzle flow and the other,
facing away from the nozzle flow arranged alternatively around the circumference of
the core nozzle as shown in Figure 2.6. Koch compared the experimental results
obtained with 12 chevrons from the PIV measurements with the WIND and CRAFT
CFD solver results. Kurbatskii (2009) [25] came up with a comprehensive set of
results on using various turbulence models and corrections given for the prediction
of the mean flow field of cold and hot subsonic jets from chevron nozzles using the
ANSYS FLUENT version 12.0 flow solver.
Research shows that the presence of chevrons significantly increases the entrain-
ment of the primary flow into the secondary stream and vice versa. This mixing
results in a reduction of the primary stream velocity and lower levels of turbulent
kinetic energy far downstream in the jet plume. However, increased mixing also
causes increased levels of turbulence immediately downstream of the nozzle resulting
in additional high frequency noise. Therefore, there should be a trade-off between
the mixing of two streams and its acoustics benefits. This trade-off can be controlled
by parameters such as the chevron count, the chevron penetration, and the shear
velocity.
Figure 2.6 shows the mixing between two streams and the formation of streamwise
vortices. Low penetration chevrons cause higher mixing between adjacent streams
compared to the baseline axisymmetric nozzle and results in the jet noise reduction,
particularly at low frequencies. With high penetration chevrons, this mixing is further
enhanced and higher noise reduction of low frequency noise is achieved. This more
aggressive mixing, however causes additional turbulence immediately downstream of
the chevron nozzle, which can result in a high frequency penalty. A similar trend can
18
Figure 2.6. Mixing of two streams of the chevron nozzle and stream-
wise vortex formation.
be observed with shear velocity. High shear velocity (high velocity difference between
two streams) results in higher mixing causing higher noise reduction of low frequency
noise, but with a penalty. Further study by Rask et al. (2007) [26] shows that the
reduction on overall noise levels with the application of chevrons on fan and core
nozzle are additive.
There is always a great demand for well-validated computational tools that would
allow aerospace engineers to parametrically design and evaluate new low noise nozzle
concepts. Ideally, these tools should provide accurate aerodynamic and aeroacoustic
predictions for a wide variety of nozzle geometries operating through a range of flow
conditions. Coaxial jets consist of two or more flow streams and therefore may have
more than one potential core length. In general, these lengths can be different. In
the case of a chevron nozzle, if chevrons are fitted only on the primary jet, as shown
in Figure 2.6, the mixing between the primary stream and the secondary stream is
much faster than the mixing between the secondary stream and the atmospheric air.
Therefore, the primary core length is relatively short and the jet centerline velocity
starts to decrease well before the shear layer between the fan stream and the external
flow start to merge.
19
In the present study, the ANSYS FLUENT [29] version 6.3.26 computational
solver was used to solve Reynolds-averaged Navier-Stokes (RANS) equations and the
system of governing equations was closed using turbulence models. The FLUENT
code is a finite volume numerical solver that can be used with both structured and
unstructured meshes. Both the pressure-based segregated and the density-based cou-
pled solvers are available in the FLUENT code.
In general, flow features of high speed jet flows are predicted using density-based
numerical approaches. In density-based coupled formulations, the continuity, mo-
mentum and energy governing equations are solved simultaneously. Density-based
algorithms are efficient for high subsonic, transonic or supersonic flows. However,
they require modifications, such as the preconditioning in low Mach number flow
regions, to overcome the problem of the system matrix becoming singular in the in-
compressible limit. In the density-based algorithm, convergence is mainly governed
by the Courant number. It should be noted that numerical stability considerations
for large grids mandate a low Courant number, such as 1.0.
As an alternative, the coupled pressure-based numerical solver [30] can be used
for problems where strong coupling among equations exist. Unlike a segregated al-
20
gorithm, in which momentum equations and pressure correction equations are solved
one after another in a decoupled manner, a pressure-based coupled algorithm solves
a coupled system of equations comprised of momentum equations and the pressure
equation. Since momentum and pressure equations are solved in a strongly coupled
manner, the rate of solution convergence significantly improves when compared to
the segregated solver.
The pressure-based coupled approach can also be used for supersonic and hyper-
sonic problems that cannot be tackled by a segregated approach. In all solvers, the
convergence is determined by the residual drop in mean flow variables. The iteration
history of a flow variable at some location in the computational domain is also used
to check the convergence. When the value of the flow variable reaches a relatively
steady state, the solution is considered to be converged.
The convergence of the pressure-based algorithm is mainly governed by the un-
derrelaxation factors and hence, a high value for the Courant number is preferred.
In general, the pressure-based coupled solver is faster than then the density-based
coupled solver. Because of the above mentioned advantages, the pressure-based seg-
regated numerical solver is used for all low subsonic CFD computations and the
density-based explicit coupled numerical solver is used for transonic or supersonic
CFD simulations.
Accurate modeling of turbulence in jet flows is essential for the accurate predic-
tion of mean flow profiles, turbulence quantities and the generated noise. This can be
achieved by two methods. One approach involves solving both the mean and acoustic
field directly in the same transient simulation. Turbulence quantities in this method
can be predicted using different methods such as Unsteady Reynolds-averaged Navier-
Stokes (URANS), detached eddy simulation (DES), large eddy simulation (LES) or
direct numerical simulation (DNS) techniques which are mentioned in order of in-
21
creasing computational cost and accuracy. Another method is to calculate the mean
flow profiles and turbulence fields using RANS and use them as input into a sepa-
rate computational aeroacoustics (CAA) code or semi-empirical noise model for the
prediction of acoustic fields such as the JeNo and MGBK code [31], the Goldstein-
Lieb model [32], and the broadband shock-associated noise (BBSAN) code [33]. Both
approaches have advantages and disadvantages. RANS-based semi-empirical models
require less computational time for the estimation of jet noise. However, they are less
accurate when compared with the time resolved methods or the direct CAA approach.
For jet noise predictions, LES is preferable because it estimates the noise more
accurately by giving a good prediction of small and large scale turbulent eddies when
compared with the RANS-based methods [34]. For LES simulations, the initial wall
boundary layers at the nozzle exit need to be thin as well as turbulent (implies a
considerably high jet Reynolds numbers). Large computational resources and fine
mesh requirement limit the application of LES to research level. RANS has matured
greatly over the last decade and is widely used in the aircraft design process. There-
fore, in the present study, only RANS-based jet flow field prediction using turbulence
models was considered.
Nallasamy (1999) [35] studied the performance of various turbulence models for the
computation of turbulent jet flows and noise in round jets. He concluded that length
and time scales should be predicted accurately for the estimation of sound pressure
levels. He found that the Sarkar’s compressibility correction for supersonic jets and
Pope’s vortex stretching correction for axisymmetric jets give correct spreading rates
for round jets.
Birch et al. (2003) [27] listed the turbulence modeling requirements for a chevron
nozzle flow and proposed a two-equation zonal turbulence model for better mean
and turbulent flow predictions. A zonal turbulence model essentially means that one
model is used for mixing layers upstream of the end of the jet potential core and
a second model is used for the axisymmetric region downstream of the end of the
potential core. He concluded that though the vortex stretching term proposed by
22
Pope (1978) improves the prediction of axisymmetric jets, it is not suitable for jets
emerging from a chevron nozzle or other mixing nozzles. In these cases, the vortex
stretching term becomes large and introduces errors [27].
Georgiadis et al. (2006) [28] did a survey of various two-equation turbulence models
and corrections proposed in predicting the flow fields of heated and unheated jets.
He found that for the round jet flow calculations, all the modified turbulence models
considered in the study such as the Tam-Ganesan k- formulation, the standard k-
turbulence model with the temperature correction, and the k- turbulence model
with variable diffusion offer improved mean flow predictions relative to the unmodified
standard turbulence models viz. the Chien k- turbulence model and Menter’s k-ω
SST turbulence model. None of the standard and modified turbulence models is
capable of giving a good estimate of turbulent kinetic energy fields when compared
with the experimental data.
In the following chapters, the CFD simulation results involve the application of
various turbulence models. Hence a brief introduction of various turbulence models is
presented here. Reynolds-averaging of instantaneous Navier-Stokes equations results
in additional unknowns, known as Reynolds stresses. These unknowns are related
to mean velocity gradients using the Boussinesq hypothesis. The advantage of this
approach is a relatively low computational cost associated with the computation
of the turbulent viscosity. Additional transport equations are required to calculate
the turbulent viscosity. In the case of two-equation turbulence models, an accurate
TKE k-equation and an approximate TKE dissipation rate -equation are used to
independently determine turbulent length and velocity scales. These scales are used
to calculate the turbulent viscosity. This is the primary motivation for the two-
equation turbulent models discussed below.
23
The standard k- turbulence model was proposed by Launder and Spalding (1972)
[36]. The model has become the workhorse of practical engineering flow calculations.
RANS equations are closed using the k and equations. In the past, it was observed
that for jet flows this turbulence model underpredicts the length of the potential core
and turbulence levels.
As the advantages and disadvantages of the standard k- turbulence model became
known, several improvements were proposed to further refine the model. One problem
associated with the standard k- model is the overprediction of the TKE in high
strain-rate regions. The realizable model differs from the standard k- model in two
ways:
1. A new formulation is used for the turbulent eddy viscosity involving variable
Cµ originally proposed by Reynolds [37], and
2. A new transport equation is used for the dissipation rate , which has been
derived from the exact equation for the transport of the mean square fluctuating
vorticity.
The term realizable means that the turbulence model satisfies certain mathemati-
cal constraints on Reynolds stresses, consistent with the physics of the turbulent flow.
An immediate benefit of the realizable k- turbulence model is that it accurately pre-
dicts the spreading rate of both planar and round jets. The version of the turbulence
model implemented in the FLUENT flow solver corresponds to Shih (1995) [38].
24
Table 2.1 Thies and Tam’s k- turbulence model constants [39].
This model was proposed by Thies and Tam (1996) [39]. In this model, some of
the coefficients of the standard k- turbulence model were changed specifically for
jet flows and correction terms were added to resolve the planar/axisymmetric jet
problem and for compressibility effects. They validated the performance of these
modified constants with a variety of experimental tests involving round, elliptic and
rectangular jets. It was found that the model gives accurate mean flow predictions
for the Mach number range of 0.4 ∼ 2.0. The modified constants are presented in
Table 2.1.
Wilcox (1988) [40] postulated a new two-equation turbulence model based on his
study of the optimum choice of dependent variables and replaced the scale determining
-equation with the specific turbulent kinetic energy dissipation rate ω-equation. This
model shows improvements in the case of boundary layer flows with adverse pressure
gradients. However, it was found that this model is very sensitive to the inflow
turbulence levels and requires an accurate inflow turbulence boundary condition for
an accurate flow prediction. Wilcox has since revised the k-ω model (1998, 2006) [41];
however, the updated models are not available in the FLUENT flow solver.
25
The k-ω shear stress transport (SST) turbulence model was developed by Menter
(1994) [42] to effectively blend the robustness and accurate formulation of the k-ω
model in the near wall region and the free stream independence of the k- model
in the far field. In order to achieve this, the k- model was converted into a k-ω
formulation. The k-ω SST model is similar to the standard k-ω model, but differs in
the followings aspects:
1. The standard k-ω model and the transformed k- model are combined by using
a blending function. The SST model includes a cross-diffusion derivative term
in the ω-equation. This extra term is the difference in the two formulations. It
is turned off near a wall and turned on away from walls.
The above mentioned features make the k-ω SST model more accurate and robust
for a wide class of flows.
The Reynolds Stress Model (RSM) [43] is a higher level, more elaborate turbulence
model, known as a Second-Order Closure model. RSM deviates from the isotropic
eddy viscosity hypothesis and solves equations for Reynolds stresses, together with an
equation for the dissipation rate. This essentially means that five additional transport
equations for 2-D flows and seven for 3-D flows are required. The RSM is generally
considered to be more accurate compared to one and two-equation turbulence models,
as it accounts for streamline curvature, swirl, rotation and rapid changes in strain-
rate in a more rigorous manner. Hence use of the RSM is suggested when the flow
26
features of interest are the result of anisotropy in the Reynolds stresses such as with
highly swirling and rotating flows.
As mentioned in Section 1.4, the present work involved the validation and design
improvement of a new noise suppression nozzle, designed and experimentally tested
by Jones [44]. Hence, there was a requirement to follow a well defined design and
analysis methodology. Figure 2.7 explains the methodology followed in the current
work. The various steps involved were as follows:
3. Grid Generation: The computational meshes necessary for the CFD simula-
tions were created using the Pointwise GRIDGEN version 15.10. Both struc-
tured and unstructured grids were created in the present study. The interface
between the structured and unstructured block consisted of pyramid cells. The
boundary layers consisted of prisms for the unstructured mesh and hexahedral
cells for the structured mesh.
4. CFD Simulation: The ANSYS FLUENT version 6.3.26 flow solver was used
for all the CFD simulations. Both the pressure-based and the density-based
27
Figure 2.7. Methodology for the jet engine exhaust nozzle design and analysis.
6. Validation: All the data analysis and line plots involved in various validation
studies were done using MATLAB version 2009. The variation of the centerline
velocity magnitude and the total temperature with the axial distance were used
for studying the nozzle jet characteristics.
28
3. Chevron Nozzles
3.1 Introduction
One of the widely used techniques to reduce jet noise from high BPR turbofan
nozzles is to enhance the mixing of the core and the fan jet streams by adding serra-
tions at the trailing edges of the jet nozzle. They are known as Chevrons. Modern
turbofan engines, such as the Trent-1000 on the Boeing-787 and the GE NX-2B67 on
the Boeing 747-800, make use of these chevrons for jet noise reduction. Chevrons are
a viable design feature for the noise suppression of jet engine exhaust nozzle systems
because of their ability to reduce noise with a small percentage of thrust penalty and
easy manufacturability.
The present work served as a preliminary validation task for the design and anal-
ysis of chevron nozzles. We considered the design of chevrons studied by Janardan
et al. [21] and performed the CFD simulations similar to the work of Kenzakowski et
al. [20]. They used a baseline three-stream separate-flow axisymmetric plug nozzle
(3BB) and studied the performance of passive mixing devices such as 12 alternating
chevrons (3A12B) (as explained in Section 2.2.2) and 24 tabs (3T24B).
In the chevron nozzle case, Kenzakowski et al. (2000) found that the inward-facing
chevron enhanced the penetration of the fan stream into the core stream downstream
of the nozzle exit and the outward-facing chevron deflected the core stream into the
fan stream. This resulted in an enhanced mixing and the growth of the shear layer
between the core stream and the fan stream was greatly affected. However, the
growth of the shear layer between the fan stream and freestream did not appear to
be significantly altered by the presence of chevrons.
It was observed that the fan stream and the freestream mixing layer was the dom-
inant turbulent region, but the peaks obtained in the turbulence fields were further
29
upstream compared to the baseline axisymmetric nozzle. CFD simulations using the
WIND-CFD code showed an underprediction of the mixing rate, leading to a longer
potential core compared to the experimental data [20]. As part of the present work, it
was decided to analyze the 3BB and 3A12B nozzle geometries and perform the CFD
simulations using the ANSYS FLUENT flow solver. Since experimental and WIND-
CFD results were available in the literature, this work was beneficial in studying the
potential of the FLUENT flow solver and various turbulence models in the prediction
of the chevron jet flows.
3.2 Objectives
The objective of the present task was to study the performance of the ANSYS
FLUENT flow solver and its available turbulence models in the prediction of three-
stream separate-flow axisymmetric plug and chevron-based jet flows. Results obtained
from the CFD simulations were compared to the WIND-CFD results and validated
using the experimental PIV measurements available in the literature.
3.3.1 Introduction
The computational mesh was created using the Pointwise GRIDGEN version
15.10. As the nozzle was axisymmetric, a multiblock 2-D structured mesh, as shown
in Figure 3.2 was created. The grid consisted of a total of 0.115 million cells. A
variable wall-normal grid spacing was used for all viscous wall boundaries to yield a
y + value in the range of 30 ∼ 300 and hence wall functions were used for the near
wall turbulence.
31
The ANSYS FLUENT version 6.3.26 flow solver was used for all the CFD sim-
ulations. Since the flow conditions were high subsonic, the pressure-based explicit
coupled solver was used to solve the continuity and momentum equations. As dis-
cussed in Section 2.3.1, the convergence of the pressure-based coupled algorithm was
governed by the underrelaxation factors. Hence a high underrelaxation factor of 1.0
was used for the density, pressure, and momentum equations and 0.8 was used for the
turbulence equations. A high value of 5.0 was used for the Courant-Friedrich-Lewy
32
Table 3.1 Boundary conditions for the CFD simulation of the three-
stream separate-flow axisymmetric plug nozzle (3BB) [20].
(CFL) number. The viscous walls were assumed adiabatic and the inflow boundary
layer effects were not included.
Various turbulence models, discussed in Section 2.3.2, were used with walls func-
tions for near wall resolution. In the current work, the focus was to study the potential
of two-equation turbulence models in predicting jet flows accurately. Hence, CFD sim-
ulation of the nozzle was performed using turbulence models viz. the standard k-,
the realizable k-, Thies and Tam’s k-, the standard k-ω and the k-ω SST turbulence
models. In addition to this, the more accurate, second-order closure RSM was also
used. Convergence was greatly affected in the case of RSM and it took more compu-
tational time to get a converged solution when compared to two-equation turbulence
models.
In order to check the convergence, along with the default solver residual moni-
tor, the iteration history of the mass-balance and velocity magnitude at some axial
location downstream of the nozzle throat was monitored. When the mass-balance
reached 1 x 10−5 and the velocity magnitude reached a steady-value, the solution
was considered to be converged. The solutions presented in the present chapter are
second-order accurate.
33
3.3.4 Results
Contours of the axial velocity and turbulent kinetic energy for the baseline three-
stream separate-flow axisymmetric plug nozzle, obtained using the standard k- tur-
bulence model, are illustrated in Figures 3.4 and 3.10, respectively. The standard
k- turbulence model gives a good prediction of the mean flow variables with better
convergence. However, it fails to predict the turbulence quantities in complex flows
such as jets. When compared with the experiments, it is evident that the standard k-
34
turbulence model predicts the mean velocity contours very well but with an under-
prediction of the length of the potential core. The classical problem of the unphysical
overprediction of the turbulent kinetic energy associated with the standard k- tur-
bulence model in the high shear region is shown in Figure 3.10. The standard k-
turbulence model overpredicts the TKE near the fan stream nozzle exit plane, when
compared with the experimental data.
It has been suggested that the overprediction in the TKE contours associated with
the standard k- turbulence model is because of the unphysical values of Cµ in the
definition of turbulent viscosity,
k2
µt = ρCµ . (3.1)
Figures 3.6 and 3.12 show the axial velocity and turbulent kinetic energy contours
corresponding to the Wilcox’s k-ω (1988) turbulence model. The mean flow velocity
contour is similar to the realizable k- turbulence model with an overprediction of the
length of the potential core. However, the TKE contours are diffusive compared to
35
Menter’s k-ω SST turbulence model is an effort to make use of the robustness of
the standard k- turbulence model and the accuracy of the standard k-ω turbulence
model for flows with boundary layers using blending functions. Because of the com-
bined advantages, the k-ω SST turbulence model gives a very good prediction of the
turbulent kinetic energy and the mean axial velocity. Figure 3.7 shows the contours
of the mean axial velocity and Figure 3.13 shows the contours of the turbulent ki-
netic energy. The contours are in close agreement with the results obtained using
the realizable k- turbulence model. But as with the case of realizable k- turbulence
model, the k-ω SST turbulence model overpredicts the length of the potential core
and underpredicts the levels of TKE when compared with the experiment data.
The RSM belongs to the second-order closure class of turbulence models and is
considered to be more accurate than most of the two-equation turbulence models for
strongly swirling and rotational flows. However, as we observed, the improvement in
the accurate prediction of the three-stream separate-flow axisymmetric plug nozzle
jet mean flow is not considerable when compared with the realizable k- and the k-ω
SST turbulence model. Figure 3.8 shows the contours of the mean axial velocity. The
contours of the turbulent kinetic energy are shown in Figure 3.14.
It is evident that the contours match closely with the realizable k- and the k-ω
SST turbulence model. The potential core is somewhat longer than with the k-ω SST
turbulence model and approximately equal to the one obtained using the realizable
k- turbulence model. The contours of the turbulent kinetic energy shows that the
peak TKE regions are shifted downstream in the case of the RSM when compared
36
with the k-ω SST turbulence model. The RSM predicts a higher magnitude of the
peak TKE and are in better comparison with the experiment data when compared
with all the turbulence models used in the present study.
Figure 3.15 shows the variation of the centerline velocity magnitude (UCL ) with
axial distance from the nozzle exit. This plot summarizes the findings discussed in
the results for each turbulence model. It is evident that the standard k- turbulence
model gives a good prediction of the mean flow, following the trend of the experimental
measurements, but underpredicts the potential core length. Also, it is not accurate
for the prediction of TKE. The realizable k-, the k-ω SST and the Reynolds stress
turbulence models perform much better in the prediction of the mean axial velocity
and turbulent kinetic energy. However, they suffer from the overprediction of the jet
potential core length.
Figure 3.16 shows the variation of the total temperature (Tt ) along the centerline.
The potential core region can be identified by the constant temperature region. It
is observed that the standard k- and standard k-ω (1988) turbulence models give a
reasonable prediction of the variation of the total temperature with some underpre-
diction and overprediction, respectively.
37
1600
1400
1000
800
600
PIV experiment (NASA/CR−2000−210039)
WIND k−ω SST (Koch, 2004)
400 Standard k−ε turbulence model
Realizable k−ε turbulence model
Standard k−ω turbulence model
200
k−ω SST turbulence model
Reynolds stress model
0
0 5 10 15
X/Dfan
1200
Experiment (Kenzakowski et al. 2000)
1100 Standard k−ε turbulence model
Realizable k−ε turbulence model
Standard k−ω turbulence model
1000
k−ω SST turbulence model
Total Temperature T° K
800
700
600
500
400
300
0 5 10 15
X/Dfan
3.4.1 Introduction
The nozzle geometry used in the experimental and computational studies of Koch
and Bridges (2004) [24] was used in the present work. A CAD model was created for
the nozzle geometry using the Pro/Engineer Wildfire 4.0. Dimensions of the baseline
nozzle were the same as the three-stream separate-flow axisymmetric plug nozzle,
discussed in Section 3.3 and were extracted from figures available in the literature [24].
In addition to the baseline nozzle, the core nozzle of this configuration consisted of
12 chevrons arranged in an alternating pattern, facing inward and outward of the
core flow. The fundamental dimensions involved in the chevron design are shown in
Figure 3.17. The alternating chevron arrangement consisted of half of the chevrons
being bent towards the core flow by 4.5◦ and the other half were bent towards the
fan stream by 8◦ . A CAD model of the nozzle geometry is shown in Figure 3.18.
The computational mesh for the CFD simulation was created using the Pointwise
GRIDGEN version 15.10 grid generation code. Since the nozzle with chevrons was
43
Basic Chevron Design Parameters All dimensions in inches unless otherwise specified.
S. No Parameter Value
1 N = Number of chevrons 12
9 θ = 360o/N (degrees) 30
The flow conditions were the same as used in the CFD simulation of the baseline
three-stream separate-flow axisymmetric plug nozzle, with a NPR of 1.7. Therefore,
similar pressure-based boundary conditions were used for the CFD simulations. The
boundary conditions were pressure-inlet at the inflow boundary, inviscid wall along
the symmetry planes, pressure far field along the outer freestream boundary and
pressure-outlet at the exit. The numerical values of the flow boundary conditions are
given in Table 3.2.
Table 3.2 Boundary conditions for the CFD simulation of the three-
stream separate-flow chevron nozzle (3A12B) [24].
The ANSYS FLUENT version 6.3.26 flow solver was used for all the CFD simu-
lations. The density-based explicit solver was used for high subsonic flow conditions.
RANS-based flow governing equations were solved in a coupled manner and the sys-
tem was closed using turbulence equations. The density was defined by the ideal gas
equation and Sutherland’s three coefficient method was used to calculate the value
of the molecular viscosity. The convergence of the density-based explicit algorithm
is defined primarily by the CFL number. A low CFL value is required for stability
reasons. Hence a CFL value of 1.0 was used to get the converged first-order solution.
This was different from the three-stream separate-flow axisymmetric plug nozzle case
46
where a CFL of 5.0 was used because of the pressure-based explicit solver. The
solutions presented here are second-order accurate.
The ANSYS FLUENT flow solver was run using 5 nodes of the Booster cluster
of the School of Aeronautics & Astronautics, Purdue University. Booster is a Linux
based computing cluster with a total of 56 nodes. Each node consists of 4 AMD 64
bit based processors. Booster used a portable batch system called Torque for job
scheduling.
3.4.4 Results
3.37. Similarly, the turbulent kinetic energy contour plots on the outward-facing
chevron mid-plane are shown in Figure 3.38 for experiments, Figure 3.39 for WIND-
CFD result, and Figures 3.40-3.43 for various turbulence models.
Figures 3.23 and 3.35 show the distribution of the axial velocity and turbulent
kinetic energy for the standard k- turbulence model on the inward-facing chevron
mid-plane. It is evident that the length of the potential core is shorter compared to
the experiments. This is consistent with the results obtained from the CFD simulation
of the three-stream baseline axisymmetric plug nozzle. Figure 3.35 shows nonphysical
overprediction of the turbulent kinetic energy near the core nozzle exit because of the
high shear-rate. The problem is known and addressed by using a realizable limit in
the definition of the turbulent viscosity, as discussed in Section 3.3.4. Figures 3.29
and 3.41 show the distribution of the axial velocity and turbulent kinetic energy for
the standard k- turbulence model on the outward-facing chevron mid-plane.
The realizability limit, used in the definition of the turbulent viscosity, results in
overcoming the overprediction of the turbulent kinetic energy corresponding to the
standard k- turbulence model. The realizable k- turbulence model gives a good
prediction of the mean flow field and the turbulent kinetic energy contours are in
better comparison with the experimental results. Figures 3.24 and 3.36 show the
contours of the axial velocity and the TKE on the inward-facing chevron mid-plane.
Also the contours of the axial velocity and TKE corresponding to the outward-facing
chevron are shown in Figures 3.29 and 3.41, respectively. The realizable k- turbulence
model is also recommended in the ANSYS FLUENT flow solver’s best practice guide
for CFD simulations involving external flows [29].
48
As discussed in Section 2.3.2, Thies and Tam proposed few modifications in the
standard k- turbulence model constants for the better prediction of the turbulent
jet flows. These constants can be implemented in the ANSYS FLUENT flow solver
using the turbulence model selection user interface. Figure 3.25 and 3.31 show the
mean velocity contours on the inward-facing and outward-facing chevron mid-planes,
respectively. Similarly, Figures 3.37 and 3.43 show the turbulent kinetic energy con-
tours on the inward-facing and outward-facing chevron mid-planes, respectively. It
can be easily observed that, even though the above mentioned modifications to the
model constant result in a better prediction of the mean flow fields, it is very diffusive
for the prediction of the turbulent kinetic energy, resulting in smeared contours.
1600
1400
1000
800
600
900
Experiment (Birch, 2003)
Zonal k−ε (Birch, 2003)
Standard k−ε turbulence model
800 Realizable k−ε turbulence model
Standard k−ε with Thies−Tam’s correction
k−ω SST turbulence model
Total Temperature To K
700
600
500
400
300
0 5 10 15 20 25
X/D
fan
3.5 Conclusion
4. Ejector Nozzles
4.1 Introduction
Noise is a major problem associated with the high speed propulsion system design.
Although there are no clear cut restrictions on the noise levels for high speed aircraft,
it is reasonable to assume that the noise levels in the terminal area will be governed
by the restrictions similar to that of the subsonic aircrafts. Supersonic jet noise is an
important contributor to the overall propulsion system noise. It is essential to mini-
mize the jet noise during take-off and landing where it is more pronounced and tough
noise regulations apply. These requirements pose additional design requirements on
the exhaust nozzle design. Nozzles such as the separate-flow nozzle, the plug noz-
zle, the chevron nozzle and the mixer-ejector nozzle are examples for addressing this
challenge. In recent years, chevron and ejector nozzle designs received special atten-
tion because of their improved noise reduction characteristics and low thrust penalty.
The history and performance of ejector nozzles are discussed in Section 2.2.1. The
present chapter is dedicated to the discussion of the results obtained from the CFD
simulation of the 3-D ejector nozzle with and without clamshell doors at low-speed
experimental and high-speed take-off conditions.
4.2 Objectives
The primary objective of the present study was to perform the CFD simulations
of the 3-D ejector nozzle with clamshell doors, whose experimental performance at
subsonic conditions was studied by [44]. Initially, the computational study of the
NASA 2-D ejector nozzle test case was carried out to perform the validation of the
CFD tool. Next, the CFD simulation of the 3-D ejector nozzle without clamshell
60
doors as well as with clamshell doors at 11.5◦ were performed and the results are
compared with the experiments. The analysis of the 3-D ejector nozzle was extended
by the application of different turbulence models and performing the CFD simulation
at higher nozzle pressure ratios corresponding to the take-off conditions.
4.3.1 Introduction
Gilbert and Hill (1973) [46] investigated a turbulent, two-dimensional ejector noz-
zle flow through a rectangular section experimentally. Figure 4.1 shows the experi-
mental geometry with primary nozzle and mixing section. The setup consisted of a
discharge slot as the primary nozzle, opening into a rectangular area mixing section
of constant width of 8 in. through a pair of contoured walls placed symmetrically on
either side of the primary nozzle. The experimental data used in the present study
61
−1
Primary Nozzle
−2
−3
−4
−5 0 5 10 15 20
Axial Length (in)
Figure 4.1. Experimental setup for the 2-D ejector nozzle. (Repro-
duced courtesy of [46].)
Variables Value
Primary nozzle total pressure 246 kPa
Primary nozzle total temperature 358 K
Ambient pressure 101 kPa
Ambient temperature 305 K
Table 4.1 shows boundary conditions used in the CFD simulation which corre-
sponded to run nine in reference [46]. The pressure-based CFD boundary conditions
were used for the CFD simulation because of high nozzle pressure ratios in the exper-
iments. The pressure-inlet boundary condition was used for primary and secondary
flows with the stagnation pressure, the stagnation temperature and turbulence quan-
tities as input. The outflow was pressure-outlet with static pressure, measured in the
experiments, as input.
The ANSYS FLUENT version 6.3.26 flow solver was used for all the CFD simula-
tions. The operating conditions consisted of high nozzle pressure ratio and hence the
governing equations were solved using the density-based explicit solver. The details
63
of this solver are discussed in Section 2.3.1. Two turbulence models, viz. Menter’s k-ω
SST turbulence model and the Spalart-Allmaras turbulence (SA) model were used for
the turbulent flow simulation. The results presented here are second-order accurate.
Figure 4.3. Mach number contour plot of the 2-D ejector nozzle cor-
responding to the k-ω SST turbulence model.
Figure 4.4. Mach number contour plot of the 2-D ejector nozzle cor-
responding to the Spalart-Allmaras turbulence model.
64
4.3.4 Results
The velocity and stagnation temperature profiles at different axial locations down-
stream of the primary nozzle exit plane are plotted and compared with the experi-
mental and WIND-CFD results. The vertical distance is normalized using the semi-
height (H) of the rectangular channel. Figure 4.5 shows the velocity profile at 3 in.
downstream of the nozzle exit plane and its comparison with the experimental and
WIND-CFD results. It is evident that the Spallart-Allmaras turbulence model per-
forms well compared to the k-ω SST turbulence model. Similarly, Figures 4.6, 4.7,
and 4.8 show the velocity profiles at 5.0, 7.0, and 10.5 in. downstream, respectively.
The CFD predictions improve and match well with the experimental results as we
move downstream. Figures 4.9 and 4.10 show the stagnation temperature profiles
corresponding to 3.0 and 10.5 in. downstream, respectively. It is clear that the nu-
merical prediction of the temperature improves as we go downstream of the primary
nozzle exit plane. The ANSYS FLUENT results match well with the WIND-CFD
results.
65
1.5
Experiments (Gilbert and Hill, 1973)
k−ω SST turbulence model (WIND)
1 k−ω SST turbulence model
Spalart−Allmaras turbulence model
0.5
Y/H
0
−0.5
−1
−1.5
0 200 400 600 800 1000 1200
Axial Velocity ft/s
Figure 4.5. 2-D ejector nozzle axial velocity profile at X=3.0 in.
1.5
Experiments (Gilbert and Hill, 1973)
k−ω SST turbulence model (WIND)
1 k−ω SST turbulence model
Spalart−Allmaras turbulence model
0.5
Y/H
−0.5
−1
−1.5
0 200 400 600 800 1000 1200
Axial Velocity ft/s
Figure 4.6. 2-D ejector nozzle axial velocity profile at X=5.0 in.
66
1.5
Experiments (Gilbert and Hill, 1973)
k−ω SST turbulence model (WIND)
1 k−ω SST turbulence model
Spalart−Allmaras turbulence model
0.5
Y/H
0
−0.5
−1
−1.5
0 200 400 600 800 1000 1200
Axial Velocity ft/s
Figure 4.7. 2-D ejector nozzle axial velocity profile at X=7.0 in.
1.5
Experiments (Gilbert and Hill, 1973)
k−ω SST turbulence model (WIND)
k−ω SST turbulence model
1
Spalart−Allmaras turbulence model
0.5
Y/H
−0.5
−1
−1.5
0 200 400 600 800 1000 1200
Axial Velocity ft/s
Figure 4.8. 2-D ejector nozzle axial velocity profile at X=10.5 in.
67
1
Experiments (Gilbert and Hill, 1973)
0.8 k−ω SST turbulence model (WIND)
k−ω SST turbulence model
0.6 Spalart−Allmaras turbulence model
0.4
0.2
Y/H
0
−0.2
−0.4
−0.6
−0.8
−1
540 550 560 570 580 590 600 610 620 630
Total TemperatureoR
Figure 4.9. 2-D ejector nozzle total temperature profile at X=3.0 in.
1
Experiments (Gilbert and Hill, 1973)
0.8 k−ω SST turbulence model (WIND)
k−ω SST turbulence model
0.6 Spalart−Allmaras turbulence model
0.4
0.2
Y/H
−0.2
−0.4
−0.6
−0.8
−1
540 550 560 570 580 590 600 610 620 630
Total TemperatureoR
Figure 4.10. 2-D ejector nozzle total temperature profile at X=10.5 in.
68
4.4.1 Introduction
The CFD simulation of the 2-D ejector nozzle, as discussed in Section 4.3, was
necessary to perform the validation of the CFD tool before using it in the computa-
tional study of the 3-D ejector nozzle with and without clamshell doors. The objective
of this task was to carry out the CFD simulation of a 3-D supersonic cruise ejector
nozzle in subsonic ejector configuration and compare the computational results with
the experiments presented in reference [44]. Cases corresponding to the 3-D ejector
nozzle without clamshell doors and with clamshell doors at an ejector angle of 11.5◦
were considered. This study was further extended to higher pressure ratios and the
performance of the 3-D ejector nozzle was studied at the take-off conditions.
Jones (2009) [44] conducted a set of wind-tunnel experiments to study the charac-
teristics of the ejector nozzle flow with ejectors at different incident angles viz. 0, 5.0,
9.0, 11.5 and 15.0 degrees and low subsonic conditions. One of the objectives of these
experiments was to develop a unique test model which can capture some of the funda-
mental aerodynamic features of the 3-D ejector nozzle. The result of his work was a
test nozzle of scale 0.123 operated at approximately Mthroat =0.25 and ReD =760, 000.
More details about the experimental conditions, the 7-hole probe data acquisition
system mounted on an automated 2-axis traverse instrument, used for the velocity
measurement, and the wind-tunnel setup are available in reference [44]. Experimen-
tal results are available in the form of axial velocity measurements on Y -Z planes at
different axial locations downstream of the nozzle throat.
69
Figure 4.11. CAD model of the 3-D ejector nozzle without clamshell doors [44].
Almost all of the ejector nozzle concepts studied before the 1970s were limited
to the research level and never materialized in practical supersonic air transport
Figure 4.12. CAD model of the 3-D ejector nozzle with clamshell doors [44].
70
applications. The only exception to this is the exhaust nozzle system of the Rolls-
Royce Olympus-593 engine, which powered the world’s first supersonic airliner, the
Concorde [47]. This design served as the key motivation to the present study of
an ejector nozzle with clamshell doors. It consisted of a baseline nozzle with two
asymmetric clamshells serving as ejectors. In order to facilitate the application of
the automated computer numerical controlled (CNC) machining techniques for the
fabrication of the complex 3-D ejector nozzle with clamshell doors design, a high
fidelity, design table driven CAD model was created using the CATIA V6 as part of
the experimental work. The CAD geometry reads the parameters through an MS-
Excel file. This allowed the user to vary the angle of the clamshell doors without the
necessity to redesign the entire nozzle. The same parametric CAD model was used
for the CFD simulations discussed in the present chapter. As mentioned before, two
configurations of the ejector nozzle, one without clamshell doors and the other with
clamshell doors were studied. Figures 4.11 and 4.12 show the CAD geometry of the
3-D ejector nozzle without clamshell doors and with clamshell doors, respectively.
(a) (b)
Figure 4.13. Computational mesh for nozzle walls (a) 3-D ejector
nozzle without clamshell doors (Grid I), and (b) 3-D ejector nozzle
with clamshell doors (Grid II).
71
As a first step towards the computational analysis of the ejector nozzle, a 3-D
grid for the CAD geometry was required. Grids were generated using the Pointwise
GRIDGEN Version 15.10 [48]. Nonoverlapping, multiblock hybrid grids were used for
this geometry. Because of the symmetry, only one quadrant of the domain was used
to save computational cost and time. In the present numerical study, three cases, viz.
an ejector nozzle without clamshell doors (Grid I), an ejector nozzle with clamshell
doors at 11.5◦ (Grid II) and an ejector nozzle with clamshells at 11.5◦ for the CFD
simulation at take-off nozzle pressure ratios (Grid III) were considered.
The computational mesh for the viscous walls of the ejector nozzle with and with-
out clamshell doors are shown in Figure 4.13. Because of the complex design of the
clamshell doors and the support handle on the nozzle, an unstructured mesh was cre-
Figure 4.14. Computational mesh (Grid II) for the entire flow domain
of the 3-D ejector nozzle with clamshell doors for the CFD simulation
at experimental conditions.
72
ated near the nozzle and a structured mesh was used in the rest of the flow domain, as
shown in Figure 4.14. Grid I, corresponded to the ejector nozzle without clamshells
and consisted of 20 blocks including one unstructured block near the nozzle throat,
resulted in a total of 1.29 million cells. Grid II, corresponding to the ejector nozzle
with clamshell doors consisted of the same topography with 20 blocks, resulted in a
total of 2.2 million cells. A variable wall spacing (y) was used for all the viscous wall
boundaries in both the cases to yield a y + value in the range of 30 ∼ 300 and hence
wall functions were used for the near wall turbulence. y + is defined as
+ yU
y = , (4.1)
ν
Figure 4.15. Computational mesh (Grid III) for the entire flow domain
of the 3-D ejector nozzle with clamshell doors for the CFD simulation
at take-off conditions (higher NPR).
73
The CFD simulation at high nozzle pressure ratios, to simulate the take-off con-
ditions with a freestream Mach number of 0.3, required a higher number of nodes
near the walls compared to its low speed counterpart. Therefore, this mesh, termed
as Grid III, consisted of boundary layer grids, as shown in Figure 4.15 to accurately
capture the near wall boundary layers. Similar to Grid II, this mesh was a hybrid
grid with 29 blocks and 2.22 million cells. Grid III included nonuniform hexahedral
cells in the boundary layers, shear layers and freestream blocks, tetrahedral cells in
the unstructured block surrounding the clamshell doors, prism cells in the bound-
ary layers corresponding to the unstructured wall domains, and pyramid cells at the
interface between the structured and unstructured blocks.
Figure 4.16. Extent of the computational domain for the 3-D ejector
nozzle with clamshell doors.
The present study involved the CFD simulation of the jet flows and hence the
outflow and far field boundaries should be far enough such that they do not affect the
jet flow dynamics and unrestricted entrainment. Hence the computational region, as
shown in Figure 4.16 extended 7 DC in the radial direction representing far field and
27 DC in the streamwise direction representing the outflow boundary, where DC is
the control diameter (outer diameter = 202.844 mm) of the nozzle plenum chamber,
as defined in reference [44].
74
Two sets of boundary conditions (BC) were used for the computational meshes
described in Section 4.4.4. BC set I corresponded to the flow conditions used in the
experiments and reported in reference [44]. The numerical values of these boundary
conditions are shown in Figure 4.17. The experimental conditions were low subsonic
of the order of M ≤ 0.25 and hence the velocity-based boundary conditions were
used in the CFD analysis. The boundary conditions were velocity inlet at the inflow
boundary, inviscid wall along the symmetry planes, velocity-inlet along the outer
freestream boundary, and pressure-outlet at the exit. BC set II corresponded to the
take-off conditions which were subsonic of the order of M ≥ 0.3 and hence pressure-
based boundary conditions were used for this CFD simulation, which was an effort to
study the performance of the ejector nozzle in flight conditions. This set of boundary
condition is shown in Figure 4.18.
Inlet Distortion
In the experiments, the inlet flow from the blower was passed into the test rig
through a flow straightener [44]. This helped in the straighting of the flow and re-
moved a lot of the flow nonuniformities. However, some nonuniformity in the velocity
distribution was still present. Figure 4.19 shows the snapshot of the axial velocity
magnitude, measured at the inside of the rig in the absence of the ejector nozzle. It
was evident that the experiments involved nonuniform velocity distribution and hence
an equivalent uniform velocity inlet was calculated for the CFD simulations using the
following procedure.
It was necessary to compute the correct mass flow rate through the rig for the
accurate comparison between the experiments and computations. The mass flow rate
of the nozzle was adjusted in the computation such that it matched the velocity profile
at X/DEQ =1.0. This corrected mass flow rate gave the magnitude of the velocity
inlet used in the CFD simulations. The corrected velocity inlet corresponded to the
least RMS difference between the experiments and computations at X/DEQ =1.0.
The RMS differences were plotted against the inlet velocity magnitude, as shown in
Figure 4.20, and the arithmetic mean of the differences in both Y =0 and Z=0 planes
were used to find the corrected inlet velocity magnitude. Numerical values for the
Simulation I are tabulated in Table 4.2.
The ANSYS FLUENT [29] version 6.3.26 computational solver was used to solve
RANS equations and the system of governing equations were closed using the turbu-
lence models. The CFD simulations of the ejector nozzle without and with clamshell
doors (Simulation I and Simulation II, respectively) at experimental conditions were
performed using the pressure-based coupled implicit solver [30]. As discussed in Sec-
77
0.115
RMS Average
0.11 RMS Difference on Y=0 plane
RMS Difference on Z=0 plane
0.1
0.095
0.09
0.085
0.08
0.075
tion 2.3.1, this algorithm solves the continuity and momentum equations in a coupled
implicit manner using a pressure-velocity coupling algorithm and the coupled alge-
braic multigrid (AMG) solver. The density-based coupled explicit solver was used
for the CFD simulation of the ejector nozzle at take-off conditions (Simulation III)
involving higher pressure ratios.
For Simulation I and Simulation II, corresponding to Grid I and Grid II respec-
tively, a constant density of 1.15 kg/m3 was used. The viscosity was calculated using
Sutherland’s three coefficient method. In the experiments, a region of flow separa-
tion and flow recirculation was identified at the inner surface of the clamshell doors.
Hence, in order to capture the separated flow well, Menter’s k-ω shear stress transport
turbulence model with wall functions was primarily used for the CFD simulations.
Shih’s realizable k- turbulence model with Thies and Tam’s jet flow correction was
also used for Simulation I. The flow domain was initiated using freestream conditions
and the final second-order accurate solution was obtained. A CFL value of 2 was
used for stability reasons. The underrelaxation factors were reduced from their de-
fault values to 0.5 (density), 0.5 (turbulent kinetic energy), 0.8 (specific dissipation
rate), 0.8 (turbulent viscosity) and 0.5 (energy) for numerical stability.
Simulation III was different from the above described two simulations because of
its higher operating nozzle pressure ratio. The density-based coupled explicit solver
was used to solve RANS equations in a coupled manner. This was because of the fact
that the Mthroat ≈ 0.75. Density and viscosity were calculated using the ideal gas
equation and Sutherland’s three coefficient method, respectively. RANS equations
were closed using Menter’s k-ω shear stress transport turbulence model with wall
functions. The flow domain was initiated using freestream conditions and the final
second-order accurate solution was obtained. A CFL value of 1.0 was used for stability
reasons. The underrelaxation factors were reduced from their default values to 0.5
(turbulent kinetic energy), 0.7 (specific dissipation rate) and 0.7 (turbulent viscosity)
for numerical stability.
In order to check the convergence, along with the default solver residual monitor,
the iteration history of the mass-balance and velocity magnitude at one equivalent
diameter downstream of the nozzle throat were monitored. When the mass-balance
reached 1 × 10−5 and the velocity magnitude reached a steady-state value, the solu-
79
tion was considered to be converged. For simulations involving flow separation and
recirculation zone, the velocity magnitude at the monitor point oscillated about a
steady-state value.
4.4.7 Results
The results from the numerical simulation of the ejector nozzle with and with-
out clamshell doors at experimental as well as take-off conditions are discussed in
the present section. The computational results are compared with the experimental
results. The quantitative velocity measurements are normalized using the upstream
plenum axial velocity UP L , which is the centerline axial velocity in the plenum cham-
ber, upstream of the nozzle and the length scale using the equivalent diameter of the
nozzle throat cross-section (DEQ = 5.642 in.), which is defined as,
2
DEQ
At = π , (4.2)
2
where At is the throat area of the nozzle. At first, computational results from the CFD
simulation of the ejector nozzle without clamshell doors using Grid I (Simulation I) is
presented and compared with the experiments. This section is followed by a discussion
on the effect of turbulence models in the prediction of velocity profiles. Results
corresponding to the ejector nozzle with clamshell doors using Grid II are presented
as Simulation II and compared with the experimental measurements, followed by a
discussion on the region of flow separation. The computational results are presented
in the form of normalized axial velocity profiles at different Y -Z planes (different
X/DEQ locations) downstream of the nozzle throat and contour plots of the jet cross-
section at these locations. Finally, the results from the CFD simulation of the ejector
nozzle with clamshell doors at take-off conditions (Simulation III) are discussed. It
is observed that the flow separation near the inner surface of the clamshell doors is
attributed to the ejector nozzle design and not to the nozzle pressure ratio (NPR).
80
Results obtained from the CFD simulation of the ejector nozzle without clamshell
doors are discussed in this section. Without the clamshell doors, the jet behaves like
an elliptic jet because of the elliptic cross-section of the nozzle throat. As the nozzle
was convergent in nature and the conditions were subsonic, the flow was accelerated
to higher velocities. The velocity contour plots of the flow field at the Z=0 plane
corresponding to the k-ω shear stress transport turbulence model is shown in Figure
4.21. The CFD simulation of this configuration was necessary for initial comparison
with the experimental results before going to the more complicated configuration of
a nozzle with clamshells. The velocity contour plot of the flow field corresponding
to the realizable k- model with Thies and Tam’s correction is shown in Figure 4.22.
The lateral spreading (spreading of the jet along Y-axis) is more pronounced in the
81
case of the k-ω SST turbulence model compared to the k- model with Thies and
Tam’s correction.
As discussed earlier, experimental results are available in the form of velocity
measurements on the Y =0 and Z=0 planes and on the Y -Z planes at different X/DEQ
locations downstream of the nozzle throat. Figures 4.23-4.30 show the comparison
of normalized axial velocity magnitude contours at different axial locations between
the experiments and computations. The velocity profiles match very well with the
experimental results at X/DEQ =1.0, 1.5 and 2.0. For the case of X/DEQ =3.0, there
is an overprediction of the axial velocity magnitude. This is believed to be due to
the inability of the turbulence model to predict the length of the potential core of the
jet. At X/DEQ =3.0, as it can be inferred from the experimental results, the potential
core ends and the shear layers start to merge with each other. Figures 4.31-4.38 show
the comparison of the normalized axial velocity profiles between the experiments and
computations.
Figure 4.23. Experimental U/UP L contour plot at X/DEQ =1.0 from the throat [44].
Figure 4.24. Computational U/UP L contour plot at X/DEQ =1.0 from the throat.
83
Figure 4.25. Experimental U/UP L contour plot at X/DEQ =1.5 from the throat [44].
Figure 4.26. Computational U/UP L contour plot at X/DEQ =1.5 from the throat.
84
Figure 4.27. Experimental U/UP L contour plot at X/DEQ =2.0 from the throat [44].
Figure 4.28. Computational U/UP L contour plot at X/DEQ =2.0 from the throat.
85
Figure 4.29. Experimental U/UP L contour plot at X/DEQ =3.0 from the throat [44].
Figure 4.30. Computational U/UP L contour plot at X/DEQ =3.0 from the throat.
86
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2.2
Experiments (Jones, 2009)
2 k−ω SST turbulence model
Realizable k−ε with Theis and Tam’s Constants
1.8
1.6
1.4
1.2
0.8
0.6
0.4
0.2
−5 0 5 10 15 20 25 30 35
X/DEQ
Effect of Turbulence Models In addition to the Menter’s k-ω shear stress trans-
port turbulence model, Shih’s realizable k- turbulence model with Thies and Tam’s
jet flow correction was used to study the effect of turbulence model in predicting the
separated ejector jet flow. Figure 4.39 shows the velocity magnitude contour of jet
cross-section at X/DEQ =3.0 corresponding to Thies and Tam’s k- turbulence model.
It is evident that the lateral spreading is more pronounced in the latter compared to
the k-ω SST model, but no significant improvement is observed in terms of the axial
velocity profiles.
92
2
Experiments
1.8 k−ω SST
Real. k−ε
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 k−ω SST
Real. k−ε
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
In addition to the case of the ejector nozzle without clamshell doors, the CFD
simulation of the ejector nozzle with clamshell doors at 11.5◦ was considered in the
present study to complement the experimental findings. In addition to the survey
planes similar to the case without clamshell doors, at X/DEQ =1.0, 1.5, 2.0, 3.0 down-
stream of the nozzle exit, experimental values were also measured at X/DEQ =0.42
which was inside the clamshell doors. This survey plane helps in the understanding
of the separated flow and provides a better comparison between the experiments and
computations.
The contours of the velocity magnitude on the Z=0 symmetry plane (ejector
plane) is shown in Figure 4.44. A region of flow separation and recirculation, as
found during the experiments, is encountered. This is because of the inability of
the resulting free shear layer to attach to the inner surface of the clamshells, as
discussed in Section 2.2.1 and thereby allows the external atmospheric flow to affect
the nozzle exhaust. Figures 4.45-4.54 show the comparison of the jet cross-section
axial velocity contours between the experiments and computations. As with the case
without clamshell doors, similar good agreement at X/DEQ =0.42, 1.0, 1.5, 2.0 and
overprediction at X/DEQ =3.0 are found. The quantitative comparisons of normalized
axial velocity in the Y =0 and Z=0 planes and at different axial locations are shown
in Figures 4.55-4.64.
95
Figure 4.46. Computational U/UP L contour plot at X/DEQ =0.42 from the throat.
96
Figure 4.47. Experimental U/UP L contour plot at X/DEQ =1.0 from the throat [44].
Figure 4.48. Computational U/UP L contour plot at X/DEQ =1.0 from the throat.
97
Figure 4.49. Experimental U/UP L contour plot at X/DEQ =1.5 from the throat [44].
Figure 4.50. Computational U/UP L contour plot at X/DEQ =1.5 from the throat.
98
Figure 4.51. Experimental U/UP L contour plot at X/DEQ =2.0 from the throat [44].
Figure 4.52. Computational U/UP L contour plot at X/DEQ =2.0 from the throat.
99
Figure 4.53. Experimental U/UP L contour plot at X/DEQ =3.0 from the throat [44].
Figure 4.54. Computational U/UP L contour plot at X/DEQ =3.0 from the throat.
100
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
−0.2
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
−0.2
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
2
Experiments
1.8 CFD
1.6
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Y/DEQ
2
Experiments
1.8 CFD
1.6
Axial Flow Speed U/UPLENUM
1.4
1.2
0.8
0.6
0.4
0.2
0
−1.5 −1 −0.5 0 0.5 1 1.5
Z/DEQ
The overall objective of the design and analysis of the 3-D ejector nozzle with
clamshell doors was to study the performance of the nozzle for subsonic take-off
conditions. As discussed in Chapter 2, the optimum performance of the exhaust
system during subsonic take-off and approach with minimum noise is as challenging
as supersonic cruise. The performance study of nozzle during take-off required a
CFD analysis with take-off conditions because experimental conditions were limited
to low subsonic Mach numbers. Hence a third simulation was performed with take-off
conditions and results are presented in this section.
Figure 4.65. Mach number contour plot on the Z=0 symmetry plane
of the 3-D ejector nozzle with clamshell doors at take-off conditions.
The primary objective of this analysis was to predict if the flow separation occurs
even at higher nozzle pressure ratios. Figure 4.65 shows the presence of the flow
separation and recirculation zone at the inner surface of the clamshell doors even
at a higher nozzle pressure ratio which is detrimental to the overall performance of
the exhaust nozzle system. The contours of Mach number at X/DEQ =0.5 plane are
106
shown in Figure 4.66. The zone of reverse flow is observed near the clamshell inner
surface.
Figure 4.66. Mach number contour plot at X/DEQ =0.5 plane down-
stream of the 3-D ejector nozzle throat at take-off conditions.
4.5 Conclusion
A comprehensive computational study of the ejector nozzle has been carried out.
At first, a 2-D ejector nozzle test case was used to perform the validation task of the
available computational tool. It was found that the CFD results from the ANSYS
FLUENT flow solver were in good agreement with experiments and WIND-CFD
results for 2-D ejector nozzle test case. The one-equation Spalart-Allmaras turbulence
model performed better than the k-ω SST turbulence model in the prediction of mean
flow variables downstream from the nozzle exit plane.
The computational analysis of the 3-D ejector nozzle with and without clamshell
doors at experimental conditions was successfully carried out and the results were
compared with the experiments. Regions of flow separation, observed in the experi-
107
ments were well captured. Menter’s k-ω SST turbulence model predicted the mean
flow field very well within the potential core. However, it deviated from the experi-
mental results away from the nozzle exit because of the overprediction of the potential
core length.
The CFD simulation of the full-scale ejector nozzle was successfully carried out
at take-off conditions with Mthroat ≈ 0.75. A new grid with boundary layer mesh
was created to accurately predict the near wall turbulence. The flow separation and
recirculation zone were also observed at higher nozzle pressure ratios. This proposes
additional challenges in the design and development of a successful noise suppression
exhaust nozzle system. In order to make use of the advantages of ejector nozzle such
as thrust augmentation and noise suppression, it was necessary to remove the flow
separation and recirculation zones. One idea to address this challenge was to make
use of additional passive mixing devices such as tabs or chevrons at the ejector slot
to introduce streamwise vorticity and thereby enhance mixing. This will be discussed
in detail in the next chapter.
108
5.1 Introduction
The experimental testing of the baseline ejector nozzle with clamshell doors was
documented in the reference [44]. Its detailed CFD analysis, as discussed in Chapter 4,
showed the presence of a zone of separation and recirculation near the inner surface of
the clamshells. This had detrimental effects on the advantages of the ejector nozzle
with clamshell doors, such as reduced thrust augmentation and noise suppression.
The flow was separated because of the inability of the free shear layers, originating
from the primary nozzle surface, to attach to the inner surface of the clamshells. This
phenomenon was studied in detail by Der [17] and is explained briefly in Section 2.2.1.
One of the proposed measures to overcome flow separation and recirculation zones,
as documented in [44], was to introduce streamwise vortices in the ejector flow by
the application of passive mixing devices such as chevrons or tabs, thereby enhancing
the mixing between the ejector flow and the nozzle flow. The concept of chevrons
is not new and has already been used in civil air-transportation powerplants such
as the Rolls-Royce Trent-1000 and the General Electric GE-NX. The application
of chevrons on the ejector nozzle results in the spreading of the jet and forces the
shear layer to attach to the inner surface of the clamshells which can reduce flow
separation. In addition to the ejector nozzle performance improvement, chevrons
have noise suppression capability in the low frequency part of the spectrum.
For the above mentioned reasons, a preliminary design work involving the design
of chevrons on the ejector nozzle with clamshell doors and its detailed CFD simulation
was carried out. This chapter summarizes the work done related to the design and
computational analysis of the ejector nozzle with clamshell doors and chevrons. It was
109
found that the extent of the flow separation was greatly reduced by the application
of chevrons.
5.2 Objectives
The objective of the current task was to modify the existing design of the baseline
ejector nozzle by placing chevrons at the throat of the primary nozzle and perform
computational analysis of the new design. The performance of these passive mixers
was studied with respect to the reduction of the separation zone encountered near the
inner surface of the clamshells and the amount of ejector mass flow. Two nozzle con-
figurations with a different number of chevrons were designed and the computational
analysis of their flow fields is presented.
The ejector nozzle is preferable for subsonic operations of a supersonic cruise jet
engine because of its thrust augmentation (achieved by increasing the nozzle mass
flow through entrainment) and noise suppression (by reducing the exhaust velocity
of the nozzle jet). For these reasons, powerplants of the SR-71 and the Concorde
aircraft use the ejector nozzle concept. The addition of chevrons on the primary
nozzle throat surface results in a complex three-dimensional flow phenomenon as
shown in Figure 5.1.
The ejector nozzle with chevrons introduces counter-rotating streamwise vortices
into the primary nozzle flow. These kidney-shaped vortices interact with the ejector
flow and the primary nozzle flow. This results in an increased mixing and outward
spreading of the shear layer which finally attaches to the inner surface of the clamshell
doors. The design of the chevron is critical from the aeroacoustic point of view. The
enhanced mixing results in additional small-scale eddies which produce high frequency
noise. Hence the design of the chevron should be such that it increases the mixing
with minimum high frequency noise penalty. This can be achieved by the use of
advanced optimization techniques.
The baseline ejector nozzle geometry used in the current design task was the
same as used in Section 2.2.1 which was scaled up (8.13 : 1) to represent the full-
scale flight geometry. The baseline CAD geometry, parametrically defined using the
CATIA CAD package was exported as a STEP file for better compatibility with the
Pro/Engineer CAD package. Chevrons were designed based on the dimensions from
the previous study of Janardan et al. and documented in reference [21]. Various
dimensional variables used in the design of the chevron are shown in Figure 3.17 and
their numerical values are given in Table 5.1.
The chevron on the nozzle surface was created using the Pro/Engineer Wildfire
4.0 CAD package. Solid modeling operations such as extrusion and subtraction were
111
used to cut the nozzle throat surface in the form of chevrons. In the current study,
two designs were implemented which differs from each other in the total number of
chevrons and their dimensions. Design I consisted of 12 chevrons resulting in the
chevron-crest on the ejector nozzle Z=0 symmetry plane. Design II was based on the
dimensions corresponding to 14 chevrons and resulted in the chevron-trough on the
ejector nozzle Z=0 symmetry plane. The chevron-crest is defined as the peak of the
chevron and the chevron-trough is defined as the middle point in between the two
chevron peaks.
Table 5.1 Dimensions of the chevron on the 3-D ejector nozzle with
clamshell doors for Design I and Design II.
The presence of the clamshell door-support allowed only 8 and 10 chevrons for
Design I and Design II, respectively. Figure 5.2 shows the X-sectional view of the
ejector nozzle at the throat with 8 chevrons (Design I). The clamshell doors were hid-
den in this view for better visualization. Similarly, Figure 5.3 shows the X-sectional
view of the ejector nozzle with 10 chevrons configuration (Design II). The isometric
views of the ejector nozzle with 8 chevrons and 10 chevrons are shown in Figures 5.4
and 5.5, respectively.
112
Figure 5.2. CAD geometry of the ejector nozzle with clamshells and
chevrons, Design I - X-section at the throat.
Figure 5.3. CAD geometry of the ejector nozzle with clamshells and
chevrons, Design II - X-section at the throat.
113
Figure 5.4. CAD geometry of the ejector nozzle with clamshells and
chevrons, Design I - Isometric view.
Figure 5.5. CAD geometry of the ejector nozzle with clamshells and
chevrons, Design II - Isometric view.
114
Grids for the computational analysis of two designs described above were created
using the Pointwise GRIDGEN version 15.10 grid generation code. The nozzle ge-
ometry was exported in the IGES format from the Pro/Engineer CAD package with
edges and surfaces as the required entities. Edges are required for the creation of con-
nectors and surfaces are required for creating databases on which GRIDGEN projects
the computational mesh.
Figure 5.6. Computational mesh for ejector nozzle walls and chevrons - Design I.
A 3-D, multiblock, nonoverlapping hybrid grid, similar to the one used in Chap-
ter 4, was created. An unstructured mesh was used on chevron surfaces which were
extruded as prisms for the boundary layer mesh. Structured blocks were used for the
far field and shear layer region. The nozzle geometry was symmetric about the Y =0
and Z=0 plane. Hence a quadrant of the geometry was used for grid generation and
CFD simulation.
115
Figure 5.7. Computational mesh for ejector nozzle walls and chevrons - Design II.
Table 5.2 Boundary conditions for the CFD simulation of the ejector
nozzle with clamshell doors and chevrons.
Design II consisted of 31 blocks resulting in a total of 3.56 million cells. This mesh is
shown in Figure 5.7.
The boundary conditions for the present CFD simulation were the take-off con-
ditions with a NPR of 1.7, as discussed in Section 4.4.5. The ANSYS FLUENT
flow solver recommends the use of pressure-based boundary conditions for high NPR
and Mach numbers for numerical stability and faster convergence. Moreover, in the
ANSYS FLUENT version 6.3.26 flow solver, the ideal gas equation (required for the
definition of density) can be used only with pressure-based boundary conditions.
For the above mentioned reasons, pressure-based boundary conditions were used
for various inlets, far field and outlet boundaries. The boundary conditions were
pressure-inlet at the inflow boundary, inviscid-wall along symmetry planes, pressure
far field along the outer freestream boundary, and pressure-outlet at the exit. The
numerical values of the boundary conditions are shown in Table 5.2.
117
Turbulence quantities were also required as boundary conditions for the simulation
of the turbulent flow. The ANSYS FLUENT flow solver allowed the use of any two of
the turbulent variables viz. the turbulent kinetic energy (k), turbulent intensity (I),
turbulent viscosity ratio (β), turbulent length scale (l), and the secondary variable.
The secondary variable depended upon the turbulence model used, such as νt , , l,
and ω for various two-equation turbulence models. In this simulation, the turbulent
intensity and the turbulence length scale were used as the turbulent inlet boundary
condition for the nozzle inlet. The turbulent intensity and the viscosity ratio values
were used for the freestream inlet and the pressure far field.
The ANSYS FLUENT version 6.3.26 flow solver was used for the 3-D, steady CFD
simulation. The solver settings were similar for the CFD simulation of both Design
I and Design II. A NPR of 1.7 results in a throat Mach number on the order of 0.8
and hence the computational simulation of compressible RANS equations mandate
a coupled solver. For this reason, the density-based explicit coupled solver was used
for the numerical stability and better convergence. In general, the explicit coupled
solver requires less computational time compared to the implicit coupled solver. The
system of RANS equations was closed using the Menter’s k-ω SST turbulence model
with wall functions.
The operating conditions were based on the absolute pressure (instead of the
gauge pressure) and the ideal gas equation was used for the definition of the density.
The viscosity of the air was calculated using Sutherland’s three coefficient method
and the temperature dependent thermal conductivity was implemented in the current
simulation. The computational solution was initiated using the freestream primary
flow variables. The first-order solution was obtained after 5000 iterations and then
the discretization schemes were changed to second-order upwind for all the primary
variables to obtain the final second-order converged solution.
118
The converged solution was obtained by using the underrelaxation values of 0.5
for the turbulent kinetic energy, 0.7 for the specific dissipation rate and 0.7 for the
turbulent viscosity. A CFL value of 1.0 was used because of the numerical stability
issues associated with the hybrid grid. The convergence was monitored using residuals
of flow variables, the iteration history of the velocity magnitude at X/DEQ =3.0, and
the mass-balance between the inflow and outflow. When the velocity magnitude
reached a steady state value, the solution was considered to be converged. In this
case, the iteration history of the velocity magnitude oscillated about a mean converged
value because of the presence of a separation bubble on the clamshells.
5.8 Results
The results from the CFD simulation of the ejector nozzle with clamshell doors and
chevrons are discussed in this section. These results are compared with the ejector
nozzle without chevrons. The CFD post-processing of the flow field showing contours
of the TKE and the Mach number corresponding to the case of 14 chevrons (Design
II) is shown in Figure 5.9. Earlier in Section 4.4.7, we found that the baseline ejector
nozzle with clamshell doors resulted in a zone of separation at take-off conditions as
shown in Figure 5.10. Mach number contours of the cross-section of the nozzle jet,
inside the clamshell doors at about X/DEQ =0.5 is shown in Figure 5.11.
5.8.1 Design I
Design I consisted of 8 actual chevrons, with the chevron-crest on the Z=0 sym-
metry plane as shown in Figure 5.2. In the chevron-trough plane, the high speed
nozzle flow entrained into the shear layer and resulted in its attachment on the inner
surface of the clamshell doors. Figure 5.12 shows the contours of the Mach number
and the attachment of the shear layers in the chevron-trough plane.
In the chevron-crest plane, which is aligned with the axis of the kidney vortex
and Z=0 symmetry plane, there is not much entrainment of the high speed flow.
119
Hence the flow separated from the clamshell doors after a certain distance along the
nozzle axis. Figure 5.13 shows the contours of the Mach number in the chevron-
crest plane. It is evident that the separation near the inner surface of the clamshells
decreased in size when compared to the baseline ejector nozzle case without chevrons
(Figure 5.10). Figure 5.14 shows the jet cross-sectional contours inside the clamshell
doors at X/DEQ = 0.5.
5.8.2 Design II
Design II was based on the dimensions for 14 chevrons. The presence of the
clamshell door-supports allowed the placement of only 10 actual chevrons. Design II
was different from Design I in the sense that the chevron-trough was aligned with the
nozzle Z=0 symmetry plane. This resulted in a clocking of the vortices so that high
speed flow was entrained into the shear layer on the plane where the maximum flow
separation was present.
Figure 5.15 shows the contours of the Mach number on the Z=0 symmetry plane
which coincides with the chevron-trough plane. It is apparent that the flow separation,
observed in the case of the baseline nozzle and Design I, is completely removed in the
chevron-trough plane because of the attachment of the shear layers.
On the chevron-crest plane, a region of flow separation is observed as shown in
Figure 5.16. Therefore, the flow separation zone observed in Design I is redistributed
and divided into two smaller zones by increasing the number of chevrons from 12 to 14.
These two zones are evident in Figure 5.17 which shows the nozzle jet cross-sectional
contours of the Mach number at X/DEQ = 0.5.
Therefore, it is observed that the extent of flow separation and the recirculation
zones is decreased considerably with the application of chevrons. Each chevron results
in the formation of a counter-rotating vortex in the streamwise direction which are
of the shape of a kidney. This causes an enhanced mixing between the nozzle flow
and the ejector flow and the shear layer spreads more outwards. This results in
120
an improvement on the nozzle performance. The nozzle flow stays attached to the
clamshell’s inner surface entirely on the chevron-trough plane and until around half
the axial length of the clamshell doors on the chevron-crest plane. The recirculation
zone is still present at the rear end of the clamshells.
Nozzle flow variables along the nozzle-axis are of utmost importance in under-
standing the characteristics of the jet. Figure 5.18 shows the distribution of the
centerline velocity magnitude with respect to the normalized axial distance along the
streamwise direction. The fundamental characteristics of jet flows such as the con-
stant velocity potential core and inverse-spreading of the jet with respect to axial
distance are well captured. The oscillations in the potential core region shows the
presence of weak Mach waves. It was observed that the length of the potential core
was longer for the chevron nozzle when compared with the baseline nozzle. This may
be because of the reason that the separated jet flow in the baseline case pushes the
streamlines closer creating a smaller jet; and therefore a shorter jet potential core
length. In conclusion, this issue of longer potential core length in the case of chevrons
when compared with the baseline design is not well understood.
The variation of the total temperature with respect to the axial distance is shown
in Figure 5.19. Increasing the number of chevrons from 12 in Design I to 14 in Design
II resulted in enhanced mixing. This is evident from the decrease in the length of the
potential core for Design II when compared with Design I.
One of the effects of the ejector nozzle is the thrust augmentation. This is because
of the addition mass flow introduced into the primary nozzle flow through the ejector
slots. Therefore, the thrust performance of the ejector nozzle is dependent on the
ejector flow. One of the objectives of this design study was to analyze the effect of
121
chevrons on the ejector flow. It is observed that the addition of chevrons resulted in
an increased nozzle-inlet mass flow by 10.7% for Design I and 9.4% for Design II. The
increased nozzle inlet mass flow is also evident in Figure 5.18 showing the centerline
variation of the velocity magnitude.
Table 5.3 shows a quantitative measure of the secondary flow, entrained into the
primary nozzle flow for the baseline ejector nozzle, the ejector nozzle with 12 chevrons
and the ejector nozzle with 14 chevrons. The ejector performance is represented by
the ratio of the secondary mass flow entrained through the ejector slot (ṁej ) to the
primary nozzle mass flow (ṁin ). It was observed that the increase in the number
of chevrons from 12 to 14 resulted in an improved mass entrainment because of the
enhanced mixing. However, the mass entrainment was diminished in the case of
12 chevrons when compared with the baseline design. The reason behind this flow
phenomenon was not well understood.
122
Figure 5.10. Mach number contour plot Z=0 symmetry plane for the
ejector nozzle without chevrons.
Figure 5.13. Mach number contours on Z=0 symmetry plane for the
ejector nozzle with chevrons - Design I.
125
Figure 5.15. Mach number contours on Z=0 symmetry plane for the
ejector nozzle with chevrons - Design II.
126
550
Without chevrons
With 12 chevrons
500
With 14 chevrons
450
350
300
250
200
150
−5 0 5 10 15 20 25 30 35
Axial Length X/Deff
900
Without chevrons
With 12 chevrons
With 14 chevrons
800
Total Temperature K
700
600
500
400
300
−5 0 5 10 15 20 25 30 35
Axial Length X/Deff
5.9 Conclusion
The preliminary design of the ejector nozzle with clamshells and chevrons was
completed and the computational results obtained by CFD simulation were discussed.
The zone of flow separation, observed on the inner surface of the clamshells in the
case of the baseline ejector nozzle, was greatly reduced by the application of chevrons
on the nozzle throat surface.
Two configurations with a different number of chevrons were designed and their
computational simulations were performed. Design I consisted of 12 chevrons which
resulted in the alignment of the chevron-crest plane with the Z=0 symmetry plane.
Design II consisted of 14 chevrons and ensured that the chevron-trough plane was
aligned with the Z=0 symmetry plane.
Centerline statistics showed that an increase in the number of chevrons from
12 to 14 resulted in enhanced mixing and a reduction in the potential core length.
However, the phenomenon of the increase in the potential core length for chevrons
when compared with the baseline nozzle was not well understood and is an issue for
future study.
It was found that the application of chevrons resulted in increased effective throat
area in Design I and Design II. The increased throat area caused a mismatch in
the mass inflow between the chevron nozzle and the baseline nozzle. Hence the CFD
simulation of the 3-D ejector nozzle with 14 chevrons (Design II) with a nozzle effective
throat area equal to the baseline nozzle was necessary for the better comparison of
centerline statistics and the ejector performance.
129
The computational study of noise suppression exhaust nozzle systems has been suc-
cessfully carried out. Three-dimensional RANS computations were performed on
exhaust nozzles such as the three-stream separate-flow axisymmetric plug nozzle, the
three-stream separate-flow chevron nozzle, the 3-D ejector nozzle with clamshells,
and the 3-D ejector nozzle with clamshells and chevrons. CFD simulations were car-
ried out at low speed wind-tunnel experimental conditions and high NPR take-off
conditions.
The accuracy of the computational prediction of jet flows and the associated noise
depends on the type of the turbulence closure method used. Large computational time
and computing resources associated with high-end turbulence prediction methods
such as DNS, LES, and DES restrict their application in the preliminary design cycle.
Hence there is a need to study the ability of two-equation turbulence models in the
prediction of the mean flow field and turbulence characteristics. CFD simulations
of the baseline three-stream separate-flow axisymmetric plug nozzle with chevrons
(3A12B) and without chevrons (3BB) were performed for this task. It was found
that even though the realizable k- turbulence model with Thies and Tam’s jet flow
correction gave better prediction of the mean flow, it failed to predict turbulent flow
quantities. The standard k- turbulence model suffered from the conventional problem
of the overprediction of the turbulent kinetic energy. Hence it was concluded that the
k-ω SST turbulence model was the preferred turbulence model for the present study.
However, it was observed that the k-ω SST turbulence model overpredicted the jet
potential core.
The CFD simulation of the 2-D ejector nozzle test case, used for the validation task
showed a good agreement with the experiments. The Spalart-Allmaras one-equation
130
turbulence model performed better than Menter’s k-ω SST turbulence model. The
prediction of the mean flow field improved away from the nozzle exit.
The RANS-based computation of the 3-D ejector nozzle with clamshell doors
using the k-ω SST turbulence model showed good agreement with the wind-tunnel
experiments. The experiments were conducted at low subsonic conditions (M ≤ 0.25).
The computational results compared well with the experiments within the potential
core. However, the overprediction of the potential core resulted in an overprediction
of mean flow quantities at X/DEQ =3.0. Flow features such as the flow separation and
the recirculation zone encountered during the flow visualization were well captured
in computations.
The CFD simulation of the 3-D ejector nozzle with clamshell doors at take-off
conditions (high NPR of the order of 1.7) showed similar flow characteristics, i.e.
flow separation and recirculation zones. This confirmed that the flow separation was
attributed to the nozzle design and not to the flow nozzle pressure ratios.
Application of chevrons on ejectors was examined. The CFD simulation of the 3-D
ejector nozzle with clamshell doors and chevrons showed some improved flow features.
The flow separation zone was decreased significantly. A zone of recirculation remained
at the trailing edge of the clamshell doors. On the chevron plane, the flow separation
was completely removed and the shear layers attached to the clamshell doors giving
improved performance.
It was found that the application of chevrons resulted in an increased effective
throat area in Design I and Design II. This caused a mismatch in the mass inflow
between the chevron nozzle and the baseline nozzle. Hence the CFD simulation of
the 3-D ejector nozzle with 14 chevrons (Design II) with a nozzle effective throat
area equal to the baseline nozzle is necessary for the better comparison of centerline
statistics and the ejector performance. A parametric study on the number of chevrons
will be necessary for further reduction of the extent of the flow separation.
Ô×ÍÌ ÑÚ ÎÛÚÛÎÛÒÝÛÍ
131
LIST OF REFERENCES
[44] J. Jones. Wind Tunnel Testing of a Supersonic Cruise Nozzle in Subsonic Ejector
Configuration. Master’s thesis, School of Aeronautics & Astronautics, Purdue
University, West Lafayette, Indiana, USA, May 2009.
[45] N. J. Georgiadis, T. Chitsomboon and J. Zhu. Modification of the Two-Equation
Turbulence Model in NPARC to a Chien Low Reynolds Number k-epsilon For-
mulation. Technical Report NASA TM-106710, NASA Glenn Research Center,
Cleveland, Ohio, September 1994.
[46] G. B. Gilbert and P. G. Hill. Analysis and Testing of Two-Dimensional Slot
Nozzle Ejectors With Variable Area Mixing Sections. Technical Report NASA
CR-2251, NASA Glenn Research Center, Cleveland, Ohio, May 1973.
[47] S. Candel. Concorde and The Future of Supersonic Transport. Journal of Propul-
sion and Power, 20(1), 2004.
[48] Pointwise, Inc., Fort Worth, TX. GRIDGEN v15.09 User Manual, 2006.