Dramjet Theis

11/23/2003
The Pennsylvania State University

The Graduate School

Department of Mechanical and Nuclear Engineering

THRUST CHAMBER DYNAMICS
AND PROPULSIVE PERFORMANCE OF
AIRBREATHING PULSE DETONATION ENGINES

A Thesis in

Mechanical Engineering

by

Fuhua Ma

2003 Fuhua Ma

Submitted in Partial Fulfillment
of the Requirements
for the Degree of

Doctor of Philosophy

December 2003

11/23/2003

The thesis of Fuhua Ma was received and approved* by the following:

Vigor Yang
Distinguished Professor of Mechanical Engineering
Thesis Advisor
Chair of Committee

Robert J. Santoro
Guillet Professor of Mechanical Engineering

John Mahaffy
Associate Professor of Mechanical Engineering

Robert F. Kunz
Associate Professor of Aerospace Engineering

Richard C. Benson
Professor of Mechanical Engineering
Head of Department of Mechanical Engineering

*Signatures are on file in the Graduate School.

iii
11/23/2003
Abstract
Pulse detonation engines (PDEs) have recently been recognized as a promising
propulsion technology that offers potential advantages in thermodynamic cycle
efficiency, hardware simplicity, and operation scalability. The present work studies the
flow dynamics and system performance of airbreathing PDEs with a stoichiometric
hydrogen/air mixture. The system includes a supersonic inlet, an air manifold, a rotary
valve, a single-tube or multitube combustor, and a convergent-divergent nozzle. The
flight condition involves an altitude of 9.3 km and a flight Mach number of 2.1.
The supersonic inlet dynamics is analyzed through axisymmetric two-dimensional
simulations based on the Harten-Yee upwind total-variation-diminishing scheme.
Turbulence closure is achieved by a two-equation model. In addition to the steady-state
inlet flow dynamics, the response of the inlet shock system to downstream disturbances is
studied by imposing periodic pressure oscillations at the exit plane. A wide range of
fluctuation frequency and amplitude are investigated. In general, the acoustic response of
the inlet flow increases with increasing amplitude of the imposed oscillation, but
decreases with the frequency.
Both one- and two-dimensional simulations based on the recently developed
Space-Time conservation element/solution element method are carried out for single-tube
PDEs. The two-dimensional code is further parallelized using the message-passing-
interface library and a domain decomposition technique for unstructured grid. The flow
dynamics, the effects of the operation timing and nozzle configuration on the propulsive
performance, and the various loss mechanisms are examined. Results show that an
iv
11/23/2003
optimum cycle frequency exists for a given configuration. For a given frequency and
purge time, a longer refilling period increases the specific thrust of PDEs considered.
The nozzle studies indicate that the convergent-divergent nozzle significantly increases
the propulsive performance. Moreover, the throat area of the convergent-divergent
nozzle plays a more important role than the length.
Effort is also expended to study the flow dynamics and propulsive performance of
multitube PDEs. Comparison with the single-tube results demonstrates that the multitube
design improves the engine performance in terms of specific impulse, operation
steadiness, and timing range. The effect of the system geometry is partially assessed by
considering a free volume between the detonation tubes and the common nozzle. Results
indicate that the free volume helps to reduce the imperfect nozzle expansion loss and
improve the engine steadiness. However, it also induces more complicated shock waves
and increases the internal flow loss. The overall effect is a decrease in the propulsive
performance.
v
11/23/2003
Table of Contents
List of Figures........................................................................................................viii
List of Tables......................................................................................................... xiv
Nomenclature ......................................................................................................... xv
Acknowledgements .............................................................................................xviii
Chapter 1 Introduction............................................................................................. 1
1.1 Background and Motivation.................................................................................... 1
1.2 Detonation Physics.................................................................................................. 3
1.2.1 CJ Theory .................................................................................................................... 4
1.2.2 ZND Detonation Wave Structure .............................................................................. 10
1.2.3 ZND Detonation Wave Propagation in a Tube.......................................................... 14
1.3 Pulse Detonation Engine Concept......................................................................... 19
1.3.1 Cycle Operation of PDE............................................................................................ 20
1.3.2 Potential Advantages of PDE.................................................................................... 22
1.3.3 Typical Structure of PDE .......................................................................................... 25
1.4 Literature Survey on PDE Studies ........................................................................ 28
1.4.1 Experimental Studies on PDEs.................................................................................. 31
1.4.1.2 Impulse Measurements ............................................................................................ 34
1.4.1.3 Detonation Initiation................................................................................................ 35
1.4.2 Numerical Studies on PDEs ...................................................................................... 39
1.4.2.2 One-Dimensional Numerical Simulations ............................................................... 42
1.4.2.3 Two-Dimensional Numerical Simulations............................................................... 44
1.4.2.4 Effect of Nozzles ..................................................................................................... 46
1.4.2.5 Multitube PDEs........................................................................................................ 50
1.4.3 Analytical Studies on PDEs....................................................................................... 51
1.5 Research Objectives.............................................................................................. 54
Chapter 2 Supersonic Inlet Dynamics ................................................................... 58
2.1 Problem Description ............................................................................................. 59
2.2 Governing Equations............................................................................................. 60
2.3 Numerical Procedures and Validations................................................................. 62
2.4 Steady-State Flow Analysis .................................................................................. 64
2.5 Shock/Acoustic-Wave Interactions....................................................................... 68
2.6 Conclusions........................................................................................................... 79
vi
11/23/2003
Chapter 3 Quasi-One-Dimensional Analyses of PDEs ......................................... 80
3.1 Space-Time CE/SE Method.................................................................................. 80
3.1.1 Introduction to Space-Time CE/SE Method.............................................................. 80
3.1.2 Space-Time CE/SE Method for One-Dimensional Problems ................................... 84
3.2 Governing Equations............................................................................................. 88
3.3 Parameter Calibration and Code Validation ......................................................... 89
3.3.1 Parameter Calibration................................................................................................ 89
3.3.2 ZND Detonation in a One-Dimensional Tube........................................................... 92
3.3.3 Flow Through a Convergent-Divergent Nozzle ........................................................ 97
3.4 Flow Dynamics ..................................................................................................... 98
3.5 Parametric Study................................................................................................. 106
3.5.1 Effect of Valve Timing............................................................................................ 106
3.5.2 Effect of Nozzle Throat Area .................................................................................. 109
Chapter 4 Two-Dimensional Analyses of Single-Tube PDEs ............................ 111
4.1 Governing Equations........................................................................................... 111
4.2 Numerical Treatment and Parallel Implementation............................................ 113
4.3 Model Validation ................................................................................................ 114
4.4 Calculation of Propulsive Performance .............................................................. 116
4.5 Straight-Tube PDE.............................................................................................. 118
4.5.1 Problem Setup ......................................................................................................... 118
4.5.1.2 Domain Decomposition ......................................................................................... 119
4.5.1.3 Boundary Conditions ............................................................................................. 121
4.5.1.4 Initial Conditions ................................................................................................... 121
4.5.1.5 Detonation Initiation.............................................................................................. 121
4.5.2 Results ..................................................................................................................... 122
4.6 Single-Tube PDE with CD Nozzle ..................................................................... 131
4.6.1 Flow Evolution........................................................................................................ 133
4.6.1.1 Effect of Ambient Flow......................................................................................... 148
4.6.2 Propulsive Performance........................................................................................... 150
4.6.3 Analytical Prediction of Propulsive Performance ................................................... 152
4.6.4 Loss Mechanisms .................................................................................................... 157
4.6.6 Effect of Nozzle Throat and Length........................................................................ 168
4.7 Summary and Conclusions.................................................................................. 174
Chapter 5 Two-Dimensional Analyses of Multitube PDEs ................................ 177
5.1 System Configuration ......................................................................................... 178
5.2 Results and Discussion........................................................................................ 181
5.2.1 Flow Evolution........................................................................................................ 182
vii
11/23/2003
5.2.2 Propulsive Performance and Loss Mechanisms ...................................................... 198
5.2.4 Effect of System Geometry ..................................................................................... 207
5.3 Summary and Conclusions.................................................................................. 210
Chapter 6 Summary and Future Work ................................................................ 213
6.1 Summary ............................................................................................................. 213
6.2 Major Contributions and Conclusions ................................................................ 216
6.3 Recommendation for Future Work ..................................................................... 217
Appendix A Thermodynamic Cycle Efficiencies of Brayton, Humphrey, and
Ideal PDE Cycles.................................................................................................. 219
A.1 Brayton cycle ..................................................................................................... 220
A.2 Humphrey cycle ................................................................................................. 221
A.3 Ideal PDE cycle.................................................................................................. 222
Appendix B Jacobian Matrices for Quasi-One- and Two-Dimensional Systems
.............................................................................................................................. 223
B.1 Quasi-One-Dimensional System........................................................................ 223
B.2 Two-Dimensional System.................................................................................. 224
Appendix C Procedures of Building x-t Diagram............................................... 226
Appendix D Space-Time CE/SE Method for Two-Dimensional Problems........ 227
D.1 Governing Equations and Their Integral Form.................................................. 227
D.2 Triangular Mesh, Conservation Element, and Solution Element....................... 228
D.3 Basic Solution Procedure ................................................................................... 230
D.4 The Extended Scheme........................................................................................ 234
D.5 Source Term Treatment...................................................................................... 237
D.5.1 Explicit Treatment of Source Term........................................................................ 238
D.5.2 Implicit Treatment of Source Term........................................................................ 239
Appendix E Reflecting Boundary Conditions on an Inviscid Solid Wall........... 241
Appendix F Determination of the Time Instant of the Arrival of the First
Rarefaction Wave on the Head End ..................................................................... 245
Bibliography......................................................................................................... 249

viii
11/23/2003
List of Figures
Fig. 1.1 Steady planar detonation wave in a tube ............................................................. 5
Fig. 1.2 Schematic of Rayleigh lines and Hugoniot curve in p
2
~ 1/
2
plane (adapted
from Kuo, 1986) ................................................................................................... 7
Fig. 1.3 Variation of physical properties through a ZND detonation wave (adapted
from Kuo, 1986) ................................................................................................. 12
Fig. 1.4 Schematic of ZND detonation structure in p
2
~ 1/
2
plane (adapted from
Kuo, 1986).......................................................................................................... 13
Fig. 1.5 Smoked-foil record of a detonation (Strehlow, 1968)....................................... 13
Fig. 1.6 ZND detonation propagation in a tube closed at one end.................................. 14
Fig. 1.7 Space-time wave diagram for a ZND detonation wave propagation in a tube.. 15
Fig. 1.8 Schematic of pressure profile for a ZND detonation propagation in a tube
closed at one end. (adapted from Bussing and Pappas, 1994)............................ 15
Fig. 1.9 Conceptual comparison of PDEs with other engines ........................................ 19
Fig. 1.10 Schematic of cycle operation for idealized PDE............................................. 20
Fig. 1.11 Anticipated specific impulses of airbreathing engines (Roy, 1999)................ 23
Fig. 1.12 Typical structure of airbreathing PDE (Bussing and Pappas, 1996) ............... 25
Fig. 1.13 Supersonic airbreathing pulse detonation engine ............................................ 55
Fig. 2.1 Configurations of a mixed-compression supersonic inlet ................................. 59
Fig. 2.2 Computational domain for inlet study............................................................... 64
Fig. 2.3 Mach-number, pressure, and vorticity contours with different back pressures
under steady-state condition............................................................................... 67
Fig. 2.4 Mach-number and pressure distributions along midline under steady-state
condition (p
b
= 2.1atm)....................................................................................... 68
Fig. 2.5 Pressure distributions along walls and midline of inlet and Mach-number
contours within one cycle of oscillation (A = 0.05, f = 500 Hz)......................... 70
Fig. 2.6 Instantaneous shock locations for different cases.............................................. 74
Fig. 2.7 Shock position ranges with different frequencies.............................................. 75
Fig. 2.8 Comparison of the amplitude of terminal shock oscillation with analytical
result (A = 0.01).................................................................................................. 75
ix
11/23/2003
Fig. 2.9 Mass response to the pressure oscillation at exit over one cycle....................... 76
Fig. 2.10 Mass averaged stagnation pressure at exit over one cycle .............................. 76
Fig. 2.11 Axial velocity, stagnation pressure profiles at exit plane (A = 0.05, f = 500
Hz) ...................................................................................................................... 77
Fig. 2.12 Comparison of time averaged axial velocity and stagnation pressure
profiles at exit plane ........................................................................................... 78
Fig. 3.1 Mesh system and CEs/SEs in Space-Time CE/SE method (adopted from
Wu, 2002). .......................................................................................................... 85
Fig. 3.2 Comparison of detonation velocity and CJ state between ZND model and
CEA.................................................................................................................... 91
Fig. 3.3 Schematic of detonation tube with spark region at closed end.......................... 92
Fig. 3.4 Time evolution of pressure field for ZND problem (grid points = 2000) ......... 94
Fig. 3.5 Time evolution of temperature field for ZND problem (grid points = 2000) ... 94
Fig. 3.6 Time evolution of reactant mass ratio field for ZND problem (grid points =
2000)................................................................................................................... 95
Fig. 3.7 Time evolution of velocity field for ZND problem (grid points = 2000).......... 95
Fig. 3.8 Snapshots of pressure field for ZND problem with different grid points.......... 96
Fig. 3.9 Schematic of convergent-divergent nozzle in the validation case..................... 97
Fig. 3.10 Comparison of pressure distributuion with analytical results ......................... 97
Fig. 3.11 Configuration of the single tube and CD nozzle in the current study ............. 98
Fig. 3.12 Time periods during one cycle of operation.................................................. 100
Fig. 3.13 Temporal evolution of pressure field within one cycle of operation (
cycle
=
3 ms,
close
= 2.1 ms,
purge
= 0.1 ms)................................................................. 103
Fig. 3.14 x-t diagram for the first cycle and time histories of flow properties at head
end under typical PDE operation with CD nozzle (stoichiometric H
2
/air
mixture,
cycle
= 3 ms,
close
= 2.1 ms,
purge
= 0.1 ms). 1 = uniform unburned
region, 2 = Taylor expansion waves, 3 = uniform region, 4 = non-simple
wave region, 5 = simple wave region. .............................................................. 104
Fig. 3.15 x-t diagram for the eighth cycle and time histories of flow properties at
head end under typical PDE operation with CD nozzle (stoichiometric
H
2
/air mixture,
cycle
= 3 ms,
close
= 2.1 ms,
purge
= 0.1 ms)............................. 105
Fig. 3.16 Effect of valve close-up time on (a) specific thrust and (b) specific
impulse at four different operation frequencies; straight tube with CD
nozzle; stoichiometric H
2
/air mixture, h = 9.3 km, M
= 2.1........................... 108
x
11/23/2003
Fig. 3.17 Effects of
close
on (a) specific thrust and (b) specific impulse at three
different frequencies for two CD nozzles with different throat areas. The
larger one has a diameter of 12 cm and the smaller one 10 cm........................ 110
Fig. 4.1 Generalized impulse curve for single-pulse detonation in straight tube with
stoichiometric H
2
/air mixture. .......................................................................... 115
Fig. 4.2 Computational domain for straight-tube PDE................................................. 119
Fig. 4.3 Schematic of the computational domain decomposed into 64 sub-domains... 120
Fig. 4.4 Number of cells of sub-domains...................................................................... 120
Fig. 4.5 x-t diagram for first cycle and time histories of flow properties at head end
under typical PDE operation (stoichiometric H
2
/air mixture,
cycle
= 3 ms,
close
= 2.4 ms,
purge
= 0.1 ms). 1 = unburned region, 2 = Taylor expansion
waves, 3 = stationary region, 4 = non-simple wave region, 5 = simple wave
region. ............................................................................................................... 123
Fig. 4.6 Snapshots of pressure, density and their gradients fields at t = 0.7 ms. .......... 126
Fig. 4.7 Time evolution of pressure distribution along centerline during first cycle of
operation (
cycle
= 3 ms,
close
= 2.4 ms,
purge
= 0.1 ms). ................................... 127
Fig. 4.8 x-t diagram for fifth cycle and time histories of flow properties at head end
2
/air mixture,
cycle
= 3 ms,
close
= 2.4 ms,
purge
= 0.1 ms)........................................................................... 128
Fig. 4.9 Performance parameters of ramjet engine for stiochiometric H
2
/air mixture . 129
Fig. 4.10 Effect of valve close-up time on specific impulse (
cycle
= 3 ms,
purge
= 0.1
ms); straight tube with stoichiometric H
2
= 2.1.. 130
Fig. 4.11 Computational domain for single-tube PDE ................................................. 132
Fig. 4.12 Schematic of the computational domain decomposed into 64 sub-domains. 132
Fig. 4.13 Number of cells of sub-domains.................................................................... 132
Fig. 4.14 Time evolution of Mach number field during the first cycle of operation
(
cycle
= 3 ms,
close
= 2.1 ms)............................................................................. 134
Fig. 4.15 Time evolution of density-gradient field during the first cycle of operation
(
cycle
= 3 ms,
close
= 2.1 ms)............................................................................. 135
Fig. 4.16 Time evolution of pressure () and Mach number (---) distributions
along centerline during first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms) . 136
Fig. 4.17 Time histories of (a) pressure and (b) Mach number at midpoints of head
end, nozzle throat, and nozzle exit during first cycle of operation (
cycle
= 3
ms,
close
= 2.1 ms) ............................................................................................ 137
xi
11/23/2003
Fig. 4.18 Enlarged views of pressure contours at (a) 0.40 ms and (b) 0.65 ms............ 140
Fig. 4.19 Time history of head-end pressure during first five cycles (
cycle
= 3 ms,
close
= 2.1ms) .................................................................................................... 145
Fig. 4.20 Specific impulse and filling length of first five cycles (
cycle
= 3 ms,
close
=
2.1 ms) .............................................................................................................. 145
Fig. 4.21 Time evolution of density-gradient field during fifth cycle of operation
(
cycle
= 3 ms,
close
= 2.1 ms)............................................................................. 146
Fig. 4.22 Time histories of pressure and Mach number at midpoint of nozzle exit
during fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1ms) ............................ 147
Fig. 4.23 Density-gradient field at t = 14.5 ms, with ambient flow,
cycle
= 3 ms,
close
= 2.1 ms .................................................................................................... 149
Fig. 4.24 Close-up view of Mach-number contours at t = 14.50 ms, with ambient
flow................................................................................................................... 149
Fig. 4.25 Instantaneous thrust and impulse during first five cycles (
cycle
= 3 ms,
close
= 2.1 ms) ................................................................................................... 151
Fig. 4.26 Instantaneous impulse of nozzle during fifth cycle (
cycle
= 3 ms,
close
=
2.1 ms) .............................................................................................................. 152
Fig. 4.27 Schematic of flow-path analysis for PDE performance prediction............... 153
Fig. 4.28 Instantaneous impulse during steady cycle (
cycle
= 3 ms,
close
= 2.1 ms) ..... 158
Fig. 4.29 Effect of valve closed time on (a) air-based specific thrust and (b) fuel-
based specific impulse at three different operation frequencies;
purge
= 0.1
ms, stoichiometric H
2
= 2.1. ............................... 162

Fig. 4.30 Instantaneous thrust and impulse during a steady cycle,
cycle
= 3 ms,
purge
= 0.1 ms. ........................................................................................................... 163
Fig. 4.31 Effect of purging time on (a) air-based specific thrust and (b) fuel-based
specific impulse,
cycle
= 3 ms,
close
= 2.1 ms. .................................................. 167
Fig. 4.32 Single-tube PDE configurations (tube length = 60 cm, tube height = 16
cm): (a) baseline case, nozzle length = 20 cm, throat height = 12 cm, (b)
nozzle length = 15 cm, (c) nozzle length = 12.4 cm, (d) throat height = 9
cm. .................................................................................................................... 169
Fig. 4.33 Instantaneous thrust and impulse during steady cycle (
cycle
= 3 ms,
close
=
2.1 ms) .............................................................................................................. 170
Fig. 4.34 Instantaneous pressure impulse during steady cycle (
cycle
= 3 ms,
close
=
2.1 ms) .............................................................................................................. 172
xii
11/23/2003
Fig. 4.35 Time history of head-end pressure during steady cycle (
cycle
= 3 ms,
close
= 2.1 ms)........................................................................................................... 172
Fig. 5.1 Supersonic airbreathing pulse detonation engine ............................................ 179
Fig. 5.2 Operation sequence of a triple-tube PDE........................................................ 179
Fig. 5.3 Computational domains for multitube PDEs................................................... 180
(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ............................................. 183
Fig. 5.5 Time evolution of pressure distribution along centerline of each tube
during the first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
=
0.1 ms) .............................................................................................................. 184
Fig. 5.6 Time histories of pressure at midpoints of head ends of the bottom, middle,
and top tubes during the first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms,
and
purge
= 0.1 ms)............................................................................................ 185
Fig. 5.7 Snapshots of pressure field showing flow interactions among tubes and
nozzle during the first cycle (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms) 187
Fig. 5.8 Specific impulse and filling length of the middle tube of the first five cycles
(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ............................................. 191
Fig. 5.9 Time evolution of density-gradient field during the fifth cycle of operation
(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ............................................. 192
Fig. 5.10 Time evolution of pressure distribution along centerline of each tube
during the fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
=
0.1 ms) .............................................................................................................. 193
Fig. 5.11 Time histories of pressure at midpoints of (a) head ends and (b) exit of the
bottom, middle, and top tubes during the fifth cycle of operation (
cycle
= 3
ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ............................................................. 194
Fig. 5.12 Time histories of Mach number at midpoints of (a) nozzle throat and (b)
nozzle exit during the fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms,
and
purge
= 0.1 ms)............................................................................................ 197
Fig. 5.13 Instantaneous thrust in (a) axial and (b) vertical directions during the
fourth and fifth cycles (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ......... 199
Fig. 5.14 Spectra of periodic instantaneous axial thrusts of single-tube and triple-
tube PDEs with operation timing of
cycle
= 3 ms,
close
= 2.1 ms, and
purge
=
0.1 ms ............................................................................................................... 200
Fig. 5.15 Instantaneous pressure thrust and impulse during the fifth cycle (
cycle
= 3
ms,
close
= 2.1 ms, and
purge
= 0.1 ms) ............................................................. 202
xiii
11/23/2003
Fig. 5.16 Axial and vertical velocity profiles at nozzle exit plane during the fifth
cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)................ 203
Fig. 5.17 Effect of valve close-up time on (a) air-based specific thrust and (b) fuel-
based specific impulse;
purge
= 0.1 ms, stoichiometric H
2
/air mixture, h =
9.3 km, M
= 2.1. ............................................................................................. 206

(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms), with free volume ................ 208
Fig. 5.19 Pressure contours and streamlines at t = 13.50 ms (
cycle
= 3 ms,
close
= 2.1
ms, and
purge
= 0.1 ms), with free volume........................................................ 209
Fig. 5.20 Instantaneous axial thrust during the fifth cycle (
cycle
= 3 ms,
close
= 2.1
ms, and
purge
= 0.1 ms) ..................................................................................... 210
Fig. A.1 Temperature-entropy diagram........................................................................ 219
Fig. D.1 A triangular cell and its three neighbours....................................................... 229
Fig. D.2 Schematic of CE and SE associated with solution point (j, n) ....................... 229
Fig. E.1 A boundary cell and its mirror image with respect to the solid wall ............... 241
Fig. F.1 Space-time diagram for detonation wave propagation.................................... 245

xiv
11/23/2003
List of Tables
Table 1.1 Survey of single-pulse experimental investigations of PDEs......................... 32
Table 1.2 Survey of multicycle experimental investigations of PDEs ........................... 33
Table 1.3 Survey of one-dimensional numerical simulations of PDEs .......................... 40
Table 1.4 Survey of two-dimensional numerical simulations of PDEs .......................... 41
Table 1.5 Survey of analytical studies on PDEs ............................................................. 52
Table 3.1 Comparison of numerical results of different grid resolutions with
analytical solutions for the ZND problem.......................................................... 96
Table 4.1 Flow properties at the CJ point and in the uniform region (p
1
= 0.29 atm,
T
1
= 228 K, = 1.29, R = 368.9 J/(kgK), q = 2.72010
6
J/kg)........................ 138
Table 4.2 Analytically predicted PDE propulsive performance under various
refilling Mach numbers and purge-to-open time ratios.................................... 156
Table 4.3 Loss summary for baseline case .................................................................... 160
Table 4.4 Performance comparison among different nozzles........................................ 171

xv
11/23/2003
Nomenclature
A
e
area of engine exit plane
c speed of sound
c
p
constant-pressure specific heat
E
a
activation energy per unit mass of reactants
e
t
specific total energy
F axial thrust
F thrust vector
F
sp
specific thrust (air-based)
f fuel-to-air mass ratio of reactants
f fuel-to-air mass ratio of the mixture of reactants and purged air
g gravitational acceleration
I
sp
specific impulse (fuel-based)
i unit vector in axial direction
K pre-exponential factor
L length of detonation tube
L
driv
length of driver gas region
M Mach number
a
m mass flow rate of air delivered to the engine
f
m mass flow rate of fuel delivered to the combustor
n unit vector normal to surface
xvi
11/23/2003
p pressure
p
e
area-averaged pressure at engine exit plane
p
t1
total pressure at combustor entrance
q heat release per unit mass of reactants
q heat release per unit mass of mixture of reactants and purged air
R gas constant
s specific entropy
T temperature
T
driv
temperature of driver gas
T
t1
total temperature at combustor entrance
t time
u axial velocity
u velocity vector
u
e
mass-averaged axial velocity at engine exit plane
e
u mass-averaged velocity vector at engine exit plane
v vertical velocity
x axial coordinate
y vertical coordinate
Z mass fraction of reactants
purge-to-open time ratio
= specific heat ratio
density
xvii
11/23/2003
time period
mass production rate of reactants

Subscripts
close stage during which valve is closed
cycle PDE operation cycle
D detonation wave
e engine exit plane
i engine entrance plane
open stages during which valve is open
purge purging stage
refill refilling stage
ambient condition
1 fresh reactant upstream of detonation wave front
2 Chapman-Jouguet state
3 uniform region between the head end and the rear of Taylor wave
xviii
11/23/2003
Acknowledgements
With a deep sense of gratitude, first I wish to express my sincere thanks to my
advisor, Dr. Vigor Yang, for kindly providing guidance throughout the development of
this study. His support, encouragement, and comments have been of greatest help at all
times.
Sincere thanks also go to Dr. Robert J. Santoro, Dr. John Mahaffy, and Dr. Robert
F. Kunz for serving as my committee members. They monitored my work and took
effort in reading and giving me with valuable suggestions on the earlier versions of this
thesis.
This work was supported by the Department of Defense Multidisciplinary
University Research Initiative under the Office of Naval Research Grant N00014-02-1-
0589, with Gabriel Roy serving as the program manager. The financial support is greatly
appreciated.
Sincere thanks also go to Dr. Jeong-Yeol Choi for his help on this research, and to
Dr. Xiaoyen Wang and Dr. Yuhui Wu for their help on the space-time method.
I wish I would never forget the company I had from all the members of our
research lab. In particular, I am thankful to Shanwu Wang, Danning You, Ying Huang,
Nan Zong, and Yanxing Wang for their help.
I take this opportunity to express my love and sincere thanks to my parents for
their endless love and support. My appreciation and love also go to my brother and
sisters. They taught me the value of hard work by their own examples and provided me
enormous support throughout my life.
Finally I wish to give a very special thank and all my love to my wife, Chunxia.
Her support, love, and understanding have been invaluable to me.
1
11/23/2003
Chapter 1
Introduction
1.1 Background and Motivation
The combustion process is a vital mechanism in most propulsion systems. A
combustion process can be characterized as either a deflagration or a detonation. The
deflagration is mainly governed by mass and thermal diffusion and has a flame speed of
one or more meters per second. Usually, a deflagration process produces a slight
decrease in pressure and can be modeled as a constant-pressure combustion process.
Engines based on the deflagration process can be constructed to operate at steady state
and are easy for design optimization with modular analyses of each subsystem. Most
conventional engines, such as turbofans, turbojets, ramjets, and rocket engines, utilize a
steady deflagration process.
In contrast to deflagration, the detonation process takes place much more rapidly
and produces a supersonic combustion wave, or a detonation wave, propagating at around
two thousand meters per second toward the unburned reactants. The detonation wave can
be described as a strong shock wave coupled to a reaction zone. The shock wave
compresses the reactants, acting like a valve between the reactants and products. The
reaction thus takes place at a much higher pressure. Since there is not enough time for
the pressure to reach its equilibrium, a detonation process closely approximates to a
constant-volume combustion process. It is well known that an engine based on a
constant-volume combustion process has higher thermodynamic efficiency than that
2
11/23/2003
based on a constant-pressure combustion process. The potential advantage in
thermodynamic cycle efficiency becomes the primary reason that has been driving
peoples interests in developing engines that employ detonation processes. Examples of
these engine concepts include those employing standing detonation waves, such as the
detonation thrusters, the detonation ramjet (dramjet), and the oblique detonation wave
engine (ODWE), or those employing intermittent traveling detonation waves, such as the
pulse detonation engine (PDE). The engine concepts employing standing detonation
waves are hampered by the difficulty in stabilizing the detonation wave. The PDE thus
becomes a more practical candidate nowadays.
Early work on PDEs dates back to the 1940s (Hoffman, 1940; Nicholls et al.,
1957; Dunlap et al., 1958; Krzycki, 1962) and was suspended possibly due to the lack of
funding in the late 1960s. Interest in PDEs was reignited in the late 1980s by the work of
Helman et al. (1986). Recently, the PDEs have been recognized worldwide as a
promising propulsion technology that offers potential advantages in thermodynamic cycle
efficiency, hardware simplicity, and operation scalability and reliability (Bussing and
Pappas, 1996; Kailasanath, 2002; Wu et al., 2003). Several PDE concepts have been
proposed and experimentally investigated in the past. The main challenges are to develop
an ignition system for repetitive detonation initiation, to integrate the inlet with the
unsteady-operated detonation chamber, and to choose an optimized nozzle to achieve
high performance. Although substantial efforts have been made to overcome these
challenges, a real PDE-based vehicle has not emerged yet.
One of the most important concerns is to estimate the propulsive performance of
PDE concepts and to compare it with those of the conventional engines. In parallel to
3
11/23/2003
experimental investigations, attempts were made both theoretically and numerically to
estimate the performance of PDEs. However, a general agreement on the performance
has not been reached yet. Most previous numerical simulations were either limited to
single-pulse operation with one- and two- dimensional models, or multicycle operation
with only one-dimensional models. Very few efforts have been expended to simulate a
PDE with multicycle operations using multidimensional models. On the other hand,
since the unsteady nature of the PDE is dominated by the unsteady flow dynamics during
a steady periodic cycle, it is thus very important to understand the flow dynamics prior to
calculating the performance. Also, this understanding may help in improving the PDE
designs. Unfortunately, none of the previous studies have ever presented the detailed
multidimensional flow evolution within a whole cycle. Hence, the main motivation of
this thesis is to establish a multidimensional numerical analysis to examine the flow
dynamics involved in PDEs with multicycle operations, to provide a more convincible
estimation of the propulsive performance of PDEs, and to investigate the effect of the
various design parameters on the performance.
1.2 Detonation Physics
Studies of detonation phenomena date back to the end of the 19th century. The
simplest theory was proposed by Chapman (1889) and Jouguet (1905), usually referred to
as the CJ theory. It treats the detonation wave as a discontinuity plane in one dimension.
This theory can be used to predict the detonation wave velocity without the need to know
the details of the chemical reaction and the detonation wave structure. A significant
4
11/23/2003
advance in the understanding of the detonation wave structure was made independently
by Zeldovich (1940) in Russia, von Neumann (1942) in the United States, and Dring
(1943) in Germany. They considered the detonation wave as a leading planar shock
wave with a chemical reaction zone behind the shock. Their treatment has come to be
called the ZND model of detonation, and the corresponding detonation wave structure is
called the ZND detonation wave structure. Although all the experimentally observed
detonation waves have much more complex cellular three-dimensional structures
resulting from the strong nonlinear coupling between gasdynamics and chemical kinetics
(Glassman, 1996), the CJ theory and the ZND model, which assume a planar one-
dimensional detonation wave, are still very useful. An overview of them is hence given
in this subsection to provide some basic knowledge on detonation physics. More detailed
and extended discussions about detonation physics and phenomena can be found from
several textbooks (Fickett and David, 1979; Kuo, 1986; Glassman, 1996).
1.2.1 CJ Theory
For a steady, planar, one-dimensional detonation wave (see Fig. 1.1), with both
the reactants and products modeled as the same perfect gas and the detonation wave
modeled as a discontinuity plane at which heat addition occurs, the conservations of
mass, momentum, and energy in a coordinate system fixed at the wave front give:

1 2 2
( )
D D
u u u = (1.1)

2 2
1 1 2 2 2
( )
D D
p u p u u + = + (1.2)

2 2 1 2
2
1 2
1 1
( )
1 2 1 2
D D
p p
u q u u

+ + = +

(1.3)
5
11/23/2003
where the velocities relative to the wave front have already been expressed out with the
velocities relative to the tube; u
D
is the detonation wave velocity, q the heat release per
unit mass of reactants due to chemical reaction, and the specific heat ratio; , p, and u
represent the density, pressure, and velocity, respectively; subscript 1 refers to the state of
unburned gas or reactants and subscript 2 for the final state, i.e., the state of the burned
gas immediately behind the detonation wave. Since there are only three equations, one
extra equation is needed to solve for the four unknowns p
2
,
2
, u
2
and u
D
. This extra
equation will be revealed through the following analysis.

burned gas (p
2
,
2
) unburned gas (p
1
,
1
)
(relative velocity)
2 2 D
u u u =
0 1 D
u u =
2
u
D
u
1
0 u = (actual velocity)
wave front

Fig. 1.1 Steady planar detonation wave in a tube

Combining the mass and momentum conservation equations leads to the so-called
Rayleigh relation,

2 2
2 1 1
2 1
1 1
( ) /( )
D
p p u

= (1.4)
Manipulations of Eqs. (1.1) (1.3) yield the following Hugoniot relation,

2 1
2 1
2 1 2 1
1 1 1
( ) ( )( )
1 2
p p
p p q

+ =
(1.5)
If these two relations are plotted in the p
2
~ 1/
2
plane, a Rayleigh line and a
Hugoniot curve can be constructed. Figure 1.2 shows the schematic of the Hugoniot
6
11/23/2003
curve and the Rayleigh line. The point corresponding to the unburned gas state is
denoted by A. Apparently, all the Rayleigh lines pass through point A. The possible
final states are defined by the intersection of the Rayleigh line and the Hugoniot curve.
Among all the straight lines passing through point A, there are two which are tangent to
the Hugoniot curve. The corresponding tangent points are generally called as the CJ
points, denoted in the figure by point U for the upper CJ point and L for the lower CJ
point. The horizontal and vertical lines passing through point A correspond to a constant-
pressure and a constant-volume process, respectively. The Hugoniot curve is divided into
five regions, i.e., regions I~V, by the two tangent lines and the horizontal and vertical
lines. Region V is unphysical since the Rayleigh lines defined by Eq. (1.4) cannot have
positive slope. Regions I and II are called detonation branch, within which the velocity
of the wave front is supersonic; regions III and IV are called deflagration branch, within
which the velocity of the wave front is subsonic. The upper CJ point corresponds to a
minimum detonation velocity, whereas the lower CJ point corresponds to a maximum
deflagration velocity.
Through simple mathematical derivations, i.e., equating the slope of the Hugoniot
curve to that of the Rayleigh line, the following relation, usually referred to as the CJ
condition, can be obtained at the CJ points,

2 2 2 2 2
/
D
u u u p c = = = or
2 2 2
/ 1 M u c = = (1.6)
where
2
c is the sound speed of the burned gas,
2
u and
2
M the velocity and Mach number
of the burned gas relative to the wave front, respectively. The flow velocity relative to
the wave front at the CJ points equaling to the local sound speed is one of the notable
7
11/23/2003
characteristics of the CJ points. At the upper CJ point, since
2 2 D
u c u + = , any rarefaction
waves arising behind the wave front will not overtake the detonation wave and thus a
self-sustained steady detonation wave can be established.

p
2
I (strong detonation)
II (weak detonation)
III (weak deflagration)
IV (strong deflagration)
upper CJ point
V
L
lower CJ point
1/
2
1/
1
p
1
A
U

Fig. 1.2 Schematic of Rayleigh lines and Hugoniot curve in p
2
~ 1/
2
plane (adapted
from Kuo, 1986)

Region I is called the strong-detonation region. In this region, the velocity of the
burned gas relative to the wave front is subsonic, i.e.,
2 2 D
u c u + > , thus, any rarefaction
waves arising behind the wave front will overtake and weaken the detonation wave. As a
matter of fact, a strong detonation, also called overdriven detonation, is not stable and is
thus seldom observed experimentally. It may, however, appear during a transient process
or be generated with a driving piston.
Region II is called the weak-detonation region. In this region, the velocity of the
burned gas relative to the wave front is supersonic, i.e.,
2 2 D
u c u + < . If the ZND
detonation wave structure is adopted, the detonation wave can be considered as a shock
8
11/23/2003
wave and a following heat addition zone. The gas velocity immediately behind the shock
relative to the wave front is known to be subsonic from classical shock dynamics theory.
On the other hand, it is also well known that for a steady flow in a constant-area tube, the
fluid cannot be accelerated from subsonic to supersonic by heat addition. This means
that the velocity of the burned gas relative to the wave front cannot be supersonic. Thus,
region II is physically impossible as long as the ZND detonation wave structure is
assumed. Another discussion leading to the same conclusion can be found in the
textbook of Glassman (1996).
Region III is called the weak-detonation region. The weak deflagration, or simply
the deflagration, is often observed in experiments. A deflagration wave propagates
toward the unburned gas at a subsonic velocity. Across a deflagration wave, the velocity
of the gas relative to the wave front is accelerated within the subsonic regime, and the
pressure is reduced.
Region IV is called the strong-deflagration region. Passing through a strong-
deflagration wave, the gas velocity relative to the wave front is to be accelerated from
subsonic to supersonic. Similar to the discussion for region II, this violates the common
conclusion that for a steady flow in a constant-area tube, the fluid cannot be accelerated
from subsonic to supersonic by heat addition. Thus, region IV is physically impossible,
and a strong deflagration is never observed experimentally.
Based on the above discussions, the upper CJ point is the only possible state for a
self-sustained steady detonation wave commonly observed. Thus, the CJ condition, Eq.
(1.6), can be used as an additional equation along with Eqs. (1.1) ~ (1.3) to solve for the
four unknowns aforementioned, i.e., p
2
,
2
, u
2
and u
D
. It is convenient to find the
9
11/23/2003
detonation wave Mach number
D
M first and then express the other unknowns with
respect to it. For concision, the corresponding mathematical manipulations are skipped,
and the solutions for the unknowns as well as several other properties of the final state are
listed as below.

2 2
1 1 1
1 1
1
2 2
D
D
u q q
M
c RT RT

= + + (1.7)

1 D D
u M RT = (1.8)

2
2 1
1
(1 )
D
D
M
u RT
M

=
+
(1.9)

2
2 1
1
(1 )
D
D
M
c RT
M
+
=
+
(1.10)

2
2 2
1
1
D
D
M
M
M

=
+
(1.11)

+
+
=
1
1
2
1
2 D
M
p
p
(1.12)

2
2
1
2
1
) 1 (
D
D
M
M
+
+
= (1.13)

2
2
2 2 2
1 1 1
/( ) 1
/( ) (1 )
D
D
T p R M
T p R M

| | +
=
|
+
\ .
(1.14)
where R is the gas constant; T
1
and c
1
are the temperature and sound speed of the
unburned gas, respectively; T
2
, c
2
, and M
2
, are the temperature, sound speed, and Mach
number of the unburned gas, respectively.
10
11/23/2003
In most cases, the square of the detonation wave Mach number,
2
D
M , is much
larger than 1. Some of the above expressions can thus be approximately reduced to the
following ones.

2
1
1
D
u u
+
(1.15)

D
u c
1
2
(1.16)

2
1
M
(1.17)

2 2
1

1
D
p
M
p

+
(1.18)

2
1
1

+
(1.19)

2
2
1
1
D
T M
T

| |
|
+
\ .
(1.20)
1.2.2 ZND Detonation Wave Structure
The CJ theory brought great success in predicting the detonation wave velocity.
It, however, gives no information about the details of the detonation wave structure. In
the early 1940s, Zeldovich (1940), von Neumann (1942), and Dring (1943)
independently extended the CJ theory to consider the detonation wave structure that has
become the well-known ZND detonation structure. Their treatment is referred to as the
ZND model. According to them, the detonation wave is interpreted as a strong planar
shock wave propagating at the CJ detonation velocity, with a chemical reaction region
11
11/23/2003
followed and coupled to the shock wave. The shock wave compresses and heats the
reactants to a temperature at which a reaction takes place at a rate high enough for the
ensuring deflagration to propagate as fast as the shock wave. From the energy point of
view, the shock wave provides activation energy to ignite reaction, whereas the energy
released by reaction keeps the shock moving. Their assumption that no reaction takes
place in the shock wave region was based on the fact that the width of the shock wave is
in the order of a few mean free paths of the gas molecules, whereas the width of the
reaction region is in the order of one centimeter (Kuo, 1986).
Figure 1.3 shows schematically the variation of physical properties through a
ZND detonation wave. Plane 1 denotes the state of unburned gas. Plane 1' denotes the
state immediately after the shock wave. Chemical reaction starts at plane 1' and finishes
at plane 2, at which the CJ state reaches. If a single variable is used to represent the
reaction progress or the degree of reaction, it will have a value of 0 at plane 1' and a value
of 1 at plane 2. Generally, the reaction rate, such as that that follows the Arrhenius law,
increases with temperature, and the chemical reaction region can be further divided into
an induction and a heat addition zone. In the induction zone directly behind the shock
wave where the temperature is not very high, the reaction rate is relatively slow and as a
result, the temperature, pressure, and density profiles are relatively flat. In the heat
addition zone behind the induction zone, the reaction rate increases drastically to high
values along with a large amount of heat release from the reaction so that the gas
properties change sharply.

12
11/23/2003
T
,

p
,

x
1 1' 1'' 2
induction
zone
heat addition
zone
shock
wave
deflagration
T
p
T
,

p
,

x
1 1' 1'' 2
induction
zone
heat addition
zone
shock
wave
deflagration
T
p

Fig. 1.3 Variation of physical properties through a ZND detonation wave (adapted from
Kuo, 1986)

The ZND detonation wave structure can be also interpreted by Hugoniot curves
shown in Fig. 1.4. There are many paths, such as a, b, c, and d, by which a reacting
mixture may pass through the detonation wave from the unburned state to the burned
state (Kuo, 1986). In the limit of zero chemical-energy release in the shock, a path will
reach point s, the intersection of the shock Hugoniot curve and the Rayleigh line, and
then the upper CJ point. The point s is referred to as the von Neumann spike. The von
Neumann spike pressure can be determined from a normal shock relation:
1
1
) 1 ( 2
2
1

+
+
=
D s
M
p
p
(1.21)
Using Eq. (1.12), the von Neumann spike pressure relates to the CJ pressure as follows:
1 2
1
2
1
=
p
p
p
p
s
(1.22)
Hence, the von Neumann spike pressure is about twice of the CJ pressure.
13
11/23/2003
The ZND model marks a great advance from the CJ theory in recognizing the
detonation wave structure. However, all the experimentally observed detonation waves
exhibit much more complex cellular three-dimensional structures. The smoked-foil
record of a detonation shown in Fig. 1.5 displays the typical cellular structure (Strehlow,
1968). More detailed discussions about these structures can be obtained from the
textbooks of Kuo (1986) and Glassman (1996).

b
p
2
1/
2
von Neumann spike
Hugoniot curve (q>0)
s
d
c
a
p
1
1/
1
L
shock Hugoniot (q=0)
U
A
b
b
p
2
1/
2
von Neumann spike
Hugoniot curve (q>0)
s
d
c
a
p
1
1/
1
L
shock Hugoniot (q=0)
U
A
b

Fig. 1.4 Schematic of ZND detonation structure in p
2
~ 1/
2
plane (adapted from Kuo,
1986)

Fig. 1.5 Smoked-foil record of a detonation (Strehlow, 1968)

14
11/23/2003
1.2.3 ZND Detonation Wave Propagation in a Tube
This subsection considers the ZND detonation propagation in a constant-area tube
that is closed at one end and open at the other, as shown schematically in Fig. 1.6. The
tube is initially filled with a static premixed detonable mixture. Detonation is initiated at
the closed end and propagates downstream toward the open end. Following the
detonation wave is a centered rarefaction wave, known as the Taylor wave, emanating
from the closed end to satisfy the stationary boundary condition there. After the passage
of the Taylor wave, a uniform region forms. The corresponding wave diagram in the
space-time plane is given in Fig. 1.7. Figure 1.8 schematically shows the pressure profile
within the tube. The width of the detonation wave is enlarged for visualization. The
states 1, s, 2, and 3 denote the unburned gas state, the von Neumann spike state, the CJ
state, and the uniform region state, respectively. The von Neumann spike state and the
CJ state can be readily determined using the equations derived in the previous subsection.
The focus of this subsection is hence on the solution of the Taylor wave and uniform
regions.

burned gas
unburned gas
D
u
wave front
burned gas
unburned gas
D
u
wave front

Fig. 1.6 ZND detonation propagation in a tube closed at one end

15
11/23/2003
unburned gas
uniform region
u
3
= 0
front
tail
Taylor
wave
x
t
detonation wave
2
1
3
unburned gas
uniform region
u
3
= 0
front
tail
Taylor
wave
x
t
detonation wave
2
1
3

Fig. 1.7 Space-time wave diagram for a ZND detonation wave propagation in a tube

detonation
wave
rarefaction
waves
heat
addition
ignition
delay
shock front
unburned
gas
p
1
, T
1
von Neumann
pressure spike s
2
st nt state regi n
p
s
, T
s
p
2
, T
2
p
3
, T
3
moves at u
D
3
1
(CJ point)
uniform region
Taylor
wave
detonation
wave
rarefaction
waves
heat
addition
ignition
delay
shock front
unburned
gas
p
1
, T
1
von Neumann
pressure spike s
2
st nt state regi n
p
s
, T
s
p
2
, T
2
p
3
, T
3
moves at u
D
3
1
(CJ point)
uniform region
Taylor
wave

Fig. 1.8 Schematic of pressure profile for a ZND detonation propagation in a tube closed
at one end. (adapted from Bussing and Pappas, 1994)

The properties of the uniform region can be obtained as follows. Applying the
Riemann invariant relation along the characteristic line passing through the Taylor wave
from state 2 to state 3 gets

3 3 2 2
2 2
1 1
u c u c

=

(1.23)
16
11/23/2003
where u
3
and c
3
are the velocity and sound speed in the uniform region, respectively.
Since u
3
= 0, the above equation yields,

2
3 2
2 2
2 2
1 1 1 1
1 1 1
2 2 2 1
D
D
c u M
M
c c M

= = =
+
(1.24)
where M
2
is the Mach number of the gas at state 2 expressed with Eq. (1.11).
Consequently, the temperature in the uniform region is

2
2
3
2
2
1 1
1
2 1
D
D
T M
T M
| |
=
|
+
\ .
(1.25)
Through an isentropic relation from state 2 to state 3, the pressure can be determined as:

2
2
1
3
2
2
1 1
1
2 1
D
D
p M
p M

| |
=
|
+
\ .
(1.26)
The relation of the sound speed c
3
with the detonation wave velocity u
D
can be obtained
by combining Eqs. (1.24), (1.10), and (1.8):

2
3
2
1
2
D
D D
c M
u M
+
= (1.27)
With the sound speed c
3
available, the length of the uniform region, L
const
, at a
particular instant of time t, can also be determined. Since the rear of the Taylor wave
propagates at c
3
, thus,
t c L
const 3
= (1.28)
Similar to the previous subsection, considering that the square of the detonation
wave Mach number,
2
D
M , is much larger than 1, the above expressions, Eqs. (1.24)~
(1.27) can be further simplified with approximation:
17
11/23/2003

3
2
1
2
c
c

+
(1.29)

2
3
2
1
2
T
T

| | +
|
\ .
(1.30)

2
1
3
2
1
2
p
p

| | +
|
\ .
(1.31)

3
1
2
D
c
u
(1.32)
The length of the uniform region can thus be approximated as:
t u t c L
D const
2
1
3
= (1.33)
It indicates that the length of the uniform region is about halfway between the detonation
wave front and the head end (Fickett and Davis, 1979).
The flow properties within the Taylor wave region,
3 D
c t x u t , can also be
derived. The Riemann invariant relation along the characteristic line from state 3 to the
point (x, t) gives:

3
2 2
0
1 1
c u c

=

(1.34)
where u and c are the velocity and sound speed at point (x, t). On the other hand, since
the forward characteristic lines are straight, thus,
c u
t
x
+ = (1.35)
Combination of the above two equations leads to the following solutions of u and c.
18
11/23/2003

3 3
2 2
1 1
u x
c c t
=
+ +
(1.36)

3 3
1 2
1 1
c x
c c t
= +
+ +
(1.37)
Other properties such as the temperature and pressure at point (x, t) are:

2
3 3
1 2
1 1
T x
T c t

| |
= +
|
+ +
\ .
(1.38)

2
1
3 3
1 2
1 1
p x
p c t

| |
= +
|
+ +
\ .
(1.39)

19
11/23/2003
1.3 Pulse Detonation Engine Concept
The pulse detonation engine (PDE) is an unsteady propulsion device that
produces periodic impulse by utilizing repetitive detonations. Based on whether or not to
use the ambient air as working fluid, it can be classified as an airbreathing version
referred to as the airbreathing PDE and a rocket version referred to as the pulse
detonation rocket engine (PDRE). The PDE differs from the conventional engines by its
two distinguishing characteristics: unsteady operation and detonation process. It is
noteworthy that the PDE concept is also different from the widely known pulsejets, such
as the German V-1 buzz bomb used in World War II. These pulsejets employ a
deflagration combustion process although they are unsteady engines. A comparison of
the PDE with the existing engines is briefly summarized in Fig. 1.9. The cycle operation,
potential advantages, and typical structure of the PDE are discussed in detail in the
following paragraphs.

conventional engines
(turbojet, ramjet, etc.)
pulsejet (e.g., German V-1
buzz bomb)
Pulse Detonation Engine
steady
unsteady
deflagration
detonation
combustion process operation process
Engines
conventional engines
(turbojet, ramjet, etc.)
pulsejet (e.g., German V-1
buzz bomb)
Pulse Detonation Engine
steady
unsteady
deflagration
detonation
combustion process operation process
Engines

Fig. 1.9 Conceptual comparison of PDEs with other engines

20
11/23/2003
1.3.1 Cycle Operation of PDE
A typical PDE cycle operation includes four basic processes: initiation of
detonation wave, propagation of detonation wave, exhaust of combustion products or the
blowdown process, and refilling of fresh reactants (Bussing and Pappas, 1994). These
four processes are schematically shown in Fig. 1.10 for an idealized PDE which involves
a straight detonation tube along with a valve located at the head end.

e. blowdown finished
products
p
0
u = 0
a. valve closed
reactants
p
1
p
0
c. detonation propagation
p
1
p
0
u
D
c
3
p
3
u
3
= 0
b. detonation initiation
p
1
p
0
detonation wave
Taylor wave
f. refilling
p
1
p
0
u
d. blowdown
c
3
p
3
u
3
= 0
p
2
e. blowdown finished
products
p
0
u = 0
a. valve closed
reactants
p
1
p
0
c. detonation propagation
p
1
p
0
u
D
c
3
p
3
u
3
= 0
b. detonation initiation
p
1
p
0
detonation wave
Taylor wave
f. refilling
p
1
p
0
u
d. blowdown
c
3
p
3
u
3
= 0
p
2

Fig. 1.10 Schematic of cycle operation for idealized PDE

The cycle begins as the tube is filled with reactants at pressure p
1
. The valve is
then closed and the detonation is initiated either directly or after a deflagration-to-
detonation transition (DDT) by an ignition source near the closed end. The detonation
wave propagates toward the open end at the CJ detonation velocity u
D
, usually in the
order of 2000 m/s. Following the detonation wave is a set of rarefaction waves known as
the Taylor wave emanating from the closed end to ensure the zero velocity condition at
21
11/23/2003
the closed end. The tail of the Taylor wave propagates toward the open end at the local
sound speed c
3
which is in the order of 1000 m/s. Between the tail of the Taylor wave
and the closed end is a uniform region, as discussed in Section 1.2.3. The Taylor wave
reduces the CJ pressure (p
2
) immediately behind the detonation wave to a relatively lower
level (p
3
) in the uniform region. This pressure, usually referred to as the plateau pressure,
is still much higher than the ambient pressure (p
0
) so that thrust is produced on the closed
end.
When the detonation wave exits the tube, another set of rarefaction waves are
generated and propagate back into the tube to reduce the pressure thereby, marking the
beginning of the blowdown process. The reflection of these rarefaction waves off the
closed end forms another set of rarefaction waves propagating downstream toward the
open end, further reducing the chamber. The unsteady blowdown process is
characterized by a series of compression and rarefaction waves which are alternately
generated from the closed and open ends (Bussing and Pappas, 1994). Eventually, the
chamber pressure decays to the ambient level and the blowdown process is finished.
Upon the end of the blowdown process, the valve opens to allow for the fresh
reactants to be charged into the tube. The valve timing is controlled so that no fresh
reactants escape from the open end to the ambient. This requires that the leading fresh
reactants should be caught by the detonation wave of the next cycle somewhere within
the detonation tube, or ideally, at the exit plane of the detonation tube. After the refilling
process finishes, the valve closes and the next cycle begins.
In a more practical cycle operation, the refilling process may start when the
closed-end pressure decays to a prespecified pressure level instead of the ambient value,
22
11/23/2003
to avoid the very low or even negative thrust produced on the closed end during the later
period of the blowdown process. In addition, the temperature of the products near the
closed end may be still very high and ignite the refilling fresh reactants if they contact
directly. This kind of preignition may cause engine unstart. Thus, a purging process, i.e.,
filling a little inert gas or cold air before filling the reactants, would be necessary to
prevent the preignition.
1.3.2 Potential Advantages of PDE
Due to the much more rapid burning or material conversion rate of a detonation
process than a deflagration process, PDEs possess several potential advantages over those
conventional engines based on the deflagration process. The first and also the most
attractive one is the higher thermodynamic cycle efficiency, which is attributed to the
facts that a detonation process approximately closes to a constant-volume combustion
process (Kailasanath, 2002) and that a constant-volume combustion based engine cycle
(the Humphrey cycle) has higher thermodynamic cycle efficiency than a constant-
pressure combustion based engine cycle (the Brayton cycle). The thermodynamic cycle
efficiencies of the Brayton, Humphrey, and ideal PDE cycle are given in Appendix A. It
has been noted that the Humphrey cycle efficiency may be 30% to 50% higher than that
of Brayton cycle for stoichiometric hydrogen/air mixture (Bussing and Pappas, 1994).
The potentially higher thermodynamic cycle efficiency of PDEs directly translates to
their potentially higher specific impulse. Figure 1.11, provided by Adroit System Inc.,
shows the anticipated specific impulse and operational envelope of various airbreathing
engines including PDEs, turbojets, ramjets, and scramjets using hydrogen or hydrocarbon
23
11/23/2003
fuels (Roy, 1999). Clearly from this figure, PDEs have a performance advantage over
their counterpart ramjet within their mutual flight regime. The lower performance of
PDEs than turbojets within the regime of flight Mach number less than 2 is due to the fact
that the turbojets utilize an additional compressor for precompression of the incoming air.

S
p
e
c
i
f
i
c

I
m
p
u
l
s
e
,

s
e
c

Flight Mach Number
0
1000
2000
3000
4000
5000
6000
7000
8000
0 2 4 6 8 10 12 14 16
Ramjets
Turbojets
Ramjets
Turbojets
PDE
PDE
HYDROGEN
HYDROCARBONS
Note:
- PDE Performance Estimated
Using ASI Performance Deck
- Turbojet, Ramjet and Scramjet
Performance Levels Reflect
Well-Designed Systems for
Man-Rated Thrust Classes
Pre-ignition
Limit
Scramjets
Scramjets

Fig. 1.11 Anticipated specific impulses of airbreathing engines (Roy, 1999)

In addition to their potentially higher thermodynamic cycle efficiency and specific
impulse, PDEs have a wider operation range in terms of flight Mach number. The
turbojets used nowadays in most commercial and military aircrafts are suitable for
subsonic and low supersonic flight speed because of their high performance resulting
from the precompression of the incoming air by the compressor. They become expensive
and inefficient as the flight Mach number increases and cannot be used at a flight Mach
number of greater than about 3 due to the cooling problem of the compressor blade.
Ramjets, on the other hand, are designed for supersonic operation with a flight Mach
24
11/23/2003
number greater than 2. They usually require solid propellant rocket boosters for
accelerating to the ramjet take over speed. PDEs, however, can produce thrust at static
conditions and thus have the potential to operate in both subsonic and supersonic flight
regimes with competitive efficiencies.
Other desirable features of PDEs over current propulsion systems include
hardware simplicity and configuration scalability. PDEs do not necessarily require
compressors to precompress the incoming air, nor turbines to produce work, nor boosters
to produce thrust at static condition, thus significantly reducing the hardware complexity
and design cost of the engine. Unlike pulsejets, which are tuned to the acoustical
resonances of the combustion chamber, PDEs are uncoupled from the acoustical chamber
resonance (Eidelman et al., 1990). It is theoretically possible to construct PDEs of a large
range of sizes and thrust levels. In addition, the detonation frequency is physically
restricted by the filling velocity of the fresh reactants into the detonation tube and is thus
also scalable.
In summary, PDEs offer the potential advantages of higher thermodynamics cycle
efficiency and specific impulse, wider operating region, hardware simplicity, and
operation scalability over other existing propulsion systems. However, it should be
mentioned that, in spite of these advantages, PDEs have also encountered a lot of
challenging design issues such as the low-energy repetitive detonation initiation, inlet
integration, nozzle optimization, and so forth. A real PDE-powered vehicle will not
emerge until all these challenging issues are resolved.
25
11/23/2003
1.3.3 Typical Structure of PDE
Figure 1.12 shows schematically the typical structure of an airbreathing PDE,
which includes the following components (Bussing and Pappas, 1996): inlet, fuel and/or
oxidizer sources, distribution manifolds, mixer, initiator, detonation tube, nozzle, and
interfaces. The incoming air delivered by the inlet is distributed by the air manifold and
mixes with the fuel distributed by the fuel manifold from the fuel source. The resultant
mixture is fed into the detonation tube for detonation. The detonation products exhaust
through the nozzle to generate thrust.

M
i
x
e
r
I
n
i
t
i
a
t
o
r
Detonation
Tube
I
n
t
e
r
f
a
c
e
Nozzle
I
n
t
e
r
f
a
c
e
D
i
s
t
r
i
b
u
t
i
o
n
M
a
n
i
f
o
l
d
Inlet
Distribution
Manifold
Distribution
Manifold
Fuel
Source
Oxidizer
Source

Fig. 1.12 Typical structure of airbreathing PDE (Bussing and Pappas, 1996)

All conventional airbreathing engines require an inlet to compress the air from the
freestream velocity and pressure to lower velocity and higher pressure for further
processing by other engine components. So do the airbreathing PDEs. The inlet of
conventional airbreathing engines is designed to provide stable airflow to the combustor
26
11/23/2003
with the highest possible recovery of total pressure and lowest possible external drag and
weight (Yang and Cappuccio, 1991). In airbreathing PDEs, because of the strong
pressure waves arising from the unsteady operation of the detonation tube, additional
concerns should be placed in the inlet design. Several approaches have been proposed.
The first is to design an inlet manifold large enough to dampen the pressure waves. It is,
however, not practical for volume-limited applications (Bussing and Pappas, 1996). The
second approach is to bleed excess air between detonation cycles through a well-designed
inlet/detonation tube manifold (Bussing and Pappas, 1996). A more practical approach
would be to utilize a combustor consisting of multiple detonation tubes that operate out of
phase to enable continuous airflow to the combustor (Bussing and Pappas, 1996) so that
the pressure waves propagating to the inlet are significantly reduced to a level within the
stable margin of the inlet.
The inlet/detonation tube interface serves both to isolate the detonation tube
flowfield from the inlet flowfield during the detonation initiation and propagation and the
blowdown processes, and to allow for the air and reactants to be delivered to the
detonation tube during the purging and refilling processes. An example of this interface
is a mechanical valve located at the head end of the detonation tube as mentioned in
Section 1.3.1. Another way to achieve the same purpose would be through some kind of
aerodynamic means.
The fuel source supplies fuel to the mixer through the fuel distribution manifold.
In some cases, a separate oxidizer source and oxidizer manifold may be required to
reduce the time and physical distance required for detonation initiation. The mixer is to
provide efficient mixing of fuel and air, and/or fuel and oxidizer for further detonation.
27
11/23/2003
The initiator is used to initiate the detonation wave in the detonation tube. The
detonation can be initiated either directly through a large amount of energy deposition, or
indirectly through a low-energy deposition along with a deflagration-to-detonation
transition (DDT) process. In general, the energy required for a direct initiation of fuel/air
mixture is impractical for repetitive operation. On the other hand, the DDT length
associated with the indirect initiation might be quite large. One of the possible ways to
reduce the DDT length is to fill the initiator section with the fuel/oxidizer mixture which
is known to be more detonable than its corresponding fuel/air mixture. The detonation is
first initiated in the initiator section, usually referred to as the predetonator, and then
propagates into the detonation tube to initiate the detonation in the fuel/air mixture.
The detonation tube is the kernel part designed to sustain and confine the
detonation process. Since the tube wall is exposed periodically to high-temperature,
high-pressure detonation products and low-temperature, low-pressure reactants, the
selection of proper material and the cooling of the detonation tube are two major design
considerations. In addition, the size of the detonation tube should be scaled to satisfy the
performance requirement such as the thrust level and the operation frequency.
The nozzle plays an important role in determining the PDE performance. In
conventional steady engines, the nozzle is simply optimized by matching the exhaust
pressure with the ambient pressure. This approach is, however, not available for PDEs
since the exhaust flow is unsteady and the exit pressure is not constant. The situation is
further compounded by the presence of the strong shock waves arising from the
detonation process. Therefore, the nozzle design of PDEs represents a challenging task.
28
11/23/2003
1.4 Literature Survey on PDE Studies
Studies of PDEs have been conducted for several decades. The first reported
work on intermittent or pulse detonation engines is attributed to Hoffman (1940) in
Germany. The device involved no valves. Holes were bored in a long detonation tube to
allow gas jets to escape. The heights of the gas jets escaping through the holes were used
to indicate the progress of the combustion wave and the onset of the detonation. Both
gaseous acetylene and liquid benzene fuels were employed with oxygen in his
experiments. He found that a continuous injection of the combustible mixture leads to
only a narrow range of ignition frequencies that produce an intermittent detonation cycle.
He also pointed out the importance of the spark plug location with respect to the tube
length. World Wall II prevented further work by Hoffmann and colleagues.
A substantial effort was made by Nicholls et al. (1957) at the University of
Michigan in the United States. They performed a series of single- and multiple-cycle
detonation experiments with hydrogen/oxygen, hydrogen/air, acetylene/oxygen, and
acetylene/oxygen mixtures in a stainless-steel detonation tube mounted on a pendulum
platform suspended by support wires. The tube has a length of 182.9 cm (6 feet) and an
internal diameter of 2.54 cm (1 inch). Fuel and oxidizer were injected coannually from
the head end of the tube and ignited by a spark plug located 25.4 cm (10 inches)
downstream. A maximum frequency of 35 Hz was obtained in their tests. The most
promising results were demonstrated for the hydrogen/air mixture, where fuel-based
specific impulses up to 2100 s were reached and agreed well with the predictions from
their simplified theoretical analyses. However, the agreement was partly fortuitous since
29
11/23/2003
the measured thrust-time history was significantly different from the theoretical results.
Moreover, because a low-energy spark ignitor was used in their experiments and no
deflagration-to-detonation transition (DDT) augmentation device was utilized, it is not
clear whether full detonation waves were realized.
In a setup similar to Nicholls, Krzycki (1962) performed experimental
investigations of PDEs at the US Naval Ordinance Test Station. A gaseous propane/air
mixture was continuously injected at the head end of the detonation tube through a
reverse-flow injector and ignited by an automobile spark plug at frequencies up to 60 Hz.
The power output of the spark plug varied inversely with the ignition frequency and was
only 0.65 J at the frequency of 60 Hz. This fact implies that a substantial part of the
combustion process occurring in the detonation tube was deflagration. Krzycki also used
the method of characteristics to calculate the pressure history at the head end of the
detonation tube, with the effect of the blowdown process taken into account. Based on
his well-documented experimental results, Krzycki concluded that the PDEs are not
promising for propulsion application. Possibly due to this conclusion, most experimental
works related to the PDE concept were stopped in the late 1960s.
A different design concept was proposed by Nicholls et al. (1966). They
examined the feasibility of a rocket motor that utilized a rotating detonation wave
propagating in an annular combustor chamber. Both hydrogen/oxygen and
methane/oxygen mixtures were employed in their experiments. A unidirectional
detonation wave was created after transition from a deflagration wave. However,
multicycle operations were not achieved. Their experiments and accompanying analyses
30
11/23/2003
indicated that the ideal performance of such a rocket motor is essentially the same as that
of a conventional rocket engine.
Applications of the pulse detonation devices to fields rather than propulsion were
reported in Russia in the 1980s. Korovin et al. (1981) examined the efficiency of thermal
oxidation of nitrogen in a detonation reactor involving intermittent detonation waves with
initiation frequencies from 2 to 16 Hz. The reactor was reported to be able to operate
without significant changes for 2000 hours. Smirnov and Boichenko (1986) studied
intermittent detonations of a gasoline/air mixture in a 3-m-long tube with frequencies up
to 8 Hz, aiming at improving the efficiency of a commercial rock-crushing apparatus.
In the late 1980s, Helman and colleagues (1986) reexamined the PDE concept at
the US Naval Postgraduate School. They carried out a series of experiments with an
ethylene/air mixture, showing the first successful self-aspirating airbreathing PDE. The
system was operated at frequencies up to 25 Hz, the maximum frequency allowed by the
solenoid valve that was used to control the gas flow. An important new idea employed in
their experiments is the use of a predetonator to overcome the energy requirement for
detonation initiation. The detonation was first initiated in the predetonator, a small tube
containing an ethylene/oxygen mixture, and then transmitted to the primary detonation
tube. The volume of the predetonator is only 2% of that of the primary detonation tube.
Based on the experimental results, they suggested that frequencies of 150 Hz and specific
impulse in the range of 1000-1400 s might be obtained in a practical PDE.
Since the renewed work of Helman et al., there has been a growing interest in the
PDE as an advanced propulsion system. A number of reviews can be obtained from the
literature. Eidelman et al. (1991) reviewed some of the early research as well as that
31
11/23/2003
done in the late 1980s. Bussing and Pappas (1996) discussed the basic theory and design
concepts. Kailasanath (2000) reviewed the applications of detonations to propulsion. A
comprehensive review of the more recent progress in PDE research was provided by
Kailasanath (2003). Wu, Ma, and Yang (2003) summarized the experimental work up to
date. For clarity, the research on PDEs since 1986 is divided into three categories:
experimental, numerical, and analytical studies, and is reviewed in the following
subsections.
1.4.1 Experimental Studies on PDEs
Much effort has been expended in studying various important aspects of PDEs
experimentally. These studies can be divided into single-pulse and multicycle
experiments and are briefly summarized in Tables 1.1 and 1.2. Single-pulse experiments
involve only the detonation initiation and propagation and the blowdown processes,
whereas multicycle experiments include the additional purging and refilling processes.
Single-pulse experiments can be used to determine the detonation initiation energy
required for a given mixture, to measure the detonation wave properties, and to validate
the concepts, serving as the basis for the more complex multicycle experiments. Both
hydrogen and hydrocarbon fuels were involved in the experiments. The hydrocarbon
fuels include both gaseous fuels such as ethylene (C
2
H
4
) and propane (C
3
H
8
) and liquid
fuels such as JP10 (C
10
H
16
). Ethylene (C
2
H
4
) was selected by many researchers because
of its well-documented detonation properties and as a common decomposition of some
typical heavy hydrocarbon fuels.

32
11/23/2003
Table 1.1 Survey of single-pulse experimental investigations of PDEs

reference propellants tube length
ignitor
energy
DDT
augmentation
impulse
measurement
Hinkey et al. (1995)
H
2
/air
H
2
/O
2

unknown 1.7 J Shchelkin spiral
pressure history
load cell
Sterling et al. (1996) H
2
/air 175.9 cm N/A unspecified device N/A
Broda et al. (1999) C
2
H
4
/air 182.9 cm 3.5 J half-disk protrusion N/A
Daniau et al. (2000) C
2
H
4
/O
2
6.5~10 cm 30 J
direct initiation
Shchelkin spiral
ballistic pendulum
Sanders et al. (2000)
C
2
H
4
/air
C
2
H
4
/O
2

135 cm N/A Shchelkin spiral N/A
Sinibaldi et al. (2000) C
2
H
4
/O
2
/N
2
190.5 cm 0.33~8.31 J none N/A
Litchford (2001) H
2
/O
2
90 cm 0.11 J Shchelkin spiral load cell
Sinibaldi et al. (2001)
C
2
H
4
/air
C
2
H
4
/O
2
C
3
H
8
/O
2

120 cm N/A N/A N/A
Cooper et al. (2002)
C
2
H
4
/O
2
/N
2

C
3
H
8
/O
2
/N
2

60.9~150 cm 0.03 J
Shchelkin spiral
blockage plate
orifice plate
ballistic pendulum
Lieberman et al. (2002) C
3
H
8
/O
2
/N
2
100 cm 0.03 J hot jet initiation ballistic pendulum
Meyer et al. (2002) H
2
/air 91.4 cm N/A
Shchelkin spiral
extended cavity
coannulus
N/A
Cooper et al. (2003) C
2
H
4
/O
2
105.7 cm 0.03 J none ballistic pendulum

33
11/23/2003
Table 1.2 Survey of multicycle experimental investigations of PDEs

reference propellants tube length
ignitor
energy
DDT
augmentation
impulse
measurement
frequency
Aarnio et al. (1996) H
2
/air 121.9 cm 1.7 J N/A
pressure history
load cell
5 Hz
Sterling et al. (1996)
H
2
/O
2

C
2
H
4
/O
2

15.2 cm
50.8 cm
N/A
unspecified
device
N/A
33 Hz
100 Hz
Stuessy and Wilson
(1996)
H
2
/O
2
53.3 cm N/A N/A N/A 10~12 Hz
Hinkey et al. (1997) H
2
/air 91.4 cm 1.5 J N/A N/A 10 Hz
Stuessy and Wilson
(1997)
C
3
H
8
/O
2
53.3 cm N/A N/A N/A
20.4~28.5
Hz
Brophy et al. (1998) JP10/O
2
15.2~76.2 cm 1.4 J N/A N/A 5 Hz
Schauer et al. (1999) H
2
/air
91.4 cm
(1~4 tubes)
N/A Shchelkin spiral N/A
0.5~100 Hz
(per tube)
Broda et al. (1999) C
2
H
4
/air 182.9 cm 4~8 J obstacle N/A 8~10 Hz
Brophy and Netzer
(1999)
JP10/O
2
29 cm 0.5 J N/A N/A 10 Hz
Zitoun and Desbordes
(1999)
C
2
H
4
/O
2
6.1~43.6 cm 35 J none pressure history 1~15 Hz
Schauer et al. (2000) H
2
/air 91.4 cm N/A Shchelkin spiral pressure history 14~40 Hz
Watts et al. (2000) C
2
H
4
/air N/A 25 J obstacle N/A 10 Hz
Falempin et al. (2001) C
2
H
4
/O
2
5.0~42.6 cm N/A N/A N/A 80 Hz
Litchford (2001) H
2
/O
2
90 cm 0.11 J Shchelkin spiral load cell 5 Hz
McManus et al. (2001) H
2
/air 25.4 cm 0.02 J N/A load cell 10~35Hz
Frankey et al. (2002) H
2
/air 182.88 cm N/A Shchelkin spiral pressure history 11~21 Hz
Shimo et al. (2002) C
2
H
4
/air 82.2 cm N/A Shchelkin spiral N/A 15 Hz
Brophy et al. (2002)
C
2
H
4
/air
JP10/O
2

25 cm N/A N/A pressure history
80 Hz
30 Hz
Farinaccio et al. (2002) C
3
H
8
/O
2
40 cm N/A N/A load cell 10~25 Hz
Brophy et al. (2003)
C
2
H
4
/air
C
3
H
8
/air
100 cm N/A obstacle N/A 30 Hz
Meyers et al. (2003) C
3
H
8
/O
2
7.62~30.5 cm N/A Shchelkin spiral pressure history 4.4~20 Hz
Rasheed et al. (2003) H
2
/air 100 cm N/A orifice plate spring-damper 10 Hz
Shehadeh et al. (2003) C
2
H
4
/O
2
/N
2
184 cm N/A N/A spring-damper 10 Hz

34
11/23/2003
1.4.1.2 Impulse Measurements
Impulse is one of the key performance measures of a PDE. Several techniques
have been proposed for the impulse measurement. The most straightforward one is to
integrate the pressure history at the closed end of the detonation tube. This technique
doesnt require complex measurement setup but can be implemented only to a simple
detonation tube without internal obstacles. Another popular technique used in single-
pulse experiments is known as ballistic pendulum technique, in which the detonation tube
is suspended as a pendulum by support wires and the impulse is determined by measuring
the maximum horizontal deflection of the tube (Cooper et al., 2003). In multicycle
experiments, the load cell technique is often used. In this technique, the force history is
directly measured by the load cell attached to the detonation tube through a load cell cage
(Hinkey et al., 1995). Since the negative thrust cannot be recorded, the impulse may be
overestimated in this technique. In addition, the response of the structure must be taken
into account. Other reported techniques include using damped thrust stand (Schaller et
al., 2001) and spring-damper system (Shehadeh et al., 2003). Due to the limitation of
each technique, a combination of these techniques may be required to obtain reliable
impulse measurements. Hinkey et al. (1995) measured the impulse in their single-pulse
experiments using both the pressure history and load cell techniques and found that the
impulse from the pressure history is about 20% lower than that from the load cell data.
With the measured impulse and the weight of fuel to generate this impulse, the
propulsive performance parameter such as the specific impulse can be readily obtained.
It should be mentioned that the specific impulse calculated from these direct-connect
experiments doesnt represent the specific impulse of PDEs at flight condition, under
35
11/23/2003
which the engine inlet and the momentum of the incoming air must be taken into account.
On the other hand, because the purging and refilling processes are not included in the
single-pulse experiments and the negative thrust may appear during these two processes,
the impulses from single-pulse experiments are usually lower than those from multicycle
experiments.
1.4.1.3 Detonation Initiation
Detonation initiation is one of the major challenges in the PDE design. A
detonation can be initiated either directly through a large amount of energy deposition or
indirectly through a low-energy deposition along with a deflagration-to-detonation
(DDT) process. Typical values for direct initiation energy for hydrocarbon fuel/air
mixtures are of the order of Kilo-Joules to Meg-Joules (Benedick et al., 1986). The
deposition of such high initiation energies is impractical for repetitive initiations. Most
PDE experiments have thus relied on a DDT process for detonation initiation.
According to the literature (Oppenheim, 1963; Lee and Moen, 1980; Kuo, 1986),
a DDT process consists of the following sequence of events: 1) deflagration initiation
a deflagration combustion is initiated by a low-energy deposition; 2) shock wave
formation the energy released by the deflagration increases the volume of the products
and generates a train of weak compression waves that propagate into the reactants ahead
of the flame and finally merge into a shock wave; 3) onset of an explosion in an
explosion the shock wave heats and compresses the reactants ahead of the flame,
creates a turbulent reaction zone within the flame front, and eventually cause one or more
explosive centers formed behind the shock front; 4) overdriven detonation formation
36
11/23/2003
strong shock waves are produced by the explosions and couple with the reaction zone to
form a overdriven detonation; 5) stable detonation establishment the overdriven
detonation wave decreases to a steady speed at around the CJ detonation velocity.
The distance from the ignition to the detonation formation point is referred to as
the DDT length, which is in general a function of the fuel and oxidizer, the tube diameter
and geometry, the tube wall surface roughness, and the method used to ignite the mixture.
Sinibaldi et al. (2000) investigated the dependence of the DDT length on the ignition
energy, ignition location, and mixture stoichiometry for a C
2
H
4
/O
2
/N
2
mixture. They
found that ignition energies above 0.28 J had little effect on DDT lengths. The ignition
location tests revealed that when the ignitor was placed 1.33 tube diameter from the head
wall, the DDT length could be reduced by up to 32%. Their results also showed that the
mixture equivalence ratio significantly affects the DDT length. The minimum DDT
length of around 7.5 cm for the C
2
H
4
/O
2
mixture was obtained with an equivalence ratio
of 1.2. A drastic increase in DDT length was observed when the equivalence ratio is less
than 0.75.
In general, the DDT length could be large compared to the tube length used in
PDE experiments. Hinkey et al. (1995) carried out a series of tests with H
2
/O
2
mixtures
of various equivalence ratios and found that the DDT lengths are on the order of 30 to
100 cm. They thus suggested using some DDT augmentation devices to enhance the
DDT process and reduce the DDT length, which was adopted in most late PDE
experiments. In the early multicycle experimental work of Nicholls (1957) and Krzycki
(1962), it is not clear whether full detonation waves were realized because a low-energy
37
11/23/2003
spark ignitor was used in their experiments and no DDT augmentation devices were
implemented.
A classical approach for DDT enhancement is to place a spiral, known as the
Shchelkin spiral (Shchelkin, 1940), into the detonation tube. Hinkey et al. (1995) first
applied this approach to their single-pulse PDE experiments with H
2
/O
2
mixture and
found that the Shchelkin spiral reduced the DDT length by a factor of about 3. In
addition to the Shchelkin spirals, other internal obstacles such as half-disk protrusions
(Broda et al., 1999), blockage plates and orifice plates (Cooper et al., 2002), and
coannulus (Mayer et al., 2002) have also been used by various researchers. It should be
noted, however, while enhancing the DDT processes, all these obstacles result in
significant total pressure loss and degrade the propulsive performance. Cooper et al.
(2002) reported that the DDT lengths could be reduced by an average of 65% in various
C
3
H
8
/O
2
/N
2
and C
2
H
4
/O
2
/N
2
mixtures using obstacles with a blockage ratio of 0.43,
whereas the impulse was reduced by up to 25%.
Another traditional detonation initiation concept involves using a predetonator
(Helman et al., 1986), which is in essence a detonation-to-detonation initiator. A
detonation is initiated in a more easily detonable mixture called as the driver gas and then
propagates into and initiates a detonation in the primary mixture. A simple example of
this concept is to fill a fuel/oxygen mixture or the driver gas in an initiation section near
the closed end of the detonation tube (Hinkey et al., 1995; Sanders et al., 2000). The
minimum length of the initiation section is the DDT length of the driver gas. The
aforementioned DDT augmentation devices can be further implemented within the
initiation section to achieve more rapid initiation. A disadvantage of this concept is the
38
11/23/2003
need to carry an additional driver gas or oxygen generator, which increases the system
weight. The additional driver gas also lowers the specific impulse since the weight flow
rates of both the fuel and the driver gas should be taken into account in calculating the
specific impulse. To mitigate this disadvantage, the amount of the driver gas or the
volume of the driver gas region must be as small as possible. A practical way to reduce
the volume of the driver gas region is to utilize an additional smaller tube for the driver
gas. This additional tube, usually with a volume on the order of 1% of that of the main
detonation tube, is called a predetonator. The detonation transmission from the
predetonator to the main detonation tube is thus a key issue in the predetonator
applications (Sinibaldi et al., 2001; Santoro et al., 2003).
Diffraction of detonation from a small tube into an unconfined space has been
extensively investigated in the past (Lee, 1984; Desbordes, 1988). According to the
literature, a successful detonation transmission happens if the tube diameter is larger than
the critical diameter. This critical diameter is usually expressed in terms of the
detonation cell size of the mixture. It is now commonly accepted that the critical
diameter is about thirteen times of the detonation cell size for smooth circular tubes. The
detonation transmission from the predetonator to the main detonation tube has also been
investigated recently (Sinibaldi et al., 2001; Santoro et al., 2003; Brophy, et al., 2003).
Successful detonation transmission from the predetonator to the main detonation tube
could be achieved at predetonator tube diameters less than the critical diameter because
of the confinement of the transition region and main detonation tube (Santoro et al.,
2003).
39
11/23/2003
Other techniques proposed for promptly achieving detonation initiation include
hot jet initiation, detonation wave focusing, etc. Hot jet initiation was observed by
Knystautas et al. (1979) for sensitive fuel-oxygen mixtures. Lieberman et al. (2002)
recently demonstrated the possibility of using a hot jet to initiate a detonation in a short
detonation tube filled with the C
3
H
8
/O
2
/N
2
mixture. The idea of the detonation wave
focusing is to initiate the detonation in the main detonation tube through the merging of
the detonation waves from a bunch of small tubes (Jackson and Shepherd, 2002).
1.4.2 Numerical Studies on PDEs
In addition to experimental investigations, substantial attempts were made to
numerically study the single-pulse and multicycle operations of various PDEs consisting
of single or multiple detonation tubes with or without nozzles and ejectors. Based on the
spatial dimensions used, these numerical studies can be classified into one-dimensional or
quasi-one-dimensional and two-dimensional or axisymmetric simulations, as summarized
in Tables 1.3 and 1.4, respectively.
In contrast to experiments in which ethylene (C
2
H
4
) is often used as the fuel, most
numerical studies use hydrogen (H
2
) as the fuel because of its relatively simpler chemical
kinetics. Another significant difference between experiments and simulations is the
detonation initiation. In experiments, as discussed in Sec. 1.4.1.3, the detonation is
usually initiated indirectly through a low-energy deposition along with a DDT process
since the direct initiation with high-energy deposition is not practical. In numerical
simulations, however, a small spark region with high temperature and pressure is
commonly implemented to directly initiate the detonation. In addition, because of the
40
11/23/2003
lack of the DDT process and because there is no frequency limitation arising from the
hardware such as the solenoid valves, the operating frequencies are much higher in
numerical simulations than those that can be achieved in experiments. In spite of many
simplifications used in the numerical simulation, its capabilities in providing the detailed
unsteady flowfield make it a practical and efficient methodology for PDE performance
and flow dynamic analyses.

Table 1.3 Survey of one-dimensional numerical simulations of PDEs

reference exit B.C. propellants
chemical
kinetics
tube
length
spark
length
nozzle
frequency
(cycles)
Cambier and
Adelman (1988)
N/A H
2
/air 16-step 7-species 50 cm 2 cm divergent
667 Hz
(3 cycles)
Bussing (1994) N/A
H
2
/O
2

H
2
/air
18-step 7-species 20 cm 0.2 cm none single pulse
Sterling et al.
(1995)
constant
pressure
H
2
/air
equilibrium
chemistry
100 cm 0.5 cm none
220 Hz
(6 cycles)
Cambier and
Tegner (1998)
N/A H
2
/air N/A 10 cm 0.4 cm divergent
1000~2600
Hz
Kailasanath et al.
(1999)
relaxing
pressure
H
2
/air 48-step 8-species 20 cm 0.2 cm none single pulse
Fong and Nalim
(2000)
constant
pressure
H
2
/air N/A 10.2 cm N/A none 30~150Hz
Mohanraj and
Merkle (2000)
constant
pressure
H
2
/O
2
1-step 2-species 50 cm 0.12 cm
divergent
CD
230 Hz
Ebrahimi and
Merkle (2002)
constant
pressure
H
2
/O
2
16-step 8-species 20 cm 0.2 cm none
400 Hz
(3 cycles)
Ebrahimi et al.
(2002)
varying
pressure
with 2D
correction
H
2
/O
2
1-step 2-species 10.2 cm 0.1 cm none
200~1000
Hz
Wu, Ma, and
Yang (2003)
zero-
gradient
H
2
/air 1-step 2-species 60 cm 0.02cm CD 200~400 Hz

41
11/23/2003
Table 1.4 Survey of two-dimensional numerical simulations of PDEs

reference
num.
of
tubes
propellants
chemical
kinetics
tube
length
spark
length
nozzle
cycles/
frequency
Eidelman et al.
(1990)
1 C
2
H
4
/air
energy release
behind detonation
8cm
16 cm
N/A none single pulse
Lynch and
Edelman (1996)
1 H
2
/air 7-step 6-species
8cm
16 cm
1~2 cm none single pulse
Cambier and
Tegner (1998)
1 H
2
/air N/A 10 cm 0.4 cm divergent single pulse
Eidelman and
Yang (1998)
1 C
2
H
2
/air 1-step 16 cm N/A
convergent
divergent
single pulse
Li et al. (2000) 1 C
2
H
4
/air
2-step induction-
parameter
50 cm 1.25 cm none single pulse
Li and Kailasanath
(2001)
1 C
2
H
4
/O
2

2-step induction-
parameter
135 cm 2 cm none single pulse
Allgood et al.
(2002)
1 H
2
/O
2
/Ar
2-step induction-
parameter
50.8 cm N/A ejector single pulse
Ebrahimi et al.
(2002)
1 H
2
/O
2
16-step 8-species 10.2 cm 0.1 cm none single pulse
Kawai and
Fujiwara (2002)
1 H
2
/O
2
/Ar
2-step induction-
parameter
10~40 cm N/A none 2 cycles
Li and Kailasanath
(2002)
1 C
2
H
4
/O
2

2-step induction-
parameter
10~500
cm
1 cm none single pulse
Yungster and
Perkins (2002)
1 H
2
/O
2
19-step 9-species 182.9 cm N/A ejector
5 cycles
(122 Hz)
Li and Kailasanath
(2003)
1
C
2
H
4
/O
2

C
2
H
4
/air
2-step induction-
parameter
50 cm
jet
initiation
none single pulse
Povinelli and
Yungster (2003)
1 C
2
H
4
/air 36-step 20-species 100 cm 0.5~1 cm none single pulse
Tangirala et al.
(2003)
1
H
2
/air
C
2
H
4
/O
2

23-step 7-species
38-step 21-species
100 cm 0.5 cm none single pulse
Wu, Ma, and
Yang (2003)
1 H
2
/air 1-step 2-species 60 cm 0.02 cm none 200~400 Hz
Yungster (2003) 1
H
2
/O
2

H
2
/air
19-step 9-species 100 cm 0.5 cm divergent
3 cycles
(110Hz)
Ebrahimi et al.
(2001)
2 H
2
/O
2
16-step 8-species 10 cm N/A CD single pulse
Ebrahimi and
Merkle (2002)
2 H
2
/O
2
16-step 8-species 10 cm N/A CD single pulse
Ebrahimi et al.
(2003)
1~3 H
2
/O
2
16-step 8-species 10~30 cm N/A CD single pulse
Ma et al. (2003) 1; 3 H
2
/air 1-step 2-species 60 cm 0.02 cm CD 250~333 Hz
42
11/23/2003
1.4.2.2 One-Dimensional Numerical Simulations
One of the first numerical studies on PDE reported in the literature was attributed
to Cambier and Adelman (1988). Quasi-one dimensional simulations with multi-step
finite rate chemical kinetics were carried out for a 50-cm-long detonation tube attached to
a 43-cm-long diverging nozzle with stoichiometric H
2
/air mixture. The detonation was
initiated at the closed end by a 2-cm-long spark region with a temperature of 1500 K.
The refilling process started when the head-end pressure fell to 3.5 atm. The fresh
reactants moved at speeds up to 350 m/s. The engine reached steady cyclic operation by
the third cycle with a cycle frequency of about 667 Hz. Specific impulses up to 6507 s
were reported. These very high specific impulses may be due to the facts that their
calculations were based on the gross thrust from the exit plane and that the contribution
from the spark region on the impulse was not accounted for.
Sterling et al. (1995) also conducted one-dimensional simulations with H
2
/air
mixture. The detonation tube is 100 cm long. The spark region has a length of 0.5 cm
with a temperature of 3000 K and a pressure of 50 atm. The refilling process began when
the head-end pressure decayed to the atmospheric pressure. They calculated the specific
impulse for the sixth cycle based on the head-end pressure history and obtained a value of
5152 s, which is much smaller than that of Cambier and Adelman (1988).
Cambier and Tegner (1998) examined the effect of the spark region on the
performance in their quasi-one-dimensional and two-dimensional simulations with H
2
/air
mixture for a 10-com-long detonation tube with various divergent nozzles. The spark
region spans 0.4 cm, with a temperature of 2500 K and a pressure of 50 atm. They
observed that the contribution from the spark region on the single-pulse peak impulse
43
11/23/2003
ranges from 17% to 27%. After subtracting this contribution, they obtained specific
impulses of 3500 s ~ 4100 s based on the head-end pressure history of their quasi-one-
dimensional multicycle results.
In addition to the differences in the operation conditions and in the calculation
methods for the specific impulse, the exit boundary condition plays another important
factor that cause the differences in the reported specific impulses from one-dimensional
simulations (Kailasanath et al., 1999). Choosing appropriate exit boundary conditions
represents a major challenge in the one-dimensional simulations for PDEs. In general, if
the outflow is supersonic, extrapolation of the condition from the inside can be used.
However, for subsonic outflow, the flow inside the domain near the exit relates to the
flowfield outside the domain, which is unknown from the one-dimensional simulations.
The situation is further compounded by the existence of the unsteady detonation waves
and other associated waves. Strictly, there are no accurate exit boundary conditions for
this kind of complex unsteady subsonic outflow.
Most one-dimensional simulations have relied on a fixed-pressure boundary
condition for the outflow (Cambier and Adelman, 1988; Starling, 1995). Ebrahimi et al.
(2002) used the fixed-boundary condition as well as a varying-pressure condition based
on corrections from two-dimensional simulations. They found that with the fixed-
pressure condition, the flow chokes as soon as the detonation wave arrives at the exit,
whereas the varying-pressure result shows a duration of a subsonic outflow before
choking. Kailasanath et al. (1999) utilized a relaxing-pressure boundary condition. The
pressure is set to reach the ambient value in a relaxation length from the end of the
detonation tube. A larger value of the relaxation length implies a slower relaxation of
44
11/23/2003
pressure to ambient. Based on single-pulse operation, they observed that the specific
impulse is larger for slower relation cases. They then concluded that the variations in
specific impulses reported in the literature could be explained on the basis of exit
boundary conditions used.
1.4.2.3 Two-Dimensional Numerical Simulations
In spite of the computational efficiency of the one-dimensional simulations for
performance predictions, the difficulty in specifying the exit boundary condition makes
them less attractive. Multi-dimensional simulations with computation domain including
both the detonation tube and the external region are required in order to faithfully
describe the system dynamics, especially in the near field of the tube exit where the flow
is intrinsically multi-dimensional, and to provide more accurate performance predictions.
This subsection will provide a brief review on some of the two-dimensional simulations.
Issues related to nozzle and multitube will be discussed in other subsections.
Eidelman et al. (1990) carried out two-dimensional axisymmetric simulations for
a valveless PDE with C
2
H
4
/air mixture. Both internal and external flowfields were
included in their computational domain. The system simulated consisted of a cylindrical
detonation tube with a length of 8 or 16 cm. Its operation is a little different from that
described in Sec. 1.3.1. The detonation is initiated at the aft open end and travels toward
the head end where the thrust wall, fuel injector, and air inlet are located, ejecting burned
gases through both the open end and the air inlet. Air is entrained into the detonation
tube when the pressure falls below the atmospheric pressure during part of the cycle.
Since no valves are required to separate the air inlet and the detonation tube, this kind of
45
11/23/2003
PDE is self-aspirating and is called valveless PDE. From a scaling study based on single-
pulse operation, they concluded that the thrust of the detonation tube increases linearly
with the detonation tube length when the other parameters are kept constant. This
conclusion, however, is incorrect or misleading, since the thrust they calculated was
based on only a portion of a cycle.
Lynch and Edelman (1996) conducted two-dimensional axisymmetric simulations
with multi-step finite rate chemical kinetics for H
2
/air mixture for a valveless PDE
similar to that analyzed by Eidelman et al. (1990). They investigated the flow dynamics
with two types of air inlet. In the straight inlet case, a large recirculation zone is formed
behind the thrust surface upon the inflow of the fresh air. This recirculation zone may
cause an increase in the time required to entrain the air and to purge the products. By
using a scoop inlet, the recirculation zone is reduced. They suggested additional inlet
shaping and viscous entrainment effects to obtain optimizing design.
Most of the recent two-dimensional studies focused on single-tube PDEs with
detonation initiated at the closed end. Ebrahimi et al. (2002) carried out two-dimensional
simulations with a H
2
/O
2
mixture to provide exit boundary condition corrections for their
one-dimensional simulations. Li and Kailasanath (2001) looked at the single-pulse
flowfield evolution under full or partial fills for a C
2
H
4
/O
2
mixture. The two-step
induction-parameter chemical kinetic model was adopted, in which combustion radicals
are formed in the first step, which is represented by an induction parameter, and the
energy is released in the second step. This model has also been used by other researchers
for H
2
/O
2
mixtures (Allgood et al., 2002; Kawai and Fujiwara, 2002). In a later work of
Li and Kailasanath (2002), they further investigated the effect of partial filling on the
46
11/23/2003
propulsive performance and observed that the fuel-based specific impulse increases as the
partial filling ratio decreases for single-pulse operation. A scaling relation of the specific
impulses between the partial and full fillings was then proposed.
Many of the two-dimensional simulations are limited to single-pulse operations.
However, significant differences exist between single-pulse and multicycle computations
and the conclusions from single-pulse studies may not be applied to multicycle cases
directly (Cambier and Tegner, 1998; Yungster, 2003; Wu, Ma, and Yang, 2003). To
date, two- dimensional simulations with multicycle operations are rather sparse. Kawai
and Fujiwara (2002) carried out two-dimensional simulations for the first two cycles of a
straight tube PDE containing an Ar-diluted stoichiometric H
2
/O
2
mixture, with attention
paid to the injection behavior during the 2nd cycle. Yungster and Perkins (2002)
conducted multicycle simulations of various PDE-ejector configurations utilizing H
2
/O
2

mixture. The thrust and mass flow rate over five cycles are presented. They stated that
more cycles might be required to establish some kind of limit cycle or steady periodic
cycle. Up to now, two-dimensional simulations with steady periodic cycle attained have
been reported only by Ma and colleagues (Wu, Ma, and Yang, 2003; Ma et al., 2003).
1.4.2.4 Effect of Nozzles
The detonation products exhausting from the detonation tube carry a large amount
of internal energy. A nozzle is thus needed to convert this internal energy to kinetic
energy to improve the performance. In conventional steady engines, the nozzle is
optimized by matching the exit pressure with the ambient pressure. The nozzle designs
for PDEs, however, are complicated by the unsteady nature of the PDEs, especially the
47
11/23/2003
complex shock waves resulting from the repetitive detonations. To date, no theory for
PDE nozzle designs has been developed. Several numerical and experimental studies
that focused on the effect of nozzles on the PDE performance are reviewed in the
following paragraphs.
Cambier and Adelman (1988) considered a 43-cm-long divergent nozzle attached
to a 50-cm-long detonation tube in their quasi-one-dimension multicycle simulations with
a H
2
/air mixture. They stated that other designs with a convergent section and throat
cause the detonation waves to reflect as shock waves from the converging section. The
reflected shock waves elevate the head-end pressure, interfere with the refilling process,
and lower the cycle frequency. On the other hand, a divergent nozzle causes reflected
rarefaction waves that propagate back into the tube to allow rapid refilling processes. A
cycle frequency of as high as 667 Hz was reported in their studies with a divergent
nozzle. However, they didnt compare the propulsive performance numbers between the
divergent nozzle and other nozzles. It is thus unclear which kind of nozzle is better for
achieving higher propulsive performance.
Cambier and Tegner (1998) investigated the effects of five different divergent
nozzles on the performance based on quasi-one-dimensional multicycle simulations and
two-dimensional single-pulse simulations. The lengths of the detonation tube and the
nozzle are 10 cm and 5 cm, respectively. Both the tube and nozzles were initially filled
with a H
2
/air mixture. Their single-pulse results indicate that the divergent nozzles can
increase the impulse and that the bell-shaped nozzle produces a higher impulse than those
with shapes of positive curvatures. In addition, significant differences were observed
48
11/23/2003
between single-pulse and multicycle results, and between quasi-one-dimensional and
two-dimensional results.
Eidelman and Yang (1998) also considered different nozzle shapes, i.e. three
divergent and three convergent nozzles, in their two-dimensional single-pulse
simulations. The detonation tube was filled with a C
2
H
2
/air mixture, whereas the nozzle
was filled with air. Their results showed that both the convergent and divergent nozzles
drastically increase the PDE performance. Convergent nozzles lead to significant
increases in impulse generation time and rapid changes in thrust due to shock wave
reflections and relative later times in attaining the peak impulses. Divergent nozzles tend
to generate impulse rapidly during detonation product expansion due to the increase in
effective thrust wall area so that the peak impulses are reached earlier. These peak
impulses, however, are followed by rapid reductions due to the negative thrust generated
during the overexpansion of the detonation products. Among the three divergent nozzles,
the bell-shaped nozzle produces the highest peak impulse.
Mohanraj and Merkle (2000) used quasi-one-dimension multicycle simulations to
study the effects of the divergent and convergent-divergent (CD) nozzles at different back
pressures. The injection pressure is fixed at one atmosphere. They found that both types
of nozzles provide performance benefit at low back pressures, while the divergent nozzle
causes performance deterioration at high back pressures. Their results also showed that
divergent nozzles lead to higher cycle frequency.
More recently, Yungster (2003) examined the effect of divergent nozzles on PDE
performance in his two-dimensional simulations with operations up to 3 cycles. Both
H
2
/O
2
and H
2
/air mixtures were employed. The detonation tube and nozzles are 100 cm
49
11/23/2003
and 40 cm long, respectively. In the single-pulse operation, the nozzles can utilize some
of the energy carried by the detonation products to significantly augment the impulse
obtained from the detonation tube alone. Impulse augmentation ratios of up to 2.34 and
1.41 were obtained with the bell-shaped nozzle for H
2
/O
2
and H
2
/air mixtures,
respectively. In multicycle operation, however, the performance augmentation is very
limited or even negative because of the nozzle flow overexpansion during most of the
purging and refilling periods. He stated that in order for the nozzle to remain effective
after the first cycle, the combustion products must be purged from the nozzle before the
next cycle is started. This will, however, reduce the cycle frequency.
The effect of nozzles was also investigated in several experiments. Stuessy and
Wilson (1997) conducted multicycle experiments with and without conical exhaust
nozzles. They noted that the inclusion of the nozzle improves the performance. Daniau
et al. (2000) investigated the effects of divergent nozzles of different shapes and lengths
on the performance of a PDE based on single-pulse experiments. They also observed that
bell-shaped nozzles are especially efficient. Cooper et al. (2002) carried out single-pulse
experiments and investigated the effect of a 30-cm-long divergent nozzles attached to a
101.6-cm-long detonation tube. They found that the divergent nozzle had a negligible
effect, increasing the specific impulse by an average of 1% which is within the
experimental uncertainty.
The observations and conclusions reported in these previous numerical and
experimental studies indicate the nozzle effect on the PDE performance is far from
mature. More studies based on multicycle operations need to be conducted.
50
11/23/2003
1.4.2.5 Multitube PDEs
In addition to the nozzle, another way that may improve the PDE performance is
the use of multiple detonation tubes (Ma et al., 2003). The concept of multitube designs
is not new. In as early as 1950, Goddard considered a valved multiple combustor
deflagration-based pulse jet engine (Goddard, 1950). Bussing proposed a rotary valve
multitube pulse detonation engine concept in 1995 by combining some aspects of the
Goddard design and the detonation process (Bussing, 1995). It consists of several
detonation tubes coupled to an air inlet and fuel source via a rotary valve. The rotary
valve isolates the steady operation of the air inlet and fuel system from the unsteady
operation of the detonation tubes and allows the filling of some of the detonation tubes
while detonation occurs in other tubes. A simple performance model, including the
contributions from the inlet, the mixer, the combustor, and the nozzle, was proposed by
Bratkovich and Bussing (1995) to examine the aforementioned multitube PDE
performance characteristics over a wide range of flight regime. Recently, an in-house
four-detonation-tube research PDE was built at the US Air Force Research Laboratory,
serving as a test-bed for detonation initiation concept, high frequency operation, heat
transfer studies, multitube detonation engine operation, and pulse jet research (Schauer,
2001). It used a valve system based on the Quad-4, a sixteen-valve, four-cylinder
automobile engine from General Motors. A rotary position sensor was adapted to the
intake camshaft to provide both an index of the valve timing sequence and the relative
position of the valves. Each detonation tube can operate at frequencies of up to 100 Hz.
Numerical investigations on multitube PDEs also commenced recently. Mohanraj
et al. (2001) presented an approximate model for a PDE with five detonation tubes. Time
51
11/23/2003
accurate one-dimensional solution of the flowfield was obtained for only one tube. Time
delayed versions of this solution were used to model the other tubes. Their results
showed that the filling process in a multitube PDE can be markedly different from that in
a single-tube PDE. For example, increasing the fill time does not affect the fill fraction in
the tubes at some conditions. They also observed that the number of cycles needed to
attain steady periodic operation in a multitube configuration is typically larger. Although
this approach saves computational effort, the resultant error due to the model
approximation is difficult to estimate and may likely be quite large.
Ebrahimi et al. (2001) conducted two-dimensional simulations for a dual-tube
PDE, but with only the first-pulse operation. They found that the pressure induced by the
detonation in the neighboring tube is nearly as large as that produced by the detonation
itself and that the shock wave produced by the detonation is sufficient to initiate
combustion in the adjacent tube filled with fresh reactants. In a more recent work
(Ebrahimi et al., 2003), the effects of the number of tubes and the tube length on the
flowfield interaction among tubes were investigated based on the first-pulse operation.
The results indicate that the flowfield interaction among three tubes is approximately a
factor of three smaller than that between two tubes, but that increased tube length has no
significant effect. The first multidimensional numerical study on multitube PDEs
involving multicycle operations was conducted by Ma et al. (2003).
1.4.3 Analytical Studies on PDEs
The intrinsic unsteady nature of the PDE operation process makes the theoretical
analyses rather difficult or even impossible. On the other hand, it is desirable to develop
52
11/23/2003
simple analytical models that can be used to rapidly and reliably predict the PDE
performance and to conduct parametric studies. Several analytical models have been
proposed in the past and are summarized in Table 1.5. Based on how the impulse is
obtained, these models fall into two classes. One class employs unsteady gasdynamic
analysis to determine the instantaneous pressures and forces acting on the thrust wall and
calculate the impulse as the integration of them (Nicholls, et al., 1957; Wintenberger et
al., 2002; Wintenberger and Shepherd, 2003). The other class, without the need to
consider the unsteady wave process within the detonation tube, obtains the impulse or
specific impulse of the engine by deriving the flow properties at the exit plane (Talley
and Coy, 2002; Heiser and Pratt, 2002; Wu, Ma, and Yang, 2003).

Table 1.5 Survey of analytical studies on PDEs

classification based on
impulse calculation
reference remarks
Nicholls et al.
(1957)
Only constant pressure portion was
considered.
Wintenberger et al.
(2003)
Decaying pressure portion was modeled
through dimensional analysis and
experimental calibration.
head-end pressure
Wintenberger and
Shepherd (2003)
Filling process was incorporated.
Talley and Coy
(2002)
The cycle consists of a constant-volume
combustion, a constant-volume blowdown,
and a constant-pressure filling process.
Heiser and Pratt
(2002)
Classical thermodynamic cycle analysis was
applied to an ideal PDE cycle to predict the
upper limit of the PDE performance.
exit-plane properties
Wu, Ma, and Yang
(2002)
The Heiser and Pratt (2002) approach was
extended to accommodate property variations
across the detonation wave front.
53
11/23/2003
The most simplified model was reported by Nicholls et al. (1957) in conjunction
with their straight-tube experiments. The pressure at the head end remains constant until
the arrival of the first rarefaction wave generated as the detonation wave emerges from
the open end. The head-end pressure then decays due to the continuing arriving of more
rarefaction waves. Only the contribution of the constant pressure portion was considered
and the contribution of the decaying pressure period was neglected. Consequently, this
model underpredicted the impulse.
Recently, Wintenberger et al. (2003) presented a semi-analytical model for a
single-pulse detonation tube. Both the constant pressure and decaying pressure portions
are considered. The duration and amplitude of the constant pressure portion is
determined by analyzing the gasdynamics of the self-similar flow behind a steadily
moving detonation wave within the tube. The decaying pressure portion is analyzed
using dimensional analysis, leaving a nondimensional parameter to be obtained through
experimental calibrations. Wintenberger and Shepherd (2003) later extended this model
to multicycle operations by accommodating the refilling process. These models,
however, are only useful for predicting the impulse from a simple straight detonation
tube.
The constant volume limit model proposed by Talley and Coy (2002) is based on
a cycle consisting of a constant-volume combustion process, a constant-volume
blowdown process, and a constant-pressure filling process. The characteristic wave
transit times are assumed to be much shorter than the blowdown time. During the
blowdown process, the gases in the chamber are modeled as being time varying but
spatially uniform while venting occurs through an infinitely thin nozzle with quasi-steady
54
11/23/2003
flow. When the chamber pressure reaches the fill pressure, the filling process begins and
is modeled as a constant-pressure process.
Heiser and Pratt (2002) applied the classical thermodynamic cycle analysis to an
ideal PDE thermodynamic cycle. The detonation wave converts the reactants to the
products with CJ properties. These products are then assumed to isentropically expand to
the atmospheric pressure. They recognized that no simple device can fulfill this unsteady
isentropic expansion process but the upper limit of the performance can be established
through this assumption. After the thermodynamic cycle efficiency is obtained, the
specific impulse can be readily determined through the control-volume energy balance.
This model was later extended by Wu, Ma, and Yang (2003) to accommodate property
variations across the detonation wave front. The advantage of this model over that
proposed by Tally and Coy (2002) is that the detonation combustion is considered.
However, since all thermodynamic cycle analyses are based on an assumption that all the
fluid elements experience the same processes at same sequence, the purging and refilling
processes involved in a PDE can thus not be incorporated. Moreover, comparisons with
more accurate numerical simulations should be made in order to judge its accuracy in
predicting the upper performance limit.
1.5 Research Objectives
The primary objective of this research is to evaluate the concept of airbreathing
PDEs, to investigate their flow dynamics, to determine their performance, to identify the
55
11/23/2003
various loss mechanisms, and to examine the effects of the operation timing and system
geometry.
Figure 1.13 schematically shows the airbreathing PDE studied herein. It includes
a supersonic inlet with mixed compression, an air manifold/acoustic cavity, a rotary
valve, a combustion chamber containing single or multiple detonation tubes, and a
convergent-divergent nozzle. Hydrogen is used as the fuel. As a specific example, the
flight condition with a flight altitude of 9.3 km and a flight Mach number of 2.1 is
considered. The static pressure and temperature are 0.29 atm and 228 K, respectively,
corresponding to a total pressure of 2.65 atm and a total temperature of 428 K.

inlet
air diffuser
& distributor
fuel manifold
& injector
multiple-tube
combustor
controller
nozzle

combustor inlet
isolator
nozzle
rotary valve manifold
inlet
air diffuser
& distributor
fuel manifold
& injector
multiple-tube
combustor
controller
nozzle
inlet
air diffuser
& distributor
fuel manifold
& injector
multiple-tube
combustor
controller
nozzle

combustor inlet
isolator
nozzle

Fig. 1.13 Supersonic airbreathing pulse detonation engine

In Chapter 2, the supersonic inlet dynamics is studied. The time-dependent
Favre-averaged Navier-Stokes equations along with a two-equation turbulence model are
solved by a finite-volume method, with a four-stage Runge-Kutta time integration
technique for temporal discretization and the Harten-Yee upwind TVD scheme for spatial
56
11/23/2003
discretization. The steady-state inlet flow dynamics and the response of the inlet shock
system to downstream pressure disturbance are investigated.
Chapter 3 introduces the recently developed space-time conservation
element/solution element method and utilizes it to the quasi-one-dimensional study of a
single-tube PDE. Chemical kinetics is simulated using a simple one-progress-variable
model for a stoichiometric hydrogen/air mixture. The flow dynamics is examined
through numerically obtained x-t diagrams. The effects of operating frequency and valve
close-up time on the propulsive performance are investigated.
In Chapter 4, two-dimensional analyses are carried out for single-tube PDEs with
or without CD nozzles under multicycle operations. The two-dimensional code is
efficiently parallelized by implementing the message-passing-interface (MPI) library and
a domain decomposition technique. The flow dynamics involved in the PDE operation is
examined in detail. A flow-path based performance prediction model is proposed to
estimate the upper performance limit of the current PDE system. The various
performance loss mechanisms are identified. The effects of operation timing including
the cycle time or frequency, valve close-up time, and purge time on the propulsive
performance are comprehensively studied. Moreover, the effects of nozzle
configurations in terms of the nozzle length and throat area are examined.
In Chapter 5, multitube airbreathing PDEs with repetitive operations are studied
by means of two-dimensional simulations. The combustor consists of three detonation
tubes. The flow evolution and tube interactions are carefully examined and the various
loss mechanisms are identified. The effects of operating timing such as the cycle time
and the valve close-up time were investigated systematically. Comparisons with the
57
11/23/2003
single-tube results are made. The effect of the system geometry is partially assessed by
considering a free volume between the detonation tubes and the common nozzle.
Finally, Chapter 6 summarizes the present work and provides several
recommendations for the future work.

58
11/23/2003
Chapter 2
Supersonic Inlet Dynamics
The inlet and its interaction with combustor represent a crucial aspect in the
development of any airbreathing engine, including PDEs. The inlet is designed to
capture and supply stable airflow at a rate demanded by the combustor, and to maintain
high pressure recovery and stability margin at various engine operating conditions. The
overall vehicle performance depends greatly on the energy level and flow quality of the
incoming air. A small loss in inlet efficiency may translate to a substantial penalty in
engine thrust. Moreover, any change in the inlet flow structure may modify the
downstream combustion characteristics and subsequently lead to undesirable behaviors,
such as flame blow-off and flashback. Thus, matching inlet behavior to engine
requirements is of fundamental importance to designers (Yang and Cappuccio, 1990).
In addition to its primary function of supplying air, an inlet has a determining
influence on the dynamics of the entire system through its intrinsic unsteadiness and
interactions with the combustion chamber. Typically, pressure waves are produced in the
combustion chamber and propagate upstream to interact with the inlet flow through a
manifold where mixing of air and fuel occurs. The resultant flow oscillations in the inlet
diffuser then either propagate downstream in the form of acoustic waves or are convected
downstream with the mean flow in the form of vorticity and entropy waves, and further
reinforce the unsteady motions in the combustor. A feedback loop is thus established
between the inlet and combustor. The situation is much more complicated in a
supersonic airbreathing PDE due to its pulsed operation.
59
11/23/2003
2.1 Problem Description
In this chapter, axisymmetric two-dimensional simulations are conducted for a
mixed-compression supersonic inlet as shown in Fig 2.1. The cowl radius is R
c
= 3.4 cm.
This inlet is designed for a flight Mach number of 2.1 and a flight altitude of 9.3 km. The
static pressure and temperature of the freestream are 0.29 atm and 228 K, respectively.
The corresponding total pressure and temperature are 2.65 atm and 428 K, respectively.
The time-dependent Favre-averaged Navier-Stokes equations are solved by a finite-
volume method, with a four-stage Runge-Kutta time integration technique for temporal
discretization and the Harten-Yee upwind TVD scheme (Harten, 1983; Yee, 1989) for
spatial discretization. The specific objectives of this work are: 1) to understand the
overall shock and flow structures in a realistic supersonic inlet, 2) to examine the effects
of downstream disturbances on inlet dynamics, 3) to investigate the influences of the
viscous boundary layer and flow separations downstream of the terminal shock, and 4) to
provide entrance information for further combustor simulations.

31.90
3.40
3.89 20
31.25
2.84
20.25
25
unit: cm
8

Fig. 2.1 Configurations of a mixed-compression supersonic inlet

60
11/23/2003
2.2 Governing Equations
The analysis for the supersonic inlet flows is based on the axisymmetric, Favre-
averaged conservation equations of mass, momentum, and energy. In a vector notation,
this set of equations becomes
( ) ( )
v v
t x r

+ + =

Q
E E F F H (2.1)
where x and r stand for the axial and radial coordinates, respectively. The conserved
variable vector Q, convective flux vectors E and F, diffusion flux vectors E
v
and F
v
, and
source vector H are defined as
[ , , , ]
T
t
r u v e = Q (2.2)

2
[ , , , ( ) ]
T
t
r u u p uv e p u = + + E (2.3)

2
[ , , , ( ) ]
T
t
r v uv v p e p v = + + F (2.4)
[0, , , ]
T
v xx xr xx xr x
r u v q = + E (2.5)
[0, , , ]
T
v xr rr xr rr r
r u v q = + F (2.6)
[0, 0, , 0]
T
r p

= H (2.7)
In the above equations, standard notations in fluid mechanics are used. The stress tensor
and the heat flux vector q are given by
(2 / 2/ 3 )
xx
u x = u (2.8)
(2 / 2/ 3 )
rr
v r = u (2.9)
( / / )
xr
v x u r = + (2.10)
(2 / 2/ 3 ) v r
= u (2.11)
61
11/23/2003
/
x
q T x = (2.12)
/
r
q T r = (2.13)
The divergence in cylindrical coordinates is defined as

1 ( ) u rv
x r r

= +

u (2.14)
The pressure p and temperature T are obtained through the equation of state for a perfect
gas as

2 2
1
( 1)[ ( )]
2
t
p e u v = + (2.15)
/( ) T p R = (2.16)
where and R are the specific heat ratio and gas constant, respectively. These two
parameters are taken as: 1.4 = and 287 J/kg K R = for air.
The viscosity and thermal conductivity contain both laminar and turbulent
components,
,
l t l t
= + = + (2.17)
The laminar component of viscosity
l
is obtained from the Sutherland law (White,
1974),

1.5
0
0 0
l
T S T
T S T
| | +
=
|
+
\ .
(2.18)
where
0
is the Sutherland law reference viscosity, T
0
the Sutherland law reference
temperature, and S the Sutherland law temperature. For the current problem, these
constants are:
5
0 0
300 K, 1.8464 10 kg/m s, 110 K T S

= = = .
62
11/23/2003
The turbulent viscosity
t
is evaluated with a two-layer model of Rodi (1991).
This model combines the standard high-Reynolds-number - k model in the bulk flow
region with a one-equation model near the wall. In the bulk flow region, the distributions
of the turbulent kinetic energy k and its dissipation rate are calculated from the
turbulence transport equations. In the viscosity-affected near-wall region, the turbulent
kinetic energy is still determined from the transport equation, but the dissipation rate is
determined from a prescribed length-scale distribution. The details of the application of
this model can be obtained from the thesis of Oh (1994).
The laminar and turbulent thermal conductivities
l
and
t
are calculated as

t
t p
t
l
l p
l
C C
Pr
,
Pr

= = (2.19)
where
p
C is the specific heat at constant pressure. The laminar Prandtl number
l
Pr and
the turbulent Prandtl number
t
Pr are taken to be 0.73 and 0.9, respectively.
2.3 Numerical Procedures and Validations
The governing equations outlined above are solved numerically by means of a
finite-volume methodology. Temporal discretization is obtained using a four-stage
Runge-Kutta integration method. The spatial discretization employs an upwind total-
variation-diminishing (TVD) scheme developed by Harten (1983) and Yee (1989) in
generalized coordinates for the convective terms and second-order central-differencing
method for the diffusion terms. Specific details of the numerical algorithm can be
obtained from Oh (1994).
63
11/23/2003
The boundaries encountered in the internal and external flowfields of an
axisymmetric supersonic inlet include inflow, outflow, symmetry, wall, and far-field
conditions. Since the inflow is supersonic along the boundary, the flow variables are
fixed at their corresponding free stream values. At the exit boundary of the inlet duct, a
constant back pressure is prescribed with other flow variables deduced from the interior
points for the steady-state calculations. For the unsteady calculations, a sinusoidal
pressure fluctuation is applied at the outflow boundary of the inlet duct, with the
implementation of the non-reflective boundary conditions proposed by Watson and
Myers (1991). For the symmetry boundary, the normal velocity and the normal gradients
of the axial velocity, pressure, and temperature are set to zero. A non-slip boundary
condition is applied along the wall, together with zero normal gradients of pressure and
temperature. Finally, the flow variables at the far-field boundary are extrapolated from
the interior along the characteristic lines based on the solution of a simple wave (Roache,
1982) to avoid shock reflections.
In order to assess the accuracy of the aforementioned numerical method, a wide
variety of flow problems were tested by Oh (1994). A calculation was carried out for a
supersonic flow over a cone with a semi-vertex cone angle of 20 degrees. The free
stream Mach number and pressure is 2.1 and 0.29 atm, respectively. Both Mach number
and pressure distributions agree very well with the exact numerical solution of the
ordinary differential equation derived by Taylor and Maccoll (1929). The turbulence
model is tested for a turbulent boundary layer on a flat plate. The calculated velocity
profile at 1410 Re =
agrees very well with the DNS data (Spapart, 1988).

64
11/23/2003
2.4 Steady-State Flow Analysis
The steady-state flowfields are first studied to understand the overall flow
structures in a realistic supersonic inlet and to provide a baseline for studying the
response of the inlet flow to downstream disturbances.
The computational domain, as shown in Fig. 2.2, comprises the internal flow
region (inner domain) that contains most of the essential flow structure, and the external
flow region (outer domain), which becomes important when flow spillage over the cowl
lip occurs at subcritical operating conditions. Two sizes of grids are used in order to
obtain grid independent solution. The fine grid consists of 601101 points for the inner
domain and 20181 points for the outer domain, while the coarse grid consists of 40181
points for the inner domain and 16181 points for the outer domain. Both grids are
stretched toward the walls to provide sufficient resolution of turbulent boundary layer.
Results from the steady-state calculations using these two grids show that the relative
difference in terms of the terminal shock position is less than 2%, which demonstrates
that grid independence is satisfied. To save CPU time, the coarse grid will be used in the
later unsteady-flow calculations. A more strict grid-independent study may be conducted
based on the Richardson extrapolation (Roache, 1997).

inner domain
outer domain

Fig. 2.2 Computational domain for inlet study

65
11/23/2003
Figure 2.3 presents the Mach-number, pressure, and vorticity contours at two
different back pressures (p
b
= 2.1 and 2.2 atm), which are carefully chosen such that the
engine operates at a supercritical condition to provide a sufficient shock stability margin.
Under these conditions, the two leading conical shocks generated by the double-cone
centerbody compress the airflow externally, merge slightly above the cowl lip, and form
a strong oblique shock, which extends into the external-flow region. In addition, a shock
stemming from the cowl inner surface continues downstream, hitting and reflecting from
both cowl and centerbody walls, and finally leading to a terminal normal shock. The
flow in this region undergoes a compression-expansion cycle, being compressed by
reflecting shocks and expanded by expansion waves. The wavy Mach number and
pressure distributions along the middle line of the inlet duct for the case of p
b
= 2.1 atm,
as shown in Fig. 2.4, clearly depict this feature. The flow finally becomes subsonic after
passing through the normal shock located in the divergent section of the diffuser. During
this process, the inlet recovers a high percentage of the free stream total pressure by
decelerating the air flow through the shock train. The total pressure recovery coefficients
are 84% for the case of p
b
= 2.1 atm and 88% for the case of p
b
= 2.2 atm, and the Mach
number immediately in front of the terminal shock are 1.42 and 1.32, respectively. At the
same time, the flow direction, which is originally deflected away by the leading shocks,
is adjusted back to the axial direction.
Due to the viscous effect, boundary layers exist on both the cowl and centerbody
walls. The growth of the boundary layer and shock-wave/boundary-layer interactions
play important roles in dictating the inlet flow structure. The presence of the boundary
layer prohibits a discontinuous change in pressure across the shock near the wall since the
66
11/23/2003
flow in the inner part of the boundary layer is subsonic. The overall pressure rise across
the shock is partly transmitted upstream through the boundary layer, causing the
streamlines in that region to diverge (Green, 1970). Consequently, the boundary layer
thickens and may be separated from the wall if the pressure rise across the shock is
sufficiently large. Due to its interaction with the boundary layer, the terminal shock is no
longer a strictly normal shock. An oblique shock is formed due to the abrupt thickening
of the boundary layer and runs into the terminal shock. If the flow deflects away enough
after the leading oblique shock, a rear oblique shock may be formed from the intersection
of the leading oblique shock and the main shock to form the lambda-shock (Dyke, 1997).
On the other hand, the flow passing through the stronger normal shock will be
slower than that passing through the oblique shock systems. Thus, vortex sheets emanate
from the shock bifurcation points, convect downstream, and maybe interact with each
other. This phenomenon is clearly depicted in the vorticity contours in Fig. 2.3.
Furthermore, the vorticity displays more complicated structure due to the different
compression histories of the flow just upstream of the terminal shock. However, the
pressure distribution in the transverse direction behind the terminal shock is quite
uniform, in order to satisfy the condition of equal static pressure in that direction. This
implies a non-uniformity of total pressure in the transverse direction.
The abrupt thickening of the boundary layer downstream of the normal shock acts
like a convergent nozzle by reducing the effective duct area, accelerating the subsonic
flow after the shock with an associated pressure drop. The sudden decrease in pressure
(or sudden increase in Mach number) after the terminal shock in the core flow region, as
shown in Fig. 2.4, is attributed to this effect.
67
11/23/2003

tmp

tmp
p
b
=2.1 atm
Mach-number Contours
p
b
=2.2 atm

tmp

tmp
p
b
=2.1 atm
p
b
=2.2 atm
p
b
=2.1 atm
p
b
=2.2 atm

tmp

tmp
p
b
=2.1 atm
Pressure Contours
p
b
=2.2 atm

tmp

tmp
p
b
=2.1 atm
Pressure Contours
p
b
=2.2 atm
p
b
=2.1 atm
Pressure Contours
p
b
=2.2 atm

p
b
=2.1 atm
Vorticity Contours
p
b
=2.2 atm
p
b
=2.1 atm
Vorticity Contours
p
b
=2.2 atm

Fig. 2.3 Mach-number, pressure, and vorticity contours with different back pressures
under steady-state condition

68
11/23/2003
x, m
M
a
c
h
n
u
m
b
e
r
p
,
a
t
m
0.0 0.1 0.2 0.3
0.0
0.5
1.0
1.5
2.0
0.0
0.5
1.0
1.5
2.0

Fig. 2.4 Mach-number and pressure distributions along midline under steady-state
condition (p
b
= 2.1atm)
2.5 Shock/Acoustic-Wave Interactions
The steady-state flowfields with back pressure of 2.1 atm are used to initialize the
unsteady flow calculations, considering that the terminal shock in this case is further
downstream, which means it has a larger stability margin. The downstream disturbance
is simulated by imposing a sinusoidal pressure oscillation at the exit plane:
) sin( t Ap p
b
= (2.20)
where is the angular frequency, f 2 , with f the frequency. A wide range of A and f
are considered with A = 0.05 and f = 500 Hz as the baseline case.
Figure 2.5 shows the pressure distributions along both cowl and centerbody walls
and the midline over one cycle of oscillation for the baseline case. The corresponding
Mach-number contours near the terminal shock are also presented. Different time
69
11/23/2003
intervals are chosen in order to display more key structures of the unsteady flowfield.
The oscillation cycle begins when the terminal shock is at its steady-state position. As
the terminal shock progresses upstream, the Mach number before the shock decreases.
However, that does not necessarily mean that the terminal shock weakens. It is the
relative Mach number with respect to the terminal shock that determines the strength of
the shock. The numerical results indicate that the moving shock has a velocity as large as
40 m/s as it moves upstream for the baseline case. With this in consideration, the
terminal shock may strengthen first and weaken then. The terminal shock reverses its
direction after it reaches the farthest upstream position by the time of about = t .
Similarly, as the terminal shock moves downstream, it first strengthens but soon becomes
weaker, now that it has a larger downstream moving velocity and the relative Mach
number becomes smaller. As the velocity of the shock increases, the shock may finally
turns into a pressure pulse. On the other hand, a strong adverse pressure gradient wave
gradually forms farther downstream and eventually steepens into a secondary shock.
This secondary shock becomes stronger as it moves upstream and finally combines into
the primary shock (or pressure pulse) to form a stronger shock. The shape of the terminal
shock also changes as it moves upstream or downstream. The S in Fig. 2.5 stands for
the separation point on the centerbody wall. It can be seen that the terminal shock will
always induce boundary layer separation as long as it reaches certain strength. It is worth
mentioning that in the last plot of Fig. 2.5, two separation points can be found, the first
one is induced by the terminal shock, and the second one is induced by the pressure
gradient resulting from the acoustic wave and area changing.

70
11/23/2003
p
/
p
0
,
0.2
0.4
0.6
0.8
1.0
t = /2 (t = 11.67 ms)
0.728
p
/
p
0
,
0.2
0.4
0.6
0.8
1.0
t = (t = 12.17 ms)
0.722
p
/
p
0
,
0.2
0.4
0.6
0.8
1.0
t = 7/6 (t = 12.34 ms)
0.742
p
/
p
0
,
0.2
0.4
0.6
0.8
1.0
t = 3/2 (t = 12.67 ms)
0.791
p
/
p
0
,
0.2
0.4
0.6
0.8
1.0
top wall
midline
bottom wall
t = 0 ( t = 11.17 ms )
0.782
S
Ms = 1.43
S
Ms = 1.30
S
Ms = 1.30
Ms = 1.38
S
Ms = 1.46, 1.17
Pressure Mach Number ( 0 M 1.74, M= 0.042 )
p
/
p
0
,
2.0 3.0 4.0 5.0 6.0

0.2
0.4
0.6
0.8
1.0
t = 5/3 (t = 12.84 ms)
0.802
x/R
c
2.0 2.5 3.0 3.5 4.0
S
1
S
2
Ms = 1.46
x/R
c

Fig. 2.5 Pressure distributions along walls and midline of inlet and Mach-number
contours within one cycle of oscillation (A = 0.05, f = 500 Hz)

71
11/23/2003
Figure 2.6 shows the instantaneous terminal shock locations for several cases.
The shock location is determined by finding a position with Mach number equal to 1.
The terminal shock begins to respond after the time required for the disturbance to travel
upstream to it. It then may oscillate around its mean position and be stabilized after one
or two cycles, or be disgorged out of the inlet as for the case of A = 0.1 and f = 500 Hz. It
is evidently shown that the shock exhibits larger upstream and downstream excursions for
lower frequency and higher amplitude disturbance. It can also be seen that the farthest
upstream excursion of the shock moves gradually from upstream to downstream during
the initial response stage, which can be rationalized as follows. When the acoustic
pressure wave (left-running) hits the normal shock, part of the pressure wave is absorbed
by the shock and the remainder is reflected as a right-running wave. This reflected wave
could either strengthen the left-running wave or weaken it, depending on the phase angle
difference between the two. Thus, for the first few cycles in which the reflected wave
does not fully interact with the left-running wave in the entire subsonic diffuser section,
the shock responds primarily to the left-running wave at first, and then gradually
responds to the combined left- and right-running waves. In the present study, the
reflecting wave seems to weaken the left-running wave, causing the phenomena
aforementioned. The shock oscillation displays a nearly sinusoidal shape for the cases
with small disturbance, and the mean position of the terminal shock is very close to the
steady-state position. However, when the disturbance becomes larger, many nonlinear
phenomena emerge. The oscillation of the terminal shock is no longer sinusoidal. The
shock moves faster downstream than upstream. A significant characteristic is that a
secondary shock forms and the primary shock disappears sometime within the cycle. The
72
11/23/2003
discontinuous line sections in Fig. 2.6(b) are due to this phenomenon. In the extreme
case, say, for the case of A = 0.1 and f = 500 Hz, the terminal shock is disgorged out of
the inlet, and engine unstart (Seddon and Goldsmith, 1999) may happen.
In Fig. 2.7, the position ranges of the terminal shock are given. For the cases with
A = 0.05, the upmost point is the farthest downstream position of the secondary shock. It
is quite clear that the range decreases as the frequency increases. The amplitude of the
shock oscillation has been derived by Culick and Rogers (1983) under one dimensional,
linear, and inviscid assumption. In the present study, the flowfield displays multi-
dimensional effects as mentioned before, however, the motion of the terminal shock is
mainly determined by the pressure wave behind it, which is nearly one dimensional.
Figure 2.8 displays good agreements of the amplitudes of terminal shock oscillations with
the analytical results.
Air flow matching is an important issue for the inlet. The mass flow rate entering
the inlet does not change during the supercritical operating conditions, but oscillations of
mass flow rate do change at the exit of inlet due to shock motion. Figure 2.9 presents the
ratio of relative mass flow rate fluctuation to the relative amplitude of pressure
oscillation. It is observed that the oscillation of pressure causes larger oscillation of mass
flow rate. For an engine with constant fuel injection rate, this implies that the
equivalence ratio would fluctuate more severely, perhaps enhancing combustion
instability. The phase angle differences between the mass and pressure oscillations are
1.16, 1.06, 0.95, and 0.97 for the four cases given in Fig. 2.9, respectively.
Total pressure recovery is another important issue of inlet. A small loss of inlet
total pressure may cause considerable loss in engine thrust. Figure 2.10 presents the mass
73
11/23/2003
averaged total pressure at exit over one cycle of oscillation for several cases. It is found
that the averaged total pressures appear nearly sinusoidal oscillations. The main effect on
the averaged total pressures is from the amplitude of pressure oscillation, while the
frequency has a minor effect. The flow distributions at the exit plane of the inlet provide
the coupling between the inlet and combustor. Figure 2.11 shows the axial velocity and
total pressure profile at the exit plane over one cycle of oscillation for the baseline case.
The steady-state results are also plotted for comparison. It can be seen that the unsteady
profiles have similar shapes with the steady ones. However, the pressure oscillations
cause very large velocity oscillations, which can reach 60% near the centerbody wall.
This may exert substantial influence on the combustor environments. Figure 2.12
compares the time-averaged axial velocity and stagnation pressure profiles with those of
steady-state, indicating that they are very close.
74
11/23/2003

time, ms
x
s
/
R
c
0
0
2
2
4
4
6
6
8
8
10
10
2.7 2.7
2.8 2.8
2.9 2.9
3.0 3.0
3.1 3.1
f = 250 Hz
f = 500 Hz
f = 1000 Hz

(a) A = 0.01

time, ms
x
s
/
R
c
0
0
2
2
4
4
6
6
8
8
10
10
2.0 2.0
2.5 2.5
3.0 3.0
3.5 3.5
4.0 4.0
A = 0.05, f = 500 Hz
A = 0.05, f = 1000 Hz
A = 0.10, f = 500 Hz

(b) A = 0.05, 0.10

Fig. 2.6 Instantaneous shock locations for different cases

75
11/23/2003

frequency, Hz
x
s
/
R
c
0
0
500
500
1000
1000
1500
1500
2000
2000
2500
2500
3000
3000
3500
3500
4000
4000
1.8 1.8
2.0 2.0
2.2 2.2
2.4 2.4
2.6 2.6
2.8 2.8
3.0 3.0
3.2 3.2
3.4 3.4
3.6 3.6
A=0.01, farthest upstream point
A=0.01, farthest downstream point
A=0.05, farthest upstream point
A=0.05, farthest downstream point
terminal shock
at steady state

Fig. 2.7 Shock position ranges with different frequencies

frequency, Hz
a
m
p
l
i
t
u
d
e
,
|
x
s
|
/
R
c
0
0
1000
1000
2000
2000
3000
3000
4000
4000
0.00 0.00
0.02 0.02
0.04 0.04
0.06 0.06
0.08 0.08
0.10 0.10
0.12 0.12
0.14 0.14
0.16 0.16
numerical result
analytical result

Fig. 2.8 Comparison of the amplitude of terminal shock oscillation with analytical result
(A = 0.01)

76
11/23/2003

t
(
m
e
'
/
m
e
)
/
A
-2.0 -2.0
-1.0 -1.0
0.0 0.0
1.0 1.0
2.0 2.0
A = 0.05, f = 250 Hz
A = 0.05, f = 500 Hz
A = 0.05, f = 1000 Hz
A = 0.10, f = 1000 Hz
.
.
|
0
1/2
2
3/2
0 1/2 3/2 2

Fig. 2.9 Mass response to the pressure oscillation at exit over one cycle

t
p
0
,
e
/
p
0
,
0.78 0.78
0.80 0.80
0.82 0.82
0.84 0.84
0.86 0.86
0.88 0.88
0.90 0.90
0.92 0.92
A = 0.05, f = 250 Hz
A = 0.05, f = 500 Hz
A = 0.05, f = 1000 Hz
A = 0.10, f = 1000 Hz
|
0 1/2 2 3/2

Fig. 2.10 Mass averaged stagnation pressure at exit over one cycle

77
11/23/2003

u
e
, m/s
y
/
R
c
0
0
50
50
100
100
150
150
0.0 0.0
0.2 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1.0 1.0
1.2 1.2
steady state
= 0/6 T
= 1/6 T
= 2/6 T
= 3/6 T
= 4/6 T
= 5/6 T
0 1
4
3 5 2

(a) Axial velocity

p
0,e
/p
0,
y
/
R
c
0.75
0.75
0.80
0.80
0.85
0.85
0.90
0.90
0.0 0.0
0.2 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1.0 1.0
1.2 1.2
steady state
= 0/6 T
= 1/6 T
= 2/6 T
= 3/6 T
= 4/6 T
= 5/6 T
0 1
4
3 5
2

(b) Stagnation pressure

Fig. 2.11 Axial velocity, stagnation pressure profiles at exit plane (A = 0.05, f = 500 Hz)

78
11/23/2003

u
e
, m/s
y
/
R
c
0 50 100 150
0.0
0.2
0.4
0.6
0.8
1.0
1.2
A = 0.05, f = 250
A = 0.05, f = 500
A = 0.05, f = 1000
A = 0.10, f = 1000
steady state
_

(a) Axial velocity

p
0,e
/p
0,
y
/
R
c
0.80 0.82 0.84 0.86 0.88
0.0
0.2
0.4
0.6
0.8
1.0
1.2
A = 0.05, f = 250
A = 0.05, f = 500
A = 0.05, f = 1000
A = 0.10, f = 1000
steady state
_

(b) Stagnation pressure

Fig. 2.12 Comparison of time averaged axial velocity and stagnation pressure profiles at
exit plane

79
11/23/2003
2.6 Conclusions
Interactions between the shock waves and pressure waves were numerically
investigated for a viscous flowfield in an axisymmetric, mixed-compression supersonic
inlet diffuser by solving the unsteady Navier-Stokes equations with a two-equation
turbulence model in cylindrical coordinates. A 500 Hz sinusoidal pressure disturbance
with amplitude of 5% of the mean static pressure was applied to the exit plane as a
baseline case. A number of notable features observed in this study are given below.
1) The acoustic response of the inlet flow increases with increasing amplitude of
the imposed oscillation, but decreases with frequency.
2) For a relatively large pressure oscillations (such as the case of A=5%), the
adverse pressure gradient farther downstream of the terminal shock can steepen to form a
secondary shock, while the primary shock may turn into a pressure pulse. Under an
extreme case, the terminal shock may be disgorged out of the inlet to cause engine
unstart.
3) Vorticities generated from the intersection points of the terminal shock and
oblique shocks change their position in both axial and transverse directions as the
terminal shock moves, resulting in a significant multi-dimensional effect.
4) Shock-induced flow-separation pockets on both walls form and disappear as
the terminal shock moves.
5) The pressure oscillation causes a larger mass flow rate oscillation and severe
axial velocity oscillation, which may exert strong influence on the combustor flowfield.
80
11/23/2003
Chapter 3
Quasi-One-Dimensional Analyses of PDEs
It has been pointed out that one-dimensional or quasi-one-dimensional
simulations are not enough for the PDE analysis since they cannot correctly account for
the effect of the exit boundary conditions. However, for a PDE with a convergent-
divergent (CD) nozzle, the exhaust flow may exhibit supersonic behavior during most of
the cycle period. Therefore, quasi-one-dimensional simulations are still acceptable and
are conducted in this chapter as a preliminary step. Two-dimensional analyses will be
considered in next chapters.
The recently developed space-time conservation element/solution element
(CE/SE) method is implemented because it circumvents the deficiencies of existing
numerical methods for treating detonation waves and shock discontinuities. Chemical
kinetics is modeled by a one-progress-variable scheme which is calibrated with NASA
CEA code. The diffusive transport is neglected in the current study because of its minor
role in determining detonation dynamics and system performance.
3.1 Space-Time CE/SE Method
3.1.1 Introduction to Space-Time CE/SE Method
The space-time conservation element and solution element (CE/SE) method was
originally developed by S.C. Chang at NASA Glenn Research Center and was first
published out in 1995 (Chang, 1995). It is a high-resolution and genuinely
81
11/23/2003
multidimensional numerical method for solving conservation laws. This method differs
substantially in both concept and methodology from other well-established methods, such
as finite-difference, finite-volume, finite-element, and spectral methods. It has many
nontraditional features, such as a unified treatment of space and time, enforcement of flux
conservation in the space-time domain, introduction of solution and conservation
elements to construct simple stencil, treatment of independent variables and their spatial
derivatives as unknowns to be solved simultaneously, and no interpolation or
extrapolation required to evaluate fluxes at cell interfaces.
The development of the CE/SE method was motivated by the fact that physical
conservation laws are actually a collection of statements of flux conservation in space-
time. Mathematically, these laws are represented by a set of integral equations, and the
differential form of these laws is obtained from the integral form by assuming that the
physical solution is smooth. For a physical solution in a region with rapid variations, this
smoothness assumption is difficult to be realized by a numerical approximation that uses
only a limited number of discrete variables. The situation becomes even worse in the
presence of large gradients or discontinuities. Thus, a method designed to obtain
numerical solutions to the differential form without enforcing flux conservation is a
fundamental disadvantage in modeling physical phenomena with steep gradients. In
contrast, a numerical solution obtained from a method that enforces flux conservation
both locally and globally will always retain the basic physical reality of flux conservation
even in a region involving discontinuities (Chang, 1995).
With this consideration, the CE/SE method was designed to treat space and time
as a single entity, and the integral form of the conservation laws are solved to ensure both
82
11/23/2003
local and global flux conservation in space and time. In comparison, finite-difference,
finite-element, and spectral methods deal with the differential form of the conservation
laws. The traditional finite-volume method attempts to enforce flux conservation in
space and/or time, but it treats space and time separately in the sense that flux
conservation cells are defined only in the space domain. In the finite-volume method, a
flux must be assigned at any interface separating two neighboring cells, and the flux is
typically evaluated by extrapolating or interpolating the mesh values at the neighboring
cells. This evaluation generally requires ad hoc techniques such as flux limiters or slope
limiters. These ad hoc techniques usually involve the calculation of the eigenvector
matrices of the Jacobian matrices, which may complicate the coding and increase the
computational time.
The CE/SE method introduces two types of cells in the space-time domain:
conservation elements (CEs) and solution elements (SEs). The union of the CEs covers
the whole computational domain. Physical conservation laws are enforced on each CE.
The SEs are designed to facilitate the evaluation of the fluxes at interfaces separating
neighboring CEs. Several unique features are adopted for this purpose. (1) Each SE is
chosen to cover some interfaces of neighboring CEs. (2) A solution point is defined
within each SE to hold both the basic independent variables and their spatial derivatives
as unknowns to be solved simultaneously, while the time derivatives are obtained by
satisfying the differential form of the governing equations at the solution point. (3) Each
physical flux vector is approximated in terms of some simple smooth functions within
each SE. With the above strategies, the CE/SE method is able to enforce both local and
global flux conservation in space and time with flux evaluation being an integral part of
83
11/23/2003
the solution procedure and requiring no interpolation or extrapolation. As a result, the
CE/SE method is more efficient than the traditional finite-volume method in terms of
coding complexity and computational burden.
A distinguished feature of the Space-Time CE/SE method is that it can fully
control numerical dissipation, even down to zero if necessary. This is important in
simulating flows involving small disturbances such as sound waves. Relatively large
numerical dissipation introduced by numerical schemes can lead to the annihilation of
small disturbances. Unexpected numerical dissipation may overwhelm physical
dissipation, causing a complete distortion of the solution. Thus, a solver that can fully
control its numerical dissipation is desirable. For conventional methods, numerical
dissipation is adjusted by varying the magnitude of added artificial dissipation terms.
With a few exceptions, numerical dissipation in almost every traditional scheme cannot
be reduced to zero even without artificial dissipation terms. A study of finite-difference
analogues of a simple convection equation shows that a numerical analogue is free of
numerical dissipation only if it does not violate certain space-time invariant properties of
the convection equation (Chang, 1992). In other words, numerical dissipation may be
considered as a result of symmetry-breaking by a numerical scheme. An ideal scheme
must be able to reserve this symmetry property in order to avoid numerical dissipation.
Because of the intrinsic nature of space-time unity, the Space-Time CE/SE method is
perfectly suited to constructing such schemes. Its ability to accurately simulate both
small disturbances and strong shock waves has been proven in treating various problems
involving interactions between shock waves and sound waves (Loh et al., 1996).
84
11/23/2003
3.1.2 Space-Time CE/SE Method for One-Dimensional Problems
This subsection, adopted from Wu (2002) with minor modifications, describes the
solution procedure of the Space-Time CE/SE method for one-dimensional systems
governed by

Q E
H
t x

+ =

(3.1)
Since the Space-Time CE/SE method treats space and time as a single entity, the
computational domain for spatially one-dimensional problems becomes a two-
dimensional Euclidean space ) , (
2
t x E = . Define ( , ) E Q = h as the current vector in E
2
.
The Gaussian divergence theorem in E
2
leads to the integral form, which is a more
physically basic form:

( ) s V V
d HdV =

h s
v
(3.2)
The second step is to construct the CEs and SEs. The computational domain E
2
is
discretized equally in space and time, and the dimension of each cell is assumed to be x
by t. Following the suggestion of Yu and Chang (1997), the conservation element
CE(j, n) and the solution element SE(j, n) are both defined to cover the rectangular area
ABCDEFA with a line segment AG sticking out on top of the rectangle, as shown in
Fig. 3.1. Each CE(j, n) can be divided into two sub-elements: CE
_
(j, n) and CE
+
(j, n).
The boundaries of CE
_
(j, n) are AB , BC , CD, and DA, while CE
+
(j, n) is enclosed by
FA, AD, DE , and EF . Note that the line segments BC and CD belong to the solution
element SE(j-1/2, n-1/2) and the line segments DE and EF to SE(j+1/2, n-1/2).
85
11/23/2003
j-1 j-1/2 j j+1/2 j+1
n-1
n-1/2
n
n+1/2
t
x
A B
C
D
E
F
G

Fig. 3.1 Mesh system and CEs/SEs in Space-Time CE/SE method (adopted from Wu,
2002).

For any point (x, t) SE(j, n), the dependent variable vector Q(x, t), flux vector
E(x, t), and source term vector S(x, t) are approximated by Q
*
(x, t; j, n), E
*
(x, t; j, n), and
H
*
(x, t; j, n), respectively, whose definitions are given below.
) ( ) ( ) ( ) ( ) , ; , (
* n n
j t j
n
j x
n
j
t t Q x x Q Q n j t x Q + + = (3.3)
) ( ) ( ) ( ) ( ) , ; , (
* n n
j t j
n
j x
n
j
t t E x x E E n j t x E + + = (3.4)

*
( , ; , ) ( ) ( ) ( ) ( )
n n n n
j x j j t j
H x t j n H H x x H t t = + + (3.5)
where the label
n
j
) ( denotes the corresponding quantities evaluated at mesh point (j, n)
whose coordinate is (x
j
, t
n
). Q
*
(x, t; j, n), E
*
(x, t; j, n), and H
*
(x, t; j, n) are actually first-
order Taylor's expansions of Q(x, t), E(x, t), and H(x, t) at point (j, n), respectively.
Moreover, all terms on the right-hand sides in the above approximations are only
functions of
n
j
Q ,
n
j x
Q ) ( , and
n
j t
Q ) ( because 1) both
n
j
E and
n
j
H are only functions of
n
j
Q , and 2)
n
j x
E ) ( , ( )
n
x j
H
n
j t
E ) ( , and ( )
n
t j
H are only functions of
n
j
Q ,
n
j x
Q ) ( , and
n
j t
Q ) (
due to the following relations:
CE
+
(j, n)
CE
(j, n)
86
11/23/2003
( ) ( ) , ( ) ( )
n n n n n n
x j j x j x j j x j
E A Q H D Q = = (3.6)
( ) ( ) , ( ) ( )
n n n n n n
t j j t j t j j t j
E A Q H D Q = = (3.7)
where
n
j
n
j
Q
E
A ) (
and ( )
n n
j j
H
D
Q
are the Jacobian matrices of the convective flux

vector E and source term vector H at (x
j
, t
n
) , respectively.
The Space-Time CE/SE method further requires the above solution
approximations to satisfy the following differential form of the governing equations
within SE(j, n) except the line segment AG that sticks out on the top of the solution
element.

* *
*
Q E
H
t x

+ =

(3.8)
Substitution of equations (3.3)-(3.5) into the above equation leads to
( ) ( )
n n n
t j x j j
Q E H + = (3.9)
From which
n
j t
Q ) ( can be expressed as function of
n
j
Q and
n
j x
Q ) ( .
Define
* * *
( , ; , ) ( ( , ; , ), ( , ; , )) x t j n E x t j n Q x t j n h as the approximation of the
current density vector ( , ) E Q = h within the conservation element CE(j, n). Then the
fluxes at AB and FA can be evaluated in terms of
n
j
Q and
n
j x
Q ) ( . Similarly, the fluxes
at other boundaries of CE(j, n) (i.e., BC , CD, DE , and EF ) can be obtained by
applying equations (3.3)-(3.9) to the solution elements SE(j1/2, n-1/2).
With the fluxes at all the boundaries of CE(j, n) being evaluated, the integral form
of the conservation laws, equation (3.2), can be readily enforced on CE(j, n).
87
11/23/2003

1/ 2 1/ 2 1/ 2 1/ 2
1/ 2 1/ 2 1/ 2 1/ 2
( , )
1 1
= + +
2
n n n n n
j j j j j
CE j n
Q HdV Q Q R R RHS
x

+ +
(

(3.10)
where R is defined as

n
j x
n
j
n
j
n
j x
n
j
E A
x
t
E
x
t
Q
x
R ) (
4
) (
+ ) (
4
2
(3.11)
After
n
j
Q are solved from equation (3.10),
n
j x
Q ) ( can be evaluated using the
following oscillation-suppressing procedure. First, the solutions at point (j1/2, n) are
approximated using the first-order Taylor series expansions as shown below.

2 /
'
) (
2 / 1
x
Q Q
Q
n
j
n
j n
j x
(3.12)

2 / 1
2 / 1
2 / 1
2 / 1 2 / 1
) (
2
'

+ =
n
j t
n
j
n
j
Q
t
Q Q (3.13)
where
n
j x
Q ) (
+
+
and
n
j x
Q ) (
+
are two numerical analogues of

n
j x
Q ) ( at point (j, n) evaluated
from the right and left sides, respectively. Define function W as

+
+ +
+
+
+
=
x x
x x x x
x x W ) ; , ( (3.14)
Then
n
j x
Q ) ( can be calculated using
) ; ) ( , ) (( ) (
n
j x
n
j x
n
j x
Q Q W Q
+
+
+
= (3.15)
Usually, is chosen as an integer of 2.
Before closing this section, there are two points to be emphasized. First, even
though numerical dissipation has been introduced into the scheme during the above
oscillation-suppressing procedure, conservation laws are strictly enforced on CE(j, n).
Moreover, since the union of all conservation elements covers the entire computational
88
11/23/2003
domain, global conservation is also enforced strictly. Second, although the order of the
accuracy of the solution approximations, equations (3.3)-(3.5), is second order, the actual
accuracy of the above scheme is comparable to that of 4-6th order compact difference
schemes when nonlinear Euler problems are modeled. This is because second order is its
nominal order of accuracy, and is only meaningful for linearized equations. For non-
linear problems, the flux conservation property ensures that the space-time method
possesses an order of accuracy much higher than its nominal one. In addition, the space-
time method has surprisingly small dispersive errors in nonlinear calculations, which also
contributes to the enhancement of its accuracy (Wang, et al., 1994).
The conservation equations of mass, momentum, energy, and species
concentration in a quasi-one-dimensional system can be written in the following vector
form.

t x

+ =

Q E
H (3.16)
The conserved variable vector Q, convective flux vector E, and source vector H are
defined as

t
u
e
Z
(
(
(
=
(
(

Q ,
2
( )
t
u
u p
u e p
uZ
(
(
+
(
=
+ (
(

E ,
2
0
0
1
( ) 0
t
u
u
dA
u e p A dx
uZ

( (
( (
( (
= +
+ ( (
( (

H
(3.17)
89
11/23/2003
In the above equations, , u ,
t
e , Z and A represent the density, velocity, specific total
energy, mass fraction of reactant, and area of cross section, respectively. The pressure p
is obtained through the equation of state for a perfect gas as,
] 2 / [ ) 1 (
2
Zq u e p
t
= (3.18)
with the specific heat ratio, q the heat release per unit mass of reactant. The mass
production rate is obtained using the Arrhenius expression,
) / exp( RT E Z K
a
= (3.19)
with K the pre-exponential reaction rate factor, Ea the activation energy per unit mass of
reactant, R the gas constant, and T the temperature which can be obtained using perfect
gas state equation:
R p T / = (3.20)
It is noted that there are five parameters ( , R, q, Ea and K) associated with the above
governing equations. Determination of these parameters is discussed in the following
section. The Jacobian matrices of the flux and source vectors are given in Appendix B.
3.3 Parameter Calibration and Code Validation
3.3.1 Parameter Calibration
Among the five parameters , R, q, Ea, and K, the first three are thermodynamic
parameters that determine the detonation wave properties such as the CJ state and wave
speed. These three parameters are calibrated with the NASA CEA code (McBride and
Gordon, 1996) for a stoichiometric H
2
/air mixture, as described below.
90
11/23/2003
Two sets of CJ detonation solutions, one from the CEA code and the other from
analytical formulation given in Sec. 1.2.1, are first obtained for a series of different
unburned pressures and temperatures: p
1
= 1, 2 atm, T
1
= 300, 350, 400, 450, 500 K. The
simulated annealing (SA) optimization algorithm (Belegundu and Chandrupatla, 1999) is
then applied to find the optimum values of , R, and q by minimizing the relative
difference in terms of T
CJ
/T
1
, p
CJ
/p
1
and u
D
between these two sets of solutions. The
following values are obtained.
= 1.290, R = 368.9 J/(kgK), q = 2.72010
6
J/kg (3.21)
The analytical CJ detonation velocity and CJ state based on these parameters are
compared with those from CEA code and shown in Fig. 3.2. It is observed that both the
CJ temperature and pressure agree with the CEA results very well, with a maximum
relative error less than 8%, and have the same trends as T
1
increases. The detonation
velocity differs in the trend. For example, the analytical CJ detonation velocity increases
with T
1
, while the detonation velocity from CEA code decreases with T
1
. Although the
trend is different, the relative errors for the detonation velocity are still small, e.g., less
than 3%. These small relative errors indicate that the current one-progress-variable
model represents a good approximation in predicting the detonation wave properties.
The other two parameters, Ea and K, are chemical kinetic parameters. Their
primary influence appears in the internal structure of a detonation wave front, and the
effect on the overall flow evolution and propulsion performance are relatively minor.
The following values, adopted from Mohanraj and Merkle (2000), are used for these two
parameters.
Ea = 4.79410
6
J/kg, K = 7.510
9
s
1
(3.22)
91
11/23/2003
T
1
(K)
T
C
J
/
T
1
,
p
C
J
/
p
1
D
C
J
(
m
/
s
)
300 350 400 450 500
300 350 400 450 500
4
6
8
10
12
14
16
18
20
1920
1940
1960
1980
2000
2020
ZND
T
CJ
/ T
1
p
CJ
/ p
1
D
CJ
D
CJ
CEA
p
1
= 1 atm

(a) p
1
= 1 atm

T
1
(K)
T
C
J
/
T
1
,
p
C
J
/
p
1
D
C
J
(
m
/
s
)
300 350 400 450 500
300 350 400 450 500
4
6
8
10
12
14
16
18
20
1920
1940
1960
1980
2000
2020
ZND
T
CJ
/ T
1
p
CJ
/ p
1
D
CJ
D
CJ
CEA
p
1
= 2 atm

(b) p
1
= 2 atm

Fig. 3.2 Comparison of detonation velocity and CJ state between ZND model and CEA

92
11/23/2003
3.3.2 ZND Detonation in a One-Dimensional Tube
The code is first validated for a ZND detonation in a one-dimensional tube as
discussed in Sec. 1.2.3. Figure 3.3 shows the schematic of the physical problem. The 20
cm-long tube is closed at the head end and open at the other. Initially, it is filled with
static, premixed stoichiometric hydrogen/air at temperature T
1
= 300 K and pressure p
1
=
1 atm. A small energetic spark region spanning 0.02 cm with temperature T
i
= 2000 K
and pressure p
i
= 30 atm is placed at the head end to initiate detonation.

stoichiometric hydrogen/air at
p
1
= 1 atm and T
1
= 300 K
spark region (p
i
= 30 atm and T
i
= 2000 K)
stoichiometric hydrogen/air at
p
1
= 1 atm and T
1
= 300 K
spark region (p
i
= 30 atm and T
i
= 2000 K)

Fig. 3.3 Schematic of detonation tube with spark region at closed end

A series of calculations are carried out with five different grid resolutions to
conduct grid-dependence study as well as validation. The grid points of these grids are
1000, 2000, 4000, 8000, 16000, respectively. Figures 3.4~3.7 show the time evolutions
of pressure, temperature, mass fraction and velocity, respectively, for the case of grid
points of 2000. These figures clearly depict the propagation of a self-sustained and fully-
developed detonation wave. The leading shock, the rarefaction wave region, and the
uniform region can also be observed. The small non-uniform region of temperature near
the head end is due to the remaining effect of the initial spark.
Figure 3.8 shows the snapshots of pressure profiles at t = 80 s with different grid
resolutions. These profiles collapse into a single one, demonstrating nearly perfect
93
11/23/2003
satisfaction of grid dependence. The only exception is that the pressure immediately after
the shock (spike pressure) increases with grid resolution. This is due to the fact that the
higher the resoltuion, the less the degree of reaction across the shock front, thus leading
to a higher spike pressure. Theoretically, the von Neummann spike pressure can be
reached by using tremendously high resolution grid. However, significant round-errors
appear with high reslolution grid, leaving the spike pressure unreachable in practice. In a
worst case, too fine grid may cause unphysical soltutions. It is found that the detonation
is not ignited with the finest grid (grid points = 16000). Similar phenomenon was also
reported by Ebrahimi et al. (1999).
Table 3.1 compares the numerical solutions of different grid resolutions in terms
of detonation velocity, von Neummann spike pressure, CJ state, and head-end pressure.
The analytical solutions, obtained from the formulation given in Sections 1.2.1~1.2.3, are
also listed for comparison. All these values are in perfect agreement with each other
except for the von Neumann spike pressure.

94
11/23/2003
x, m
p
,
a
t
m
0
0
0.05
0.05
0.1
0.1
0.15
0.15
0.2
0.2
0 0
5 5
10 10
15 15
20 20
25 25
t = 20 40 60 80 s

Fig. 3.4 Time evolution of pressure field for ZND problem (grid points = 2000)

x, m
T
,
K
0
0
0.05
0.05
0.1
0.1
0.15
0.15
0.2
0.2
0 0
500 500
1000 1000
1500 1500
2000 2000
2500 2500
3000 3000
t = 20 40 60 80 s

Fig. 3.5 Time evolution of temperature field for ZND problem (grid points = 2000)
95
11/23/2003

x, m
Z
0
0
0.05
0.05
0.1
0.1
0.15
0.15
0.2
0.2
0 0
0.2 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1 1
1.2 1.2
t = 20 40 60 80 s

Fig. 3.6 Time evolution of reactant mass ratio field for ZND problem (grid points =
2000)

x, m
u
,
m
/
s
0
0
0.05
0.05
0.1
0.1
0.15
0.15
0.2
0.2
0 0
200 200
400 400
600 600
800 800
1000 1000
1200 1200
t = 20 40 60 80 s

Fig. 3.7 Time evolution of velocity field for ZND problem (grid points = 2000)

96
11/23/2003

Fig. 3.8 Snapshots of pressure field for ZND problem with different grid points

Table 3.1 Comparison of numerical results of different grid resolutions with analytical
solutions for the ZND problem

grid points u
D
, m/s p
s
, atm p
CJ
, atm T
CJ
, K u
CJ
, m/s p
3
, atm
analytical 1973 30.60 15.80 2746 830.0 5.871
1000 1975 18.80.1 15.80 2746 829.9 5.872
2000 1974 22.10.4 15.78 2745 830.0 5.872
4000 1974 25.60.8 15.79 2745 830.0 5.872
8000 1973 29.21.0 15.79 2745 830.0 5.872
16000 detonation can not be initiated

97
11/23/2003
3.3.3 Flow Through a Convergent-Divergent Nozzle
A calculation is carried out for a non-reacting flow through a nozzle with a length
of 1 m and a cross-section area of A(x) = 1 + (2x1)
2
, as shown in Fig 3.9. The
stagnation pressure (P
0
), stagnation temperature (T
0
) and back pressure (p
b
) are 2 atm,
400 K, and 1.5 atm, respectively. Specific heat ratio is taken as 1.4. The analytical
solution can be easily obtained from the classical gasdynamics. Figure 3.10 compares the
pressure distribution with the analytical solution and perfect agreement is observed.
P
0
= 2 atm
T
0
= 400 K
p
b
= 1.5 atm
A(x) = 1 + ( 2x -1 )
2
= 1.4

Fig. 3.9 Schematic of convergent-divergent nozzle in the validation case

x, m
p
,
a
t
m
0
0
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
1
1
0.0 0.0
0.5 0.5
1.0 1.0
1.5 1.5
2.0 2.0
numerical
analytical

Fig. 3.10 Comparison of pressure distributuion with analytical results

98
11/23/2003
3.4 Flow Dynamics
The validated quasi-one-dimensional code is then used to study the flow
dynamics and system performance of an airbreathing PDE with a convergent-divergent
(CD) nozzle. The reason for choosing a CD nozzle is that other types of nozzles such as
convergent, divergent and plug nozzles have been studied by many researchers and
limited performance improvements on the performance were found. On the other hand,
the CD nozzle results in a supersonic exit flow during most of the cycle period, thus
reducing the error induced by the exit boundary condition.
The tube has a length of 60 cm and a diameter of 16 cm, with a valve located at
the head end as shown in Fig. 3.11. The valve is supposed to be either fully closed or
fully open. The nozzle has a length of 20 cm, with a 45 convergent angle and a 15
divergent angle. The diameters of the nozzle throat and exit are 12 cm and 20 cm,
respectively.
60 cm
16 cm 20 cm 12 cm
20 cm
valve
15 45

Fig. 3.11 Configuration of the single tube and CD nozzle in the current study

As a specific example, the flight condition involving an altitude of 9.3 km and a
Mach number of 2.1 is considered. The free stream static pressure and temperature are
0.29 atm and 228 K, respectively, corresponding to a total pressure of 2.65 atm and a
total temperature of 428 K. The supersonic inlet at this flight condition has been studied
99
11/23/2003
in Chapter 2. Based on the inlet analyses, the exit flow of the inlet has a total temperature
of 428 K and a total pressure of 2.23 atm. By assuming 5% loss of the total pressure
within the manifold, the total pressure at the entrance of the combustor becomes 2.12
atm, where the total temperature is still 428K.
The cyclic operation of the PDE is controlled by a valve located at the entrance of
the combustor. Generally, there are two modes of valve operation (Mohanraj and
Merkle, 2000). One is external mode, i.e., the timing of the valve opening and closing is
specified externally. The other is internal mode. In this mode, the opening and closing
procedures are controlled by the flow condition inside the detonation tube, e.g., using a
pre-specified threshold pressure for valve open and a chemical sensor in the detonation
tube for valve close. In the present study, the first mode is chosen and the valve response
time is neglected for simplicity, i.e., the valve is either fully closed or fully open.
Furthermore, the valve open area is equal to that of the detonation tube. The engine
operation sequence, as shown schematically in Fig. 3.12, is thus controlled by three time
periods: the valve close-up period (
close
) during which the valve is closed and the tube
undergoes detonation initiation and propagation and blowdown processes, the purging
period (
purge
) during which a small amount of cold air is injected into the tube to prevent
preignition of fresh reactants, and the refilling period (
refill
) during which the
combustible mixture is delivered to the tube. The sum of these three periods equals the
operation cycle time (
cycle
), that is,
cycle
=
close
+
purge
+
refill
.

100
11/23/2003
detonation initiation and
propagation; blow-down
purge
refill
refill
purge
close
detonation initiation and
propagation; blow-down
purge
refill
refill
purge
close

Fig. 3.12 Time periods during one cycle of operation

The boundary conditions at the head end of the detonation tube are specified
according to the engine operation. During the valve close-up stage, the head end is
modeled as a rigid wall. During the purging stage, the total temperature and total
pressure are specified as 428 K and 2.12 atm, respectively, as mentioned in Section II.
The axial velocity is obtained with extrapolation, and the reactant mass fraction is set to
zero. During the refilling stage, the same conditions are used except that the reactant
mass fraction is set to unity. At the exit of the nozzle, the boundary conditions are
determined from the flow condition at exit. If the flow is supersonic, all flow variables
are extrapolated from inside. If the flow is subsonic, the exit pressure is specified as the
ambient pressure and other flow variables are extrapolated.
A series of calculations are conducted for various operation parameters. A
particular case with
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms is first studied in
detail to investigate the flow dynamics of the airbreathing PDE. The detonation tube is
initially filled with a stoichiometric H
2
/air mixture at ambient pressure and temperature.
It takes about 6 cycles to reach a stable operation. Figure 3.13 presents the temporal
evolution of the pressure field within one cycle of operation. Detonation is initiated
immediately after the valve is closed. The detonation wave moves downstream and
catches the leading fresh reactant at about x = 36 cm. It then degenerates to a non-
101
11/23/2003
reactive shock wave. Expansion waves from the interface reduce the pressures on both
sides. When the shock wave reaches the convergent section of the nozzle, a reflected
shock wave is formed and travels upstream. This reflected shock is reflected again when
it reaches the head end of the combustor. On the other hand, the average chamber
pressure decreases as the mass flows out of the combustor and nozzle. As the purging
stage begins, a shock wave is generated due to the pressure difference across the valve. It
is observed that the nozzle throat remains choked during most of the cycle period.
A more clear and detailed wave pattern is numerically obtained through the
Eulerian approach as detailed in Appendix C. Figure 3.14 shows the x-t diagram during
the first cycle. The time history of the flow properties at head end is also presented. The
detonation wave is ignited in a very short time and propagates at a uniform CJ velocity of
1935 m/s in the unburned region (region 1), followed by the Taylor expansion waves
(region 2) to satisfy the stationary condition at head end. A uniform region (region 3) is
then formed behind the tail of Taylor waves. At t = 0.310 ms, the detonation wave
arrives the reactant-product interface (point A), which is also the nozzle inlet, and transits
to a non-reactive shock wave continuing to the nozzle exit. At the same time, expansion
waves are generated from the interface. However, these expansion waves are soon
overridden by the later formed compression waves from the convergent section of the
nozzle. A shock wave is eventually formed due to the coalescence of the compression
waves. This shock wave arrives at the head end at t = 0.949 ms (point B) and reflects
again, causing an abrupt increase of the head-end pressure, which can be clearly observed
in the time history of head-end pressure. This leads to a significant difference in the
head-end pressure history from that of Wintenberger et al. (2003) which was based on an
102
11/23/2003
ideal modeling. Expansion waves are also emanated from the nozzle throat, result in a
non-simple wave region when interacting with the Taylor waves, form a simple wave
region (region 5) after passing through the tail of the Taylor waves, and cause a gradual
decrease of head-end pressure. Inside the nozzle, flow chokes shortly after the primary
shock passing through the throat. A secondary shock is then generated due to the
interaction of the supersonic flow later induced by flow expansion from the divergent
section and the subsonic flow immediately following the primary shock. The flow within
the divergent section becomes fully supersonic after this secondary shock moves out of
the nozzle. When the purging stage begins at t = 2.1 ms, the head-end pressure increases
suddenly and two discontinuities appear: a shock wave caused by the pressure difference
across the valve and a contact surface caused by the temperature difference between the
hot product and the cold air. It can be clearly seen that the left-traveling characteristics
deflect when passing through these two discontinuities. The refilling stage begins 0.1 ms
later, causing a third discontinuity, i.e., a contact surface between reactant and air. This
contact surface travels to x = 32 cm at the end of the first cycle, representing a partial
filling for the next cycle.
Under the effects from the previous cycle, the wave pattern becomes much more
complicated during later cycles. Figure 3.15 shows the x-t diagram and time history of
head-end properties for a stable cycle. A significant difference from the first cycle is the
absence of the pressure plateau due to the effects of the shock waves and expansion
waves from the previous cycle. However, many phenomena are still similar, such as the
propagation of detonation wave, Taylor expansion waves, choked throat, contact surfaces
induced by purging and refilling, and so on.
103
11/23/2003
p
,
a
t
m
0
5
10
15
20
= 0.15 ms
(propagation stage)
p
,
a
t
m
0
2
4
6
8
10
= 0.25 ms
p
,
a
t
m
0
1
2
3
4
5
= 0.75 ms
p
,
a
t
m
0
1
2
3
4
5
= 1.25 ms
p
,
a
t
m
0
1
2
= 2.15 ms (purging stage)
p
,
a
t
m
0
1
2
= 2.25 ms (refilling stage)
x, m
p
,
a
t
m
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
1
2
= 2.50 ms

Fig. 3.13 Temporal evolution of pressure field within one cycle of operation (
cycle
= 3
ms,
close
= 2.1 ms,
purge
= 0.1 ms)
104
11/23/2003

Fig. 3.14 x-t diagram for the first cycle and time histories of flow properties at head end
under typical PDE operation with CD nozzle (stoichiometric H
2
/air mixture,
cycle
= 3 ms,
close
= 2.1 ms,
purge
= 0.1 ms). 1 = uniform unburned region, 2 = Taylor expansion waves,
3 = uniform region, 4 = non-simple wave region, 5 = simple wave region.

105
11/23/2003

Fig. 3.15 x-t diagram for the eighth cycle and time histories of flow properties at head
end under typical PDE operation with CD nozzle (stoichiometric H
2
/air mixture,
cycle
= 3
ms,
close
= 2.1 ms,
purge
= 0.1 ms).

106
11/23/2003
3.5 Parametric Study
3.5.1 Effect of Valve Timing
A parametric study is conducted to study the timing effect on system performance
by varying
cycle
and
close
. The purge time
purge
is fixed at 0.1 ms. Figure 3.16 shows the
effect of
close
on the specific thrust F
sp
, defined as the cycle-averaged thrust per unit of air
mass flow rate, and the fuel-based specific impulse, I
sp
, at four different cycle frequencies
of 200, 250, 333, and 400 Hz. The corresponding cycle periods are 5, 4, 3, and 2.5 ms,
respectively.
The specific thrust increases as
close
decreases for all of the frequencies
considered herein. This can be explained as follows. For a given
cycle
and
purge,
a
smaller
close
translates to a shorter blow-down process. The resultant higher chamber
pressure during the refilling stage increases the loading density of fresh reactants. The
increased refilling period also enhances the amount of reactants delivered to the chamber.
Combined, these two factors result in a higher cycle-averaged chamber pressure and
consequently a higher specific thrust. It should be noted, however, that the lower bound
of
close
is subject to three practical constraints. The first is concerned with inlet over-
pressurization. The head-end pressure must not exceed the stagnation pressure of the
inlet air to allow for purging and refilling when the valve is open. The second is related
to chamber over-filling. The fresh reactants should not flow out of the nozzle to the
external region before being burned completely unless after-burning is considered. The
third constraint, although commonly satisfied in practical cases, is that
close
should be
sufficiently long to cover at least the time required for detonation initiation and
107
11/23/2003
propagation throughout the entire chamber. The upper bound of
close
(or the lower
bound of
refill
) is based on the requirement that an appropriate amount of fresh reactants
be delivered to the chamber to produce thrust.
The effect of
close
on the fuel-based specific impulse follows the same trend as
that of the air-based specific thrust, except for a small range of
close
close to its lower
bound. The specific impulse and specific thrust satisfy the following relation,

purge refill
(1 / )
sp
sp
F
I
f g
+
= (3.23)
As
close
decreases, the factor (1 +
purge/
refill
) decreases and may override the increase of
F
sp
, consequently leading to a decrease in I
sp
, as shown in Fig. 3.16b.
For a given cycle period,
close
determines the filling length of fresh reactants. A
larger
close
(or smaller
refill
) leads to a smaller filling length in most cases and
consequently decreases the specific impulse. This result, however, is in contrast to the
previous experimental (Cooper and Shepherd, 2002) and numerical (Li and Kailasanath,
2002) observations for single-pulse operations, which concluded that the specific impulse
increases as the filling length decreases. One factor contributing to this discrepancy is
that in single-pulse studies, the pressure and temperature of reactants are preconditioned
to ambient values, while in the present multicycle study, the flow conditions of the
refilled mixture depend on the timing of the engine operation. The use of a choked CD
nozzle also exerted a substantial influence on the chamber dynamics. Significant
differences thus exist between single-pulse and multicycle operations. The conclusions
from single-pulse studies may not be applied to multicycle cases directly.
108
11/23/2003
close
, ms
s
p
e
c
i
f
i
c
t
h
r
u
s
t
F
s
p
,
m
/
s
1.5 2 2.5 3 3.5 4 4.5
0
200
400
600
800
1000
1200
3 ms
5 ms 4 ms
inlet over-pressurization
combustor overfilled
cycle
= 2.5 ms

(a) specific thrust

close
, ms
I
s
p
,
s
1.5 2 2.5 3 3.5 4 4.5
0
1000
2000
3000
4000
3 ms
5 ms
4 ms
cycle
= 2.5 ms

(b) specific impulse

Fig. 3.16 Effect of valve close-up time on (a) specific thrust and (b) specific impulse at
four different operation frequencies; straight tube with CD nozzle; stoichiometric H
2
/air
mixture, h = 9.3 km, M
= 2.1.

109
11/23/2003
Figure 3.16 also demonstrates the existence of an optimum frequency for
achieving a maximum performance. At a low cycle frequency, more reactants can be
recharged into the detonation tube. As a consequence, a higher chamber pressure can be
reached and the engine efficiency improves. However, a large refilling time associated
with low-frequency operation may cause chamber over-filling and thus degrade the
performance. These two conflicting effects result in an optimum frequency. In the
present study, the operating frequency of 250 Hz (
cycle
= 4 ms) offers the best
performance. The highest specific impulse is 3676 s, slightly lower than its ramjet
counterpart of 3866 s with optimum nozzle flow expansion (Wu, Ma, and Yang, 2003).
3.5.2 Effect of Nozzle Throat Area
The effect of the nozzle throat area was also studied. Figure 3.17 compares the
performances between the aforementioned CD nozzle and another CD nozzle with a
smaller throat diameter of 10 cm. A smaller throat implies higher chamber pressure,
therefore, performance gains are observed for those
close
that are larger than their
corresponding optimums. A smaller throat also leads to an earlier inlet over-
pressurization. Apparently, the operating ranges for
close
are narrower with the smaller
throat under frequencies of 333 Hz and 250 Hz. For the 200 Hz case, however, the
operating range becomes wider, since the lower limit is determined by combustor
overfilling for the larger throat under this frequency while a smaller throat can postpone
the combustor overfilling. An even smaller throat with a diameter of 8 cm was also
tested. A too small throat is not beneficial because of the inlet over-pressurization.
110
11/23/2003
close
, ms
s
p
e
c
i
f
i
c
t
h
r
u
s
t
F
s
p
,
m
/
s
1.5 2 2.5 3 3.5 4 4.5
0
200
400
600
800
1000
1200
5 ms
4 ms
cycle
= 3 ms
larger throat
smaller throat

(a) specific thrust

larger throat
close
, ms
I
s
p
,
s
1.5 2 2.5 3 3.5 4 4.5
0
1000
2000
3000
4000
5 ms
4 ms
cycle
= 3 ms
smaller throat

(b) specific impulse

Fig. 3.17 Effects of
close
on (a) specific thrust and (b) specific impulse at three different
frequencies for two CD nozzles with different throat areas. The larger one has a diameter
of 12 cm and the smaller one 10 cm.

111
11/23/2003
Chapter 4
Two-Dimensional Analyses of Single-Tube PDEs
One-dimensional simulations are computationally efficient. Nevertheless, they
cannot provide detailed multidimensional flow dynamics and cannot be applied to a PDE
involving multiple detonation tubes. Multidimensional simulations with computation
domain including both the detonation tube and the external region are required in order to
faithfully describe the system dynamics, especially in the near field of the tube exit where
the flow is intrinsically multidimensional, and to provide more accurate performance
predictions. However, the difficulties of using very fine grid to resolve the significant
flow structure and running many time steps to reach steady periodic operation have
limited previous multidimensional simulations of PDEs. Two-dimensional multicycle
simulations, to the knowledge of the author, has not been reported elsewhere in the
literature. In the current study, an efficient parallel computing technology coupling to the
high fidelity Space-Time CE/SE method is implemented to overcome these difficulties.
The analysis is based on the two-dimensional conservation equations of mass,
momentum, and energy and takes into account finite-rate chemical kinetics. Diffusive
effects are neglected in the current study because of their minor roles in determining the
overall flow dynamics and propulsive performance of PDEs. If the chemical reaction rate
112
11/23/2003
is expressed with a single progress variable, the resultant governing equations can be
written in the following vector form:

t x y

+ + =

Q E F
H (4.1)
where the dependent variable vector, Q, convective flux vectors, E and F, and source
vector, H, are defined as:

t
u
v
e
Z
(
(
(
= (
(
(
(

Q ,
2
( )
t
u
u p
uv
u e p
uZ
(
(
+
(
= (
(
+
(
(

E ,
2
( )
t
v
uv
v p
v e p
vZ
(
(
(
= + (
(
+
(
(

F ,
0
0
0
0
(
(
(
= (
(
(
(

H
(4.2)
In the above equations, , u, v, e
t
, and Z represent the density, axial velocity, vertical
velocity, specific total energy, and progress variable (i.e., mass fraction of reactant),
respectively. The pressure p is obtained through the equation of state,
] 2 / ) ( [ ) 1 (
2 2
Zq v u e p
t
+ = (4.3)
where is the specific heat ratio and q the heat release per unit mass of reactant. For a
one-step, irreversible reaction, the mass production rate of reactant is
) / exp( RT E Z K
a
= (4.4)
where K is the pre-exponential factor, T the temperature, E
a
the activation energy per unit
mass of reactant, and R the gas constant. The five parameters involved in the above
equations remain the same as those in Chapter 3. The Jacobian matrices of the flux and
source vectors of Eq. (4.1) are given in Appendix B.
113
11/23/2003
4.2 Numerical Treatment and Parallel Implementation
A two-dimensional unstructured triangular mesh non-reacting Euler solver based
on the space-time CE/SE method has been developed by Wang and Chang (1999). The
ideas and solution procedure are identical to the one-dimensional version described in
Chapter 3. The details of the space-time CE/SE method for two-dimensional problems
are given in Appendix D. In the current study, a two-dimensional code is developed and
efficiently parallelized by implementing the message-passing-interface (MPI) library and
the domain decomposition technique. The parallel code is then executed on an in-house
Beowulf cluster.
Beowulf is a multi-computer architecture for parallel computations. A Beowulf
system is a cluster of PCs (or workstations) including server nodes and client nodes
connected through network such as switch (http://www.beowulf.org). This system has
only a history of several years. The first Beowulf Cluster was built by NASA in 1994.
The current in-house Beowulf cluster was built up in 1997 and has been extended to a
large parallel system consisting of 350 workstations and several high-speed switches.
Parallel virtual machine (PVM) and message passing interface (MPI) are software
systems for writing message-passing parallel programs that run on a cluster. PVM used
to be the de facto standard until MPI appeared. The MPI is standardized by the MPI
Forum and available on all massively parallel supercomputers.
To effectively use the Beowulf system, the task should be properly distributed
among the processors. A common approach is to decompose the computational domain
into sub-domains and assign each sub-domain to a different processor. The objective of
114
11/23/2003
domain decomposition is to balance the computational workload and memory occupancy
of processing nodes while keeping the inter-node communication as small as possible. In
the current work, a software package, METIS, which is a family of programs for
partitioning unstructured graphs and hypergraphs and computing fill-reducing orderings
of sparse matrices (Karypis and Kumar, 1998), is used for domain decomposition.
4.3 Model Validation
As part of the model validation effort, a series of single-pulse calculations were
conducted for a straight tube of 60 cm in length initially filled with a stoichiometric
mixture of hydrogen and air at preconditioned pressure p
1
and temperature T
1
. A spark
region spanning 0.2 mm near the head end with a temperature of 2,000 K and a pressure
of 30 atm was employed to directly initiate the detonation wave. Four different
numerical grids with the sizes of 0.2, 0.1, 0.05, and 0.025 mm were used to check the
solution accuracy in terms of grid independence. All of the calculated pressure profiles
collapsed onto a single curve, with the CJ properties matching the analytical values
exactly. As a result, the 0.2 mm grid was chosen for the entire study to alleviate the
computational burden. For a single-pulse operation, the head-end pressure remains at a
plateau value p
3
for certain period soon after the detonation initiation, and then decays
gradually to a level lower than the ambient state. The impulse on the thrust wall can be
determined by integrating temporally the force exerted on the head end from t = 0 to the
instant when the head-end pressure reaches the ambient value. The contribution to the
impulse from the ignition source is estimated to be less than 0.5%. Figure 4.1 shows the
115
11/23/2003
impulse per unit cross-sectional area as a function of the plateau pressure p
3
and the
detonation residence time
D
(defined as the tube length L divided by the detonation wave
velocity u
D
, i.e.,
D
L/u
D
). Results can be correlated well in the following form

3 1 D
/ 4.1( ) I A p p = (4.5)

(p
3
-p
1
)
D
, Pas
I
m
p
u
l
s
e
/
A
r
e
a
,
N
s
/
m
2
0 100 200 300 400
0
400
800
1200
1600
Kailasanath (2003)
present results

Fig. 4.1 Generalized impulse curve for single-pulse detonation in straight tube with
stoichiometric H
2
/air mixture.

This expression is quite similar to those obtained from the semi-analytical
analysis of Wintenberger et al. (2003) and the experimental work of Falempin et al.
(2001). The constants of proportionality differ slightly in the various studies, suggesting
dependence on the details of experimental procedure and operating conditions. The
generalized formula proposed by Kailasanath (2003) based on his numerical simulations
for hydrogen/air, ethylene/oxygen, and propane/oxygen mixtures has a larger constant of
proportionality of 4.65. One factor contributing to this disparity may be differences in
116
11/23/2003
ignition source. The width of the detonation initiation region used by Kailasanath (2003)
is 20 mm, as opposed to 0.2 mm in the present simulations. Nonetheless, the above
parametric study demonstrates the capacity and fidelity of the present approach for the
PDE performance analysis.
4.4 Calculation of Propulsive Performance
The propulsive performance of the PDE must be calculated appropriately. There
are several ways to experimentally measure the impulse, such as integrating the pressure
force on the thrust wall, using the ballistic pendulum, load cell, damped thrust stand, and
spring-damper system, as discussed in Sec. 1.4.1.1. In numerical simulations, the
impulse or thrust can be obtained either by integrating the pressure force on the thrust
wall or through a momentum balance fro the entire system. The latter is more practical
for PDEs including both inlet and nozzle and is detailed as below.
Considering a control volume that contains the fluid within the entire engine, the
momentum conservation gives:
[ ( ) ] 0
CV CS
dV p p dS
t
+ + =

u
uu n n
v
(4.6)
where CV represents the control volume and CS the control surface which can be further
divided into three parts: the entrance plane S
i
, the exit plane S
e
, and the remaining surface
S
w
. Defining
e
m ,
e
u , and
e
p as

e
e
S
m dS =
u n (4.7)
117
11/23/2003

1
e
e
e S
dS
m
=
u uu n
(4.8)

1
e
e
e S
p pdS
A
=

(4.9)
the instantaneous thrust, in a vector form, can thus be derived:
{ } [ ] [( ) ]
e e a e e
CV
m m u p p A dV
t

= +
u
F u i i (4.10)
The second term on the right-hand side is arisen from the unsteady effect. For a steady
periodic operation, the cycle average of this term becomes zero, and the cycle-averaged
thrust becomes

( )
e e a e e
m m u p p A

( = (

F u i i (4.11)
where the bracket denotes the cycle-averaged quantities. Because of the symmetric
property of the system, the vertical component of the cycle-averaged thrust is zero, and
the axial component is

( )
e e a e e
F m u m u p p A

( = + (

(4.12)
This formulation is identical to the conventional one obtained for steady engines, except
for the averaging symbol. The two terms on the right-hand side are referred to as
momentum thrust and pressure thrust, respectively. The air-based specific thrust and
fuel-based specific impulse are then calculated as

sp
a
F
F
m
=
(4.13)

sp
f
F
I
m g
=
(4.14)
118
11/23/2003
It should be pointed out that, since the valve timing, the total pressure, and the
total temperature in front of the valve are prespecified, the air mass match between the
combustor and the inlet is achieved either by adjusting the width of the combustor, which
is a free parameter in two-dimensional simulations, or by changing the size of the inlet.
Otherwise, the computational domain should be extended more upstream to cover a
choked section so that the mass flow rate instead of the total pressure is specified at the
entrance boundary.
4.5 Straight-Tube PDE
As a first approach, a single straight tube is considered to provide direct insight
into the chamber dynamics without complications arising from the nozzle.
4.5.1 Problem Setup
The computational domain is shown schematically in Fig. 4.2. The detonation
tube measures 60 cm in length and 16 cm in height, which are similar to the dimensions
of contemporary ramjet combustors for air defense applications. An external region is
included to circumvent the difficulty of specifying boundary conditions at the tube exit.
The computational domain is discretized into 421590 unstructured triangular grid cells,
with 320000 for the detonation tube and 101590 for the external region. This dense grid
resolves the detailed detonation propagation in the axial direction and was selected after
performing a grid independence analysis of the computed solution such as pressure
profile along the axial direction. The cell size near the head end is about 0.4 mm in the
119
11/23/2003
axial direction and 2 mm in the vertical direction. It increases to about 5 mm at the
external boundary.

Fig. 4.2 Computational domain for straight-tube PDE

4.5.1.2 Domain Decomposition
The computational domain is decomposed into 64 sub-domains using the METIS
software (Karypis and Kumar, 1998) for parallel computing. This domain decomposition
technique allows for the balance of the computational workload and memory occupancy
of processors, while keeping the inter-processor communications as low as possible, and
thus obtaining a high parallel efficiency. With METIS, each cell is marked with a
number to which sub-domain (or sub-grid) it belongs. A user code is then needed to
gather this information and write out sub-grid data into different grid files. Figure 4.3
schematically shows the computational domain that decomposed into 64 sub-domains.
The number of cells for each sub-domain is given in Fig. 4.4, demonstrating a quite
evenly decomposition of grid cells.
120
11/23/2003

Fig. 4.3 Schematic of the computational domain decomposed into 64 sub-domains

sub-domain
n
u
m
b
e
r
o
f
c
e
l
l
s
8 16 24 32 40 48 56 64
0
2000
4000
6000
8000
10000
1

Fig. 4.4 Number of cells of sub-domains

121
11/23/2003
4.5.1.3 Boundary Conditions
The boundary conditions at the head end of the detonation tube are specified
according to the engine operation. During the valve close-up stage, the head end is
modeled as a rigid wall. During the purging stage, the total temperature and total
pressure are specified as 428 K and 2.12 atm, respectively. The axial velocity is obtained
with extrapolation, and the reactant mass fraction is set to zero. During the refilling
stage, the same conditions are used except that the reactant mass fraction is set to unity.
At the open boundary of the external region, a non-reflecting boundary condition is
implemented (Wang and Chang, 1999). The slip wall boundary conditions are detailed in
Appendix E.
4.5.1.4 Initial Conditions
The detonation tube is initially filled up with a stoichiometric hydrogen/air
mixture at the ambient pressure (0.29 atm) and temperature (228 K), whereas the other
region is filled with the ambient air.
4.5.1.5 Detonation Initiation
Detonation initiation is assured by a small driver region near the head end that
extends across the entire tube cross section. The temperature and pressure of the driver
gas are 2,000K and 30 atm, respectively. Because of the high energy density associated
with the driver gas, its effect on the propulsive performance could be quite large.
Cambier and Tegner reported an effect of 17 to 27% on the single-pulse impulse for a
PDE system with a 10-cm-long detonation tube and a 5-cm-long divergent nozzle
122
11/23/2003
(Cambier and Tegner, 1998). In the present study, several lengths of the driver region
were tested and the smallest one of 0.2 mm was selected to minimize the effect of the
initiation source on the propulsive performance. The thermal energy per unit area of this
initiation source is approximated as

2
[ /( 1)] 0.27 /
p driv driv driv driv driv
c T L p L J cm = (4.15)
Its effect on the propulsive performance can be estimated by comparing the thermal
energy of the initiation source with the heat that could be released from the detonation of
the reactant within the tube:

[ /( 1)]
0.5%
[ /( )]
p driv driv driv
driv driv
c T L
p L
q L q RT pL

= (4.16)
The net effect appears to be limited.
4.5.2 Results
A series of analyses are conducted over a wide range of operation parameters.
The CFL number used is 0.5, whereas the corresponding time step is about 510
-5
ms.
The typical turn-around time for one cycle of calculation is about 10 hours on an in-house
PC cluster consisting of 64 Pentium II processors.
The baseline case has
cycle
of 3 ms and
close
of 2.4 ms. It takes about 5 cycles to
reach steady cyclic operation. Figure 4.5 shows the x-t diagram for the first cycle of
operation, obtained by tracing the characteristic lines of the flowfield along the centerline
of the tube. The time histories of the flow properties at the head end are also presented.
Procedures for building the x-t diagram are given in Appendix C.

123
11/23/2003

Fig. 4.5 x-t diagram for first cycle and time histories of flow properties at head end under
typical PDE operation (stoichiometric H
2
/air mixture,
cycle
= 3 ms,
close
= 2.4 ms,
purge
=
0.1 ms). 1 = unburned region, 2 = Taylor expansion waves, 3 = stationary region, 4 =
non-simple wave region, 5 = simple wave region.

The detonation wave is directly initiated by a hot driver gas and propagates
downstream at the CJ detonation velocity toward the unburned mixture (region 1). It
then induces the Taylor expansion waves (region 2) in order to satisfy the stationary
condition at the head end, causing a uniform region (region 3) with constant flow
properties in the upstream.
124
11/23/2003
The detonation wave reaches the reactant/air interface at the tube exit at t = 0.305
ms (point A), which deviates slightly from the following analytical prediction by 0.6%
due to the effect of the externally imposed ignition source.

D
0.6 m
0.307 ms
1956 m/s
D
L
u
= = = (4.17)
The wave then degenerates to a non-reactive shock (i.e., the primary shock wave)
proceeding further downstream into the external region, followed by a contact surface
separating the ambient air and combustion products. A sonic region is gradually formed
near the tube exit due to the local flow expansion, as evidenced by the clustered
characteristic lines in the x-t diagram. Downstream of the sonic region, the flow is
expanded to become supersonic and finally leads to the formation of a secondary shock to
match with the subsonic flow behind the primary shock. This secondary shock wave
moves further downstream, meeting with expansion waves originating from the primary
shock wave. These complicated flow structures can be also observed in Fig. 4.6, which
shows the instantaneous pressure and density and their gradient fields at t = 0.7 ms.
Many salient features are clearly examined, including the expansion fans, vortices, and
the rolled-up slip lines developed as the shock diffracts over the edge of the tube exit.
As the detonation wave catches the reactant/air interface and the resultant primary
shock wave travels outside the tube, a series of expansion waves are generated and
propagate upstream, resulting in a non-simple wave region (region 4) when interacting
with the incoming Taylor waves. A simple wave region (region 5) is recovered after
passing through the Taylor waves. The first expansion wave reaches the head end at t =
0.935 ms (point B), which can be determined by considering the interaction between the
125
11/23/2003
expansion and the Taylor waves and the sound speed in region 3. A similarity solution
has been derived by Wintenberger et al. (2003) and is detailed in Appendix F. The
analytical solution gives

3
t
c
L
u
L
D
+ = (4.18)
where is function of and M
D
and can be calculated as,

1
2
2( 1)
2 2
1 1 2( 1)
(1 ) 2 1
2 ( 1)( 1)
D
D D
M
M M

( +
= +
`
(
+ +

)
(4.19)
Application of Eqs. (4.18) and (4.19) gives rise to an analytical value of 0.958 ms. The
slight difference between the numerical and the analytical solutions may be attributed to
the numerical resolution and dissipation near the tube exit.
On the arrival of the first expansion wave at the head end, the pressure begins to
decay gradually. These expansion waves reflect off the head end and form another series
of expansion waves, further reducing the chamber pressure. The downstream-traveling
expansion waves weaken the secondary shock, and eventually cause it to move upstream.
The head-end pressure decays to 0.23 atm at t = 2.4 ms, at which point the
purging stage begins. The head-end temperature is 1258 K at this instant. Because of the
pressure difference across the entrance plane, a right-running shock wave is established,
along with a series of central expansion waves and a contact surface between the burned
gas and the cold air. Another contact surface forms between the fresh reactants and
purging air when the refilling stage commences 0.1 ms later. The corresponding refilling
pressure, velocity, and Mach number are about 0.91 atm, 500 m/s, and 1.2, respectively.
126
11/23/2003
The time evolution of the pressure distribution along the centerline during the first cycle
of operation is shown in Fig. 4.7.
The flow evolution during a steady operation cycle is examined. Figure 4.8
shows the x-t diagram and time histories of flow properties at the head end for the fifth
cycle. The main flow features remain qualitatively the same as those in the first cycle.
However, the secondary shock wave disappears, since the flow behind the primary shock
wave is already supersonic. In addition, the head-end pressure and temperature begin to
decay earlier relative to the first cycle, due to the rarefaction waves produced from the
previous cycle. It should also be noted that the detonation wave catches the leading fresh
reactant at x = 51.2 cm instead of at the tube exit.

Fig. 4.6 Snapshots of pressure, density and their gradients fields at t = 0.7 ms.

127
11/23/2003
p
,
a
t
m
0
2
4
6
t = 0.20 ms (propagation)
p
,
a
t
m
0
2
4
t = 0.40 ms (blow-down)
p
,
a
t
m
0
1
2
3
p
,
a
t
m
0
1
2
p
,
a
t
m
0
1
p
,
a
t
m
0
0.5
p
,
a
t
m
0
1
t = 2.45 ms (purging)
x, m
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
p
,
a
t
m
0
1
t = 2.80 ms (refilling)

Fig. 4.7 Time evolution of pressure distribution along centerline during first cycle of
operation (
cycle
= 3 ms,
close
= 2.4 ms,
purge
= 0.1 ms).

128
11/23/2003

Fig. 4.8 x-t diagram for fifth cycle and time histories of flow properties at head end
2
/air mixture,
cycle
= 3 ms,
close
= 2.4 ms,
purge
= 0.1 ms).

The impulse of each cycle is calculated by considering the momentum balance
over a control volume enclosing the entire engine. The cycle-averaged specific thrust
(air-based) and specific impulse (fuel-based) are then obtained by dividing the impulse by
the air mass and fuel weight for each cycle, respectively. For the baseline case, the fuel-
based specific impulse is 2328 s.
129
11/23/2003
For comparison, the performance parameters and flow conditions of a
corresponding ramjet engine with perfect nozzle flow expansion are given in Fig. 4.9,
which has a specific impulse of about 3866 s. Thus, with the straight-tube configuration,
the PDE performs even worse than the ramjet engine.
combustor conditions:
M
3
= 0.1, M
4
= 0.26,
T
0,4
= 2472 K, p
0,4
= 2.05 atm
T
4
= 2458 K, p
4
= 1.97 atm
inlet exit conditions:
M
2
= 0.29, = 1.01 kg/s
T
0,2
= 428 K, p
0,2
= 2.23 atm
T
2
= 422 K, p
2
= 2.10 atm
freestream conditions:
M
= 2.1, h = 9.3 km
T
0,
= 428 K, p
0,
= 2.65 atm
T
= 228 K, p
= 0.29 atm
2
0
6
engine performance:
F = 1105 N (248 lbf)
m
= 1.0(H
2
)
I
sp
= 3866 s
combustor conditions:
M
3
= 0.1, M
4
= 0.26,
T
0,4
= 2472 K, p
0,4
= 2.05 atm
T
4
= 2458 K, p
4
= 1.97 atm
inlet exit conditions:
M
2
= 0.29, = 1.01 kg/s
T
0,2
= 428 K, p
0,2
= 2.23 atm
T
2
= 422 K, p
2
= 2.10 atm
freestream conditions:
M
= 2.1, h = 9.3 km
T
0,
= 428 K, p
0,
= 2.65 atm
T
= 228 K, p
= 0.29 atm
2
0
6
engine performance:
F = 1105 N (248 lbf)
m
= 1.0(H
2
)
I
sp
= 3866 s

Fig. 4.9 Performance parameters of ramjet engine for stiochiometric H
2
/air mixture

A parametric study was carried out to examine the effect of various operating
times on the system performance. Figure 4.10 shows the result as a function of
close
.
The highest specific impulse obtained is 2906 s. This is far lower than its theoretical
limit of 5263 s (Wu, Ma, and Yang, 2003), which assumes isentropic flow processes in
the inlet and nozzle. Although the calculated specific impulse can be improved by
optimizing the operation frequency and timing, the net gain appears to be limited with the
current design. Several fundamental mechanisms responsible for such an unacceptable
performance have been identified. First, at high altitudes, the straight-tube design fails to
preserve the chamber pressure during the refilling stage at a level sufficient to meet the
130
11/23/2003
requirements for the mass loading density of fresh reactants. Second, the low chamber
pressure in the refilling stages causes a high-speed reactant stream in the tube, and
subsequently results in a large performance loss. It is well established that the stagnation
pressure drop due to energy addition is proportional to the square of the Mach number.
In the present case, the local Mach number may reach a value of up to 1.2 during the
refilling process. The ensuing loss of thermodynamic efficiency becomes exceedingly
large compared with conventional propulsion systems with subsonic combustions. Third,
the lack of an appropriate flow expansion device downstream of the detonation tube gives
rise to an extremely complicated flow structure near the tube exit. The internal energy of
the exhaust flow cannot be effectively converted to the kinetic energy for thrust
generation, further deteriorating the situation.

close
, ms
I
s
p
,
s
1.5 2 2.5 3
0
1000
2000
3000
4000
cycle
= 2.5 ms
cycle
= 3 ms

Fig. 4.10 Effect of valve close-up time on specific impulse (
cycle
= 3 ms,
purge
= 0.1 ms);
straight tube with stoichiometric H
2
= 2.1

131
11/23/2003
4.6 Single-Tube PDE with CD Nozzle
In light of the limited performance of the straight-tube design, much effort was
expended to study the effect of nozzle configuration on the system propulsive
performance. The nozzle design for PDEs poses a serious challenge because of the
intrinsically unsteady nature of the pulse detonation process. Recent studies on the
nozzle effect, as been reviewed in Chapter 1, indicate that the nozzle configuration may
significantly change the thrust delivered by an engine. In addition to its influence on
specific impulse through modification of the gas expansion process, the nozzle affects the
chamber flow dynamics and, consequently, the timing of various phases of the engine
operation cycle, especially for high-altitude and space applications. The present work
focuses on a choked convergent-divergent (CD) nozzle because of its effectiveness in
preserving the chamber pressure during the blowdown and refilling stages. In contrast,
divergent and plug nozzles do not possess such an advantage, especially under high-
altitude conditions, in spite of their superior performance for a single-pulse operation at
sea level.
Figure 4.11 shows schematically the computational domain. The detonation tube
remains the same as in Sec. 4.5. The CD nozzle has a length of 20 cm and a throat height
of 12 cm, with a 45 convergent angle and a 15 divergent angle. The computational
domain is discretized into 554228 unstructured triangular grid cells, with 320000 for the
detonation tube, 88080 for the nozzle, and 146168 for the external region. Figures 4.12
and 4.13, respectively, show the decomposed sub-domains and the number of cells of the
sub-domains.
132
11/23/2003

Fig. 4.11 Computational domain for single-tube PDE

Fig. 4.12 Schematic of the computational domain decomposed into 64 sub-domains

sub-domain
n
u
m
b
e
r
o
f
c
e
l
l
s
8 16 24 32 40 48 56 64
0
2000
4000
6000
8000
10000
1

Fig. 4.13 Number of cells of sub-domains

133
11/23/2003
4.6.1 Flow Evolution
A series of analyses are conducted over a wide range of operation parameters.
The baseline case has an operation cycle period (
cycle
) of 3 ms, a valve close-up time
(
close
) of 2.1 ms, and a purge time (
purge
) of 0.1 ms. The ambient flow is treated as
stationary. Figures 4.14 and 4.15 show the time evolution of the Mach number and
density-gradient fields during the first cycle of operation, respectively. The
corresponding pressure and Mach-number distributions along the centerline of the
computational domain are displayed in Fig. 4.16. Figure 4.17 shows the time histories of
pressure and Mach number at the midpoints of the head end, nozzle throat, and nozzle
exit.
Initially, the detonation tube is filled with a stoichiometric hydrogen/air mixture at
the ambient pressure and temperature. The cycle begins with the valve closed.
Detonation is then initiated by the driver gas near the head end and propagates
downstream toward the unburned mixture. It is immediately followed by a centered
rarefaction wave known as the Taylor wave (Taylor, 1950; Fickett and Davis, 2000),
which decreases the pressure and brings the flow to rest in order to satisfy the stationary
condition at the head end. Between the head end and the rear of the Taylor wave is a
uniform region with constant flow properties. These wave structures have also been
presented in detail in Sec. 4.5 by means of a numerically obtained x-t diagram. The
detonation wave speed, u
D
, Mach number, M
D
, and the flow properties at CJ point can be
obtained analytically based on the standard CJ theory. The flow properties of the
uniform region can be derived through the Riemann invariant relation and the isentropic
134
11/23/2003

Fig. 4.14 Time evolution of Mach number field during the first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
135
11/23/2003

(
cycle
= 3 ms,
close
= 2.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
136
11/23/2003
x, m
0 0.2 0.4 0.6 0.8 1 1.2
p
,
a
t
m
M
0
4
8
0
1
2
t = 0.15 ms (propagation)
p
,
a
t
m
M
0
2
4
6
0
1
2
t = 0.40 ms
p
,
a
t
m
M
0
2
4
0
1
2
t = 0.50 ms
p
,
a
t
m
M
0
2
4
0
1
2
t = 0.65 ms (blowdown)
p
,
a
t
m
M
0
2
4
0
1
2
t = 0.80 ms
p
,
a
t
m
M
0
2
4
0
1
2
t = 1.00 ms
p
,
a
t
m
M
0
1
2
0
1
2
t = 1.45 ms
p
,
a
t
m
M
0
1
0
1
2
t = 2.15 ms (purging)
p
,
a
t
m
M
0
1
0
1
2
t = 2.50 ms (refilling)
p
,
a
t
m
M
0
1
0
1
2
t = 3.00 ms

Fig. 4.16 Time evolution of pressure () and Mach number (---) distributions along
centerline during first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms)
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
137
11/23/2003

Fig. 4.17 Time histories of (a) pressure and (b) Mach number at midpoints of head end,
nozzle throat, and nozzle exit during first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms)

(a)
(b)
138
11/23/2003
relation from the CJ point to the uniform region (Fickett and Davis, 2000; Wintenberger
et al., 2003). The resultant analytical formulations have been given in Chapter 1 and are
summarized again as below.
1
2
1
2
1
1
2
1
2
+
=
RT
q
RT
q
M
D
,
1 D D
u M RT = (4.20)

2
2 2
1
1
D
D
M
M
M

=
+
,
+
+
=
1
1
2
1
2 D
M
p
p
,
2
2
2
1
1
(1 )
D
D
T M
T M
| | +
=
|
+
\ .
(4.21)

2
2
1
3
2
2
1 1
1
2 1
D
D
p M
p M

| |
=
|
+
\ .
,
2
2
3
2
2
1 1
1
2 1
D
D
T M
T M
| |
=
|
+
\ .
(4.22)
Equation (4.20) yields a detonation wave Mach number of 5.94 and a detonation wave
speed of 1956 m/s, which agrees well with the 1950 m/s obtained from the NASA CEA
code (McBride and Gordon, 1996). The simulated flow properties at the CJ point and in
the uniform region coincide with the corresponding analytical predictions (e.g., deviation
of less than 0.1% for the CJ properties and 0.4% for the uniform region properties), as
manifested in Table 4.1.

Table 4.1 Flow properties at the CJ point and in the uniform region
(p
1
= 0.29 atm, T
1
= 228 K, = 1.29, R = 368.9 J/(kgK), q = 2.72010
6
J/kg)

numerical analytical
p
2
, atm 5.855 5.888
T
2
, K 2663 2665
M
2
0.737 0.737
P
3
, atm 2.158 2.154
T
3
, K 2133 2126
C
3
, m/s 1007 1006
139
11/23/2003
At t = 0.15 ms, the detonation wave travels to x = 29.5 cm, and the length of the
uniform region becomes 15.0 cm. These two lengths deviate slightly from the following
analytical predictions by about 0.6% due to the effect of the detonation initiation process.
1956 m/s 0.15 ms 29.3 cm
D D
x u t = = = (4.23)

3 3
1006 m/s 0.15 ms 15.1 cm x c t = = = (4.24)
where c
3
is the sound speed in the uniform region. Generally, the length of the uniform
region is about halfway between the detonation wave front and the head end (Fickett and
Davis, 2000), as demonstrated by the following relation:

2
3 3
2
1
1 1
2 2
D
D D D
RT c M
u M M RT
+
= = (4.25)
The detonation wave continues downstream and reaches the reactant/air interface
at the tube exit at t = 0.305 ms. It then degenerates to a non-reacting shock wave (i.e., the
primary shock wave) due to the lack of energy support from chemical reaction.
Meanwhile, a series of expansion waves are generated and propagate upstream. The
primary shock wave proceeds further downstream through the nozzle and reflects off the
convergent wall. Its propagation through the nozzle throat resembles the shock
diffraction over a convex curved wall (Han and Yin, 1993). At t = 0.40 ms, the primary
shock wave has reached the divergent section and is curved due to effect of the expansion
waves generated from the curved wall, and the two reflected shock waves have
intersected with each other. Along the curved wall, the flow behind the primary shock is
locally expanded to supersonic, leading to the formation of two shock waves stemming
out from the wall, as can be seen clearly in the enlarged pressure field in Fig. 4.18a. On
140
11/23/2003
the other hand, the interaction of the upstream-traveling expansion waves with the
downstream-traveling Taylor wave causes a small zero-gradient region, as evidenced by
the blank region located at about x = 53 cm from head end in Fig. 4.15b and by the
pressure contours in Fig. 4.18a.

Fig. 4.18 Enlarged views of pressure contours at (a) 0.40 ms and (b) 0.65 ms

At t = 0.50 ms, the primary shock wave arrives at the nozzle exit. The pressure
and temperature on the centerline immediately behind the shock wave are 2.76 atm and
527 K, respectively, corresponding to a shock wave Mach number of 2.93, which is much
(a)
(b)
141
11/23/2003
smaller than the original detonation wave Mach number of 5.94. Within the detonation
tube, the upstream-traveling expansion waves have passed through the downstream-
traveling Taylor wave so that the length of the uniform region begins to decrease. The
two reflected shock waves intersect with each other, propagate further upstream, and hit
and reflect off the tube walls, leading to more complicated shock wave structures.
At t = 0.65 ms, the primary shock wave has moved out of the nozzle. The shock
Mach number is 2.53 at the center and reduces to about 1 near the wall due to strong flow
expansion around the edge of nozzle exit. This is consistent with Skews study (Skews,
1967) on the shock-wave diffraction over a sharp corner, in which the shock Mach
number near the wall approaches to 1 for large corner angles. On the other hand, due to
the velocity difference between the exhaust and the ambient flow, slip lines (or vortex
sheets) are formed and roll up, as can be observed more clearly in the snapshots (e) and
(f) of Fig. 4.15. At t = 0.65 ms, the center of the upper spiral vortex sheet is positioned at
x = 0.811 m and y = 0.136 m, at which the flow is stationary and has a pressure of as low
as 0.18 atm. More quantitative work on the vorticity production in shock diffraction was
recently conducted by Sun and Takayama (2003). In the inner region near the edge, the
flow accelerates from subsonic to sonic due to expansion waves emanated from the edge,
and finally a Prandtl-Meyer expansion fan is formed, as evidenced in Fig. 4.18b. The
expanded supersonic flow is terminated by a downstream secondary shock attached to the
vortex sheet. It should be noted that although the flow structure resembles that of shock
diffraction over a sharp corner (Sun and Takayama, 2003; Skews, 1967; Korobeinikov
and Urtiew, 1986; Dyke, 1997), it is complicated by the flow non-uniformity resulting
from detonation and contact surface. Within the detonation tube, the downstream shock
142
11/23/2003
waves tend to catch up with the upstream shock wave, forming a lambda-shock structure,
as displayed in Fig. 4.15d. These shock waves continue upstream and finally lead to a
nearly normal leading shock wave (see Fig. 4.15e).
At t = 1.00 ms, part of the primary shock wave has moved out the computational
domain. The centers of the spiral vortex sheets move slowly to (0.829 m, 0.175 m),
whereas the pressure further decreases to 0.07 atm. The secondary shocks attached to the
vortex sheets have jointed together. The Prandtl-Meyer expansion fans originated from
the edges still exist, and the pressure at the nozzle exit plane is about 0.48 atm, higher
than the ambient value of 0.29 atm, demonstrating that the nozzle flow remains
underexpanded. In the vicinity of the nozzle throat, the sonic region has just grown from
near the wall to the entire throat section. The curved sonic line, as evidenced in Fig.
4.14f, starts at the wall slightly upstream of the throat and crosses the nozzle centerline
downstream of the throat (Hodge and Koenig, 1995). The Mach number at the midpoint
of the nozzle throat is less than unity (see Fig. 4.17b). Along the centerline downstream
of the sonic line, the flow is still slightly below sonic and then accelerated to supersonic
near the exit plane due to the expansion waves emanated from the edge of the nozzle exit
(see Fig. 4.16f). The Mach number at the midpoint of the nozzle exit plane is about 1.05.
Inside the detonation tube, the upstream-traveling expansion waves and leading shock
wave have reached the head end and been reflected, terminating the head-end pressure
plateau that remained for 0.935 ms. The head-end pressure thus decays gradually,
followed by an abrupt increase, as displayed in Fig. 4.17a. The zigzag shape of the head-
end pressure history during 1 to 2 ms arises from the reflection of the subsequent
upstream-traveling expansion waves and shock waves off the head-end.
143
11/23/2003
As the blowdown process continues, the pressure within the tube and nozzle
decays. At t = 1.45 ms, the nozzle exit pressure decreases to 0.13 atm at the midpoint and
0.21 atm at the wall, and the nozzle flow has developed to be overexpanded. Also, the
Prandtl-Meyer expansion fans disappear and oblique shocks form near the edges. Within
the nozzle, the flow downstream of the curved sonic line becomes supersonic, and the
Mach number at the midpoint of nozzle exit is about 2.2.
The valve opens at t = 2.10 ms, and the purging process begins. The head-end
pressure is about 0.6 atm prior to the valve opening, whereas the total pressure at the
combustor entrance is 2.12 atm from the inlet analyses in Chapter 1. As a result of this
large pressure difference, a right-running shock wave is established along with a contact
surface between the burned and the purged gases, as shown in Figs. 4.14h and 4.15h.
The shock wave and the contact surface are located at 0.06 m and 0.02 m from head end
at this time instant, respectively. The temperature increases from 370 K to 1916 K across
the contact surface, and the Mach number decreases from 1.04 to 0.46. Another contact
surface forms between the fresh reactants and the purged air when the refilling process
commences 0.1 ms later. The corresponding refilling pressure and velocity are 1.16 atm
and 423 m/s, respectively. At t = 2.50 ms, the shock wave and the two contact surfaces
travel to 0.5, 0.18, and 0.13 m, respectively. The pressure at the nozzle exit decays to
0.06 atm. The external flow structures bear a close resemblance to those of an
overexpanded nozzle flow at steady-state condition (Hodge and Koenig, 1995), e.g., the
oblique shocks intersect with each other and reflect off the slip lines (or shear layers) to
generate expansion waves, as displayed in Fig. 4.15i. A Mach intersection (Hodge and
144
11/23/2003
Koenig, 1995) of the two oblique shocks may happen as the intersection point moves
upstream with a further decay of the nozzle exit pressure.
At the end of the cycle (t = 3.0 ms), the shock wave resulting from the purging
process moves out of the nozzle and interacts with the existing waves, further
complicating the external flowfield. The fresh reactant fills the tube up to about two-
thirds. The pressure of the reactant ranges from 1.16 atm at the head end to 0.60 atm at
the leading point, which is significantly higher than the ambient pressure of 0.29 atm.
The velocity of the reactant ranges from 430 to 600 m/s, in contrast to the initial
stationary condition.
The flow tends to reach steady cyclic condition as the cycle repeats. Figure 4.19
shows the time history of the head-end pressure during the first five cycles. Significant
differences between the first and later cycles are observed. Specially, the pressure
plateau during the first 0.935 ms does not appear in later cycles because of the rarefaction
of waves from the previous cycle. On the other hand, because of the higher pressure of
the refilled reactants, the head-end pressure immediately after detonation is higher in later
cycles than in the first cycle. Figure 4.20 shows the specific impulse and the filling
length of each cycle. The filling length is defined as the length at which the detonation
wave catches the leading fresh reactant. Calculation of the specific impulse is given in
the next subsection. The specific impulse of the second cycle is much higher than that of
the first cycle due to the higher loading density of the reactant during the second cycle.
The specific impulse and the filling length reach steady values of 3402 s and 47.5 cm,
respectively. In this paper, the relative difference of specific impulse between two
145
11/23/2003
continuous cycles being less than 0.1% will be used as the quantitative criterion for
reaching steady cyclic operation.

Fig. 4.19 Time history of head-end pressure during first five cycles (
cycle
= 3 ms,
close
=
2.1ms)

cycle
I
s
p
,
s
f
i
l
l
i
n
g
l
e
n
g
t
h
,
m
1 2 3 4 5
0
1000
2000
3000
4000
0.45
0.50
0.55
0.60

Fig. 4.20 Specific impulse and filling length of first five cycles (
cycle
= 3 ms,
close
= 2.1
ms)

146
11/23/2003

Fig. 4.21 Time evolution of density-gradient field during fifth cycle of operation (
cycle
=
3 ms,
close
= 2.1 ms)

Figure 4.21 shows the time evolution of the density-gradient field during the fifth
cycle, at which the steady cyclic operation is reached. Although the time sequence is the
same as that of the first cycle, quite different flow patterns are obtained due to the effect
of flow non-uniformity arising from the previous cycle. The reactant has already had a
speed of about 500 m/s before being burned, the detonation wave thus propagates faster
to the tube exit, e.g., the detonation wave front travels 36.8 cm at t = 12.15 ms in contrast
to 29.5 cm in the first cycle. In the external region, the pressure level is lower because
(a)
(b)
(c)
(d)
(e)
(f)
147
11/23/2003
the flow remains overexpanded during a large amount of the cycle period. From the time
history of the pressure at the midpoint of the nozzle exit shown in Fig. 4.22, the flow is
overexpanded from 12.0 through 12.3 ms and from 12.9 through 15.0 ms, in contrast to
about 1.6 ms of overexpansion during the first cycle. The Mach number at the midpoint
of the nozzle exit remains supersonic throughout the whole cycle. From the time
evolution of the Mach-number field (not shown), the nozzle is choked during most of the
cycle period, thus helping preserve the chamber pressure. The results presented here
suggest again that the PDE analysis should be based on multicycle operations.

Fig. 4.22 Time histories of pressure and Mach number at midpoint of nozzle exit during
fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1ms)

148
11/23/2003
4.6.1.1 Effect of Ambient Flow
In the above simulation case, the ambient flow is neglected, i.e., the external
flowfield is initially stationary and a non-reflection boundary condition is used. In a real
flight condition, however, the ambient flow is not stationary and will interact with the
exhaust flow. To consider its effect, a supersonic inflow with a velocity of 636 m/s is
applied at the left boundary of the external region (see Fig. 4.11). Figure 4.23 shows the
density-gradient field at t = 14.5 ms, which can be compared with Fig. 4.21e for the
baseline case. A close-up view of the Mach number contours is shown in Fig. 4.24. In
spite of the drastic change of the flow structure in the external region, the flowfield
within the detonation tube and nozzle remains nearly the same. This is attributed to the
fact that the flow at the nozzle exit plane is supersonic during the whole cycle, and thus
the ambient flow effect cannot propagate into the chamber to affect the geodynamics
therein. As a result, the specific thrust and specific impulse are unchanged. On the other
hand, since the ambient flow modifies the external flowfield, it thus affects the engine
drag, which is beyond the scope of this paper. The ambient flow will be treated as
stationary for all the other simulation cases.
149
11/23/2003

Fig. 4.23 Density-gradient field at t = 14.5 ms, with ambient flow,
cycle
= 3 ms,
close
=
2.1 ms

Fig. 4.24 Close-up view of Mach-number contours at t = 14.50 ms, with ambient flow

150
11/23/2003
4.6.2 Propulsive Performance
Figure 4.25 shows the instantaneous axial thrust calculated with Eq. (4.10) by
neglecting the last term (i.e., the time-derivative term) during the first five cycles.
Initially, the thrust is negative, i.e., 636 N, since the fluid at the engine exit plane is
stationary. The exit flow starts at t = 0.47 ms when the primary shock wave moves out of
the nozzle, leading to a sudden jump in thrust. This kind of peak exists in each cycle.
During the later part of the blowdown stage, the thrust becomes negative due to the low
pressure and density at the exit plane. The thrust changes from negative to positive again
when the shock resulting from the purging process reaches the exit plane. During a
steady periodic cycle, e.g., the fifth cycle, the time duration of negative thrust is about 0.7
ms, that is, from 14.0 to 14.7 ms. With Eqs. (4.12)-(4.14), the cycle-averaged thrust
obtained is 862 N, and the specific thrust and specific impulse are 862 m/s and 3402 s,
respectively. This specific impulse is much higher than that achieved with only a straight
tube. It is noteworthy that the CD nozzle significantly increases the performance because
of its effectiveness in preserving the chamber pressure during the blowdown and refilling
stages, rather than by thrust generation on the nozzle wall itself. The impulse generated
on the nozzle wall can be obtained either by integrating the overpressure force on the
wall or by applying momentum balance between the entrance and exit planes of the
nozzle. Both are shown in Fig. 4.26. Although they dont coincide with each other at
every time instant, the cumulate impulses generated over the fifth cycle are exactly the
same, i.e., 0.716 Ns, which provides an essential validation check on the present
151
11/23/2003
performance calculations. The negative impulse of nozzle has also been reported by
Yungster (2003) for divergent nozzles.

Fig. 4.25 Instantaneous thrust and impulse during first five cycles (
cycle
= 3 ms,
close
=
2.1 ms)

(a)
(b)
152
11/23/2003

Fig. 4.26 Instantaneous impulse of nozzle during fifth cycle (
cycle
= 3 ms,
close
= 2.1 ms)

4.6.3 Analytical Prediction of Propulsive Performance
Although the CD nozzle increases the propulsive performance significantly, the
specific impulse of 3402 s for the baseline case is still below its ramjet counterpart of
3866 s with perfect nozzle flow expansion. It is desirable to develop simple analytical
models that can be used to rapidly and reliably predict the PDE performance and to
directly compare with the numerical simulation results.
Several analytical models have been proposed in the past and are summarized in
Chapter 1. Based on how the impulse is obtained, these models fall into two classes.
One employs unsteady gasdynamic analysis to determine the instantaneous pressures and
forces acting on the thrust wall and calculate the impulse as the integration of them
(Nicholls et al., 1957; Wintenberger et al., 2003). The other obtains the engine impulse
153
11/23/2003
by deriving the flow properties at the exit plane (Wu, Ma, and Yang, 2003; Heiser and
Pratt, 2002; Talley and Coy, 2002). The analytical model presented here follows the
approach of Heiser and Pratt (2002), but takes into account the effects of the refilling
velocity and the purging process to provides a more accurate prediction. Figure 4.27
schematically shows the flow path for the analysis. Note that the subscripts , 1, 2, and e
represent the states of the freestream, unburned gas, CJ point, and exit plane,
respectively. The procedure is detailed as below.

inlet combustor nozzle e
T
, p
, u
T
t1
, p
t1
M
1
T
1
, p
1
T
2
, p
2
T
e
u
e
CJ
isentropic
perfect match
energy balance
inlet combustor nozzle e
T
, p
, u
T
t1
, p
t1
M
1
T
1
, p
1
T
2
, p
2
T
e
u
e
CJ
isentropic
perfect match
energy balance

Fig. 4.27 Schematic of flow-path analysis for PDE performance prediction

i. Determine the total temperature T
t1
and total pressure p
t1
at the entrance of the
combustor from the inlet flow analysis.
ii. Obtain the reactant temperature T
1
and pressure p
1
by assuming a refilling Mach
number M
1
.
iii. Calculate the CJ temperature T
2
and pressure p
2
using Eqs. (4.20) and (4.21).
iv. Calculate the exit temperature by assuming an isentropic flow expansion from the CJ
state to the exit plane with a perfect pressure match:

1
2 2
( / )
e
T T p p

= (4.26)
154
11/23/2003
v. Deduce the exit velocity by applying the energy balance from the combustor
entrance to the engine exit plane:

1
2 ( )
e p e t
u q c T T ( =

(4.27)
vi. Obtain the specific thrust and specific impulse:
(1 )
sp e
F f u u
= + , /( )
sp sp
I F fg = (4.28)
If the purging process is further incorporated, the exit temperature can be
recalculated based on the following average:

1 purge 2 refill open
( ) /
e e e
T T T = + (4.29)
where T
e1
and T
e2
are the temperatures obtained by assuming isentropic flow expansion
from the purged-gas state (T
1
, p
1
) and the CJ state (T
2
, p
2
) to the exit plane with perfect
pressure match, respectively.

1
1 1 1
( / )
e
T T p p

= ,
1
2 2 2
( / )
e
T T p p

= (4.30)
Also, f and q need to be replaced by the overall fuel-to-air mass ratio f and the heat
addition per unit mass of the mixture of the reactants and purged gas q , respectively.

refill open
/ (1 ) f f f = = (4.31)

refill open
/ (1 ) q q q = = (4.32)
where is defined as

purge open
/ = (4.33)
The specific thrust and specific impulse then become
(1 )
sp e
F f u u
= + , /( )
sp sp
I F fg = (4.34)
155
11/23/2003
Based on the above analysis, the specific thrust and specific impulse are functions
of , R, q, f, p
, T
t1
, p
t1
, M
1
, and . If the current system is to be considered, the first
seven parameters are fixed values, that is, = 1.290, R = 368.9 J/(kgK), q = 2.72010
6

J/kg, f = 0.0292 (for stoichiometric H2/air reactants), p
= 0.29 atm, T
t1
= 428 K, and p
t1

= 2.12 atm, and M
1
and are two free parameters. The effects of these two parameters
are then considered.
Table 4.2 lists the predicted PDE performance under various M
1
and . The result
from the baseline numerical simulation case, which has an average refilling Mach
number of 0.93 and of 1/9, is also listed for comparison. The first condition represents
an ideal one which has zero refilling velocity and no purging process, and has an I
sp
of
4360 s. Both the specific thrust and specific impulse decrease as M
1
increases. This can
be directly explained from the aforementioned analytical formulas. As M
1
increases, T
1

decreases, leading to an increase in M
D
, as given by Eq. (4.20). The combination of Eqs.
(4.21), (4.26), and (4.27) indicates that T
e
increases and u
e
decreases as M
D
increases.
Thus, F
sp
and I
sp
decrease. The decrease in performance with the increase of M
1
can also
be explained from the entropy point of view, since the entropy rise across the detonation
wave is expressed as

2 2
2 1
1 1
1 1
2
2 2
2
ln ln
1 1
ln ln
(1 ) 1 1
p
D D
p p D
D
T p
s s s c R
T p
M M
c c M
M

+
= =
( (
| | | | + +
( (
=
| |
( (
+ + +
\ . \ .
( (

(4.35)
the increase in M
1
leads to increase in M
D
and s, thus degrading the performance.

156
11/23/2003
Table 4.2 Analytically predicted PDE propulsive performance under
various refilling Mach numbers and purge-to-open time ratios

Condition (p
t1
= 2.12 atm, T
t1
= 428 K) F
sp
, m/s I
sp
, s
(1) M
1
= 0, =
purge
/
open
= 0 1246 4360
(2) M
1
= 0.93, = 0 1167 4084
(3) M
1
= 0, = 1/9 1150 4527
(4) M
1
= 0.93, =1/9 1075 4235
baseline numerical simulation case 862 3402

Comparison between the first and third conditions of Table 4.2 indicates that F
sp

decreases and I
sp
increases with . The effect of is quite similar to the partial filling
effect for single-pulse operation (Cooper and Shepherd, 2002; Li and Kailasanath, 2002).
A larger translates to a larger air-to-fuel mass ratio. The heat released from combustion
thus needs to increase the kinetic energy of more extra air, leading to a decrease in exit
velocity. The specific thrust thus decreases. On the other hand, the overall fuel-to-air
mass ratio f , defined by Eq. (4.31), decreases faster than the specific thrust, leading to a
increase in specific thrust. The effect of can be seen more clearly through the Taylor
series expansion of F
sp
and I
sp
with respect to .

0
(1 ) (1 )
sp e sp
F f u u F A
= + (4.36)

0
/( ) [1 (1 ) ]
sp sp sp
I F fg I A = + (4.37)
where

0
0
1
0 0
1
1 ( ) /
1
2 1 ( ) /
p e e
e
p e t e
c T T q
u
A
c T T q u u

=

(4.38)
157
11/23/2003
and
0
sp
F ,
0
sp
I ,
0
e
T , and
0
e
u denote the specific thrust, specific impulse, exit temperature,
and exit velocity at zero purge time, respectively. For the current system, A 0.7.
Therefore, F
sp
decreases and I
sp
increases with , and since (1 ) A A > , the relative
increase in I
sp
is less than the relative decrease in F
sp
. From Table 4.2, when increases
from 0 to 1/9, F
sp
decreases by 7.7% and I
sp
increases by 3.8%, which agrees well with
the predictions from Eqs. (4.36) and (4.37).
4.6.4 Loss Mechanisms
In addition to the losses in the inlet and manifold mentioned in Chapter 2, there
are several losses in the combustor and nozzle of a PDE, including the viscous damping,
the loss due to heat transfer to the wall, the loss due to refilling velocity, nozzle
expansion loss, flow divergence loss, and internal flow loss. The viscous damping and
heat transfer losses will not be considered in the current study. The loss due to refilling
velocity has been discussed in Sec. 4.6.3. As the refilling Mach number M
1
increases
from 0 to 0.93, the specific impulse decreases by 6.5% from 4527 s to 4235 s, as shown
in Table 4.2. The performance discrepancy (about 25%) between the baseline numerical
simulation case and the fourth condition given in Table 4.2 is attributed to 1) nozzle
expansion loss or pressure-mismatch loss, 2) flow divergence loss, and 3) internal flow
loss. These three loss mechanisms are discussed in the following paragraphs.
In conventional steady engines, the nozzle configuration is optimized such that
the exit pressure matches with the ambient pressure. This condition, however, does not
exist for PDEs due to their unsteady operations. The resultant performance loss, referred
158
11/23/2003
to as the nozzle expansion loss, can be estimated through a post-processing analysis.
This post-process allows the recorded exhaust flow to be compressed or expanded
isentropically to match the ambient pressure and calculates the corresponding ideal
specific impulse. For the baseline case, this ideal specific impulse is 3604 s, and the
nozzle expansion loss is thus about 6.0%. Figure 4.28 shows the temporal variations of
momentum impulse and pressure impulse during a steady periodic cycle for the baseline
case. Although the pressure impulse is much smaller than the momentum impulse, it is
not zero. As mentioned in Sec. 4.6.1, the nozzle is overexpanded from 12.0 through 12.3
ms and from 12.9 through 15.0 ms, and is underexpanded otherwise. At the end of the
cycle, the momentum impulse and pressure impulse are 2.72 and 0.135 Ns,
respectively, demonstrating an overall effect of overexpansion of the nozzle flow.

Fig. 4.28 Instantaneous impulse during steady cycle (
cycle
= 3 ms,
close
= 2.1 ms)

159
11/23/2003
The flow divergence loss results from the exhaust velocity vector angularity
(Berton, 1991; Mattingly, 1996), and is generally characterized by a nozzle divergence
coefficient defined as the ratio of actual axial exhaust momentum to that from an ideal
nozzle where all of the exhaust flow is axial. For the current PDE system, the nozzle
divergence coefficient can be expressed by

e e
d
e e
m u
m u
=

(4.39)
where
e
u and
e
u are defined as

1
e
e
e S
u udS
m
=
u n
(4.40)

2 2
1
e
e
e S
u u v dS
m
= +
u n
(4.41)
Equation (4.39) gives rise to a value of 98.7% for the baseline case, which is slightly
lower than the analytical value for a steady two-dimensional nozzle flow with a
divergence angle of =15:

sin
98.9%
d

= = (4.42)
This means that the flow divergence loss in PDEs is about the same as that in
conventional steady engines. The nozzle divergence coefficient directly relates to the
loss in gross thrust. The loss in thrust due to flow divergence can be readily obtained
after taking into account the freestream momentum,
1
e e a
d
e e a
m u m u
m u m u

(4.43)
which leads to a value of 2.0% for the baseline case.
160
11/23/2003
The internal flow loss is mainly attributed to the shock waves and their
interactions within the internal flowfield. In the analytical performance prediction, the
flow is assumed to experience an isentropic expansion from the CJ state to the exit plane.
However, there exist complicated shock waves within the internal flowfield as described
in Sec. 4.6.1. The associated entropy increase and total pressure drop lead to a
performance loss. It is formidable to estimate the internal flow loss directly from the
flowfield. Nevertheless, since the sum of the nozzle expansion loss, the flow divergence
loss, and the internal flow loss is 24.5%, i.e., the deviation of the simulated performance
from the analytical prediction (condition 4 in Table 4.2), an internal flow loss of 16.5%
can thus be obtained for the baseline case. This large internal flow loss is unique for
PDEs and is one of the major disadvantages that degrade the PDE performance. The
quantitative values of the various losses discussed above are summarized in Table 4.3 for
the baseline case.

Table 4.3 Loss summary for baseline case

loss mechanism loss in specific impulse
a

loss due to refilling velocity 8.5%
nozzle expansion loss 6.0%
flow divergence loss 2.0%
internal flow loss 16.5%
total 33%

a
relative loss based on the baseline numerical result of 3402 s

161
11/23/2003
Parametric studies are conducted to investigate the timing effect on the propulsive
performance by varying the cycle time
cycle
, valve close-up time
close
, and purge time
purge
. Figure 4.29 shows the effect of
close
on the specific thrust F
sp
and the specific
impulse I
sp
for three different
cycle
: 2.5, 3 and 4 ms, corresponding to operation
frequencies of 400, 333, and 250 Hz, respectively. The purge time
purge
is fixed at 0.1
ms. Similar studies have been conducted in Chapter 3 based on quasi-one-dimensional
simulations. When the straight-tube design is compared at the same operating condition,
the present system with a choked CD nozzle can indeed substantially improve the engine
performance by a margin of 25%.
For each frequency considered herein, the specific thrust increases as
close

decreases. This can be explained as below. For a given
cycle
and
purge
,

a smaller
close

brings the following four positive effects. First, the blowdown process is shorter, and the
resultant higher chamber pressure during the refilling stage increases the loading density
of fresh reactant. Second, the increased refilling period enhances the amount of reactant
delivered to the chamber, thus increasing the chamber pressure. Third, the duration of
negative thrust is shorter. It has been pointed out in Sec. 4.6.3 that negative thrust may
appear at the later part of the blowdown process due to the low-energy level of the gases
in the combustor. Figure 4.30 shows the instantaneous thrust and impulse during a steady
cycle for three cases with different
close
. The solid line represents the baseline case.
When
close
decreases from 2.4 to 1.8 ms, the time duration of negative thrust decreases
162
11/23/2003

close
, ms
s
p
e
c
i
f
i
c
t
h
r
u
s
t
F
s
p
,
m
/
s
1.5 2 2.5 3 3.5
0
200
400
600
800
1000
1200
3 ms
4 ms
cycle
= 2.5 ms

close
, ms
I
s
p
,
s
1.5 2 2.5 3 3.5
1000
2000
3000
4000
3 ms
4 ms
cycle
= 2.5 ms

Fig. 4.29 Effect of valve closed time on (a) air-based specific thrust and (b) fuel-based
specific impulse at three different operation frequencies;
purge
= 0.1 ms, stoichiometric
H
2
= 2.1.

(a)
(b)
163
11/23/2003

Fig. 4.30 Instantaneous thrust and impulse during a steady cycle,
cycle
= 3 ms,
purge
=
0.1 ms.
(a)
(b)
164
11/23/2003
from about 1.5 ms to 0.4 ms. Finally, the internal flow loss decreases. For a smaller
close
,
the head-end pressure is higher as the valve opens. Correspondingly, the shock resulting
from the pressure difference across the valve is weaker. In addition, the reactant
temperature T
1
increases, leading to a weaker detonation wave, i.e., the detonation wave
number M
D
calculated with Eq. (4.20) is smaller. Since the internal flow loss is mainly
associated with the shock interactions within the flowfield, it thus decreases as the shock
waves weaken. These combined effects make F
sp
decrease rapidly as
close
increases.
It should be noted, however, the lower bound of
close
is subject to three practical
constraints, as mentioned in Chapter 3. The first is concerned with the inlet over-
pressurization. The head-end pressure must not exceed the total pressure of the inlet air
to allow for purging and refilling when the valve is open. Otherwise, reverse flow may
occur and cause engine unstart. This kind of lower bound is denoted by the open circles
on the curves of
cycle
= 2.5 and 3 ms. The second is related to chamber overfilling. The
fresh reactant should not flow out of the nozzle to the external region before being burned
completely unless afterburning is considered. This kind of lower bound is usually
reached in low-frequency operations, as denoted by the full circle on the
cycle
= 4 ms
curve. The third constraint, although commonly satisfied in practical cases, is that
close

should be sufficiently long to cover at least the time required for detonation initiation and
propagation throughout the entire chamber. The upper bound of
close
(also the lower
bound of
refill
) lies in the requirement that an appropriate amount of fresh reactant be
delivered to the chamber to produce thrust.
165
11/23/2003
The effect of
close
on the specific impulse follows the same trend as that of the
specific thrust, except for a small range of
close
near its lower bound. The specific
impulse and specific thrust satisfy the following relation:

purge refill
(1 / )
sp
sp
F
I
f g
+
= (4.44)
As
close
decreases, the factor (1 +
purge/
refill
) decreases and may override the increase of
F
sp
, consequently leading to a decrease in I
sp
, as shown in Fig. 4.29b.
Also observed in Fig. 4.29 is the existence of an optimum frequency. For a given
close
and
purge,
a lower frequency translates to a longer refilling period. As a
consequence, a higher chamber pressure can be reached, and the engine performance
increases. However, a large refilling time associated with low-frequency operation may
cause chamber overfilling and thus degrade the performance. These two conflicting
effects result in an optimum frequency. Among the three frequencies considered herein,
the 333 Hz (
cycle
= 3 ms) operation offers the best performance margin. The highest
specific impulse is 3672 s, slightly lower than its ramjet counterpart of 3866 s with
optimum nozzle flow expansion.
The effect of
purge
on the propulsive performance is also studied. Figure 4.31
shows the specific thrust and specific impulse at different
purge
with
cycle
of 3 ms and
close
of 2.1 ms. The full symbols denote the numerical simulation results, and the open
ones denote the analytical predictions based on the procedures proposed in Sec. 4.6.3. A
performance loss of about 20% (or 25% relative to the numerical results) is observed,
similar to that for the baseline case. The specific thrust decreases with
purge
, while the
166
11/23/2003
specific impulse increases with
purge,
which can be explained as follows. For given
cycle

and
close
, a larger
purge
translates to a larger air-to-fuel mass ratio. The heat released
from combustion thus needs to increase the kinetic energy of more extra air, leading to a
decrease in exit velocity. The decreased
refill
also degrades the specific thrust.
Combined, the specific thrust decreases. On the other hand, the factor (1 +
purge/
refill
) in
Eq. (4.44) increases and overrides the decrease in specific thrust. The specific impulse
thus increases. Note that since
open
is fixed (0.9 ms), the effect of
purge
is thus equivalent
to that of discussed in Sec. 4.6.3 based on analytical predictions. The increase of
specific impulse is usually at the price of more decrease in specific thrust. In the present
study, when
purge
increases from 0 to 0.4 ms, the specific impulse increases about 14%,
whereas the specific thrust decreases about 37%.
167
11/23/2003

purge
, ms
F
s
p
,
m
/
s
0 0.1 0.2 0.3 0.4
500
750
1000
1250
anlytical
numerical
cycle
= 3 ms,
close
= 2.1 ms
2
0
%
p
e
r
f
o
r
m
a
n
c
e
l
o
s
s

purge
, ms
I
s
p
,
s
0 0.1 0.2 0.3 0.4
2000
3000
4000
5000
anlytical
numerical
cycle
= 3 ms,
close
= 2.1 ms
20%performance loss

Fig. 4.31 Effect of purging time on (a) air-based specific thrust and (b) fuel-based
specific impulse,
cycle
= 3 ms,
close
= 2.1 ms.

(a)
(b)
168
11/23/2003
4.6.6 Effect of Nozzle Throat and Length
In addition to the operation timing, the nozzle configuration represents another
important factor that affects the PDE propulsive performance. In conventional steady
engines, the nozzle is optimized by matching the exit pressure to the ambient pressure.
This simple criterion, however, is not applicable to PDEs because the intrinsically
unsteady pulse detonation process leads to an unsteady pressure field at the exit plane.
Both numerical (Cambier and Tegner, 1998; Eidelman and Yang, 1998) and experimental
(Daniau et al., 2001; Cooper and Shepherd, 2002) studies on the effect of nozzles on the
PDE performance have been reported recently. Most of them, however, were limited to
single-pulse operation, and the question of nozzle optimization is far from resolved. In
general, the nozzle affects not only the performance through the modification of the gas
expansion process, but also the chamber dynamics and is coupled with the operation
timing, further compounding the problem. A complete nozzle optimization for PDE thus
requires tremendous computation resources. Instead of conducting such an optimization,
the main purpose of the present nozzle study is to investigate the qualitative effect of the
nozzle configuration on the PDE propulsive performance.
It has been demonstrated in the previous subsections that the CD nozzle
significantly increases the performance that can be obtained with only a straight tube. In
this subsection, the effect of the CD nozzle configuration in terms of the nozzle length
and the throat is further examined. Figure 4.32 shows the four nozzle configurations
considered herein. The baseline configuration, as described earlier, has a length of 20 cm
and a throat height of 12 cm. In the second configuration, the nozzle length is reduced to
15 cm. The third nozzle has an even smaller length of 12.4 cm, with an exit area equal to
169
11/23/2003
that of the tube. The fourth one has a throat height of 9 cm, which is 25% smaller than
that of the baseline nozzle.

Fig. 4.32 Single-tube PDE configurations (tube length = 60 cm, tube height = 16 cm):
(a) baseline case, nozzle length = 20 cm, throat height = 12 cm, (b) nozzle length = 15
cm, (c) nozzle length = 12.4 cm, (d) throat height = 9 cm.
(a)
(c)
(b)
(d)
170
11/23/2003

Fig. 4.33 Instantaneous thrust and impulse during steady cycle (
cycle
= 3 ms,
close
= 2.1
ms)
(a)
(b)
171
11/23/2003
Simulations have been conducted for all the configurations with the same timing:
cycle
of 3 ms,
close
of 2.1 ms, and
purge
of 0.1 ms. Figure 4.33 shows the instantaneous
thrust and impulse during a steady cycle. The corresponding performance parameters are
listed in Table 4.4. The effect of the nozzle length can be examined from the results of
the first three configurations. Both the thrust and impulse curves for these three
configurations are very close. Compared to the baseline configuration, the second and
third ones have I
sp
of 0.1% and 1% lower, respectively. Therefore, the nozzle length has
a minor effect on the propulsive performance. The pressure impulse shown in Fig. 4.34
demonstrates that the overall effect of the second and third nozzles is underexpansion
instead of overexpansion for the baseline nozzle. The underexpansion losses are 6% and
7%, respectively. The nominal perfect-expansion nozzle, in the sense of zero cycle-
averaged pressure impulse, thus has a length between 15 cm (second nozzle) and 20 cm
(baseline nozzle). Figure 4.35 shows the time history of the head-end pressure during a
steady cycle. The first three curves coincide, manifesting the fact that the nozzle length
only affects the gas expansion process within the divergent section of nozzle, but not the
gasdynamics within the detonation tube.

Table 4.4 Performance comparison among different nozzles

Nozzle F
sp
, m/s I
sp
, s
imperfect nozzle
expansion loss
flow
divergence loss
Baseline 862 3402 6%, overexpansion 2.0%
nozzle length = 15 cm 860 3393 6%, underexpansion 2.5%
nozzle length = 12.4 cm 852 3365 7%, underexpansion 2.6%
throat height = 9 cm 917 3597 4%, overexpansion 2.0%
172
11/23/2003

Fig. 4.34 Instantaneous pressure impulse during steady cycle (
cycle
= 3 ms,
close
= 2.1
ms)

Fig. 4.35 Time history of head-end pressure during steady cycle (
cycle
= 3 ms,
close
= 2.1
ms)

173
11/23/2003
The nozzle throat area, on the other hand, affects both the nozzle expansion
process and the flow dynamics within the detonation tube, and thus has a much more
significant effect on the performance. The impulse at the end of the cycle for the fourth
configuration is apparently higher than the others. The corresponding I
sp
of 3597 s is 6%
higher than that of the baseline configuration, attributed to the following factors. The
smaller throat helps preserve the chamber pressure, as evidenced in Fig. 4.35. The
refilling pressure for the fourth configuration is about 1.49 atm, higher than the 1.24 atm
for the baseline configuration. Consequently, the refilling Mach number of 0.76 is
smaller than the 0.93 for the baseline configuration. In addition, the imperfect nozzle
expansion loss is reduced, as listed in Table 4.4. All these factors lead to a higher
propulsive performance for a smaller throat. Nevertheless, the relation of the propulsive
performance and the nozzle throat area is not monotonic. As the nozzle throat becomes
exceedingly small, the internal flow loss associated with the shock interactions increases
since the shocks reflected off the nozzle convergent wall become stronger. Furthermore,
a longer blowdown process may be required to avoid inlet over-pressurization, thus
degrading the performance. These two negative effects and the aforementioned positive
effects result in the existence of an optimum nozzle throat area.
To find the optimum nozzle throat area requires conducting a large number of
simulations because of the coupling between the effects of the nozzle throat and
operation timing. A presumptive estimation is that the optimum nozzle may improve the
best performance of the baseline nozzle by 6%, that is, the highest I
sp
of about 3892 s,
which is only slightly higher than its ramjet counterpart of 3866 s with perfect nozzle
flow expansion. Further improvement of the system design and optimization thus need to
174
11/23/2003
be conducted to make the PDEs competitive with other conventional engines. One such
improvement is to use multiple detonation tubes, which will be presented in the next
Chapter.
Finally, it is noteworthy that the analyses based on single-pulse operation may
lead to totally different trends for the nozzle effect. For example, during the first cycle,
the overall effect of all four nozzles is underexpansion, and the impulse at the end of the
first cycle for the baseline configuration is higher than that for the fourth configuration.
A realistic analysis of nozzle effect on PDE performance should thus be based on
multicycle instead of single-pulse operations.
4.7 Summary and Conclusions
The thrust chamber dynamics in single-tube airbreathing PDEs with multicycle
operations has been studied by means of two-dimensional simulations. The system under
consideration includes a supersonic inlet, an air manifold, a rotary valve, a single-tube
combustor, and a convergent-divergent nozzle, which is designed for the flight condition
at an altitude of 9.3 km and a flight Mach number of 2.1. The combustion and
gasdynamics involved in the detonation initiation and propagation, blowdown, purging,
and refilling processes are examined in detail. A flow-path based performance prediction
model is proposed to estimate the upper performance limit of the current PDE system.
The various performance loss mechanisms, such as the imperfect nozzle expansion loss,
flow divergence loss, and internal flow loss, are identified. The effects of operation
timing, including the cycle time, valve close-up time, and purge time, and nozzle
175
11/23/2003
configurations in terms of nozzle length and throat area on the propulsive performance
are investigated systematically. The analyses presented in this paper can be effectively
utilized to provide guidelines for the PDE design and to identify the various loss
mechanisms limiting the PDE performance. A number of important conclusions drawn
from the present studies are given as below.
(1) The imperfect nozzle expansion loss resulted from the mismatch between the
exit and ambient pressures is unavoidable in PDEs because the intrinsically unsteady
pulse detonation process leads to an unsteady pressure field at the exit plane. The
internal flow loss is mainly attributed to the shock waves and their interactions within the
internal flowfield. These two losses reach 6% and 16.5% for the baseline case,
respectively, composing the major factors that degrade the propulsive performance and
render the PDE less attractive.
(2) There exists an optimum operation frequency for achieving the best
performance margin. At a given frequency and purge time, a smaller valve close-up time
increases the performance in most cases. On the other hand, for a given frequency and
valve close-up time, a larger purge time decreases the specific thrust and increases the
specific impulse.
(3) The nozzle length has a minor effect on the propulsive performance since it
only modifies the gas expansion process within the divergent section. The throat area, in
contrast, affects both the gas expansion process and the gasdynamics within the chamber,
thus exerting a much more significant effect. Among the four nozzle configurations
studied herein, the smaller throat improves the performance up to 6%, whereas the nozzle
length affects the performance by only 1%. A smaller throat tends to increase the
176
11/23/2003
performance by raising the chamber pressure. However, an exceedingly small throat
renders negative effects due to the related longer blowdown process and larger internal
flow loss. An optimum throat thus exists.
(4) Significant differences exist between single-pulse and multicycle operations.
A realistic analysis of PDE performance should be based on multicycle instead of single-
pulse operations.
The best specific impulses obtained for the current PDE system with a
stoichiometric hydrogen/air mixture, e.g., 3672 s for the baseline configuration and 3892
s for an optimum nozzle, are only close to or slightly higher than its ramjet counterpart of
3866 s. Further improvement of the design philosophy and optimization of the system
configuration and operation timing thus need to be conducted to make the PDEs
competitive with conventional engines.
177
11/23/2003
Chapter 5
Two-Dimensional Analyses of Multitube PDEs
Chapter 4 has focused on the thrust chamber dynamics and propulsive
performance of single-tube airbreathing PDEs. At a flight altitude of 9.3 km and Mach
number of 2.1, the best specific impulse for a baseline configuration with a stoichiometric
hydrogen/air system is 3672 s, which is slightly lower than its ramjet counterpart of 3866
s. Further improvements of the system design are thus required to make the PDE
competitive with other conventional engines. One of the improvements is to use a
combustor consisting of multiple detonation tubes operating sequentially. In principle,
this multitube design offers the following advantages:
(1) Delivering air from inlet to multiple detonation tubes reduces the inlet loss
associated with the airflow stagnation during the period when none of the tubes are being
filled. In a single-tube PDE, this time period takes up a large part of the cycle time and
may cause inlet unstart.
(2) Exhaust from multiple detonation tubes discharging into a common nozzle
provides a more stable nozzle flow and helps to increase the nozzle exit pressure which
would be quite low in a single-tube PDE during the later part of the blowdown process
and the purging and refilling processes, and thus improves the nozzle performance.
(3) The detonation wave from one tube can precompress the reactants in other
tubes.
(4) The purging and refilling processes are less coupled with the blowdown
process, thus leading to a wider range of operation timing.
178
11/23/2003
(5) The overall engine operation frequency is increased, usually by a factor equal
to the number of detonation tubes used. Meanwhile, the degree of unsteadiness is
reduced.
(6) It provides the potential of fluidic thrust vectoring.
This chapter attempts to develop a comprehensive numerical analysis dealing with
the thrust chamber dynamics in multitube PDEs with repetitive operations, to examine
the flow interaction among the detonation tubes, and to investigate the effects of
operation timing and system geometry on the propulsive performance.
5.1 System Configuration
The system under consideration is shown schematically in Fig. 5.1. It includes a
co-axial supersonic inlet with mixed compression, an air manifold, a rotary valve, a
combustor consisting of multiple detonation tubes, and a common convergent-divergent
(CD) nozzle. This PDE was designed for a flight altitude of 9.3 km and a flight Mach
number of 2.1, corresponding to the flight condition of advanced missiles. The static
pressure, static temperature, total pressure, and total temperature of the freestream are
0.29 atm, 228 K, 2.65 atm, and 428 K, respectively. The total pressure at the entrance of
the combustor is set to 2.12 atm according to the study of the inlet aerodynamics.
The cyclic operation of the PDE is controlled by a valve system located at the
entrance of the combustor. The detonation tubes operate sequentially with a fixed time
lag between each, as shown schematically in Fig. 5.2 for three detonation tubes. The left
ends (head ends) of the tubes are assumed to be fully open or fully closed for simplicity.
179
11/23/2003
The operation sequence of each tube is the same as that for a single-tube PDE, which is
controlled by three time periods: the valve close-up period (
close
), the purging period
(
purge
), and the refilling period (
refill
). The cycle period of each tube is the sum of the
above three periods. It may, however, be different from that for the engine. For clarity,
the parameter
cycle
is used to represent the cycle period based on each tube, that is,
cycle

=
close
+
purge
+
refill
.

combustor
inlet
isolator
nozzle

Fig. 5.1 Supersonic airbreathing pulse detonation engine

tube open tube close
detonation blowdown
purge refill
tube 3
tube 1
tube 2
p
p
p
b b
b b
b d
d
d f
f
f
p
b
f
d
tube open tube close
detonation blowdown
purge refill
tube 3
tube 1
tube 2
p
p
p
b b
b b
b d
d
d f
f
f
p
b
f
d
p
b
f
d

Fig. 5.2 Operation sequence of a triple-tube PDE

180
11/23/2003

(a) without free volume

(b) with free volume

Fig. 5.3 Computational domains for multitube PDEs

Practical multitube PDE configurations are three-dimensional. Nevertheless,
conducting full three-dimensional simulation of these systems requires tremendous
computational resources. The present work thus considers only planar systems to
conduct two-dimensional simulations. The resultant flow phenomena and the
performance trend are believed to be qualitatively the same as those from the three-
dimensional analysis. In addition, the two-dimensional analysis presented here provides
direct comparisons with that on single-tube PDEs in Chapter 4.
181
11/23/2003
Figure 5.3 shows the computational domains for the two configurations
considered herein. The combustor contains three detonation tubes, each with a height of
5 cm. The tubes are spaced 0.5 cm apart. The nozzle adopts the baseline shape used in
Chapter 4, which has a length of 20 cm and a throat height of 12 cm, with a 45
convergent angle and a 15 divergent angle. In the first configuration, the length of the
detonation tubes is 60 cm, and the tubes extend to the nozzle entrance. In the second
configuration, the length of the detonation tubes decreases to 45 cm, leaving a free
volume of 15 cm long between the detonation tubes and the nozzle. The same external
region is included in the computational domains for both configurations. The numbers of
unstructured grid cells for these two cases are 623254 and 664362, respectively.
5.2 Results and Discussion
A series of simulations are conducted over a wide range of operation parameters
for both configurations. The flow evolution, as well as the interactions among the
multiple detonation tubes, is first presented in detail for a baseline case, serving as a basis
for the further performance analysis and parametric studies. The engine propulsive
performance is then obtained, and the various loss mechanisms are identified. Finally,
the effects of the valve timing and the system geometry on the propulsive performance
are discussed.
182
11/23/2003
5.2.1 Flow Evolution
The baseline case for the first configuration has an operation cycle period (
cycle
)
of 3 ms, leading to a 1 ms time lag between tubes. The valve close-up time (
close
),
purging time (
purge
), and refilling time (
refill
) are 2.1 ms, 0.1 ms, and 0.8 ms,
respectively. The ambient flow is treated as stationary because of its minor effect on the
propulsive performance. Figure 5.4 shows the time evolution of the density-gradient
field during the first cycle of operation. The corresponding pressure distribution along the
centerline of each tube is given in Fig. 5.5, and the time histories of pressure at the
midpoint of the head end of each tube in Fig. 5.6.
Initially, the bottom tube is partially (75%) filled with a quiescent stoichiometric
hydrogen/air mixture at the ambient pressure (0.29 atm) and temperature (228 K),
whereas the other region is filled with the ambient air. Detonation is directly initiated in
the bottom tube by a driver gas region spanning 0.2 mm near the head end with a
temperature of 2000 K and a pressure of 30 atm. This small amount of driver gas has
been demonstrated to have neglectable contribution to the engine impulse in Chapter 4.
The detonation wave then propagates downstream at a Chapman-Jouguet (CJ) speed of
1956 m/s, followed by the Taylor wave and a uniform region as have been discussed in
Chapter 4. The CJ pressure and temperature are 5.855 atm and 2663 K, respectively,
whereas the pressure and temperature of the uniform region are 2.158 atm and 2133 K,
respectively. At t = 0.15 ms, the detonation wave has traveled approximately one half of
the tube length, and the uniform region spreads about halfway between the detonation
wave front and the head end, as also demonstrated in Chapter 4. The middle tube is in
183
11/23/2003

Fig. 5.4 Time evolution of density-gradient field during the first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
top: close
mid: purging
bot: detonation
top: close
mid: refilling
bot: blowdown
top: close
mid: refilling
bot: blowdown
top: close
mid: ignition
bot: blowdown
top: purging
mid: detonation
bot: blowdown
top: blowdown
mid: blowdown
bot: ignition
top: blowdown
mid: blowdown
bot: refilling
top: detonation
mid: blowdown
bot: purging
top: ignition
mid: blowdown
bot: blowdown
top: refilling
mid: blowdown
bot: blowdown
184
11/23/2003
p
,
a
t
m
0
2
4
6
t = 0.15 ms
top tube
middle tube
bottom tube
p
,
a
t
m
0
1
2
3
t = 0.60 ms
p
,
a
t
m
0
1
2
3
t = 0.80 ms
p
,
a
t
m
0
1
2
3
t = 1.00 ms
p
,
a
t
m
0
5
10
15
t = 1.15 ms
p
,
a
t
m
0
2
4 t = 1.50 ms
p
,
a
t
m
0
1
2
3
t = 2.00 ms
p
,
a
t
m
0
10
20
t = 2.15 ms
p
,
a
t
m
0
2
4
6
t = 2.50 ms
p
,
a
t
m
0
1
2
3
t = 3.00 ms
x, m
0 0.1 0.2 0.3 0.4 0.5 0.6

Fig. 5.5 Time evolution of pressure distribution along centerline of each tube during the
first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
185
11/23/2003

Fig. 5.6 Time histories of pressure at midpoints of head ends of the bottom, middle, and
top tubes during the first cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1
ms)

the purging stage. The two vertical lines represent the shock wave and the contact
surface induced by the pressure difference across the valve as the purging process begins.
The pressure and velocity behind the shock wave are 1.20 atm and 411 m/s, respectively.
The detonation wave in the bottom tube reaches the reactant/air interface located
at 40 cm from head end at t = 0.20 ms. It then degenerates to a non-reacting shock wave,
the primary shock wave. Meanwhile, a series of expansion waves are generated at the
interface. The expansion waves propagate both downstream along with the Taylor wave
to the tube exit and upstream to the head end. The upstream-traveling expansion waves
interact and pass through the Taylor wave and thus reduce the length of the uniform
region, as displayed in Fig. 5.5b. The first expansion wave arrives at the head end at t =
0.625 ms, and consequently, the head-end pressure begins to decay gradually. As the
186
11/23/2003
expansion waves reflect off the head end, another series of expansion waves form and
propagate downstream toward the tube exit, further reducing the pressure in the bottom
tube.
The primary shock wave reaches the bottom tube exit at t = 0.380 ms. It then
diffracts at the tube exit and reflects from the nozzle walls, causing complex waves
propagating upstream into all the three tubes and downstream into the nozzle, as
displayed in the snapshot of t = 0.60 ms in Fig. 5.4. A close-up view of these flow
developments and the flow interactions among the tubes and nozzle are given in Fig. 5.7.
The primary shock wave has been significantly weakened by the expansion waves from
the interface before emerging from the bottom tube, e.g., the pressure behind the shock
decays to 3.05 atm. Its possibility to initiate detonations in the neighbor tubes is thus
avoided. The flow structures related to the diffraction of the shock wave around the
upper edge of the bottom tube, such as the Prandtl-Meyer expansion fan, the vortex, and
the secondary shock, and the shock reflected from the nozzle wall are all clearly seen in
Fig. 5.7b. In Fig. 5.7c, the diffracted and reflected shock waves have propagated into the
middle and bottom tubes, respectively, and the pressures behind them are 0.32 and 2.40
atm. On the other hand, the upper part of the leading shock hits the edge connecting the
middle and top tubes, whereas the right part propagates in the divergent section of the
nozzle. Along the curved wall, the flow behind the leading shock is locally expanded to
supersonic, leading to the formation of a shock wave stemming out from the wall, as
evidenced in Fig. 5.7c. The upper part of the leading shock then hits (Fig. 5.7d) and
reflects off (Fig. 5.7e) the upper wall of the nozzle. In the top tube, the first shock relates
to the diffracted part and the second one to the reflected shock from the lower wall of the
187
11/23/2003
nozzle. The pressures behind these two shocks are 0.37 and 0.53 atm, respectively. The
shock waves established in all the three tubes propagate upstream and elevate the
pressure therein (see Figs. 5.5b and 5.5c). In addition, because of the relative low
strength, they will not initiate detonations if encountering fresh reactant.

Fig. 5.7 Snapshots of pressure field showing flow interactions among tubes and nozzle
during the first cycle (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

At t = 0.80 ms, the primary shock wave has emerged from the nozzle into the
external region. Vortices are formed near the edges of the nozzle exit. Except for the
asymmetrical pattern, the external flowfield exhibits quite similar structures to that of the
(a)
(b)
(c)
(d)
(e)
(f)
188
11/23/2003
single-tube case. Within the combustor, the reflected shock in the bottom tube
propagates toward the head end faster than those in the middle and top tubes, due to the
larger sound speed related to the high temperature of the products in the bottom tube.
The middle tube is in the refilling stage. The shock wave induced in the purging process
travels to the location of 45 cm and is going to meet the upstream-traveling waves
induced by the detonation of the bottom tube.
The middle tube ends the refilling process and ignites at t = 1 ms. At this time
instant, the purge-induced contact surface locates at x = 37 cm and the leading fresh
reactant at x = 32 cm. The purge-induced shock has interacted with the upstream-
traveling waves induced by the detonation of the bottom tube. At t = 1.15 ms, the
detonation wave in the middle tube reaches x = 37 cm. It is faster than the detonation
wave in the bottom tube since the reactant in the middle tube has already had a velocity
of about 411 m/s prior to detonation. The top tube undergoes the purging process. At a
slightly later time, the detonation wave in the middle tube catches the leading fresh
reactant at x = 39 cm and degenerates to a non-reacting shock wave. The resultant shock
wave proceeds further downstream and interacts with the waves induced previously by
the detonation wave of the bottom tube, causing very complicated flow structures in the
nozzle and the external regions, as displayed in Fig. 5.4f. Two Prandtl-Meyer expansion
fans are seen at the exit of the middle tube. The top tube is now in the refilling process,
with a refilling pressure of 1.20 atm and a refilling velocity of 411 m/s. The interaction
of the downstream- and upstream-traveling shock waves leads to a high pressure region
with pressure up to 2.2 atm in the top tube, as shown in Fig. 5.5f. This pressure is not
high enough to initiate a detonation in the top tube, but may interfere with the refilling
189
11/23/2003
process. The refilling process should thus be finished before the upstream-traveling
shock wave arrives at the head end. Otherwise, inlet over-pressurization may happen.
For the current case, the shock wave propagates toward the head end at a speed of as low
as about 126 m/s, and thus permits enough time for the refilling process. It reaches x =
11 cm as the refilling process finishes.
Ignition occurs in the top tube at t = 2 ms while the middle and bottom tubes
undergo blowdown process. The aforementioned upstream-traveling shock wave
interacts with the detonation wave and Taylor wave and reflects off the head end at about
t = 2.15 ms, causing an abrupt raise in head-end pressure, as evidenced in Fig. 5.6. In
Fig. 5.4h, the detonation wave has passed through the leading fresh reactant at x = 31.2
cm and degenerated to a non-reacting shock wave. At t = 2.5 ms, the shock wave has
moved out of the nozzle to further interact with the local flowfield in the external region.
Reflected shock waves can be seen in all the three tubes near their exits. The pressures
behind these shocks, from bottom to top, are 0.8, 1.8, and 4.0 atm. The bottom tube is in
the refilling stage. The refilling pressure and velocity are 1.30 atm and 380 m/s,
respectively.
The bottom tube finishes the refilling process at t = 3.0 ms, the end of the first
cycle, whereas the middle and top tubes are in the blowdown process. The leading fresh
reactant in the bottom tube reaches x = 35.5 cm. The upstream-traveling shock wave
induced by the detonation of the top tube has interacted with the purge-induced shock
wave and contact surface and propagated into the fresh reactant, denoted by the pressure
jump at about x = 32 cm in Fig. 5.5j.
190
11/23/2003
Comparing with the flow evolution described in Chapter 4 for a single-tube PDE,
the current one bears the following characteristics. First, the flowfield exhibits
asymmetry and is much more complicated. Second, the expansion pattern of the nozzle
is not as apparent as that in the single-tube case, in which the Prandtl-Meyer expansion
fan, representing an underexpanded nozzle flow, or the oblique shock wave, representing
an overexpanded nozzle flow, can be clearly observed near the nozzle exit during certain
periods of the blowdown process. Third, the pressure in a tube is raised not only by the
shock waves induced by the detonation wave of this tube but also by the shock waves
induced by the detonation waves from other tubes.
The flow tends to attain a steady periodic condition as the cycle repeats. Figure
5.8 shows the specific impulse and the middle tube filling length of each cycle. The
specific impulse is calculated based on the momentum balance over a control volume
enclosing the entire engine, as detailed in Chapter 4. The filling length is defined as the
length at which the detonation wave catches the leading fresh reactant. By using the
quantitative criterion proposed in Chapter 4, i.e., the relative deviation in cycle-averaged
specific impulses of two continuous cycles being less than 0.1%, the current baseline case
reaches steady periodic operation at the fifth cycle. The time evolution of the density-
gradient field during this cycle is shown in Fig. 5.9, and the corresponding pressure
distribution along the centerline of each tube is given in Fig. 5.10. Quite different flow
patterns are obtained due to the effect of flow non-uniformity arising from the previous
cycle. The flowfield at t = 15.00 ms is almost the same as that at t = 12.00 ms, verifying
again that the steady periodic operation has been reached. The averaged refilling
pressures of the bottom, middle, and top tubes are 1.36, 1.16, and 1.24 atm, respectively.
191
11/23/2003
The corresponding refilling Mach numbers are 0.85, 1.0, and 0.94, and the refilling
velocities are 361, 440, and 392 m/s, respectively. The filling lengths of the bottom,
middle, and top tubes are 39.0, 38.8, and 40.5 cm, respectively, slightly smaller than the
47.5 cm for the single-tube case.

cycle
I
s
p
,
s
f
i
l
l
i
n
g
l
e
n
g
t
h
,
m
1 2 3 4 5
0
1000
2000
3000
4000
0.35
0.36
0.37
0.38
0.39
0.40

Fig. 5.8 Specific impulse and filling length of the middle tube of the first five cycles
(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)
192
11/23/2003

(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
top: blowdown
mid: blowdown
bot: ignition
top: blowdown
mid: purging
bot: detonation
top: blowdown
mid: refilling
bot: blowdown
top: blowdown
mid: ignition
bot: blowdown
top: purging
mid: detonation
bot: blowdown
top: blowdown
mid: blowdown
bot: ignition
top: blowdown
mid: blowdown
bot: refilling
top: detonation
mid: blowdown
bot: purging
top: ignition
mid: blowdown
bot: blowdown
top: refilling
mid: blowdown
bot: blowdown
193
11/23/2003
x, m
0 0.1 0.2 0.3 0.4 0.5 0.6
p
,
a
t
m
0
2
4
6
t = 12.50 ms
p
,
a
t
m
0
1
2
3
t = 13.00 ms
p
,
a
t
m
0
10
20
t = 13.15 ms
p
,
a
t
m
0
1
2
t = 14.00 ms
p
,
a
t
m
0
2
4
6
t = 13.50 ms
p
,
a
t
m
0
1
2
t = 15.00 ms
p
,
a
t
m
0
10
20
t = 14.15 ms
p
,
a
t
m
0
2
4 t = 14.50 ms
p
,
a
t
m
0
1
2
t = 12.00 ms
bot.
mid.
top
p
,
a
t
m
0
10
20
t = 12.15 ms
top tube
middle tube
bottom tube

Fig. 5.10 Time evolution of pressure distribution along centerline of each tube during
the fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
194
11/23/2003

Fig. 5.11 Time histories of pressure at midpoints of (a) head ends and (b) exit of the
bottom, middle, and top tubes during the fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1
ms, and
purge
= 0.1 ms)

Figure 5.11 shows the time histories of the pressure at the midpoints of the head
end and the exit of each tube, to provide more quantitative inspection on the tube
interactions. In Fig. 5.11a, the highest peak on each trace corresponds to the initiation of
(a)
(b)
195
11/23/2003
detonation at the head end. The second peak on the bottom tube trace, point A, indicates
the shock wave induced by the detonation of the top tube arriving at the head end. The
second peak on the middle tube trace, point B, relates to the arrival of the shock wave
induced by the detonation of the bottom tube. This detonation wave also induces a shock
wave that propagates into the top tube, interferes with the refilling process, and causes a
small jump in the head-end pressure of the top tube, as evidenced by point E. The raised
pressure is still less than the total pressure of the incoming gas so that the refilling
process continues. The second peak on the top tube trace, point C, results from the
detonation wave of the middle tube. Similarly, this detonation wave induces a shock
wave in the bottom tube that leads to a small jump in the head-end pressure (point F).
The third peak, point D, corresponds to the arrival of the shock wave resulting from the
purge-induced shock in the middle tube. This purge-induced shock, however, exhibits
little effect on the bottom tube head-end pressure. The reason is that the bottom tube has
a supersonic exhaust flow when the purge-induced shock emerges from the middle tube.
It is noted that all the aforementioned peak pressures exceed the total pressure of the gas
immediately in front of the tubes. They should be thus within the period with valve
closed. Otherwise, reverse flow may occur at the entrance of the detonation tubes and
engine unstart happens.
In Fig. 5.11b, the first peaks on the three traces, points A, B, and C, correspond to
the arrival of the detonation-degenerated shock wave at the tube exit. The second peaks
on the traces of the bottom and top tubes, points D and E, correspond to the arrival of the
reflected shocks from the nozzle wall. The middle tube, nevertheless, doesnt have this
kind of peak because the middle tube exit is farther from the nozzle wall, and a
196
11/23/2003
supersonic exhaust flow is developed soon after the detonation-degenerated shock wave
emerges from the middle tube.
The flowfield within the common nozzle, as can be seen in Fig. 5.9, is quite
different from that of the single-tube PDE. Multiple asymmetrical transverse waves exist
within the common nozzle and disorganize the choke pattern which appears in the nozzle
of the single-tube PDE during most of the cycle period. The nozzle throat in the
multitube PDE seems to play a role less important to the performance than in the single-
tube PDE. The complexity of the nozzle flow is also evidenced in Fig. 5.12, in which the
time histories of the Mach number at the midpoints of the nozzle throat and exit planes
are displayed. In the single-tube PDE, the nozzle is choked during the entire cycle except
for a small period during which the detonation-degenerated shock wave sweeps the throat
region. The typical choke pattern is represented by a curved sonic line that starts at the
wall slightly upstream of the throat and crosses the nozzle centerline downstream of the
throat. As a result, the Mach number at the midpoint of the throat is less than unity and
remains quite smooth during most time of the cycle as demonstrated in Fig. 5.12a.
However, in the multitube PDE, the Mach number at the midpoint of the throat exhibits
significant variation and is either larger than or far below unity during a large portion of
the cycle. The choke effect of the nozzle throat is thus quite weak. In Fig. 5.12b, the
Mach numbers at the midpoint of the nozzle exit plane also display significant difference,
such as the three large peaks appearing in the triple-tube trace. Both Mach numbers,
however, are larger than unity during the entire cycle.

197
11/23/2003

Fig. 5.12 Time histories of Mach number at midpoints of (a) nozzle throat and (b)
nozzle exit during the fifth cycle of operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1
ms)

(a)
(b)
198
11/23/2003
5.2.2 Propulsive Performance and Loss Mechanisms
Because of the intrinsic unsteady nature of PDEs, their propulsive performance
must be calculated appropriately. A detailed description of the performance calculation
based on the momentum balance over a control volume enclosing the entire engine has
been given in Chapter 4. Figure 5.13 shows the instantaneous thrust in both axial and
vertical directions, obtained from Eq. (4.10) by negelecting the last term. The single-tube
PDE results are also included for comparison. In Fig. 5.13a, a very high peak up to
15000 N exists for the single-tube PDE, corresponding to the arrival of the primary shock
wave at the nozzle exit plane. The deviation of the peak value from the cycle-averaged
value represents a quantitative measure of the degree of unsteadiness of the engine thrust.
For the triple-tube PDE, the number of peaks increases to three in each cycle, but the
peak magnitudes are significantly reduced, e.g., by a factor of about three, rendering a
substantial improvement in engine steadiness. The three peaks represent the arrival of the
detonation-degenerataed shock waves of the bottom, middle, and top tubes at the nozzle
exit plane, respectively. The second peak related to the middle tube is higher because the
shock wave from the middle tube experiences less diffraction and reflection than those
from the other two tubes. The same reason also leads to a much lower peak related to the
middle tube in the lateral thrust (Fig. 5.13b). The increase of the peak number modifies
the spectral property of the thrust. Figure 5.14 represents the spectra of the instantaneous
axial thrusts. Clearly, the dominant frequency of the single-tube PDE is equal to the tube
operation frequency, i.e., 333 Hz. However, that of the triple-tube PDE is increased by
three times to 1000 Hz.

199
11/23/2003

Fig. 5.13 Instantaneous thrust in (a) axial and (b) vertical directions during the fourth
and fifth cycles (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

Also observed in Fig. 5.13a is that the time duration of negative thrust of triple-
tube PDE is significantly decreased to nearly zero, whereas that of the single-tube PDE is
about 0.7 ms, e.g., from 14.0 to 14.7 ms during the fifth cycle. This is due to the fact that
the exhaust flow of low energy level during the later part of the blowdown stage and the
(a)
(b)
200
11/23/2003
refilling stage of a single-tube PDE is avoided in multitube PDE by the sequential
operations of the tubes. For exapmle, when one tube is in the later part of the blowdown
stage, the other tubes may be in the earlier part of the blowdown stage.
Figure 5.13b demonstrates the existance of a considerable lateral thrust up to 1000
N in the current triple-tube PDE, thereby causing unnecessary vibration of the vehicle.
One way to mitigate this problem is to implement tube-pairs. Each tube-pair includes
two detonation tubes which are located at symmertic positions and operate synchronously
to ensure symmetric operation and consequently eliminates the lateral thrust. In some
cases, however, the lateral thrust can also be appropriately utilized for thrust vectoring.

Fig. 5.14 Spectra of periodic instantaneous axial thrusts of single-tube and triple-tube
PDEs with operation timing of
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms

The cycle-averaged specific impulse and specific thrust during a steady periodic
cycle, e.g., the fifth cycle, are 3540 s and 895 m/s, respectively. These performance
201
11/23/2003
numbers are about 4% higher than those of the baseline single-tube PDE given in Chapter
4. The multitube design thus slightly improves the propulsive performance. To identify
the various loss mechanisms, the flow-path based analytical performance prediction
model proposed in Chapter 4 is used to predict the upper performance limit. This model
requires specifying the refilling Mach number. As mentioned in Sec. III.A, the refilling
Mach numbers of the bottom, middle, and top tubes are 0.85, 1.0, and 0.94, respectively.
The average is 0.93. Interestingly, it is just the same as that for the baseline single-tube
PDE. The corresponding upper limit of the specific impulse is thus 4235 s. Comparing
the simulated specific impulse to this upper limit, a performance loss of about 19.6% is
obtained. This performance loss includes the imperfect nozzle expansion loss, the flow
divergence loss, and the internal flow loss. Based on the calculation methods provided in
Chapter 4, the three losses for the current baseline triple-tube PDE are 3%, 2.3%, and
14.3%, respectively. Compared to the baseline single-tube PDE, the nozzle expansion
loss is significantly reduced, the internal flow loss is slightly reduced, and the flow
divergence loss remains about the same. The 4% increase in propulsive performance of
the triple-tube PDE thus comes from the reduction of the imperfect nozzle expansion loss
and the internal flow loss.
Figure 5.15 shows the instantaneous pressure thrust and impulse during a steady
periodic cycle. The single-tube results are also included for comparison. In conventional
steady engines with optimized nozzle, the exit pressure is equal to the ambient pressure,
and the pressure thrust is thus zero. For unsteady engines such as PDE, the closer the
pressure thrust to zero, the less the imperfect nozzle expansion loss. It is clear in Fig.
5.15a that the pressure thrust of the triple-tube PDE remains closer to zero than that of the
202
11/23/2003
single-tube PDE, demonstrating a reduction in the imperfect nozzle expansion loss. The
cumulative pressure impulse over the cycle is less than zero, as can be seen in Fig. 5.15b,
indicating the overall effect of the nozzle flow is overexpansion.

Fig. 5.15 Instantaneous pressure thrust and impulse during the fifth cycle (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

(a)
(b)
203
11/23/2003
The flow divergence loss results from the exhaust velocity vector angularity.
Figure 5.16 compares the velocity profiles at the nozzle exit plane during a steady
periodic cycle between the triple- and single-tube PDEs. The asymmetric pattern of the
triple-tube results are clearly observed. However, the velocities from both cases are very
close, leading to almost the same flow divergence losses.

t
1
= 12.15 ms t
2
= 12.50 ms t
3
= 13.15 ms t
4
= 13.50 ms t
5
= 14.15 ms t
6
= 14.50 ms

Fig. 5.16 Axial and vertical velocity profiles at nozzle exit plane during the fifth cycle of
operation (
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms)

The internal flow loss is mainly attributed to the shock waves and their
interactions within the internal flowfield. In spite of the much more complicated shock
interactions within the multitube PDE flowfield, the strengths of the shock waves related
to the diffraction around the tube exit and the reflection from the nozzle wall are
single tube
triple tube
204
11/23/2003
relatively smaller than the strength of the shock wave related to the reflection from the
nozzle wall in the single-tube PDE, leading to a slightly smaller internal flow loss.
A parametric study is conducted to investigate the timing effect on the propulsive
performance by varying the cycle time
cycle
and the valve close-up time
close
. The purge
time
purge
is fixed at 0.1 ms. Figure 5.17 shows the effects of
close
on the specific thrust
F
sp
and the specific impulse I
sp
for two
cycle
: 3 ms and 4 ms. The single-tube results are
also included for comparison. Clearly, the multitube design helps improve the propulsive
performance.
The specific thrust increases as
close
decreases for each frequency considered
herein. The explanation of this trend for single-tube PDEs presented in Chapter 4 also
applies here. Briefly, for a given
cycle
and
purge
, a smaller
close
leads to higher loading
density of fresh reactant, a larger amount of reactant delivered to the chamber, a shorter
period of negative thrust, and less internal flow loss, and thus results in a higher specific
thrust. The lower bound of
close
is subjected to three practical constraints as mentioned
in Chapter 4.
For single-tube PDEs, it has been demonstrated in Chapter 4 that there exists an
optimum frequency for a given configuration, which is attributed to the following two
conflicting effects: 1) more reactants can be recharged into the detonation tube at a lower
cycle frequency and 2) an exceedingly large refilling time associated with a low-
frequency operation may cause chamber overfilling and thus degrade the performance. A
205
11/23/2003
similar conclusion is found for multitube PDEs. Figure 5.17b indicates that the lower
frequency of 250 Hz (
cycle
= 4 ms) offers not only a wide operating range but also higher
performance than the higher frequency of 333 Hz (
cycle
= 3 ms). The optimum frequency
is thus less than 333 Hz. However, further parametric studies are needed to determine
this.
A comparison between the triple-tube and single-tube results demonstrates the
superiority of the multitube design. For the 333 Hz operation, the triple-tube PDE has
performances of about 4% to 5% higher than those of single-tube PDEs. For the 250 Hz
operation, the triple-tube PDE possesses both a wider operation range and higher
performance. The lower bound of
close
related to the combustor overfilling in the single-
tube PDE doesnt appear in the triple-tube PDE. Instead, the performance increases as
close
decreases until the lower bound related to the inlet over-pressurization is reached.
This is attributed to the fact that in the multitube PDE, the pressure in a detonation tube
within the refilling process can be raised by the detonation-degenerated shock waves
from other tubes, as discussed in Sec. 5.2.1. In contrast, the chamber pressure during the
refilling process of the single-tube PDE may be quite low, especially for cases with very
long blowdown process, thus resulting in high refilling velocity and chamber overfilling.
The highest specific impulse obtained for the current design is 3870 s, at an operation
timing with
cycle
of 4 ms and
close
of 1.8 ms. This performance number slightly exceeds
its ramjet counterpart of 3866 s with optimum nozzle flow expansion and can be further
improved by optimizing the operation timing.

206
11/23/2003

close
, ms
s
p
e
c
i
f
i
c
t
h
r
u
s
t
F
s
p
,
m
/
s
1.5 2 2.5 3 3.5
0
200
400
600
800
1000
1200
4 ms
cycle
= 3 ms
single tube
triple tube

close
, ms
I
s
p
,
s
1.5 2 2.5 3 3.5
1000
2000
3000
4000
4 ms
triple tube
cycle
= 3 ms
single tube

Fig. 5.17 Effect of valve close-up time on (a) air-based specific thrust and (b) fuel-based
specific impulse;
purge
= 0.1 ms, stoichiometric H
2
= 2.1.

(a)
(b)
207
11/23/2003
5.2.4 Effect of System Geometry
In addition to the operation timing, the system geometry represents another
important factor that affects both the flow dynamics and the propulsive performance of a
multitube PDE. The system geometry includes many independent variables, such as tube
length, nozzle length, nozzle throat and exit areas, and so on. Generally, the study of the
geometrical effect requires much effort due to the need to regenerate the computational
grid. In Chapter 4 for single-tube PDEs, the effect of the CD nozzle configuration has
been investigated by varying the nozzle length and throat area. In this chapter, the effect
of a free volume located between the multiple tubes and the common nozzle, as depicted
in Fig. 5.3, is considered. The purpose of adding such a free volume is to provide a
buffer region to smooth the flowfield and improve the operation steadiness.
Figure 5.18 shows the time evolution of the density-gradient field during the fifth
cycle. The operation timing remains identical to that for the baseline case. The refilling
lengths for the bottom, middle, and top tubes are 33.6, 44.3, and 34.8 cm, respectively.
The averaged refilling Mach numbers are 0.88, 1.0, and 0.99, respectively. The flowfield
characteristics within the detonation tubes and the external region are quite similar in
nature to those without free volume. Within the free volume, however, the flow exhibits
very complicated structures. In addition to the diffraction and reflection of the shock
waves from the three detonation tubes, the interactions of the supersonic exhaust flow
from the detonation tubes with the subsonic flow within the corner regions near the
conjunction of the combustor and the convergent nozzle wall lead to the formation of
standing-like shock waves. These shock waves may be quite strong. For example, the
shock wave located at the middle of the free volume has a pressure ratio of as high as 5.
208
11/23/2003

(
cycle
= 3 ms,
close
= 2.1 ms, and
purge
= 0.1 ms), with free volume

These shock waves represent certain internal flow loss and have no contribution to raise
the detonation tube pressure. Only those shock waves that propagate back into the tubes
have such contribution. The interactions of the exhaust flows and the corner zones also
result in the formation of recirculation zones, as evidenced in Fig. 5.19, in which two
large and one small recirculation zones between the bottom tube exit, and the upper
corner and one large recirculation zone between the bottom tube exit and lower corner
(a)
(b)
(c)
(d)
(e)
(f)
top: blowdown
mid: purging
bot: detonation
top: blowdown
mid: refilling
bot: blowdown
top: purging
mid: detonation
bot: blowdown
top: blowdown
mid: blowdown
bot: refilling
top: detonation
mid: blowdown
bot: purging
top: refilling
mid: blowdown
bot: blowdown
209
11/23/2003
zone can be clearly seen. This kind of flow structure further increases the internal flow
loss and is thus detrimental to the propulsive performance. A plug nozzle may be a better
candidate for multitube PDE since it permits certain tube interactions through moving
shock waves within the detonation tubes while keeping away the standing-like shock
waves from the internal flowfield.

Fig. 5.19 Pressure contours and streamlines at t = 13.50 ms (
cycle
= 3 ms,
close
= 2.1 ms,
and
purge
= 0.1 ms), with free volume

Fig. 5.20 shows the instantaneous axial thrust during a steady periodic cycle for
both configurations. The second peak related to the middle tube detonation is slightly
reduced with the addition of the free volume, while the first and third peaks related to the
other two tubes remain the same. Therefore, the improvement in operation steadiness by
the free volume is very limited.
The specific impulse and specific thrust are 3372 s and 855 m/s, respectively.
These values are lower than those without free volume and are even lower than those of
the single-tube PDE, demonstrating a negative effect of the free volume on the propulsive
210
11/23/2003
performance. The performance losses are identified similarly as in Sec. III.B. The
average refilling Mach number is 0.96, slightly larger than that without free volume. The
imperfect nozzle expansion loss, flow divergence loss, and internal flow loss are 2.2%,
2.3%, and 15.5%, respectively. The free volume thus reduces the imperfect nozzle
expansion loss. However, it leads to a significant increase in the internal flow loss,
mainly due to the complicated shock waves and the recirculation zones within the free
volume as aforementioned.

Fig. 5.20 Instantaneous axial thrust during the fifth cycle (
cycle
= 3 ms,
close
= 2.1 ms,
and
purge
= 0.1 ms)
5.3 Summary and Conclusions
The thrust chamber dynamics and propulsive performance of multitube
airbreathing PDEs with repetitive operations were studied by means of two-dimensional
numerical simulations. The system under consideration includes a supersonic inlet, an air
211
11/23/2003
manifold, a rotary valve, a triple-tube combustor, and a convergent-divergent nozzle,
which is designed for a flight altitude of 9.3 km and a flight Mach number of 2.1. The
flow evolution and tube interactions were studied in detail and the various loss
mechanisms were identified. The effects of operating timing such as the cycle time and
the valve close-up time were investigated systematically. Similar trends as those for
single-tube PDEs were found. There exists an optimum frequency, and at a given
frequency, a smaller valve close-up time increases the performance in most cases.
Comparison with the single-tube results also demonstrates that the multitube design
improves the engine performance in terms of specific impulse, operation steadiness, and
timing range. The effect of the system geometry was partially assessed by considering a
free volume between the detonation tubes and the common nozzle. Results indicate that
the free volume helps to reduce the imperfect nozzle expansion loss and improve the
engine steadiness. However, it also induces more complicated shock waves and increases
the internal flow loss. The overall effect is a decrease in the propulsive performance.
The highest specific impulse obtained for the current triple-tube PDE is 3870 s at
an operation frequency of 250 Hz. It exceeds its ramjet counterpart of 3866 s with
perfect nozzle flow expansion and can be further improved by optimizing the operation
timing and the system geometry. However, it should be also noted that some loss
mechanisms not considered in the present analysis, such as the energy required for
detonation initiation and the flow losses associated with the inlet isolator, rotary valve,
and air distributor, may render the PDE less attractive. These losses should be counted in
the future to provide more accurate comparison between PDEs and conventional engines.
212
11/23/2003
The flow evolution and the system geometry effect analyses suggest that the
benefit of the convergent-divergent nozzle for a multitube PDE is not as apparent as that
for a single-tube PDE. A plug nozzle may be a better candidate for multitube PDE than
the internal flow nozzles since it permits certain tube interactions through the moving
shock waves induced by the detonations in the neighboring tubes to raise the chamber
pressure while avoiding the standing-like shock waves within the internal flowfield that
are detrimental to performance.

213
11/23/2003
Chapter 6
Summary and Future Work
6.1 Summary
The present work studied the flow dynamics and system performance of
airbreathing pulse detonation engines (PDEs) with a stoichiometric hydrogen/air mixture.
The system under consideration includes a supersonic inlet, an air manifold, a rotary
valve, a single-tube or multitube combustor, and a convergent-divergent (CD) nozzle,
which is designed for the flight condition with a flight altitude of 9.3 km and a flight
Mach number of 2.1.
Axisymmetric two-dimensional simulations are first conducted for a mixed-
compression supersonic inlet to investigate the steady-state inlet flow dynamics and the
response of the inlet shock system to downstream disturbances. Two different back
pressures (2.1 atm and 2.2 atm) are carefully chosen such that the inlet operates at a
supercritical condition to provide a sufficient shock stability margin. The response of the
inlet shock system to downstream disturbances is studied by imposing periodic pressure
oscillations at the exit plane. A wide range of fluctuation frequency and amplitude are
investigated. Important phenomena of concern include oscillations of mass flow rate,
pressure recovery, flow distribution, and terminal shock displacement. In general, the
acoustic response of the inlet flow increases with increasing amplitude of the imposed
oscillation, but decreases with the frequency.
214
11/23/2003
A quasi-one-dimensional Euler code is then established based on the recently
developed Space-Time CE/SE method, which circumvents the deficiencies of existing
numerical methods for treating detonation waves and shock discontinuities, to study the
flow dynamics and propulsive performance of an airbreathing PDE with a CD nozzle.
The chemical kinetics is simulated using a simple one-progress variable model calibrated
with the NASA CEA code. The flow dynamics is clearly depicted through numerically
obtained x-t diagrams for both the first and the steady periodic cycles. Parametric studies
are conducted by varying the cycle time (or frequency) and valve close-up time. Four
operation frequencies (200 Hz, 250 Hz, 333 Hz, and 400 Hz) are considered. At a given
frequency and purge time, a smaller valve close-up time (or a longer refilling time)
increases the performance in most cases. The lower bound of the valve close-up time is
subject to the constraints associated with inlet over-pressurization and combustion
overfilling. The results also demonstrate the existence of an optimum frequency for a
given engine configuration and flight condition.
Two-dimensional analyses are carried out for single-tube PDEs with or without
CD nozzles under multicycle operation in order to provide more detailed flow dynamics
and more accurate performance predictions. The two-dimensional code is efficiently
parallelized by implementing the message-passing-interface (MPI) library and a domain
decomposition technique. The resultant code is executed on an in-house cluster using 64
CPUs. A flow-path based performance prediction model is proposed to estimate the
upper performance limit of the current PDE system. The various performance loss
mechanisms, such as the imperfect nozzle expansion loss, flow divergence loss, and
internal flow loss, are identified. The internal flow loss, which is mainly associated with
215
11/23/2003
the shock interactions within the chamber, is one of the major factors that degrade the
performance of the current PDE system. The effects of operation timing including the
cycle time or frequency, valve close-up time, and purge time on the propulsive
performance are comprehensively studied. Results manifest the existence of an optimum
operation frequency for achieving the best performance margin. At a given frequency
and purge time, a smaller valve close-up time increases the performance in most cases.
On the other hand, for a given frequency and valve close-up time, a larger purge time
decreases the specific thrust and increases the specific impulse. The effects of nozzle
configurations in terms of nozzle length and throat area on the propulsive performance
are also examined. The throat area affects both the nozzle expansion process and the
flow dynamics within the chamber, thus exerting a much more significant effect than the
nozzle length. A smaller throat tends to increase the performance by raising the chamber
pressure. However, an exceedingly small throat renders negative effects due to the
related longer blowdown process and larger internal flow loss.
Finally, multitube airbreathing PDEs with repetitive operations are studied by
means of two-dimensional simulations. The combustor consists of three detonation
tubes. The flow evolution and tube interactions are carefully examined and the various
loss mechanisms are identified. The effects of operating timing such as the cycle time
and the valve close-up time were investigated systematically. Similar trends as those for
single-tube PDEs were found. Comparison with the single-tube results demonstrates that
the multitube design improves the engine performance in terms of specific impulse,
operation steadiness, and timing range. The effect of the system geometry is partially
assessed by considering a free volume between the detonation tubes and the common
216
11/23/2003
nozzle. Results indicate that the free volume helps to reduce the imperfect nozzle
expansion loss and improve the engine steadiness. However, it also induces more
complicated shock waves and increases the internal flow loss. The overall effect is a
decrease in the propulsive performance.
6.2 Major Contributions and Conclusions
(1) A two-dimensional, reacting, unstructured-grid-based code is developed based
on the Space-Time CE/SE method and is efficiently parallelized with the implementation
of the MPI library and the domain decomposition technique.
(2) The acoustic response of the inlet flow increases with increasing amplitude of
the imposed oscillation, but decreases with the frequency.
(3) For a given PDE configuration and flight condition, an optimum cycle
frequency exists for achieving the best performance.
(4) For a given frequency and purge time, a longer refilling period increases the
specific thrust of PDEs considered herein.
(5) A convergent-divergent (CD) nozzle can increase the propulsive performance
that can be obtained with only a straight tube by up to 25%.
(6) The length of the CD nozzle has a minor effect on the performance whereas
the throat area may modify the performance by up to 6%.
(7) A performance prediction model is proposed. The propulsive performances of
PDEs are about 20% lower than the predictions from this model, which is attributed to
217
11/23/2003
three losses: the divergence loss, the imperfect nozzle expansion loss, and the internal
flow loss.
(8) The internal flow loss, which is mainly associated with the shock waves and
their interactions within the internal flowfield, is the major factor that degrades the PDE
propulsive performance.
(9) The multitube design slightly improves the propulsive performance of a
single-tube PDE.
(10) The highest specific impulse obtained for the current triple-tube PDE is 3870
s at an operation frequency of 250 Hz, which slightly exceeds its ramjet counterpart of
3866 s with perfect nozzle flow expansion.
6.3 Recommendation for Future Work
The present inlet analysis relies on imposing periodic sinusoid pressure
oscillations at the exit plane to investigate the response of the inlet shock system to
downstream disturbances. Other types of oscillations may be considered to more
accurately represent the disturbance from the pulsed detonation process. On the other
hand, the current study didnt consider the flow dynamics within the acoustic cavity that
connects the inlet and the combustor. The interaction between the inlet, acoustic cavity,
and combustor should also be addressed.
In the current numerical simulations, the detonation is initiated directly by a small
spark region which has neglectable effect on the performance. However, in most
experiments, the detonation is initiated through a deflagration-to-detonation-transition
218
11/23/2003
(DDT) process. It might be useful to incorporate this DDT process into numerical
simulations so that more detailed comparisons with experiments could be made.
The present nozzle studies indicate that the convergent-divergent (CD) nozzle can
significantly increase the propulsive performance of single-tube PDEs. The effect of the
length and throat area of the CD nozzle is investigated. Other shape parameters such as
the convergent and divergent angles could be studied in the future to aid the PDE nozzle
design.
The multitube design is found to be able to improve the propulsive performance
of a single-tube PDE. Further research can be conducted to investigate the effect of the
number of detonation tubes and their relative positions on the flow dynamics and system
performance.
The multitube PDE studies also reveal that the benefit of the CD nozzle for a
multitube PDE is not as significant as that for a single-tube PDE. A plug nozzle may be a
better candidate for multitube PDEs than the internal flow nozzles since it permits certain
tube interactions through the moving shock waves induced by the detonations in the
neighboring tubes to raise the chamber pressure while avoiding the standing-like shock
waves within the internal flowfield that are detrimental to performance. This conjecture
needs to be approved through further studies on plug nozzles for multitube PDEs.
219
11/23/2003
Appendix A
Thermodynamic Cycle Efficiencies of
Brayton, Humphrey, and Ideal PDE Cycles
Figure A.1 shows the temperature-entropy diagram for the Brayton, Humphrey,
and ideal PDE cycles.

s
T
0
1
3
2
1a
2
2
3
3
H
u
m
p
h
r
e
y
(
v
=
c
o
n
s
t
.
)
B
r
a
y
t
o
n
(
p
=
c
o
n
s
t
.
)
P
D
E
(
d
e
t
o
n
a
t
i
o
n
)

Fig. A.1 Temperature-entropy diagram

The Brayton cycle consists of an isotropic compression process (0-1), a constant-
pressure heat addition process (1-2), an isotropic expansion process (2-3), and a constant-
pressure cooling process (3-0). The Humphrey cycle is similar to the Brayton cycle
except that the constant-pressure heat addition process (1-2) in the Brayton cycle is
replaced by a constant-volume heat addition process. In the ideal PDE cycle (Heiser and
Pratt, 2002), the heat addition process is modeled by a detonation process (1-1a-2), in
which state 2 is the CJ state. The thermodynamic cycle efficiency is defined as the
220
11/23/2003
ratio of the external work done by the cycle to the amount of thermal energy added by the
heat addition process, that is,
1
rej
q
q
= (A.1)
where q is the heat added during the heat addition process, and q
rej
is the heat released to
the surroundings during the cooling process. The expressions of for the Brayton,
Humphrey, and ideal PDE cycles are derived in the following subsections with the
assumption of constant gas properties.
A.1 Brayton cycle
For a Brayton cycle,

3 0
3 0 0 3 0
2 1 2 1 1 2 1
( )
/ 1
1 1 1 1
( ) / 1
rej p
p
q c T T
h h T T T
q h h c T T T T T

= = = =

(A.2)
and

1 1
3 3 2 1 3 2 1 2
0 2 1 0 2 1 0 1
T T T T p T p T
T T T T p T p T

| | | |
= = =
| |
\ . \ .
(A.3)
Therefore,

0
1
1
T
T
= (A.4)

221
11/23/2003
A.2 Humphrey cycle
For a Humphrey cycle,

3 0
3 0 0 3 0
2 1 2 1 1 2 1
( )
/ 1
1 1 1 1
( ) / 1
rej p
v
q c T T
h h T T T
q e e c T T T T T

= = = =

(A.5)
and

1
3 3 2 1 3 2 1
0 2 1 0 2 1 0
1 1 1
0 1 2 1 0 1 2 1 2
1 2 1 0 1 2 1 0 1
=
T T T T p T T
T T T T p T T
p p T T T T T T T
p p T T T T T T T

| |
= =
|
\ .
| | | | | |
= =
| | |
\ . \ . \ .
(A.6)
Therefore,

1/
0 2 1
1 2 1
( ) 1
1
1
T T T
T T T
(A.7)
On the other hand, since

2 1 2 1
( ) ( ) /
v p
q c T T c T T = = (A.8)
then can also be expressed as:

1/
0
1
(1 ) 1
1
T q
T q
+
=

(A.9)
where

1
/( )
p
q q c T (A.10)
222
11/23/2003
A.3 Ideal PDE cycle
For an ideal PDE cycle,

3 0
0 3 0 0 3 0
1 1 1
( )
/ 1 / 1
1 1 1 1
( )
rej p
p
q c T T
T T T T T T
q q T q c T T q

= = = =
(A.11)
The temperature ratio T
3
/T
0
can be derived as follows:

3 3 2 1
0 2 1 0
T T T T
T T T T
= (A.12)

1 1 1
3 3 0 1 0 2
2 2 1 2 1 1
T p p p T p
T p p p T p

| | | | | |
= = =
| | |
\ . \ . \ .
(A.13)
The property variations across the detonation wave have been derived in Chapter 1:

+
+
=
1
1
2
1
2 D
M
p
p
,
2
2
2
1
1
(1 )
D
D
T M
T M
| | +
=
|
+
\ .
(A.14)
where the detonation wave Mach number M
D
is determined from:

2 2
1 1
1 1 1 1
1 1
2 2 2 2
D
q q
M q q
RT RT

+ +
= + + = + + (A.15)
Substitution of Eqs. (A.13) and (A.14) into Eq. (A.12) leads to

1
2
3
2
0
1 1
1
D
D
T M
T M
+
| | +
=
|
+
\ .
(A.16)
Therefore, the ideal PDE thermodynamic cycle efficiency is

( 1) /
2
2
0
1
1 1
1
1
1
D
D
M
M T
T q

+
| | +

|
+
\ .
=
(A.17)
223
11/23/2003
Appendix B
Jacobian Matrices for Quasi-One- and Two-Dimensional Systems
B.1 Quasi-One-Dimensional System
The Jacobian matrices A

E
Q
and D

H
Q
are

2
3 2
0 1 0 0
1
[( 3) ] ( 3) 1 ( 1)
2
( / ) /
( 1)
( 1) / 2 ( 1)
0
t t
u u q
A
u e p e p
u qu
u u
uZ Z u

(
(
(

(
=
(
+ +
(
+
(
(

(B.1)

2
31 32 33 34
41 42 43 44
0 0 0 0 0 1 0 0
0 0 0 0 2 0 0
1
0 0 0 0
0
u u
dA
D
A A A A A dx
d d d d uZ Z u
( (
( (
( (
= +
( (
( (

(B.2)
where A in Eq. (B.2) represents the area, and

2
41 43
42 43
2
43 4
44 43
/
[ ]
( 1)
exp( / )
t
a
a
d d e u qZ
d d u
d h E p
d d q K E p
=
=
=
=
(B.3)
224
11/23/2003
B.2 Two-Dimensional System
Jacobian matrices A

E
Q
, B

F
Q
, and D

H
Q
are

2 2
2 2 2
0 1 0 0 0
1
[( 3) ( 1) ] ( 3) ( 1) 1 ( 1)
2
0 0
( / ) /
( 1) ( 1)
( 1) ( ) / 2 ( 1)
0 0
t t
u v u v q
A uv v u
u e p e p
uv u qu
u u v u
uZ Z u

(
(
(
(
+ (
(
(
(
=
(
(
+ +
(

(
+ +
(
(
(
(

(B.4)

2 2
2 2 2
0 0 1 0 0
0 0
1
[( 3) ( 1) ] ( 1) ( 3) 1 ( 1)
2
( / ) /
( 1) ( 1)
( 1) ( ) / 2 ( 1)
0 0
t t
uv v u
B v u u v q
v e p e p
uv v qv
v u v v
vZ Z v

(
(
(
(
(
(
(
(
= +
(
(
+ +
(

(
+ +
(
(
(
(

(B.5)
225
11/23/2003

(
(
(
(
(
(
=
55 54 53 52 51
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
d d d d d
D (B.6)
where

2 2
51 54
52 54
53 54
2
54 5
55 54
[ ( ) ]
( 1) /
exp( / )
t
a
a
d d e u v qZ
d d u
d d v
d h E p
d d q K E p
= +
=
=
=
=
(B.7)
226
11/23/2003
Appendix C
Procedures of Building x-t Diagram
The common way to obtain the x-t diagram is the Lagrangian approach, in which
a number of points are placed in the flowfield and move at characteristic velocities u,
u+c, and uc. A major drawback of this approach is the large cumulative error during the
tracing process. In addition, special handling is needed when the points move out of the
computational domain.
In this thesis, an Euler approach was proposed. The basic idea of this approach is
that the characteristic lines are equivalent to the streamlines in the (x, t) plane with a
characteristic velocity in the x direction and a unit velocity in the t direction. For
example, since
/ dx dt u = , / 1 dt dt = (C.1)
the characteristic lines of / dx dt u = are equivalent to the streamlines in the (x, t) plane
with velocity (u, 1). The x-t diagram is constructed based on the following procedures:
1) At each time step, output the following information to a file at every i
skip
(for
example, i
skip
= 10) grid points: x, t, u, u+c, uc, 1.d0.
2) Use Tecplot to draw streamlines in the (x, t) plane with velocities of (u, 1), (u+c,
1), and (uc, 1). Several trials may be needed in placing the streamlines to
determine the Taylor wave, the contact surfaces, the detonation wave front, and so
forth.
227
11/23/2003
Appendix D
Space-Time CE/SE Method for Two-Dimensional Problems
A two-dimensional, unstructured triangular mesh Euler solver for non-reacting
flows based on the space-time conservation element and solution element (CE/SE)
method has been developed by Wang and Chang (1999). The ideas and solution
procedure are similar to those for one-dimensional problems described in Sec. 3.1.2. For
the purpose of completeness, this appendix describes the space-time CE/SE method,
including the basic solution procedure, the extended scheme, and source term handling
for two-dimensional problems.
D.1 Governing Equations and Their Integral Form
For clarity, the two-dimensional governing equations without source term is first
considered, which is expressed in a vector form as below:

0

Q E F
t x y

+ + =

(D.1)
where Q is the dependent variable vector, E and F are convective flux vectors. The
governing equations with source term will be considered in Sec. D.5.
Let x
1
= x, x
2
= y, x
3
= t be the coordinates of a three-dimensional Euclidean space
E
3
. Then the integral form of Eq. (D.1) in space-time E
3
is

( )
0
S V
d =
h s
v
(D.2)
228
11/23/2003
where h = (E, F, Q) is the flux vector in E
3
, S(V) the boundary of an arbitrary space-time
region V in E
3
, and ds the surface vector of an infinitesimal surface in E
3
.
D.2 Triangular Mesh, Conservation Element, and Solution Element
The present 2-D CE/SE solver is constructed for an unstructured triangular mesh.
Figure D.1 shows an arbitrary triangular cell (V
1
V
2
V
3
) and its three neighbor cells.
Points C, C
1
, C
2
, and C
3
, as marked by the circles in Fig. D.1, are the centroids of the
center cell and the three neighbor cells, respectively. The centroids of the neighbor cells
and the vertices of the center cell form a hexagon (C
1
V
3
C
2
V
1
C
3
V
2
). The centroid of this
hexagon, point S, is referred to as the solution point of the center cell. Each cell is
associated with a unique solution point (marked by cross in Fig. D.1) at which the
dependent variable vector and its spatial derivatives are defined and solved. In general
the solution points dont coincide with the cell centroids except for the uniform mesh.
The hexagon C
1
V
3
C
2
V
1
C
3
V
2
is composed of three quadrilaterals: C
1
V
3
CV
2
, C
2
V
1
CV
3
,
and C
3
V
2
CV
1
. The centroids of these three quadrilaterals are E
1
, E
2
, and E
3
, respectively.
The conservation element and solution element associated with cell j at the nth
time level are shown schematically in Fig. D.2. The superscripts (') and (") are used to
denote the n1 and n+1th time levels, respectively. The conservation element CE(j, n) is
defined by the hexagonal cylinder C
1
V
3
C
2
V
1
C
3
V
2
C
1
'V
3
'C
2
'V
1
'C
3
'V
2
'. It is composed of
three sub-CEs: CE
r
(j, n), r = 1,2,3. Each sub-CE is a quadrilateral cylinder in the E
3
-
space. The solution element SE(j, n) is the union of three vertical plane segments and a
horizontal plane segment.
229
11/23/2003
C
C
1
C
2
C
3
V
1
V
2
V
3
S
3
S
2
S
1
E
3
E
2
E
1
P
3
P
1
P
2
S

Fig. D.1 A triangular cell and its three neighbours

C
C
1
C
2
C
3
V
1
V
2
V
3
C'
C
1
'
C
2
'
C
3
'
V
1
'
V
2
'
V
3
'
t

V
1
"
V
1
'
V
2
'
V
3
'
V
1
V
2
V
3
V
2
"
V
3
"
C
C
1
C
2
C
3
C'
C"
t
t

(a) conservation element CE(j, n) (b) solution element SE(j, n)

Fig. D.2 Schematic of CE and SE associated with solution point (j, n)

230
11/23/2003
D.3 Basic Solution Procedure
For any point (x, y, t) SE(j, n), let Q(x, y, t), E(x, y, t), and F(x, y, t),
respectively, be approximated by Q
*
(x, y, t; j, n), E
*
(x, y, t; j, n), and F
*
(x, y, t; j, n)
through the following first-order Taylors expansions,

*
( , , ; , ) ( ) ( ) ( ) ( ) ( ) ( )
n n n n n
j x j j y j j t j
Q x y t j n Q Q x x Q y y Q t t = + + + (D.3)

*
( , , ; , ) ( ) ( ) ( ) ( ) ( ) ( )
n n n n n
j x j j y j j t j
E x y t j n E E x x E y y E t t = + + + (D.4)

*
( , , ; , ) ( ) ( ) ( ) ( ) ( ) ( )
n n n n n
j x j j y j j t j
F x y t j n F F x x F y y F t t = + + + (D.5)
where the label
n
j
) ( denotes the corresponding quantities evaluated at solution point (j, n)
whose coordinate is (x
j
, y
j
, t
n
). Also note that
n
j
E and
n
j
F are functions only of
n
j
Q , and
that
n
j x
E ) ( , ( )
n
y j
E , ( )
n
t j
E , ( )
n
x j
F , ( )
n
y j
F , and ( )
n
t j
F are functions of
n
j
Q ,
n
j x
Q ) ( , ( )
n
y j
Q ,
and
n
j t
Q ) ( :
, ( ) ( ) , ( ) ( ) , ( ) ( )
n n n n n n n n n n n n
j j j x j j x j y j j y j t j j t j
E A Q E A Q E A Q E A Q = = = = (D.6)
, ( ) ( ) , ( ) ( ) , ( ) ( )
n n n n n n n n n n n n
j j j x j j x j y j j y j t j j t j
F B Q F B Q F B Q F B Q = = = = (D.7)
where
E
A
Q
and
F
B
Q
are the Jacobian matrices of the convective flux vectors E

and F, respectively.
Moreover, let Q
*
, E
*
, and F
*
satisfy the differential governing equation at the
solution point (j, n),

* * *

0

Q E F
t x y

+ + =

(D.8)
The following relation can be obtained,
231
11/23/2003
( ) ( ) ( ) ( )
n n n n
t j x j y j x y j
Q E F AQ BQ = = + (D.9)
which means that
n
j t
Q ) ( is a function of
n
j
Q ,
n
j x
Q ) ( , and ( )
n
y j
Q . Therefore, there are
three sets of unknown variables at solution point (j, n), i.e.,
n
j
Q ,
n
j x
Q ) ( , and ( )
n
y j
Q .
These unknowns can be solved from the following three flux conservation equations:

*
( ( , ))
0, 1, 2, 3
r
S CE j n
d r = =
h s
v
(D.10)
where h
*
= (E
*
, F
*
, Q
*
). Details of the solving procedure are given below.
For each CE
r
(j, n) there are six faces, as indicated in Fig. D.2. Let
( ) r
j
S and
( , ) r b
j
S ,
respectively, denote the surface vectors of the top and bottom faces, and
( , ) r i
j
S , i = 1,2,3,4,
denote the surface vectors of the four lateral faces, with i = 1 and 2 for the two lateral
faces associated with point C
r
, and i = 3 and 4 for the two lateral faces associated with
point C. All these surface vectors are knowns defined by the geometires. For example,
the six surface vectors for CE
1
(j, n) are:

(1) (1)
(0, 0, 1)
j j
S = S (D.11)

(1, ) (1)
(0, 0, 1)
b
j j
S = S (D.12)

2 1 2 1
(1,1) (1,1) (1,1)
( , , 0) ( , , 0)
V C V C
j jx jy j j j j
S S t y y x x = = + S (D.13)

3 3 1 1
(1,2) (1,2) (1,2)
( , , 0) ( , , 0)
V V C C
j jx jy j j j j
S S t y y x x = = + S (D.14)

2 2
(1,3) (1,3) (1,3)
( , , 0) ( , , 0)
V V C C
j jx jy j j j j
S S t y y x x = = + S (D.15)

3 3
(1,4) (1,4) (1,4)
( , , 0) ( , , 0)
V V C C
j jx jy j j j j
S S t y y x x = = + S (D.16)
On the other hand, because h
*
is linear in x, y, and t within each SE, the flux of h
*

leaving CE
r
(j, n) through any one of the six faces is equal to the product of the vector h
*

232
11/23/2003
evaluated at the centroid of the face and the surface vector of that face. Let point F
ri
(i =
1,2,3,4) denotes the centroid of the ith lateral face of CE
r
(j, n). Then the flux
conservation equation (D.10) can be expressed out as

'
4
* * * * ( ) ( ) ( , ) ( , )
1
( ) 0, 1, 2, 3
ri ri r r
F F E E r r r i r i
j j jx jx
i
Q S Q S E S F S r
=
+ + = =
(D.17)
where

*
( ) ( ) ( ) ( )
r r r
E E E n n n
j x j j j y j j j
Q Q Q x x Q y y = + + (D.18)

'
*
1 1 1
( ) ( ) ( ) ( )
r r r
r r r r r
E
E E n n n
j x j j j y j j j
Q Q Q x x Q y y

= + + (D.19)

*
( ) ( ) ( ) ( ) ( ) / 2, 3, 4
ri ri ri
F F F n n n n
j x j j j y j j j t j
E E E x x E y y E t i = + + = (D.20)

*
( ) ( ) ( ) ( ) ( ) / 2, 3, 4
ri ri ri
F F F n n n n
F F F x x F y y F t i = + + = (D.21)

* 1 1 1 1
( ) ( ) ( ) ( ) ( ) / 2, 1, 2
ri ri ri
r r r r r
r
F F F n n n n
E E E x x E y y E t i

= + + + = (D.22)

* 1 1 1 1
( ) ( ) ( ) ( ) ( ) / 2, 1, 2
ri ri ri
r r r r r
r
F F F n n n n
F F F x x F y y F t i

= + + + = (D.23)
Here the index j
r
represents cell js rth neighbor with point C
r
as its centroid. Using Eqs.
(D.6) and (D.7), Eq. (D.17) can be further expressed in terms of Q,
x
Q , and
y
Q as

1
( ) ( ) , 1, 2, 3
r
x y n x y n
r r x r y j r r x r y j
L Q L Q L Q R Q R Q R Q r
+ + = + + = (D.24)
where

( ) ( ,3) ( ,4)
( )
r r r
r j
L S I K K = + + (D.25)

4
( ) ( , ) ( ,3) ( ,4)
3
( ) ( ) ( )
2
ri r
F E x r r i r r
r j j j j j
i
t
L x x S I x x K K K A
=
= + + +
(D.26)

4
( ) ( , ) ( ,3) ( ,4)
3
( ) ( ) ( )
2
ri r
F E y r r i r r
r j j j j j
i
t
L y y S I y y K K K B
=
= + + +
(D.27)
233
11/23/2003

( ) ( ,1) ( ,2)
( )
r r r
r j
R S I K K = + (D.28)

2
( ) ( , ) ( ,1) ( ,2)
1
( ) ( ) ( )
2
ri r
r r
F E x r r i r r
r j j j j j
i
t
R x x S I x x K K K A
=
= + +
(D.29)

2
( ) ( , ) ( ,1) ( ,2)
1
( ) ( ) ( )
2
ri r
r r
F E y r r i r r
r j j j j j
i
t
R y y S I y y K K K B
=
= + +
(D.30)
and
( , ) r i
K is defined as:

( , ) 1 ( , ) 1 ( , )
, 1, 2
r r
r i n r i n r i
j jx j jy
K A S B S i

= + = (D.31)

( , ) ( , ) ( , )
, 3, 4
r i n r i n r i
j jx j jy
K A S B S i = + = (D.32)
It is easy to prove that,

3
(1) (2) (3) ( ,3) ( ,4)
1 2 3
1
( ) ( )
r r
j j j j
r
L L L S S S I K K S I
=
+ + = + + + + =
(D.33)

1 2 3
0
x x x
L L L + + = (D.34)

1 2 3
0
y y y
L L L + + = (D.35)
The dependent variable vector
n
j
Q can be solved by summing Eq. (D.24) over r = 1,2,3,

3
1
1
1
( )
r
n x y n
j r r x r y j
r j
Q R Q R Q R Q
S

=
= + +
(D.36)
And the spatial derivatives
n
j x
Q ) ( and ( )
n
y j
Q can then be obtained as:

1
1 1 1 1
2 2 2 2
( )
( )
n n x y
x j j
n n x y
y j j
Q RHS L Q L L
Q RHS L Q L L
| | | | | |
=
| | |
| |

\ . \ . \ .
(D.37)
where

1
( )
r
x y n
r r r x r y j
RHS R Q R Q R Q

= + + (D.38)
234
11/23/2003
D.4 The Extended Scheme
The basic solution procedure is usually referred to as the Euler a scheme (Wang
and Chang, 1999) which needs no free parameter and has nearly no numerical
dissipation. This a scheme, however, is not equipped to suppress numerical wiggles that
generally appear within the region with large gradient or discontinuity. Moreover, it is
not computationally efficient since a large part of the computation time is spent on
calculating the derivatives of the mesh variable (
x
Q and
y
Q ) as indicated by Eq. (D.37).
As remedies for these deficiencies, several extended schemes, such as the Euler a- and
a--- schemes were then proposed (Chang, 1995; Wang and Chang, 1999). In this
thesis, only one parameter, , was introduced to control the numerical dissipation. This
scheme, as described in the following paragraphs, is a special case of the a--- scheme
with 1/ 2 = and 1 = (Wang and Chang, 1999). Several remarks on the parameters ,
, and were given by Chang et al. (1999).
To proceed, some geometry-related preliminaries are first introduced. Let be a
linear function of x and y in a spatial plane P
1
P
2
P
3
, and
1
,
2
, and
3
be the function
values at point P
1
, P
2
, and P
3
, respectively, then

1 3 1 3 1 3
( ) ( )
x y
x x y y = + (D.39)

2 3 2 3 2 3
( ) ( )
x y
x x y y = + (D.40)
The partial derivatives
x
and
y
can thus be solved as:
/
x x
= (D.41)
/
y y
= (D.42)
235
11/23/2003
where

3 2 3 2
3 1 3 1
y y x x
y y x x

= (D.43)

1 3 1 3
2 3 2 3
x
y y
y y

=

(D.44)

1 3 1 3
2 3 2 3
y
x x
x x

=

(D.45)
With the above preliminaries, construct plane P
1
P
2
P
3
so that the solution point S is
its centroid. The spatial coordinates of point P
r
are

r
3
P
1
1
( )
3
r r
j j j
r
x x x x
=
= +

(D.46)

r
3
P
1
1
( )
3
r r
j j j
r
y y y y
=
= +

(D.47)
The variable Q at point P
r
is evaluated as

r r r
r r
P P P *
P P 1 1 1 1
( , , ; , 1)
1
( ) ( ) ( ) ( ) ( )
2
r r r r r r
n
r
n n n n
j x j j y j j t j
Q Q x y t j n
Q Q x x Q y y Q t

=
= + + +
(D.48)
Let planes #1, 2 and 3, respectively, be the planes containing SP
2
P
3
, SP
3
P
1
and SP
1
P
2
,
then Q
x
, and Q
y
on plane #r can be evaluated as

( ) ( ) ( )
( )
r n r r
x j x
Q = (D.49)

( ) ( ) ( )
( )
r n r r
y j y
Q = (D.50)
where

2 2
3 3
P
(1)
P
P
j j
P
j j
x x y y
x x y y

=

(D.51)
236
11/23/2003

2 2
3 2
P
(1)
P
P n
j j
x P n
j j
Q Q y y
Q Q y y

=

(D.52)

2 2
3 3
P P
(1)
P P
n
j j
y n
j j
x x Q Q
x x Q Q

=

(D.53)
and
( ) r
,
( ) r
x
, and
( ) r
y
(r = 2,3) are defined by replacing the points P
2
and P
3
in the above
equations with P
3
and P
1
, and P
1
and P
2
, respectively.
With these
( ) r
x
Q and
( ) r
y
Q available,
x
Q and
y
Q can be obtained through a simple
arrange,

3
( )
1
1
( ) ( )
3
n r n
x j x j
r
Q Q
=
=

(D.54)

3
( )
1
1
( ) ( )
3
n r n
y j y j
r
Q Q
=
=

(D.55)
Alternatively, this simple average can be further improved by the following weighted
average:

(1) (2) (3)
2 3 3 1 1 2
2 3 3 1 1 2
( ) ( ) ( )
( )
( ) ( ) ( )
n
n
x x x
x j
j
Q Q Q
Q

+ +
=
+ +
(D.56)

(1) (2) (3)
2 3 3 1 1 2
2 3 3 1 1 2
( ) ( ) ( )
( )
( ) ( ) ( )
n
y y y n
y j
j
Q Q Q
Q

+ +
=
+ +
(D.57)
where

( ) 2 ( ) 2
( ) [( ) ] [( ) ]
n r n r n
r j x j y j
Q Q = + (D.58)
237
11/23/2003
and ( 0) is an adjustable constant to control the numerical dissipation and usually
chosen to be 1 or 2. Note that to avoid dividing by zero in Eqs. (D.56) and (D.57), a
small positive number such as 10
-60
may be added to the denominators.
The a--- scheme proposed by Wang and Chang (1999) solves the spatial
derivatives as follows:
2 ( ) ( )
a c a w c
x x x x x x
Q Q Q Q Q Q = + + (D.59)
2 ( ) ( )
a c a w c
y y y y y y
Q Q Q Q Q Q = + + (D.60)
where the superscript a, c and w, respectively, denote the spatial derivatives obtained by
the basic solution procedure with Eq. (D.37), the simple average with Eqs. (D.54) and
(D.55), and the weighted average with Eqs (D.56) and (D.57). The stability condition for
the a--- scheme is
0 1, 0, 0, and 1 CFL (D.61)
D.5 Source Term Treatment
The two-dimensional governing equations with source term is expressed as:

Q E F
H
t x y

+ + =

(D.62)
where Q is the dependent variable vector, E and F are convective flux vectors, and H the
source term vector. The integral form in space-time E
3
is

( ) S V V
d HdV =

h s
v
(D.63)
where dV is an infinitesimal volume in E
3
.
238
11/23/2003
The solution procedure is similar to that without source term. Only a few
equations need to be updated. Let Q
*
, E
*
, F
*
, and H satisfy the differential governing
equation at the solution point (j, n),

* * *

n
j
Q E F
H
t x y

+ + =

(D.64)
get a updated version of Eq. (D.9),
( ) ( ) ( ) ( )
n n n n n
t j j x j y j x y j
Q H E F AQ BQ DQ = = + (D.65)
In addition, the flux conservation equation, Eq. (D.10), is updated by

*
( ( , )) ( ( , ))
, 1, 2, 3
r r
S CE j n V CE j n
d HdV r = =

h s
v
(D.66)
The basic solution procedure for the dependent variable Q slightly differs depends on
whether the source term is treated explicitly of implicitly. In general, the explicit
treatment is used for non-stiff source term while the implicit treatment is used for stiff
source term.
D.5.1 Explicit Treatment of Source Term
If the source term is treated explicitly such that,

1 ( )
( ( , ))
, 1, 2, 3
r
r
n r
j j
V CE j n
HdV H S t r
= =
(D.67)
then the dependent variable Q is solved as

3
1
1
1
( )
r
n x y n
j r r x r y j
r j
Q R Q R Q R Q
S

=
= + +
(D.68)
where
x
r
R and
y
r
R , respectively, take the same formulations given by Eqs. (D.29) and
(D.30), and
r
R is updated from Eq. (D.28) as
239
11/23/2003

( ) ( ,1) ( ,2) ( ) ( ,1) ( ,2) 1
( ) ( ) / 2
r
r r r r r r n
r j j j
R S I K K t S I K K D

( ( = + + +

(D.69)
D.5.2 Implicit Treatment of Source Term
If the source term is treated implicitly such that,

( )
( ( , ))
, 1, 2, 3
r
n r
j j
V CE j n
HdV H S t r = =
(D.70)
then the dependent variable Q is solved from

3
1
1
1
( )
r
n x y n n n
j r r x r y j j j
r j
Q R Q R Q R Q tD Q
S

=
= + + +
(D.71)
Here
x
r
R and
y
r
R , respectively, take the same formulations given by Eqs. (D.29) and
(D.30), and
r
R is defined as

( ) ( ,1) ( ,2) ( ,1) ( ,2) 1
( ) ( ) / 2
r
r r r r r n
r j j
R S I K K t K K D

( ( = + +

(D.72)
Equation (D.71) can be solved either with fully implicit approach, such as the Newton-
Raphson iteration technique, or with linearized implicit approach which solves the
following linearized equation:

1
( )
n n
j j
I tD Q RHS
= (D.73)
where

3
1
1
1
( )
r
x y n
r r x r y j
r j
RHS R Q R Q R Q
S

=
= + +
(D.74)
For extremely stiff source term such as those arising from chemical reactions, the
conventional fractional-step technique (Oran and Boris, 2001) may be employed. This
240
11/23/2003
technique has been incorporated into the CE/SE scheme recently by Wu et al. (2000) and
Im et al. (2002) for detonation problems. In this technique, Eq. (D.71) is modified as

( )
1
( )
N
n m
j j
m
t
Q RHS DQ
N
=
= +

(D.75)
where N is the number of sub-time steps within t. The above equation can be solved
using the following procedure.

(1) (1)
(2) (1) (2)
( 1) ( 2) ( 1)
( 1)
( )
( )
( )
( )
j j
j j j
N N N
j j j
n N n
j j j
t
Q RHS DQ
N
t
Q Q DQ
N
t
Q Q DQ
N
t
Q Q DQ
N

= +
= +
= +
= +
# (D.76)
The Newton-Raphson iteration technique may be employed in each sub-time step.
241
11/23/2003
Appendix E
Reflecting Boundary Conditions on an Inviscid Solid Wall
This appendix describes the numerical reflecting boundary conditions on an
inviscid solid wall for the CE/SE method. The followings are adopted from Wang and
Chang (1999) with minor modifications.
Consider a boundary cell ABC that has an edge located on the solid wall, as
shown in Fig. E.1. The solid wall is aligned with an angle of from the x-axis. The
ghost cell ABD is a mirror image of the boundary cell ABC with respect to the wall.
The solution points of both cells are marked by cross. Assuming that the index of the
boundary cell ABC is j, the index of the ghost cell will be set as j.
x
y
A

(j)
(-j)
x
y
B
D
C
w
a
l
l
(x
0
, y
0
)

Fig. E.1 A boundary cell and its mirror image with respect to the solid wall

To proceed, a new spatial coordinate system - x y is introduced. Let the origin of
the - x y coordinate system, point A, be the point (x
0
, y
0
) in the x-y coordinate system.
These two coordinate systems then satisfy:

0
cos sin x x x y = (E.1)
242
11/23/2003

0
sin cos y y x y = + (E.2)
which can be written in a matrix form:

0
0
x x x
T
y y y

( (
=
( (

(E.3)
where

cos sin
sin cos
T

(
=
(

(E.4)
The inverse of matrix T is

1
cos sin
sin cos
T

(
=
(

(E.5)
Let Q denotes the dependant variable in the - x y coordinate system for the two-
dimensional Euler equations with one species equation,
| |
1 2 3 4 5
= , , , ,
T
Q q q q q q (E.6)
The slip solid wall boundary conditions are then constructed by assuming that, at any
time, the flowfields within the boundary cell (ABC) and the ghost cell (ABD) are
mirror images of each other in the - x y coordinate system. That is,
for the dependant variables,
( ) ( ) , ,2,4,5
i j i j
q q i
= =1 (E.7)
( ) ( ) , 3
i j i j
q q i
= = (E.8)
for the spatial derivatives,
, , 1, 2, 4, 5
i i i i
j j
j j
q q q q
i
x x y y

| | | |
| | | |
= = =
| | | |

\ . \ .
\ . \ .
(E.9)
243
11/23/2003
, , 3
i i i i
j j
j j
q q q q
i
x x y y

| | | |
| | | |
= = =
| | | |

\ . \ .
\ . \ .
(E.10)
In order to obtain the expressions in the x-y coordinate system, the relation
between Q and Q and between their spatial derivatives need to be derived first. It can be
proved that,
, 1,4,5
i i
q q i = = (E.11)

2 2 1
3 3
q q
T
q q
( (
=
( (

(E.12)
and

1
, 1, 4, 5
i i
i
i
q q
x x
T i
q
q
y
y
(
(
(
(

(
= = (

(
(
(
(

(E.13)

2 3 2 3
1
2 3 2 3
q q q q
x x x x
T
q q q q
y y y y

( (
( (

= ( (

( (
( (

(E.14)
With these relations available, the slip wall boundary conditions can then be
expressed as:
for Q,
( ) ( ) , ,4,5
i j i j
q q i
= =1 (E.15)

2 2
3 3
j j
q q
M
q q
( (
=
( (

(E.16)
for Q
x
and Q
y
,
244
11/23/2003
, 1, 4, 5
i i
i
i
j
j
q q
x x
M i
q
q
y
y
(
(
(
(

(
= = (
(
(
(
(

(E.17)

2 3 2 3
2 3 2 3
j j
q q q q
x x x x
M M
q q q q
y y y y

( (
( (

= ( (

( (
( (

(E.18)
where

1
1 0 cos2 sin2
0 1 sin2 cos2
M T T

( (
=
( (

(E.19)
245
11/23/2003
Appendix F
Determination of the Time Instant of the Arrival of the First
Rarefaction Wave on the Head End
Figure F.1 schematically shows the x-t diagram for detonation wave propagation
through a constant-area tube. The detonation wave is initiated at the head end (x = 0) and
propagates downstream, followed by the Taylor wave and a uniform region. As the
detonation wave arrives at the tube exit (x = L), a series of rarefaction waves are
generated and propagate upstream. The first rarefaction wave arrives at the head end at
time t
B
, which can be obtained analytically. The analysis detailed below basically
follows that given by Wintenberger et al. (2003).

x
t
E
t
L
unburned gas
uniform region 3 3
B
E
A
t
B
O
1 1
22
reflected rarefaction waves constant surface shock wave
t
a
i
l

o
f

T
a
y
l
o
r

w
a
v
e

Fig. F.1 Space-time diagram for detonation wave propagation

246
11/23/2003
Within the Taylor wave region (region OAE), the Riemann invariant relation
along the characteristic line from state 3 to the point (x, t) gives

3
2 2
0
1 1
c u c

=

(F.1)
where u and c are the velocity and sound speed at point (x, t), respectively. Since the
forward characteristic lines are straight, the relationship between x and t can be
represented by
c u
t
x
+ = (F.2)
Combination of the above two equations leads to the following solutions of u and c:

3 3
2 2
1 1
u x
c c t
=
+ +
(F.3)

3 3
1 2
1 1
c x
c c t
= +
+ +
(F.4)
And as given in Eq. (1.27), c
3
can be related to u
D
as

2
3 2
1
2
D
D
D
M
c u
M
+
= (F.5)
On the other hand, portion AE of the first reflected rarefaction wave can be
described by

dx
u c
dt
= (F.6)
Substituting Eqs. (F.3), (F.4), and (F.5) into the above equation leads to the following
ordinary differential equation:

2
2
3 2( 1)
0
1 ( 1)
D
D
D
dx x M
u
dt t M

+
+ + =
+ +
(F.7)
247
11/23/2003
With the initial condition:
at
D
x L t L u = = (F.8)
the solution to Eq. (F.7) can be obtained:

2( 1)
2 2
1
2 2
1 1
( )
( 1) ( 1)
D D
D
D D D
M M L
x t u t
M M u t
+
(
| | + +
(
= +
|
(

\ .
(

(F.9)
Considering that point E is the intersection of OE and AE, thus,

2( 1)
2 2
1
3 2 2
1 1
( )
( 1) ( 1)
D D
E E D E
D D D E
M M L
c t x t u t
M M u t
+
(
| | + +
(
= = +
|
(

\ .
(

(F.10)
Solving for
E
t ,

1
2
2( 1)
2
2( 1)
( 1)( 1)
D
E
D D
L M
t
u M
( +
=
(
+ +

(F.11)
Since the characteristic line OE and EB have the same wave speed magnitude c
3
,
B
t and
E
t satisfy the following relation:
2
B E
t t = (F.12)
Therefore,

1
2
2( 1)
2
2 2( 1)
( 1)( 1)
D
B
D D
L M
t
u M
( +
=
(
+ +

(F.13)
In addition, to obtain the used in Wintenberger et al. (2003), let

3
B
D
L L
t
u c
= + (F.14)
then it can be derived that,
248
11/23/2003

1
2
2( 1)
2 2
1 1 2( 1)
(1 ) 2 1
2 ( 1)( 1)
D
D D
M
M M

( +
= +
`
(
+ +

)
(F.15)
which is equivalent to but simpler than Eq. (20) in Wintenberger et al. (2003).
Considering that
2
1
D
M , Eqs. (F.13) and (F.15) can be further simplified with
the following approximation:

1
2( 1)
2 2
1
B
D
L
t
u
(
+

(F.16)

1
2( 1)
1 2
2( ) 1
2 1
r
r
r
r
(

(
+
(

(F.17)
Note that Eq. (F.16) was also given by Zitoun and Desbordes (1999).
249
11/23/2003
Bibliography
Aarnio, M.J., Hinkey, J.B., and Bussing, T.R.A., Multiple Cycle Detonation
Experiments during the Development of a Pulse Detonation Engine, AIAA Paper
1996-3263, July 1996.
Allgood, D., Gutmark, E., and Katta, V., Effects of Exit Geometry on the Performance
of a Pulse Detonation Engine, AIAA Paper 2002-0616, January 2002.
Belegundu, A.D., and Chandrupatla T.R., Optimization Concepts and Applications in
Engineering, Prentice-Hall, Inc., New Jersey, 1999.
Benedick, W., Guirao, C., Knystautas, R., and Lee, J.H.S., Critical Charge for Direct
Initiation of Detonation in Gaseous Fuel/Air Mixtures, Dynamics of Explosions,
Edited by Bowen, J.R. and Leyer, J.C., Vol. 106, Progress in Astronautics and
Aeronautics, AIAA, New York, 1986, pp. 181-202.
Berton, J.J., Divergence Thrust Loss Calculations for Convergent-Divergent Nozzles:
Extensions to the Classical Case, NASA TM-105176, 1991.
Bratkovich, T.E., and Bussing, T.R.A., A Pulse Detonation Engine Performance
Model, AIAA Paper 1995-3155, July 1995.
Broda, J.C., Conrad, C., Pal, S., Woodward, R.D., and Santoro, R.J., Experimental
Results on Air-Breathing Pulse Detonation Studies, Proceedings of 11th Annual
Symposium on Propulsion, Propulsion Engineering Research Center, Pennsylvania
State University, State College, PA, October 1999.
Brophy, C.M. and Netzer, D.W., Effects of Ignition Characteristics and Geometry on the
Performance of a JP-10/O2 Fueled Pulse Detonation Engine, AIAA Paper 1999-
2635, June 1999.
Brophy, C.M., Netzer, D.W., and Forster, L.D., Detonation Studies of JP-10 with
Oxygen and Air for Pulse Detonation Engine Development, AIAA Paper 1998-4003,
July 1998.
Brophy, C.M., Sinibaldi, J.O., and Damphousse, P., Initiator Performance for Liquid-
Fueled Pulse Detonation Engines, AIAA Paper 2002-0472, January 2002.
Brophy, C.M., Werner, L.S., and Sinibaldi, J.O., Performance Characterization of a
Valveless Pulse Detonation Engine, AIAA Paper 2003-1344, January 2003.
Bussing, T.R.A. and Pappas, G., An Introduction to Pulse Detonation Engines, AIAA
Paper 1994-0263, January 1994.
250
11/23/2003
Bussing, T.R.A. and Pappas, G., Pulse Detonation Engine Theory and Concepts,
Developments in High-Speed-Vehicle Propulsion Systems, Edited by Murthy, S.N.B.
and Curran, E.T., Vol. 165, Progress in Astronautics and Aeronautics, AIAA, Reston,
VA, 1996, pp. 421-472.
Bussing, T.R.A., A Rotary Valve Multiple Pulse Detonation Engine (RVMPDE),
AIAA Paper 1995-2577, July 1995.
Bussing, T.R.A., Hinkey, J.B., and Kaye, L., Pulse Detonation Engine Preliminary
Design Considerations, AIAA Paper 1994-3220, June 1994.
Cambier, J.L., and Adelman, H.G., Preliminary Numerical Simulations of a Pulsed
Detonation Wave Engine, AIAA Paper 1988-2960, July 1988.
Cambier, J.L. and Tegner, J.K, Strategies for Pulsed Detonation Engine Performance
Optimization, Journal of Propulsion and Power, Vol. 14, No. 4, 1998, pp. 489-498.
Chang, S.C., The Method of Space-Time Conservation Element and Solution Element -
A New Approach for Solving the Navier-Stokes and Euler Equations, Journal of
Computational Physics, Vol. 119, No. 2, 1995, pp. 295-324.
Chang, S.C., Loh, C.Y., Yu, S.T., Himansu, A., Wang, X.Y., Jorgenson, P., Robust and
Simple Non-Reflecting Boundary Conditions for the Space-Time Conservation
Element and Solution Element Method, AIAA Paper 1997-2077, June-July, 1997.
Chang, S.C., Wang, X.Y., and Chow, C.Y., The Space-Time Conservation Element and
Solution Element: A New High Resolution and Genuinely Multidimensional
Paradigm for Solving Conservation Laws, Journal of Computational Physics, Vol.
156, No. 1, 1999, pp. 89-136.
Chang, S.C., Wu, Y.H., Wang, X.Y., and Yang, V., Local Mesh Refinement in Space-
Time Conservation Element and Solution Element Method, Proceedings of 1st
International Conference on Computational Fluid Dynamics, Springer-Verlag,
Berlin, 2000.
Cooper, M. and Shepherd, J.E., The Effect of Nozzles and Extensions on Detonation
Tube Performance, AIAA Paper 2002-3628, July 2002.
Cooper, M., Jackson, S., Austin, J., Wintenberger, E., and Shepherd, J.E., Direct
Experimental Impulse Measurements for Detonation and Deflagrations, Journal of
Propulsion and Power, Vol. 18, No. 5, 2002, pp. 1033-1041.
Culick, F. E. C. and Rogers, T., The Response of Normal Shocks in Diffusers, AIAA
Journal, Vol. 21, No. 10, 1983, pp. 1382-1390.
251
11/23/2003
Daniau, D., Zitoun, R., Couquet, C., and Desbordes, D., Effects of Nozzles of Different
Length and Shape on the Propulsion Performance of Pulsed Detonation Engines,
High Speed Deflagration and Detonation, Fundamentals and Control, Edited by Roy,
G., Frolov, S., Netzer, D., and Borisov, A., ELEX-KM Publishers, Moscow, Russia,
2001, pp. 251-262.
Desbordes, D., Transmission of Overdriven plane detonations: Critical Diameter as a
Function of Cell Regularity and Size, Dynamics of Reactive Systems Part I: Flames;
Part II: Homogeneous Combustion and Applications, Edited by Kuhl, A.L., Bowen,
J.R., and Borisov, A.A., Vol. 114, Progress in Astronautics and Aeronautics, AIAA,
New York, 1988, pp.170-185.
Dunlap, R., Brehm, R.L., and Nicholls, J.A., A Preliminary Study of the Application of
Steady State Detonative Combustion of a Reaction Engine, Jet Propulsion, Vol. 28,
No. 7, 1958, pp. 451-456.
Dyke, M.V., An Album of Fluid Motion, The Parabolic Press, Stanford, CA, 1997, pp.
145-148.
Ebrahimi, H.B. and Merkle, C.L., A Numerical Simulation of a Pulse Detonation Engine
with Hydrogen Fuels, Journal of Propulsion and Power, Vol. 18, No. 5, 2002, pp.
1042-1048.
Ebrahimi, H.B., Mohanraj, R., and Merkle, C.L., Modeling of Multi-Tube Pulse
Detonation Engine Operation, AIAA Paper 2001-3813, July 2001.
Ebrahimi, H.B., Mohanraj, R., and Merkle, C.L., Multilevel Analysis of Pulsed
Detonation Engines, Journal of Propulsion and Power, Vol. 18, No. 2, 2002, pp.
225-232.
Ebrahimi, H.B. and Merkle, C.L., Optimization of Multitube Pulse Detonation Engine
Configuration, AIAA Paper 2002-4228, July 2002.
Ebrahimi, H.B., Malo-Molina, F.J., and Merkle, C.L., Numerical Investigation of Multi-
Tube Pulse Detonation, AIAA Paper 2003-0718, January 2003.
Eidelman, S. and Yang, X., Analysis of the Pulse Detonation Engine Efficiency, AIAA
Paper 1998-3877, July 1998.
Eidelman, S., Grossman, W., and Lottati, I., Air-Breathing Pulsed Detonation Engine
Concept; A Numerical Study, AIAA Paper 1990-2420, July 1990.
Eidelman, S., Grossman, W., and Lottati, I., Review of Propulsion Applications and
Numerical Simulations of The Pulse Detonation Engine Concept, Journal of
Propulsion and Power, Vol. 7, No. 6, 1991, pp. 857-865.
252
11/23/2003
Falempin, F., Bouchaud, D., Forrat, B., Desbordes, D., and Daniau, E., Pulsed
Detonation Engine: Possible Application to Low Cost Tactical Missile and to Space
Launcher, AIAA Paper 2001-3815, July 2001.
Farinaccio, R., Harris, P., Stowe, R.A., and Akbar, R., Multi-Pulse Detonation
Experiments with Propane-Oxygen, AIAA Paper 2002-4070, July 2002.
Fickett, W., and Davis, W.C., Detonation Theory and Experiment, Dover Publications,
New York, 2000.
Fong K.K., and Nalim M.R., Gas Dynamics Limits and Optimization of Pulsed
Detonation Static Thrust, AIAA Paper 2000-3471, July 2000.
Frankey, B., Schauer, F., Bradley, R., and Hoke, J., Evaluation of a Hybrid-Piston
Pulsed Detonation Engine, AIAA Paper 2002-0474, January 2002.
Glassman, I., Combustion, Third Edition, Chapter 5, Academic Press, CA, 1996.
Goddard, R.H., Rotating Valve for Multiple Resonance Combustion Chambers, U.S.
Patent 2515644, July 18, 1950.
Green, J. E., Interactions between Shock Waves and Turbulent Boundary Layers,
Progress in Aerospace Sciences, Vol. 11, 1970, pp. 235-340.
Han, Z.Y. and Yin, X.Z., Shock Dynamics, Kluwer Academic Publishers/Science Press,
1993, pp. 50-63.
Harris, P.G., Farinaccio, R., and Stowe, R.A., The Effect of DDT Distance on Impulse
in a Detonation Tube, AIAA Paper 2001-3467, July 2001.
Harten, A., High Resolution Schemes for Hypersonic Conservation Laws, Journal of
Heiser, W.H. and Pratt, D.T., Thermodynamic Cycle Analysis of Pulse Detonation
Engines, Journal of Propulsion and Power, Vol. 18, No. 1, 2002, pp. 68-76.
Helman, D., Shreeve, R.P., and Eidelman, S., Detonation Pulse Engine, AIAA Paper
1986-1683, June 1986.
Hinkey, J.B., Bussing T.R.A., and Kaye, L., Shock Tube Experiments for the
Development of a Hydrogen-Fueled Pulse Detonation Engine, AIAA Paper 1995-
2578, July 1995.
Hinkey, J.B., Williams, J.T., Henderson, S.E., and Bussing, T.R.A., Rotary-Valved,
Multiple-Cycle, Pulse Detonation Engine Experimental Demonstration, AIAA Paper
1997-2746, July 1997.
253
11/23/2003
Hodge, B.K. and Koenig, K., Compressible Fluid Dynamics with Personal Computer
Applications, Prentice Hall, Englewood Cliffs, New Jersey, 1995, pp. 98-99, p. 187,
pp. 467-486.
Hoffman, N., Reaction Propulsion by Intermittent Detonative Combustion, Ministry of
Supply, Volkenrode Translation, 1940.
Im, K.S., Yu, S.T., Kim, C.K., Chang, S.C., and Jorgenson, P.C.E., Application of the
CE/SE Method to Detonation with Realistic Finite-Rate Chemistry, AIAA Paper
2002-1020, January 2002.
Kailasanath, K., Patnaik, G., and Li, C., Computational Studies of Pulse Detonation
Engines: A Status Report, AIAA Paper 1999-2634, June 1999.
Kailasanath, K., Review of Propulsion Applications of Detonation Waves, AIAA
Journal, Vol. 38, No. 9, 2000, pp. 1698-1708.
Kailasanath, K., A Review of PDE Research Performance Estimates, AIAA Paper
2001-0474, January 2001.
Kailasanath, K., A Review of Research on Pulse Detonation Engine Nozzles, AIAA
Paper 2001-3932, July 2001.
Kailasanath, K., Recent Developments in the Research on Pulse Detonation Engines,
AIAA Paper 2002-0470, January 2002.
Kailasanath, K., Recent Developments in the Research on Pulse Detonation Engines,
AIAA Journal, Vol. 41, No. 2, 2003, pp. 145-159.
Karypis, G. and Kumar, V., Multilevel K-Way Partitioning Scheme for Irregular
Graphs, Journal of Parallel and Distributed Computing, Vol. 48, No. 1, 1998, pp.
96-129.
Kawai, S. and Fujiwara, T., Numerical Analysis of 1st and 2nd Cycles of Oxyhydrogen
PDE, AIAA Paper 2002-0929, January 2002.
Kentfield, J.A.C., The Fundamentals of Idealized Air-Breathing Pulse-Detonation
Engines, Journal of Propulsion and Power, Vol. 18, No. 1, 2002, pp. 77-83.
Knystautas, R., Lee, J.H.S., Moen, I., and Wagner, H.G., Direct Initiation of Spherical
Detonation by a Hot Turbulent Gas Jet, Proceedings of 17th International
Symposium on Combustion, The Combustion Institute, Pittsburgh, 1979, pp. 1235-
1245.
254
11/23/2003
Korobeinikov, V.P., and Urtiew, P.A. (English-Edition Editor), Unsteady Interaction of
Shock and Detonation Waves in Gases, Hemisphere Publishing Corporation, 1986,
Chapter 6, pp. 80-96.
Korovin, L.N., Losev, A., Ruban, S.G., and Smekhov, G.D., Combustion of Natural Gas
in a Commercial Detonation Reactor, Fizika Gor. Vzryva, Vol. 17, No. 3, 1981, p.
86. (cited in Eidelman et al., 1991)
Krzycki, L.J., Performance Characteristics of an Intermittent-Detonation Device, U.S.
Naval Ordnance Test Station, China Lake, CA, NAVWEPS Report 7655, 1962.
Kuo, K., Principles of Combustion, Chapter 4, John Wiley and Sons, 1986.
Lee, J.H.S., Dynamic Parameters of Gaseous Detonations, Annual Review of Fluid
Mechanics, Vol. 16, 1984, pp. 311-336.
Lee, J.H.S. and Moen, I.O., The Mechanism of Transition from Deflagration to
Detonation in Vapor Cloud Explosions, Progress in Energy and Combustion
Science, Vol. 6, 1980, pp. 359-389.
Lee, S.Y., Conrad, C., Watts, J., Woodward, R., Pal, S., and Santoro, R.J., Deflagration
to Detonation Transition Study Using Simultaneous Schlieren and OH PLIF Images,
Li, C., and Kailasanath, K., A Numerical Study of Reactive Flows in Pulse Detonation
Engines, AIAA Paper 2001-3933, July 2001.
Li, C., and Kailasanath, K., Performance Analysis of Pulse Detonation Engines with
Partial Fuel Filling, AIAA Paper 2002-0610, January 2002.
Li, C., Kailasanath, K., and Patnaik, G., A Numerical Study of Flow Field Evolution in a
Pulsed Detonation Engine, AIAA Paper 2000-0314, January 2000.
Lieberman, D.H., Parkin, K.L., and Shepherd, J.E., Detonation Initiation by a Hot
Turbulent Jet for Use in Pulse Detonation Engines, AIAA Paper 2002-3909, July
2002.
Litchford, R.J., Development of a Gas-Fed Pulse Detonation Research Engine, AIAA
Paper 2001-3814, July 2001.
Lynch, E.D. and Edelman, R.B., Analysis of the Pulse Detonation Wave Engine,
Developments in High-Speed-Vehicle Propulsion Systems, Edited by Murthy, S.N.B.
and Curran, E.T., Vol. 165, Progress in Astronautics and Aeronautics, AIAA, Reston,
VA, 1996, pp. 473-516.
255
11/23/2003
Ma, F.H., Choi, J.Y., Wu, Y.H., and Yang, V., Modeling of Multitube Pulse Detonation
Engine Dynamics, Confined Detonations and Pulse Detonation Engines, Edited by
Roy, G., Frolov, S., Santoro, R., and Tsyganov, S., Torus Press, Moscow, Russia,
2003, pp. 233-248.
Mattingly, J.D., Elements of Gas Turbine Propulsion, McGraw-Hill, 1996, p. 805.
McBride, B.J. and Gordon, S., Computer Program for Calculation of Complex Chemical
Equilibrium Compositions and Applications, NASA Reference Publication 1311,
June 1996.
McManus, K., Furlong, E., Leyva, I., and Sanderson, S., MEMS-Based Pulse
Detonation Engine for Small-Scale Propulsion Applications, AIAA Paper 2001-
3469, July 2001.
Meyer, T.R., Hoke, J.L., Brown, M.S., Gord, J.R., and Schauer, F.R., Experimental
Study of Deflagration Enhancement Techniques in a H2/Air Pulsed-Detonation
Engine, AIAA Paper 2002-3720, July 2002.
Meyers, J.M., Lu, F.K., and Wilson, D.R., Performance Enhancements on a Pulsed
Detonation Engine, AIAA Paper 2003-1173, January 2003.
Mohanraj, R., and Merkle, C.L., A Numerical Study of Pulse Detonation Engine
Performance, AIAA Paper 2000-0315, January 2000.
Nicholls, J.A., Wilkinson, H.R., and Morrison, R.B., Intermittent Detonation as a
Thrust-Producing Mechanism, Jet Propulsion, Vol. 27, No. 5, 1957, pp. 534-541.
Oh, J. Y., Numerical Study of Steady and Oscillatory Flow Structures in an
Axisymmetric Supersonic Inlet, Ph.D. Thesis, Department of Mechanical
Engineering, The Pennsylvania State University, University Park, PA, August 1994.
Oran, E.S., and Boris, J.P., Numerical Simulation of Reactive Flow, Second Edition,
Cambridge University Press, 2001, p. 405.
Oppenheim, A.K., Manson, N., and Wagner, H.G., Recent Progress in Detonation
Research, AIAA Journal, Vol. 1, No. 10, 1963, pp. 2243-2252.
Rasheed, A., Tangirala, V., Pinard, P.F., Dean, A.J., Experimental and Numerical
Investigations of Ejectors for PDE Applications, AIAA Paper 2003-4971, July 2003.
Roache, P.J., Quantification of Uncertainty in Computational Fluid Dynamics, Annual
Review of Fluid Mechanics, Vol. 29, 1997, pp. 126-160.
Rodi, W., Experience with Two-Layer Models Combining the k- Model with a One-
Equation Model Near the Wall, AIAA Paper 1991-0216, January 1991.
256
11/23/2003
Sanders, S.T., Jenkins, T.P., and Hanson, R.K., Diode Laser Sensor System for Multi-
Parameter Measurements in Pulse Detonation Engine Flows, AIAA Paper 2000-
3593, July 2000.
Santoro, R.J., Lee, S.Y., Conrad, C., Brumberg, J., Saretto, S., Pal, S., and Woodward,
R.D., DDT Studies for Multicycle PDE Applications, Confined Detonations and
Pulse Detonation Engines, Edited by Roy, G., Frolov, S., Santoro, R.J., and
Tsyganov, S., Torus Press, Moscow, Russia, 2003, pp. 259-270.
Schauer, F., Stutrud, J., and Bradley, R., AFRLs In-House Research Pulse Detonation
Engine, Proceedings of 11th Annual Symposium on Propulsion, Propulsion
Engineering Research Center, Pennsylvania State University, State College, PA,
October 1999.
Schauer, F., Stutrud, J., and Bradley, R., Detonation Initiation Studies and Performance
Results for Pulsed Detonation Engine Applications, AIAA Paper 2001-1129,
January 2001.
Schauer, F., Stutrud, J., Bradley, R., and Katta, V., AFRL/PRSC Pulse Detonation
Engine Research Program, Proceedings of 12th Annual Symposium on Propulsion,
Propulsion Engineering Research Center, Ohio Aerospace Institute, Cleveland, OH,
October 2000.
Seddon, J. and Goldsmith, E.L., Intake Aerodynamics, Second Edition, Blackwell
Science, Oxford, 1999, p. 137.
Shchelkin, K.I., Soviet Journal of Technical Physics, Vol. 10, 1940, pp. 823-827.
Shehadeh, R., Saretto, S., Lee, S.Y., Pal, S., and Santoro, R.J., Experimental Study of a
Pulse Detonation Engine Ejector, AIAA Paper 2003-4972, July 2003.
Shepherd, J.E., Pintgen, F., Austin, J.M., and Eckett, C.A., The Structure of the
Detonation Front in Gases, AIAA Paper 2002-0773, January 2002.
Shimo, M., Meyer, S.E., Heister, S.D., Weng, C.S., Ji, J., and Gore, J.P., An
Experimental and Computational Study of Pulsed Detonations in a Single Tube,
Sinibaldi, J.O., Brophy, C.M., and Robinson, L.J.P., Ignition Effects on Deflagration-to-
Detonation Transition Distance in Gaseous Mixtures, AIAA Paper 2000-3590, July
2000.
Sinibaldi, J.O., Brophy, C.M., Li, C., and Kailasanath, K., Initiator Detonation
Diffraction Studies in Pulsed Detonation Engines, AIAA Paper 2001-3466, July
2001.
257
11/23/2003
Skews, B.W., The Perturbed Region Behind a Diffracting Shock Wave, Journal of
Fluid Mechanics, Vol. 29, No. 4, 1967, pp. 705-719.
Skews, B.W., The Shape of a Diffracting Shock Wave, Journal of Fluid Mechanics,
Vol. 29, No. 2, 1967, pp. 297-305.
Smirnov, V.N. and Boichenko, A.P., Transition from Deflagration to Detonation in
Gasoline-Air Mixtures, Fizika Gor. Vzryva, Vol. 22, No. 2, 1986, pp. 65-67. (Cited
in Eidelman et al., 1991)
Spapart, P. R., Direct Simulation of a Turbulent Boundary Layer up to Re

= 1410,
Journal of Fluid Mechanics, Vol. 187, pp. 61-98, 1988.
Strehlow, P.A., Fundamentals of Combustion, Chapter 9, International Textbook
Company, 1968.
Sterling, J., Ghorbanian, K., Humphrey, J., and Sobota, T., Enhanced Combustion
Pulsejet Engines for Mach 0 to 3 Applications, AIAA Paper 1996-2687, July 1996.
Sterling, J., Ghorbanian, K., Humphrey, J., and Sobota, T., Numerical Investigations of
Pulse Detonation Wave Engines, AIAA Paper 1995-2479, July 1995.
Stuessy W.S. and Wilson, D.R., Experimental Investigation of an Annual Multi-Cycle
Pulsed Detonation Wave Engine, AIAA Paper 1996-0346, January 1996.
Sun, M., and Takayama, K., Vorticity Production in Shock Diffraction, Journal of
Fluid Mechanics, Vol. 478, 2003, pp. 237-256.
Talley, D.G. and Coy, E.B., Constant Volume Limit of Pulsed Propulsion for a Constant
Ideal Gas, Journal of Propulsion and Power, Vol. 18, No. 2, 2002, pp. 400-406.
Taylor, G.I., The Dynamics of the Combustion Products Behind Plane and Spherical
Detonation Fronts in Explosives, Proceedings of the Royal Society of London, Series
A: Mathematical and Physical Sciences, Vol. 200, 1950, pp. 235-247.
Taylor, G. I. and Maccoll, J. W., The Air Pressure on a Cone Moving at High Speed,
Proceedings of the Royal Society of London, Series A: Mathematical and Physical
Sciences, Vol. 139, 1929, p. 496.
Tew, D.E., Ideal Polytropic Ramjet Performance, Proceedings of 11th Annual
Symposium on Propulsion, Propulsion Engineering Research Center, Pennsylvania
State University, State College, PA, October 1999.
Wang, X.Y. and Chang, S.C, A 2D Non-Splitting Unstructured Triangular Mesh Euler
Solver Based on the Space-Time Conservation Element and Solution Element
Method, Computational Fluid Dynamic Journal, Vol. 8, No. 2, 1999, pp. 309-325.
258
11/23/2003
Watson, W. R. and Myers, M. K., Inflow-Outflow Boundary Conditions for Two-
Dimensional Acoustic Waves in Channels with Flow, AIAA Journal, Vol. 29, No. 9,
1991, pp. 1383-1389.
Watts, J., Conrad, C., Lee, S.Y., Woodward, R.D., Pal, S., Santoro, R.J., Fundamental
Studies of the DDT Process in Pulse Detonation Engines Based on Single/Multi-
Cycle Operation, Proceedings of 12th Annual Symposium on Propulsion, Propulsion
Engineering Research Center, Ohio Aerospace Institute, Cleveland, OH, October
2000.
White, F.M., Viscous Fluid Flow, McGraw Hill Book Co., New York, 1974.
Wintenberger, E., Austin, J.M., Cooper, M., Jackson, S., and Shepherd, J.E., An
Analytical Model for the Impulse of a Single-Cycle Pulse Detonation Engine,
Journal of Propulsion and Power, Vol. 19, No. 1, 2003, pp. 22-38.
Wu, Y.H., System Performance and Thermodynamic Cycle Analysis of Air-Breathing
Pulse Detonation Engines, Ph.D. Thesis, Department of Mechanical and Nuclear
Engineering, The Pennsylvania State University, University Park, PA, May 2002.
Wu, Y.H., Ma, F.H., and Yang, V., System Performance and Thermodynamic Cycle
Analysis of Airbreathing Pulse Detonation Engines, Journal of Propulsion and
Power, Vol. 19, No. 4, 2003, pp. 556-567.
Wu, Y.H., Yang, V., and Chang, S.C., Numerical Simulation of Chemically Reacting
Flows with Detailed Kinetics Using the Space-Time Method, Proceedings of 1st
International Conference on Computational Fluid Dynamics, Springer-Verlag,
Berlin, 2000, pp. 207-212.
Yang, V. and Cappuccio, M., Supersonic Inlet Design for Missiles, Research Report,
Department of Mechanical Engineering, the Pennsylvania State University,
University Park, PA, 1990.
Yang, V. and Culick, F.E.C., Analysis of Unsteady Inviscid Diffuser Flow with a Shock
Wave, Journal of Propulsion and Power, Vol. 1, 1985, pp. 222-228.
Yee, H. C., A Class of High-Resolution Explicit and Implicit Shock-Capturing
Methods, Lecture Series for Computational Fluid Dynamics, von Karman Institute
for Fluid Dynamics, March 6-10, 1989, Rhode-St-Genese, Belgium; NASA TM-
101088, 1989.
Yu, S.T. and Chang, S.C., Treatment of Stiff Source Terms in Conservation Laws by the
Method of Space-Time Conservation Element and Solution Element, AIAA Paper
1997-0435, January 1997.
259
11/23/2003
Yungster, S. and Perkins, H.D., Multiple-Cycle Simulation of a Pulse Detonation
Engine Ejector, AIAA Paper 2002-3630, July 2002.
Yungster, S., Analysis of Nozzle Effects on Pulse Detonation Engine Performance,
AIAA Paper 2003-1316, January 2003.
Zhang, Z.C., Yu, S.T., and Chang, S.C., A Space-Time Conservation Element and
Solution Element Method for Solving the Two- and Three-Dimensional Unsteady
Euler Equations Using Quadrilateral and Hexahedral Meshes, Journal of
Zitoun, R., and Desbordes, D., Propulsive Performance of Pulsed Detonations,
Combustion Science and Technology, Vol. 144, 1999, pp. 93-114.

11/23/2003
Vita
Fuhua Ma was born in Hubei, China. He received his B.S. degree in Mechanical
Engineering from the University of Science and Technology of China in 1996, and his
M.S. degree in 1999. He started his Ph.D. study in the Department of Mechanical
Engineering at the Pennsylvania State University from August 1999, focusing on the
modeling and simulation of pulse detonation engines.

Dramjet Theis

Uploaded by

Copyright:

Available Formats

Dramjet Theis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dramjet Theis

Uploaded by

Copyright:

Available Formats

11/23/2003

The Pennsylvania State University

= 2.1. ............................... 162

= 2.1. ............................................................................................. 206

agrees very well with the DNS data (Spapart, 1988).

2.0 3.0 4.0 5.0 6.0

are the Jacobian matrices of the convective flux

are two numerical analogues of

are the Jacobian matrices of the convective flux vectors E

You might also like