Thesis Hyslop

DEPARTMENT OF ENGINEERING
AUSTRALIAN NATIONAL UNIVERSITY
1998
CFD MODELLING OF SUPERSONIC COMBUSTION

IN A SCRAMJET ENGINE
FINAL REPORT
BY PETER HYSLOP
Supervisors: Dr Frank Houwing

Aerophysics and Laser Diagnostics Research Laboratory
Department of Physics and Theoretical Physics
Australian National University
Dr Keith Lovegrove
Department of Engineering
Australian National University
ABSTRACT
This project was aimed at modelling the supersonic combusting flow inside the ANU’s experimental
Scramjet engine using the Computational Fluid Dynamics (CFD) program CFD-ACE.
CFD models were initially verified with results from previous work and subsequently with results
obtained in this project from experiments performed on the Scramjet in the T3 free-piston shock
tunnel. In these experiments pressure measurements were obtained and compared with the CFD
results. CFD results were then used to optimise the Scramjet for maximum thrust by varying two
geometric parameters.
Several investigations were firstly performed on Scramjet configurations used in previous work to
develop an accurate CFD model of the flow. Turbulent and laminar flow, grid resolution, inlet
conditions and a number of combustion models were all investigated. It was found that pressure trends
along the floor of the Scramjet were well predicted for non-combusting flow, however for the
associated combusting flows pressure was somewhat under-predicted.
CFD models were then developed for Scramjet configurations investigated in the current project with
a higher grid resolution and a more accurate calculation of the inlet conditions. CFD results showed a
much better correlation with pressure measurements determined in the experimental phase of this
project. Both the position and magnitude of shock and expansion waves and the general trends of
pressure in the Scramjet duct were surprisingly well predicted.
Values of the two geometric parameters were varied in both the CFD and experimental models. The
range of parameters investigated was based on results found in previous work. For one of the
parameters the values investigated were outside the range that produced maximum thrust. For the
other, an optimum value was determined and the physical processes that cause the optimum identified
Overall, CFD-ACE was found to be a flexible and accurate tool for modelling the supersonic,
combusting flow in a Scramjet engine. While the initial development of CFD models was time
consuming, once an accurate model was determined, modifications of flow conditions and geometric
parameters could be easily and quickly made.
i
ACKNOWLEDGMENTS
There are a number of people that I would like to thank who have been associated with this project.
Firstly, my supervisor Dr Frank Houwing for providing encouragement, valuable direction and the
time for my numerous unscheduled meetings. I would also like to thank Matthew Gaston for his help
with CFD-ACE, the setup of shock-tube experiments and general knowledge of Scramjets.
Thankyou also to Sean O’Byrne, Dr Paul Danehy, Phil Palmer and Jodie Fox and the rest of the
ALDiR group for their encouragement and numerous queries answered.
None of my work in the T3 shock tunnel would have been possible without the technical expertise of
Paul Walsh and Paul Tant who helped with modifications of the Scramjet and ensured the smooth
running of the tunnel.
Thankyou also to mother nature, for providing a bad snow year – surely without which I would have
spent more time on the snow than in front of the computer and this thesis would have been ten pages
shorter.
Finally, thankyou to my parents for the years of encouragement and support.
Thankyou all.
ii
TABLE OF CONTENTS
ABSTRACT i
ACKNOWLEDGMENTS ii
1. INTRODUCTION 1
1.1 Scramjet Engines 1
1.2 ANU’s Scramjet 3
1.3 Aims 3
2. BACKGROUND 4
2.1 ANU’s Scramjet 4
2.2 The T3 free-piston shock tunnel 5
2.3 Previous work 6
2.4 CFD Modelling 8
2.4.1 Solution Methodology and Governing Equations 8
2.4.2 Turbulence Models 9
2.4.3 Reaction Models 9
2.5 Supersonic flow theory 11
2.5.1 Oblique shockwave relation 11
2.5.2 Expansion waves 12
2.5.3 Normal shockwaves 12
2.6 Equivalence Ratio 13
3. PROCEDURES 14
3.1 Overview 14
3.2 Verification of CFD-ACE with previous work 14
3.2.1 Overall procedure 15
3.2.2 Numerical procedures 16
3.2.3 CFD model input parameters 17
3.3 Optimisation of the Scramjet geometry using CFD-ACE 19
3.3.1 Numerical procedures 19
3.3.2 CFD model input parameters 20
3.3.3 Post Processing 21
3.4 Experimental verification of the optimum configuration 21
3.4.1 Experimental Procedure 22
3.4.2 Post processing 22
4. EXPERIMENTAL SETUP 23
4.1 Scramjet modifications 23
4.2 Pressure transducers 24
4.3 Injection system 24
4.4 Pressure Measurement system 25
5. RESULTS AND PRELIMINARY ANALYSIS 27

5.1 Verification of CFD-ACE with previous results 27
5.1.1 Turbulent or laminar flow 27
5.1.2 Grid resolution 27
5.1.3 Combustion model 28
5.1.4 Inlet pressure 30
5.1.5 CFD image features 30
5.1.6 Overall Results 31
5.1.7 Comparison of CFD and previous experimental results 34
5.1.8 Summary 35
5.2 Comparison of CFD and experimental results 35
6. RESULTS ANALYSIS AND DISCUSSION 39

6.1 Comparison of experimental and CFD thrust calculations 39
6.2 Effect of thrust surface angle 39
6.3 Effect of flat duct length 41
6.4 Effect of thrust surface length 43
6.5 Effect of combustion 45
7. SUMMARY AND CONCLUSIONS 46
8. FURTHER WORK 48
APPENDIX A – CFD-ACE GRIDS 51
APPENDIX B – LIST OF SHOCK TUNNEL SHOTS 52
APPENDIX C – SCRAMJET MODIFICATION DRAWINGS 53
APPENDIX D – ADDITIONAL CFD AND EXPERIMENTAL RESULTS 62

LIST OF FIGURES
Figure 1: Generic Scramjet engine.......................................................................................................... 2
Figure 2: Schematic of ANU’s Scramjet showing flat duct length, L and thrust surface angle, θ.......... 4
Figure 3: Inner sections of ANU’s Scramjet ........................................................................................... 5
Figure 4: T3 shock tunnel schematic (taken from ALDiR website) ....................................................... 5
Figure 5: Changing the effective flat duct length via (a) changing the flat duct length, (b) changing the
injector length.................................................................................................................................. 7
Figure 6: Shock and expansion waves .................................................................................................. 11
Figure 7: Oblique shockwaves .............................................................................................................. 12
Figure 8: Generation of a normal shockwave ....................................................................................... 13
Figure 9: Scramjet configurations used in verification of Doolans experiments .................................. 15
Figure 10: Sample grid (Note: for viewing purposes the grid resolution is half of what was generally
used) .............................................................................................................................................. 17
Figure 12: Modifications to the Scramjet – example of the 200mm flat floor configuration ............... 23
Figure 13: Typical injection pressure trace ........................................................................................... 25
Figure 14: Pressure data acquisition system.......................................................................................... 26
Figure 15: Grid comparison .................................................................................................................. 28
Figure 16: Combustion comparison ...................................................................................................... 29
Figure 17: Equilibrium combustion - water mass fraction.................................................................... 29
Figure 18: 2-step finite rate combustion – water mass fraction ............................................................ 29
Figure 19: 7-step finite rate combustion – water mass fraction ............................................................ 30
Figure 20: Pressure around the injection............................................................................................... 30
Figure 21: Floor pressure – long injector, 0mm flat duct...................................................................... 31
Figure 22: Floor pressure – long injector, 50mm flat duct.................................................................... 32
Figure 23: Floor pressure – long injector, 87mm flat duct.................................................................... 32
Figure 24: Floor pressure – short injector, 80mm flat duct................................................................... 33
Figure 25: Floor pressure – short injector, 1170mm flat duct............................................................... 33
Figure 26: Scramjet pressure – 200 mm, 1.75° thrust surface............................................................... 35
Figure 27: Scramjet pressure – 150 mm flat floor, 7° thrust surface .................................................... 36
Figure 28: Scramjet pressure – 100 mm flat floor, 1.75° thrust surface ............................................... 36
Figure 29: Scramjet Pressure – 0 mm flat duct, 7° thrust surface ......................................................... 37
Figure 30: Comparison of CFD and experimental thrust calculations: (a) 7° thrust surface, (b) 1.75°
thrust surface ................................................................................................................................. 39
Figure 31: Effect of thrust surface angle on thrust generated (combustion) ......................................... 40
Figure 32: Thrust generated by 15° thrust surface ................................................................................ 41
Figure 33: Effect of flat duct length on thrust generated (combustion) ................................................ 41
Figure 34: Variation of flat duct length for 7° thrust surface angle ...................................................... 42
Figure 35: H2O mass fraction – 50 mm flat duct, 7° thrust surface ...................................................... 42
Figure 36: H2O mass fraction – 100 mm flat duct, 7° thrust surface .................................................... 42
Figure 37: Temperature around the expansion region (50 mm flat duct, 7° thrust surface) ................. 43
Figure 38: Incremental thrust / Thrust surface length ........................................................................... 44
Figure 39: Scramjet floor pressure for 7° thrust surface: (a) 200 mm flat duct, (b) 100 mm flat duct . 45
Figure 40: Incremental thrust ................................................................................................................ 45
CHAPTER ONE - INTRODUCTION
1. INTRODUCTION
1.1 SCRAMJET ENGINES
One of the current interests in the space vehicle arena is in the development of aerospaceplanes –
reusable space vehicles with plane like characteristics. For these vehicles to be operationally viable an
air breathing propulsion system is needed. Unlike current rocket powered space vehicles, an air
breathing propulsion system does not require its own oxidiser to be carried. The obvious benefit is the
minimization of the amount of oxidiser that must be carried on the vehicle by utilizing the oxygen
available in the atmosphere. The weight of propellants that can be carried can be increased and, in
principle, the gross takeoff weight of the vehicle can be reduced1.
Over the past 30 years considerable effort has been directed to the development of a functional air-
breathing engine. The most viable engine to be studied is the Supersonic Combustion Ramjet engine
(or Scramjet engine). Large scale projects to include the Scramjet engine design have included the
British Aerospace HOTOL (Horizontal Take-Off and Landing) project and the NASA NASP
(National Aerospace Project) project.
The Scramjet engine design is an extension of the Ramjet. The difference between the two lies in flow
state inside the engine. Both are designed to be used for supersonic flight, however a Scramjet allows
the flow through the engine to remain supersonic, whereas in a Ramjet the flow is slowed to subsonic
levels before it enters the combustor. Up to flight Mach numbers of 3 – 6 Ramjet engines are optimal.
After Mach 6 various factors contribute to decreasing the efficiency of the engine. Slowing the flow to
subsonic levels becomes unrealistic because this causes the combustor entrance temperature and
pressure to become too high and causes the flow to dissociate. Combustion in dissociated flow is
extremely inefficient because the heat released by exothermic combustion reactions is negated by the
heat absorbed through endothermic dissociation reactions.
The Scramjet engine is fundamentally simple in concept but surprisingly difficult in realization. Figure
1 shows a basic generic Scramjet design. At the most fundamental level it works by injecting fuel
(typically hydrogen) into a flow of supersonic air. The air is at sufficiently high temperature and
pressure for the fuel to combust, and the resulting mixture is expelled from the engine at a higher
pressure2.
The Scramjet is composed of four main sections: the inlet, isolator, combustor and exhaust nozzle.
These sections can be seen for the generic Scramjet shown in Figure 1.
Chapter One – Introduction 1

inlet combustor
fuel
thrust surface
exhast nozzle
isolator
Figure 1: Generic Scramjet engine
The inlet heats and slows the flow through a series of oblique shockwaves. This “ram” portion of the
cycle means the engine cannot be operated statically. The isolator serves to separate the combustor
from the inlet of the engine, allowing further slowing of the flow. Combustion is achieved through the
continuous injection of fuel (usually hydrogen) into the supersonic flow. In the above diagram the fuel
is injected streamwise, however, other injection techniques (eg. wall mounted or transverse) can be
used. The fuel mixes and combusts, increasing the pressure and temperature of the flow. Finally the
flow is expanded via the nozzle. This serves two purposes: to allow the flow to accelerate to the
external speed, and to provide a mechanism by which the increase in pressure can be converted into
forward thrust. It should be noted that although the above engine is symmetric, this is only one
possible configuration, the design may also be asymmetric.
One of the major differences of Scramjets from conventional engines is that they have no moving
parts. Also, because of their very high flow speeds, proposed designs are much longer than
conventional engines and must be integrated into the airframe, rather than a separate attachment. This
allows the fuel/air mixture sufficient time to mix and combust. As an example, the thrust surface
usually includes the entire rear of the aircraft. It should be noted that the uniqueness of the Scramjet
engine lies in its supersonic combustion – other air-breathing engines can propel a vehicle to
supersonic speeds but none maintains the supersonic flow throughout the engine.
At present no aerospace vehicles use the Scramjet engine for propulsion. Most research to date has
been conducted in ground based research facilities. More recently small scale test engines have been
mounted to larger conventionally powered vehicles. One of the main difficulties in the development of
the Scramjet engine is the reproduction of a Scramjet’s operating conditions. Free piston shock tunnel
facilities, such as the one at the ANU, generate the required test conditions for very short periods of
time (1 – 4 ms). This short test time is a disadvantage in producing steady state ‘real world’
conditions, however they are the only facilities capable of producing the required pressure and
temperature. (Shock tunnel facilities use shock waves and a number of pressurised phases to develop
the desired pressure, temperature and flow speed). Blow down facilities generate flow for longer
periods of time, but cannot generate the high pressure and high temperature required at high Mach
numbers.
One hindrance to Scramjet engine concept lies in the fact that they are only operational when above a
Mach number of 6. This means that a dual phase propulsion system is required for flight up to Mach 6.

Kerrebrock3 suggests the possibility of using a turbine engine followed by a Ramjet and finally a
Scramjet. Another suggestion for outer orbit aerospace applications involves the initial use of a
conventional rocket, then Scramjet and again a rocket when upper atmosphere is reached. The
eventual configuration would depend on the specific application of the vehicle.
In addition to the space launch applications, air-breathing propulsion is also being considered for
hypersonic cruise vehicle applications (Moore and Ronald (1996)). Currently only satellite-servicing
concepts are on the horizon, however commercial vehicles with speeds double and triple those of the
Concorde are still the dream of many.
1.2 ANU’S SCRAMJET
The ANU’s Scramjet is a small scale version of the generic Scramjet with the inlet removed. It’s
design is based on a Scramjet used at the University of Queensland and has been in use for research at
the ANU since 1991. The Scramjet is designed to be used in the ANU’s T3 Free Piston Shock Tunnel.
It has been designed so that study can be performed on different supersonic flow processes and
properties. Previous work with the Scramjet has involved investigation of the effect of Mach number,
fuel fraction and other inlet properties on combustion; investigation of a number of different injector
shapes; and flow visualisation.
1.3 AIMS
This project was very much a continuation of work performed at the ANU in the T3 free piston shock
tunnel since 1991. Its aims were twofold:
• To model the flow inside the ANU’s model Scramjet engine using the Computational Fluid
Dynamics (CFD) program, CFD-ACE, and
• To optimise the combustor geometry of the ANU’s Scramjet engine for maximum thrust at a
single operating condition (Mach number and equivalence ratio*). The configuration of the
Scramjet was optimised by varying two geometric parameters: the length of the flat section
downstream of the fuel injector and the angle of the thrust surface.
The optimisation was performed both experimentally and using CFD. Pressure measurements on the
walls of both the physical and computation models were then compared.
This project stems from the Department of Physics recent purchase of the CFD-ACE program. In a
concise form, the overall goal of this project was to gain confidence with the use of a CFD program
for modelling supersonic flow and to apply this program to optimise the Scramjet for maximum thrust.
*
equivalence ratio is a measure of the “richness” of the fuel-air mixture.(see section 2.6)

CHAPTER TWO - BACKGROUND
2. BACKGROUND
2.1 ANU’S SCRAMJET
ANU’s Scramjet is designed to be used in the T3 Free Piston Shock Tunnel4. This shock tunnel is
capable of producing supersonic flow at a high temperature, the conditions that would be prevalent at
the inlet of a real Scramjet.
The ANU’s Scramjet is a duct approximately 500mm long, 25 mm high and 52 mm wide. It has a
plane base injector (flat ended) injecting fuel parallel to the flow and thrust surface on the bottom side.
The two geometric parameters, flat duct length, L and thrust surface angle, θ should be noted. For the
purposes of this thesis the flat duct length is defined as the distance between the end of the injector and
the start of the thrust surface. Twelve pressure transducers are mounted on the floor of the Scramjet.
Figure 2 shows a schematic of the ANU’s Scramjet. Several modification were made to the Scramjet
during this project. These are outline in section 4.1
Figure 2: Schematic of ANU’s Scramjet showing flat duct length, L and thrust surface angle, θ.
It should be noted that the ANU’s Scramjet differs from the generic Scramjet, shown in Figure 1, in
two ways. Firstly, the ANU’s Scramjet does not have a converging inlet and therefore only simulates
the flow downstream of the inlet.
The second difference is that the ANU’s Scramjet is not symmetric, there is a thrust surface on only
one side of the duct.
The injector has a leading edge as designed by the National Aerospace Laboratory in Japan. It is a
double wedge with a 10° half angle, rounded at the leading edge with a 0.5 mm radius.
Chapter Two – Background 4

flat floor sections pressure transducers
side plates
Figure 3: Inner sections of ANU’s Scramjet
The Scramjet was originally based on a design used in the University of Queensland’s shock tunnel
T4, but since then has had numerous modifications.
Figure 3 shows the inside sections of the Scramjet. The inner sections of the Scramjet is used to
modify the geometry of the combustor and thrust surface. Different sized and shaped plates can be
placed in this section to change the geometry. Different style injectors can also be inserted by
removing the front section. Pressure transducers are mounted in the floor of the duct. The transducers
mountings in the unconfined region behind the end of the roof were not used.
The injection system was contained in the bottom of the Scramjet. Pressurised hydrogen (1400 kPa for
this project) is stored in a coiled copper tube (called a Ludweig tube) which is connected to a solenoid.
When the solenoid is triggered, it activates a fast acting valve system contained in the front section of
the Scramjet and the hydrogen is injected into the flow.
2.2 THE T3 FREE-PISTON SHOCK TUNNEL
All experimental work was performed in the T3 free-piston shock tunnel at the ANU physics
department. The shock tunnel can generate supersonic flow for 1 – 4 ms at high pressure and
temperature. This short duration is one of the disadvantages of a shock tunnel, however it is the only
type of facility capable of producing supersonic flow with ‘real world’ Scramjet operating conditions.
Figure 4: T3 shock tunnel schematic (taken from ALDiR website)

As seen in Figure 4, the T3 shock tunnel is composed of a number of sections. The main section are:
the compression tube, in which the high pressure driver gas is generated; the shock tube, in which the
test gas is situated and shockwaves are developed; the nozzle, which accelerates the flow to the desired
test conditions; the test section, in which the apparatus being tested is placed; and the dump tank. A
steel diaphragm is situated between the compression and shock tubes and a Mylar diaphragm is placed
between the shock tube and nozzle.
Prior to a ‘shot’ in the tunnel, the test section, compression tube and shock tube are evacuated. The
shock tube, compression tube and high pressure reservoir are then filled with gases determined by the
desired flow conditions at the nozzle exit. The high pressure reservoir is air and the driver gas in the
compression tube is a mixture of helium and argon. The shock tube is filled with either air or nitrogen
depending on whether combustion or non-combustion is being tested in the Scramjet.
To produce supersonic flow in the tunnel the following events occur:
1. A high pressure is built up in the high pressure reservoir

2. The piston is released, driving it into the compression tube and compressing the driver gas.
3. The diaphragm bursts and a shockwave propagates through the shock tube increasing the pressure
and temperature of the test gas. The driver gas pushes the test gas in front of it.
4. The shockwave reflects from the end wall of the shock tube, raising the pressure and temperature
further and causing the Mylar diaphragm to burst.
5. The test gas exits through the nozzle which is shaped to produce the desired Mach number.
6. The test gas travels through the Scramjet, mounted in the test section, and into the dump tank.
2.3 PREVIOUS WORK
Investigation in Scramjet design and related activity has been performed for thirty years since the
research interest in supersonic combustion was first expressed. A review of the previous research is
given in Heiser and Pratt2,5.
Much of the previous work on the Scramjet at the ANU has involved flow visualisation, for which the
Aerophysics and Laser Diagnostic Research Laboratory (ALDiR) at the Department of Physics, is well
known. Various visualisation techniques have been used. These include Schlieren, shadowgraph and
Planar Laser Induced Fluorescence (PLIF) systems6. In addition to visualising the flow with the above
techniques, static pressure measurements have also been made inside the Scramjet.
The most relevant work to this project is the work performed in 1997 by Doolan7 in his Engineering
Honours Project, Optimising the Combustor Geometry in a Supersonic Combustion Ramjet. In this
project the flat duct length in the Scramjet was varied for a single thrust surface angle of 3.5°. The
angle of 3.5° was proposed by researchers at the Japanese National Aerospace Laboratories (NAL) for
the condition:
Mach number, M = 2.5

Temperature, T = 1100 K
Pressure, P = 1.4 atm

Their value of thrust surface angle was not determined on the basis of optimising thrust but on other
considerations which are discussed in detail by O’Byrne4. Essentially, this value was the minimum
value necessary to avoid an undesirable phenomena known as ‘thermal choking’, which reduces the
Mach number inside the combustor to subsonic values so that it no longer operates as a true Scramjet
engine.
The flat duct length was varied in two ways: by increasing the distance between the injector and start
of the thrust surface; and by decreasing the injector length (effectively increasing the distance between
the end of injector and the thrust surface).
L L
L L
(a) (b)
Figure 5: Changing the effective flat duct length via (a) changing the flat duct length, (b) changing the injector
length
This second method of increasing flat duct length had several shortcomings, the most significant of
these being that different shockwave patterns and flow conditions were obtained in the duct using the
different length injectors. Decreasing the injector length could not therefore be considered equivalent
to increasing the flat floor length. Regardless of this problem, the conclusion of this project was that,
for a given total length of Scramjet, increasing the flat duct length produced an increase in thrust. The
maximum flat duct length investigated (87 mm) was found to produce the greatest thrust.
Several different configurations of flat duct length and injector length were investigated. The pressure
on the floor of the Scramjet was measured using 12 PCB pressure transducers. From these pressure
measurements one dimensional calculations were performed using a Visual Basic program, Scramjet.
This program calculated several parameters including: Mach number, temperature and fraction of fuel
burnt. The results were used to compare the flow properties at the different configurations and to
calculate the thrust generated. Additionally, images of the flow in the region just behind the injector
were developed using the Shadowgraph technique4 and used for a qualitative analysis of the flow.
This project continued directly from Doolan’s work. It is aimed at determining to what extent
increasing flat duct length increases thrust and combining this with an investigation of thrust surface
angle. It also extends the previous work through the use of Computational Fluid Dynamics.
Other recent Scramjet studies at the ANU (by other Physics and Engineering students) have included:
comparison of transverse and parallel injection schemes, investigation of parallel injector geometry on
mixing performance and the investigation of a transient pressure rise in the duct8,9,10.

2.4 CFD MODELLING
Computational Fluid Dynamics (CFD) packages are very powerful tools for analysing any type of
fluid flow. They are capable of calculating a large number of flow parameters that are often difficult or
impossible to determine experimentally. For optimisation purposes, they allow easy manipulation of
geometry and flow conditions.
The program used in this project was CFD-ACE11. CFD-ACE can be used to model a large variety of
different types of flow: subsonic or supersonic; turbulent or laminar; incompressible or compressible;
mixing and reacting; and steady state or transient.
It should be noted that other programs exist that model supersonic flows better than CFD-ACE (in
particular shockwaves) but due to price, ease of use and software support CFD-ACE was preferred for
the current project. Also, CFD-ACE was currently being used by Gaston12 to model a number of
Scramjet injectors, and so advice could be given on its use.
CFD-ACE actually consists of 3 modules; CFD-GEOM13, used to generate the geometry and the grid;
CFD-ACE, used to define the remainder of the model conditions and control the solver; and CFD-
VIEW14, used to analyse the results. All these modules are controlled using user friendly Graphical
User Interfaces (GUIs), although more powerful controls can be invoked using a command language.
2.4.1 Solution Methodology and Governing Equations
CFD-ACE uses a control volume approach in calculating flow parameters. The region of interest in the
flow is divided into a grid. Each grid element is considered as a control volume with the properties
constant over its volume. For each control volume, fluid flow is simulated by numerically solving
partial differential equations that govern the transport of flow quantities, also know as flow variables.
The variables include mass, momentum, energy, turbulence quantities, mixture fractions and species
concentrations. The variables for which transport equations have to be solved will depend on the
nature of the flow problem.
The three equations common to all fluid dynamics problems are the conservation of mass, momentum
and energy equations. In differential form these are:
Conservation of mass:
∂ρ ∂
+ ( ρu j ) = 0 (2.1)
∂t ∂x j
where uj is the jth Cartesian component of the instantaneous velocity and ρ is the fluid density.
Conservation of momentum:
∂ ∂ ∂p ∂τ ij
( ρui ) + ( ρui u j ) = − + + ρf i (2.2)
∂t ∂x j ∂xk ∂x j

where p is the static pressure, τij is the viscous stress tensor and fi is the body force.
Conservation of energy:
∂ ∂ ∂q ∂p ∂p ∂u
( ρh) + ( ρu j h) = − j + + uj + τ ij i (2.3)
∂t ∂x j ∂x j ∂t ∂x j ∂x j i
where qj is the j-component of the heat flux and h is the static enthalpy.
The three Partial Differential Equations (PDE’s), along with any others dependant on the specific flow
problem, are discretized on the computational grid, a set of algebraic equations are formed, and the
solution of the algebraic equations determined. This method generates the flow variables at each grid
point.
An iterative solution scheme is used by CFD-ACE to solve the algebraic equations. The equations are
solved sequentially and repeatedly with the goal of improving the solution at each iteration. The
solution is monitored by viewing global residuals (the difference between the current and previous
solution average over the entire domain). A solution is generally considered “converged” when the
residuals have decrease by 4-5 orders of magnitude.
The most important point to consider when using CFD-ACE (or any CFD program) is that the quality
of its output is only as good as the quality of its input so care has to be taken to make sure that inputs,
such as boundary conditions, fluid properties and fluid models are as accurate for (or applicable to) the
specific problem, as possible.
2.4.2 Turbulence Models
The turbulence model used for the CFD models analysed in this project was the standard k-ε model. It
was used because it is well known and applicable to high Reynolds number flows15. The model
equations can be found in the CFD-ACE theory manual16.
The parameters associated with this model are k and L. The physical significance of these parameters
is discussed in the CFD-ACE theory manual16. k can be calculated using the equation:
2
 µ laminar 
k =   (2.4)
 0.016432 ⋅ Lρ 
where µlaminar is the laminar viscosity and ρ the density.
For all CFD models L was 25 mm, the height of the Scramjet duct.
2.4.3 Reaction Models
During this project three different reaction models were used. The fundamental concepts behind these
are outlined below.

The instantaneous model
The instantaneous reaction model assumes that a single chemical reaction occurs and proceeds
instantaneously to completion. The reaction used for the Scramjet was the hydrogen-water reaction:
2H2 + O2 → 2H2O.
The equilibrium model
The equilibrium model requires the specification of all the chemical species that might exist in the
reacting mixture. No specific reactions need to be specified. This reaction model calculates the species
concentrations at its equilibrium condition. The species specified for the reaction mixture were: H2,
O2, N2, H2O, OH, O and NO.
The multi-step finite rate model
The multi-step finite rate reaction model uses chemical rate equations to model any number reaction
occurring in the system. The reaction rates are calculated using the Arrhenius equation:
k = ApT n e ( − Ea /RT ) (2.5)
where: k is the reaction rate coefficient

Ap is the pre-exponential constant
Ea/R is the activation temperature
n is the temperature exponent
Ap, Ea/R and n are determine experimentally for a particular reaction.
Two different reaction sets were used during the course of this project, a two-step model and a seven-
step model. The reactions and rate coefficients are shown in the tables below.
2-step finite rate reaction
No. Reaction Ap n Ea/R

[m /kmol⋅s]
3
[K]
1 H2 + O2 → OH + OH 1.14× 1045 -10 24373
2 2OH +H2 → 2H2O 2.5× 1050 -13 21402
Note: this 2 step chemical rate equation was taken from Rogers and Chinitz17.
7-step finite rate reaction

No. Reaction Ap n Ea/R
[m /kmol⋅s]
3
[K]
1 H2 + O2 → OH + OH 0.17 × 1011 0 2.424 × 104
2 H + O2 → OH + O 0.142 × 1012 0 8.251 × 103
3 OH + H2 → H2O + H 0.316 × 105 1.8 1.537 × 103
4 O + H2 → OH + H 0.207 × 1012 0 6.916 × 103
5 OH + OH → H2O + O 0.55 × 1011 0 3.524 × 103
6 H + OH → H2O + M 0.221 × 1020 -2 0
7 H + H → H2 + M 0.653 × 1015 -1 0
2.5 SUPERSONIC FLOW THEORY
The most important consideration with supersonic flow is that the flow is compressible. A
compressible flow is one for which the density cannot be considered constant (for flow below M = 0.3
the fluid can be considered to have a constant density). Compressibility leads to two phenomena
unique to supersonic flow - shockwaves and expansion waves (see Figure 6).
There are two types of shockwaves: oblique and normal. Oblique shockwaves and expansion waves
are generated when a supersonic flow changes direction – a shockwave when the flow converges and
an expansion wave when the flow expands.
expansion wave normal shock wave

shock wave
M2<M1 M3>M2 M4<M3<1
M1
Figure 6: Shock and expansion waves
2.5.1 Oblique shockwave relation
An oblique shockwave is generated when the direction of a supersonic flow changes in a convergent
way. The relative conditions after a oblique shock are the same as for the normal shock - Mach
number decreases and pressure, temperature, density and entropy increase. The flow can however
remain supersonic. The simplest case is flow over a half wedge.

shockwave
M2
β
M1
θ
Figure 7: Oblique shockwaves
The angle of the resulting shockwave is a function of both the wedge angle and free stream Mach
number and is given by the relation18:
 M 2 sin 2 β − 1 
tan θ = 2 cot β  2 1  (2.6)
 M 1 (γ + cos 2 β ) + 2 
If this function is viewed graphically, it becomes apparent that for a fixed deflection angle θ, as the
free stream Mach number is decreased the shockwave angle, β increases.
2.5.2 Expansion waves
The overall effect of an expansion wave is the opposite of a shockwave: Mach number increases and
temperature, pressure and density decrease. Entropy, however remains constant. Unlike a shockwave,
the flow condition across an expansion wave change gradually and they are represented schematically
as an expansion fan.
2.5.3 Normal shockwaves
Shockwaves are established in supersonic flow as a solution to the problem of disturbance propagating
through a flow.
The properties of a flow such as pressure, temperature and density propagate through a flow at the
speed of sound, a, which for an ideal gas is given by:
a = γRT (2.7)
where γ is the ratio of specific heats cp/cv, R is the universal gas constant and T the temperature.
Consider the flow around a blunt body such as the one shown below. When the flow is subsonic,
disturbances or ‘information’ may be transmitted upstream since the speed of sound is greater than the
flow velocity. The flow can be ‘warned’ about the upcoming body and the flow condition varied
accordingly.

Figure 8: Generation of a normal shockwave
For the case where the flow is supersonic, the speed of sound is less than the velocity of the fluid.
Flow disturbances cannot therefore be transmitted upstream and they tend to coalesce a short distance
in front of the body. Since these disturbances cannot propagate upstream, ahead of the normal shock
the flow has no idea about the upcoming body and acts as if the body is not there. After the shock the
flow becomes subsonic and moves around the object.
Both normal and oblique shockwaves can be considered as discontinuities in the flow. Flow condition
change across them over a very small distance (typically 10-5 m for air at standard conditions).
Across a normal shockwave, the Mach number decreases (to below 1 (subsonic)) and pressure,
temperature, density and entropy increase.
2.6 EQUIVALENCE RATIO
When referring to the Scramjet engine the term ‘equivalence ratio’ is often used. The equivalence ratio
is a measure of the ‘richness’ of the air-fuel mixture. It is defined as the mass flux of fuel divided by
the mass flux of air, all divided by this same ratio for a stoichiometric mixture.
m fuel
m air
φ= (2.8)
m 
 fuel m

 air  stoic.

CHAPTER THREE – PROCEDURES
3. PROCEDURES
3.1 OVERVIEW
Due to the multiple aims associated with this project it was divided into three phases:
1. Verification of CFD-ACE with previous work

2. Optimisation of the Scramjet geometry using CFD-ACE
3. Experimental verification of the optimum configuration
In the first phase, experimental results obtained by Doolan7 in 1997 were compared with models
generated using CFD-ACE. Pressure measurements on the thrust surface of the Scramjet provided a
basis for this comparison. This phase took approximately half the time allocated for the project. It
involved a building a CFD model gradually - from the simplest case, to an increasing level of
complexity. When this was done, this model was used as a basis for the remaining configurations.
The second phase was to model the Scramjet at a particular operating condition (free stream pressure,
temperature, Mach number and equivalence ratio) and to vary the two geometric parameters to find the
maximum thrust. The models generated in phase one were used as a basis for these models.
The final phase was to perform experiments in the T3 free piston shock tunnel using the ANU’s
Scramjet. Several of the configurations modelled using CFD were investigated by measuring the
pressure inside the Scramjet combustor. This involved several modification to the Scramjet to
accommodate new configurations and pressure transducer mounting positions.
It should be noted that the phases in this project were not performed consecutively. Phase one was
completed first, however components of phases two and three were performed simultaneously. There
were several reasons for this. First of all, work using the T3 shock tunnel had to fit in with other work
in the tunnel. The work was scheduled in the tunnel mid-way through second semester and this time
had to be adhered to. Ideally CFD modelling would have been finished before this time but as the
project proceeded this became unrealistic. The second reason was that the CFD modelling in phase
one took longer than expected and so the other phases were pushed back.
3.2 VERIFICATION OF CFD-ACE WITH PREVIOUS WORK
Before using CFD-ACE to model new configurations of the Scramjet, the program was used to model
experiments that were performed in 1997 by Doolan7. Pressure measurements on the thrust surface of
the Scramjet were compared for these experiments with the CFD results obtained in this project.
Five different configurations of the Scramjet were investigated by Doolan. In three of the conditions, a
78 mm long injector was used with 0, 50 and 87 mm flat duct sections. In the other two, a shorter 48
mm long injector was used with 80 and 117 mm flat duct sections. In all configurations, a thrust
surface angle of 3.5° was used.
Chapter Three – Procedures 14

Figure 9: Scramjet configurations used in verification of Doolans experiments
Pressure was measured at twelve positions on the thrust surface of the Scramjet. The positions with
respect to the end of the injector are shown below:
Transducer Position (mm) Transducer Position (mm)

1 55 7 175
2 75 8 195
3 95 9 215
4 115 10 255
5 135 11 295
6 155 12 355
Experiment were performed for two different flow condition: combustion and non-combustion. In the
combustion cases the free stream gas was air, in the non-combustion case the free stream was nitrogen.
Nitrogen has a molecular mass similar to air and does not cause the hydrogen fuel to combust.
The flow condition used were as follows:
Mach number - 2.5

Pressure - 90 kPa
Temperature - 1217 K
Equivalence Ratio - 1.7
3.2.1 Overall procedure
Two dimensional CFD models of all 5 configurations were constructed. There were several reasons
why three dimensional models were not used. Firstly, CFD had not been used to model the flow in the
ANU’s Scramjet and it was thought best to start from the simplest case - two dimensions. Secondly,
3D models require significantly more computational time and finally it is easier to analyse the results
in only two dimensions.
The approach taken in building the five models was to start from the simplest possible configuration
and flow conditions and to gradually build on this, increasing complexity until the desired level of
detail and accuracy was achieved. Two models were gradually constructed using this procedure and
the remaining 3 configurations based on these. As the models increased in complexity and different

features were added, pressure measurements along the floor were monitored and compared with
experimental results. The results can be found in section 5.1.6..
In building up an accurate first model the following issues were addressed:
• What grid resolution is required ?

• Should a turbulent or laminar flow model be used ?
• What mixing model should be used ?
• What reaction model should be used ?
To address these issues the model was built up with the following configurations:
1. flat, empty Scramjet duct with laminar flow

2. flat, empty duct with turbulent flow
3. empty duct (no injector) with thrust surface and turbulent flow
4. duct with thrust surface, injector and turbulent flow
5. duct with thrust surface, injector, turbulent flow and injection and mixing
6. duct with thrust surface, injector, turbulent flow, injection, mixing and instantaneous
combustion
7. duct with thrust surface, injector, turbulent flow, injection, mixing and equilibrium combustion
8. duct with thrust surface, injector, turbulent flow, injection, mixing and 2-step finite-rate
combustion
9. duct with thrust surface, injector, turbulent flow, injection, mixing and 7-step finite-rate
combustion
Also, for step 5 a grid analysis was performed to see what effect grid resolution had on results.
48 different models were generated during this phase. This large number was due to the variety of
different flow conditions that were experimented with.
3.2.2 Numerical procedures
For each model generated in this phase there were several common steps. These steps are outlined
below:
1. Generate the grid

The grids were generated using CFD-GEOM. This program first requires the definition of the overall
geometry. The geometry is divided into several ‘domains’ which depend on the geometry of the
model. For each wall of a domain the number of grid points are defined. After all the walls are defined
the internal cells are generated. Grid points can be distributed according to a power law so that
changes in grid resolution occur gradually (grid spacing should not change between two cells by more
than a factor of 0.3). The output of this program is used by the program CFD-ACE.

Figure 10: Sample grid (Note: for viewing purposes the grid resolution is half of what was generally used)
A more detailed figure of a typical grid can be found in Appendix A.
2. Generate the model

CFD-ACE is used to generate and modify the CFD model. There are two sections to this program:
‘model’ and ‘solve’. The model section is used to define fluid properties, mixing, reacting and
turbulence models and the boundary conditions. The solve section is used the control the iteration,
relaxation, output and to start the solver.
The first step in defining a model is to import the grid. The different option of fluid models are then
selected (these can include: compressible, incompressible, turbulence, heat transfer, mixing and
reacting flows). Fluid properties are then defined. For each of the flow models specified, associated
parameter are also defined. The boundary conditions and overall initial conditions are then defined
(initial conditions generated from previously run models can be used). In the solver section of the
program, parameters such as the number of iterations, relaxation and limits are specified. Also, the
parameters that are to be recorded are selected. Finally the model is ‘submitted’ and the residuals can
be viewed. Generally the solution was deemed to be converged when the residuals had decreased by 4
– 5 orders of magnitude.
3. View the results

After the solution has converged and the run stopped, the results can be viewed using CFD-VIEW.
Two methods of representing the data were used in this project: viewing a particular property over the
entire duct using a specified colour palette, and graphing a property along a particular cell line. The
first of these gives an overall, qualitative analysis of the flow, the second a quantitative analysis.
3.2.3 CFD model input parameters
For a larger number of the models generated in this phase the parameters input to CFD-ACE were the
same. A summary of these are shown below.
Fluid properties
Density - Ideal gas law

Viscosity - Sutherland’s law
Specific heat - JANNAF method
Conductivity - Prandtl Number, 0.7
Mass diffusion - Schmidt Number, 0.9
Air composition - mass fraction

O2 - 0.21
N2 - 0.79
Boundary conditions
Inlet (free stream)

Pressure - 70 kPa
Temperature - 1217 K
Velocity - 1697 m/s
k - 0.1934
L - 0.025 m
Injector
Pressure - 54.90 kPa
Temperature - 151.5 K
Velocity - 2048 m/s
k - 31.49
L - 0.0032 m
Wall
Isothermal - T = 298 K
Outlet
Extrapolated
Models
Turbulence
k-ε - Prandtl number 0.9

Schmidt number 0.5
Reaction – Instantaneous
Reaction - 2H2 + O2 → 2H2O
Reaction - Equilibrium
Species present - H2, O2, N2, H2O, OH, O, NO
Reaction – 2-step finite rate
See section 2.4.3 for reaction rate coefficients.
Reaction – 7-step finite rate
See section 2.4.3 for reaction rate coefficients.

3.3 OPTIMISATION OF THE SCRAMJET GEOMETRY USING CFD-ACE
The second phase of this project was the optimisation of the Scramjet combustor geometry for
maximum thrust using CFD-ACE. This phase was dependant on the completion of phase one since the
correct operation of the CFD-ACE program and models had to be verified before any subsequent
models could be made.
The effect on thrust was investigated by varying the flat floor length, L and thrust surface angle, θ (see
Figure 2) for a given total length of Scramjet. Previous work by Doolan showed that in his
configuration the thrust increased as flat floor length increased (for a given total length of duct). The
increase in thrust is due to an increase in pressure as the floor length is increased. The pressure
increases because there is more time (distance) for the flow to mix and combust before the cooling and
quenching effects of the expansion are caused by the angled thrust surface. For Doolan’s
configurations, no limit to this increase was found. It was anticipated that there would be a limit to the
increase in thrust since an increasing flat floor length (producing increasing thrust) would be
counteracted by a short length of thrust surface for which the pressure can be converted into thrust.
Also, variation of thrust surface angle should have some effect on the maximum thrust attainable. It
was anticipated there would be some trade-off between the effects of pressure decrease due to a larger
θ (larger expansion) and more efficient conversion of pressure into thrust at larger θ (due to a larger
component of the thrust surface being perpendicular to the flow).
3.3.1 Numerical procedures
A total length of Scramjet of 500 mm was considered. This corresponds to the maximum length of the
ANU Scramjet.
The free stream conditions used were as follows:
Mach number - 3.9

Pressure - 104 kPa
Temperature - 938 K
Equivalence Ratio - 0.5
The Scramjet models were all based on the ANU’s Scramjet with the longer 78 mm strut injector. Flat
duct length was varied between 50 and 200 mm and thrust surface angle between 1.75° and 7°
degrees. The flat duct limits were chosen based on Doolan’s conclusions. The thrust surface angles
were chosen based on an angle determined by the National Aerospace Laboratory (NAL), Japan. This
angle was 3.5°. Angles were chosen at increments above and below this angle. All the limits were also
chosen such that the ANU’s Scramjet could be modified to accommodate the changes.
Initially a matrix of configurations within these limits was investigated. These are shown in the table
below.

Flat duct length Thrust surface angle
1.75° 3.5° 5.25° 7°
50 mm 50, 1.75 50, 3.5 50, 5.25 50, 7
100 mm 100, 1.75 100, 3.5 100, 5.25 100, 7
150 mm 150, 1.75 150, 3.5 150, 5.25 150, 7
200mm 200, 1.75 200, 3.5 200, 5.25 200, 7
Due to time constraints a number of these configurations were not modelled. These are shown in
italics. Two other configurations not shown, one with 0 mm flat duct at 7° thrust surface angle and one
with 100 mm flat duct and 15° thrust surface angle, were also modelled.
All configurations were modelled for both combustion and non-combustion cases.
3.3.2 CFD model input parameters
Numerous different parameters and properties were input to CFD-ACE. For all the models these were
the same, the only changes were to the geometry. The model properties and parameters input to the
program are listed below.
Fluid properties
Density - Ideal gas law

Viscosity - Sutherland’s law
Specific heat - JANNAF method
Conductivity - Prandtl Number, 0.7
Mass diffusion - Schmidt Number, 0.9
Air composition - mass fraction
O2 - 0.1952
N2 - 0.7272
Ar - 0.0132
NO - 0.06
O - 0.0044
Boundary conditions
Inlet (free stream)

Temperature - 938 K
Velocity - 2365 m/s
k - 0.061
L - 0.025 m
Injector
Temperature - 151.5 K
Velocity - 2055 m/s

k - 3.94
L - 0.0032 m
Wall
Isothermal - T = 298 K
Outlet
Extrapolated
Models
Turbulence
k-ε - Prandtl number 0.9

Schmidt number 0.5
Reaction - 7 step finite rate
See section 3.2.3 for details
3.3.3 Post Processing
CFD data was collected in the form of pressure traces along the floor of the Scramjet and colour
images of pressure and H2O mass fraction in the duct. H2O mass fraction gives an indication of where
and to what degree combustion is occurring in the duct.
A pressure force summary was also generated for each model. This contained pressure and shear
forces resolved into x and y components for each of the walls and boundaries in the models. These
forces could be summed to give the total thrust generated by the Scramjet.
3.4 EXPERIMENTAL VERIFICATION OF THE OPTIMUM CONFIGURATION
The third phase of this project was the verification of the optimum configuration of the combustor
geometry by performing experiments using the ANU’s Scramjet in the T3 shock tunnel. This phase
served to both further verify the operation of CFD-ACE and to determine how the optimum condition
predicted from CFD-ACE results compared with actual experimental results.
Before experiments could be performed a number of modifications to the Scramjet were made so that
the different configurations could be tested and so pressure transducers could be placed in appropriate
sections. These modifications are outlined in section 4.1.
A total of 33 shots in the T3 shock tunnel were used in this project for the testing of 14 different
configurations (see Appendix B for shot details). The configurations tested ranged from 100 – 200 mm
in flat duct length and from 1.75° to 7° thrust surface angle. Originally the 50 mm flat duct length was
going to be tested as well (as was modelled using CFD) however time constraints prevented this.
Additionally 2 configurations at 0 and 50 mm flat duct length and 7° thrust surface angle were tested.
For each configuration both combustion and non-combustion cases were investigated by using air and
nitrogen in the free stream respectively.

3.4.1 Experimental Procedure
Before the experiments were started, the injection system (see section 4.3) was calibrated by firing the
shock tunnel (with Scramjet mounted in the test section) and measuring the injector pressure trace.
The delay on the injector timing system was adjusted so that the correct equivalence ratio was
obtained. These shots were also used to see if the pressure acquisition system was working.
The first configuration tested was the 200 mm flat floor, 1.75° thrust surface angle. This had the most
constrictive geometry and was tested first to see if the flow choked* - which it did not. The remaining
configuration were then tested. For each of the flat floor configurations the thrust surface angle was
modified from to 1.75°, 3.5°, 5.25° and 7°. The flat floor length was then reduced to the next
increment. After all these configurations were tested the two additional configurations of 0 and 50 mm
flat floor with 7° angle were tested.
For each of the configurations 2 shots were fired. After every two shots the test section side plates
were removed, the Scramjet disassembled and the configuration changed. Pressure transducer
positions were changed every time the flat duct length was decreased. This process usually took
between 1½ and 2 hours. For each shot there was also 1 – 2 hours to pump down and then pressurise
the shock tunnel. Shock tunnel operation was performed by Paul Walsh and Paul Tant. An average of
3 shots were performed each day over a period of 14 days.
3.4.2 Post processing
Pressure data was recorded over a period of 10 ms. The pressure measurements used in this project
were taken at 1.4 ms after the flow began. This figure has been determined as the time at which the
flow is steady in previous experiments4. The pressures at this time were smoothed over a period of 0.2
ms.
Thrust was calculated by fitting a linear spline to the pressure data and integrating this pressure over
the distance of the thrust surface using Simpson’s rule (up to a total Scramjet length of 500 mm). The
force was then resolved to the x direction. Experimental error was determined by calculating thrust for
several configurations using the trapeziodal rule and Simpson’s rule. This yielded values consisted to
within 10%. By allowing 3% accuracy for the pressure measurements the accuracy of the thrust
measurements is estimated as ±13%.
*
A ‘choked’ flow is a subsonic flow that has been caused by too much heat being released in combustion. It can
be avoided by expanding and thus cooling the flow.

CHAPTER FOUR – EXPERIMENTAL SETUP
4. EXPERIMENTAL SETUP
4.1 SCRAMJET MODIFICATIONS
There were several modification made to the Scramjet. The drawings for these can be found in
Appendix C.
The section after the injector is composed of two sections, the flat floor and the thrust surface. In order
to change the geometry to the desired configurations flat floor sections and side plates to hold these up
were constructed. To change the thrust surface angle, side plates that hold the long, thrust surface
section were made.
flat floor sections

B C D
E F
G
side plates
Figure 11: Modifications to the Scramjet – example of the 200mm flat floor configuration
In order to change the Scramjet geometry over equal floor increments, 50 and 25 mm floor sections
were fabricated. These sections contained 2 and 1 PCB pressure transducer mountings respectively. A
100 mm floor section already existed, so 4 transducer mounts were incorporated in this (D). A single
50 mm floor section also existed. This mount had been constructed is such a way so that it could fit
immediately behind the injector, the other sections could not (B). With the 50, 100 and 50 mm
sections the configurations with flat floor length between 100 and 200 mm could be constructed. The
25 mm section was made in anticipation of smaller increments of floor length being investigated but
time constraints prevented this.
For each of the floor section made, two side plates were fabricated to serve as structural supports.
These are shown as E and F in Figure 11.
Side plates for the 3.5 degree thrust surface already existed. Plates for 1.75°, 5.25° and 7° were
constructed. A 7° side plate is shown as G in Figure 11.
Modifications were also made to the roof section of the Scramjet to accommodate two pressure
transducers immediately behind the injector (A). Ideally these transducers would have been placed on
the floor behind the injector however they would have interfered with another pressure transducer that
is mounted in the injector and has a lead that exits just below the first floor section. Consequently they
Chapter Four – Experimental Setup 23

were placed in the roof, and since the Scramjet is symmetric up until the thrust surface, the readings
should be the same as if they were placed on the floor.
4.2 PRESSURE TRANSDUCERS
Twelve PCB transducers were used to measure the pressure in the duct. One other PCB was used to
record the pressure just after the injector fast-acting valve. The PCB details are shown in the table
below.
Transducer Type Calibration Transducer Type Calibration

[mV/kPa] [mV/kPa]
1 113M165 7.66 7 113M165 7.53
2 113M165 7.56 8 113M165 7.49
3 113M165 7.56 9 113M165 7.25
4 113M165 7.44 10 113M165 7.62
5 113M165 7.49 11 113M165 7.43
6 113M165 7.53 12 113AA21 4.23
Injector 112A21 7.35
The positions of the pressure transducers on the Scramjet roof and floor varied depending on the flat
floor length used. The positions are shown in the table below.
100 mm floor 150 mm floor 200 mm floor

Transducer Position Transducer Position Transducer Position
[mm] [mm] [mm]
1 25 1 25 1 25
2 45 2 45 2 45
3 60 3 65 3 60
4 80 4 85 4 80
5 115 5 105 5 150
6 135 6 125 6 170
7 155 7 200 7 190
8 175 8 220 8 210
9 250 9 240 9 230
10 270 10 260 10 250
11 290 11 280 11 270
12 370 12 360 12 370
4.3 INJECTION SYSTEM
The injection system is configured so that the injection is initiated well before the supersonic flow
reaches the Scramjet and finished well after the flow has past. Hydrogen fuel is stored in a coiled
copper Ludweig tube. When a solenoid, situated inside the Scramjet, is triggered a fast acting valve is
activated injecting the hydrogen into the flow via the injector. A pressure transducer is mounted in a
cavity behind the valve which measures the static pressure.

1200 120
Injector
1000 Stagnation 100
Pressure [kPa]
800
Pressure [Pa]
80
600
60
400
200 40
0 20
0 20 40 60 80 100
Time [ms]
Figure 12: Typical injection pressure trace
Figure 12 shows a typical injection pressure trace. The red line shows the time at which the flow
passes through the Scramjet (for 1 ms) and the blue the injector pressure. The mass flow rate of
hydrogen out of the injector, and therefore the equivalence ratio, is based on the static pressure in the
injector. Even though this pressure is decreasing over time, during the short period that flow passes
through the Scramjet, it can be considered constant. To calibrate the injection to the correct
equivalence ratio, both the initial reservoir pressure and the timing of the injector can be modified.
A He-Ne laser attached to the side of the shock tunnel is used to trigger the injection solenoid valve.
When the shock tunnel is fired the whole tube, from the high pressure reservoir to the nozzle, moves
backward approximately 3 cm. When this occurs the laser moves out of line from a photo-diode which
is stationary and a –ve pulse is produced. This pulse is isolated, attenuated and delayed and sent to the
solenoid power supply to start injection.
4.4 PRESSURE MEASUREMENT SYSTEM
The pressure measurement system is outlined in Figure 13. There were a total of 15 PCB pressure
transducers – 12 measuring pressure in the Scramjet, one measuring pressure just before the injector
and two measuring the timing of the shockwave (stagnation and timing PCB’s).
Data was recorded on one LeCroy waveform digitizer and 3 Tektronix digital oscilloscopes. Eight of
the pressure channels were recorded on the LeCroy and the remaining 4 on the 4-channel oscilloscope.
Data was stored on the LeCroy and could be downloaded later via a Macintosh computer. Data from
the cro could be recorded directly to floppy disk. The LeCroy and cro had a sampling rates of 20 µs
and 10 µs respectively (this data was later smoothed).
The first of the 2-channel oscilloscope’s was used to record the injection pressure trace and the
stagnation pulse. The stagnation pulse showed at what time the flow passed through the Scramjet and
from this the injector pressure, and thus equivalence ratio, could be determined (see section 4.3).
The second of the 2-channel oscilloscope’s was used to record the timing and stagnation PCB traces.
These showed the relative time at which the shockwave passed. They were used to determine the
consistency of the shock speed.

The order of events involved in a single acquisition was as follows. The shockwave in the shock tube
generated two signals – one from the timing PCB and one from the stagnation PCB. These two signal
caused the timer to start and stop respectively (measuring the time between the two PCB and thus the
shock speed). The stagnation signal was also passed to a pulse generator, the output of which triggered
both the LeCroy and 4-channel oscilloscope. Data was recorded on the LeCroy and the oscilloscope as
the supersonic flow passed through the Scramjet.
nozzle shock tube

Scramjet
stagnation timing
PCB PCB
PCB pressure transducers

Injector PCB
PCB power supply PCB power supply
Group 3
attenuation unit timer
PCB’s 1 - 8
PCB’s 9 -12
Lecroy Bank 4 Farnell Pulse
Waveform digitizer Generator
stagnation pulse
injector stagnation
Tektronix Tektronix Tektronix
4 channel 2 channel 2 channel

CRO CRO CRO
timing
Figure 13: Pressure data acquisition system

CHAPTER FIVE – RESULTS AND PRELIMINARY ANALYSIS
5. RESULTS AND PRELIMINARY ANALYSIS
5.1 VERIFICATION OF CFD-ACE WITH PREVIOUS RESULTS
Results in this section are in the form of pressures along the Scramjet floor and CFD colour images of
pressure and H2O mass fraction. H2O mass fraction gives an indication of where and to what degree
combustion is occurring in the duct.
5.1.1 Turbulent or laminar flow
The first two CFD models generated where flat, straight ducts, 350 mm long with no injector and with
turbulent and laminar flow respectively. The boundary layer thickness in each of these models was
measured at the outlet by determining the point at which the flow velocity was 99% of the free stream
velocity (this was easily determined using CFD-VIEW). The results were:
laminar model, boundary layer thickness - 1 mm

turbulent model, boundary layer thickness - 6 mm
The boundary layer thickness was also calculated using 1-D isentropic flow theory and pressures at the
inlet and outlet of the duct (obtained from an empty duct shot in the Scramjet) to calculate the
effective area reduction in the flow (as the flow travels downstream the boundary layer grows and
effectively reduces the area of the duct). The result was a boundary layer thickness of 7 mm. This
thickness corresponds most closely to the turbulent model. As a result, turbulent flow was selected for
the remainder of the models.
5.1.2 Grid resolution
For a Scramjet duct with thrust surface, an injector injecting hydrogen, and mixing (no combustion), a
model was constructed and a test performed to determine the effect of grid resolution on the pressures
on the floor of the duct. The grids were based on the grid that was used for all other models. The grid
at the original resolution (11294 cells) was compared with two other grids, one with the number of
grid points halved (2618 cells) and one with the grid points doubled (45664 cells). The results of this
test can be found below.
Chapter Five – Results and Preliminary Analysis 27

300000
250000
Fine Grid
Medium Grid
Coarse Grid
200000
Pressure [Pa]
150000
100000
50000
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Distance from SCRAMJET inlet [m]
Figure 14: Grid comparison
Figure 14 shows that the higher the grid resolution, the better the model resolves the shock and
expansion waves (peaks and troughs). The overall trends for all grids however are the same and on the
thrust surface floor, where the shock waves have weakened, the pressures are very similar. The shock
and expansion waves are generally in the same place (position in the flow) for all grids. Apart from
result accuracy, the other important issue in this test was the computational time. The coarse grid took
approximately 15 minutes to converge, the medium 45 minutes and the fine 3 hours. Considering the
accuracy of the three grids and the computational time it was decided the medium grid was most
appropriate.
5.1.3 Combustion model
For the two long injector configurations with 0 and 50 mm flat ducts, instantaneous, equilibrium, 2-
step finite rate and 7-step finite rate combustion models were used. A comparison of these reaction
models can be found below. Figure 15 shows the pressures along the floor of the Scramjet. It is
apparent that the instantaneous and equilibrium combustion models produce very similar results and
that the pressure rise associated with combustion for both occurs soon after the fuel exits the injector.
This is also seen in Figure 16 where the water combustion product is visible from immediately behind
the injector. Figure 17 and Figure 18 show that, for the finite rate models, there is a delay in the
ignition. This ignition delay is apparent in experimental pressure results (see later) and is a well
documented phenomena for the hydrogen – oxygen reaction. The hydrogen – oxygen reaction actually
occurs in two stages: the first step, where the temperature does not change substantially but production
of radicals (O, H and OH) occurs; and the second step, where heat is released and the temperature rises
rapidly. The time involved in the first step is significant and is called the ignition delay time. Figure 18
shows that for the 7-step model the ignition delay is longer than the 2-step.
For the remainder of the configurations only the 7-step finite rate model was used. This was because it
better represented the actual reactions occurring in the flow. Also, in some configurations, the two step

model was quite difficult to converge and required a ‘stiff’ reaction solver, which took longer to
converge than the 7-step.
Floor Pressure - Long injector, 50mm flat duct (li-50f)
200000
Instantaneous Combustion
180000 Equilibrium Combustion
7-step finite rate combustion
160000 2-step finite rate combustion
Non-combustion
140000
120000
Pressure [Pa]
100000
80000
60000
40000
20000
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Distance from duct entrance [m]
Figure 15: Combustion comparison
Figure 16: Equilibrium combustion - water mass fraction
Figure 17: 2-step finite rate combustion – water mass fraction

Figure 18: 7-step finite rate combustion – water mass fraction
5.1.4 Inlet pressure
Originally the inlet pressure used was 90 kPa. This figure was determined in a previous project19. As
more configurations were modelled it became apparent that overall the pressures were too high (when
compared with experimental results). The inlet pressure was therefore gradually reduced until the CFD
non-combustion results matched up with experimental non-combustion result. (It was thought that the
non-combustion model were predicting the flow more accurately than the combustion models). This
was further justified by analysing the pressure at the inlet and around the injector.
The 90 kPa inlet pressure was based on a PCB pressure transducer measurement (positioned above the
injector) of 140 kPa. This difference is due to the flow passing through a shockwave before it reaches
the transducer. When a inlet pressure of 90 kPa was used the CFD model predicted a pressure of 180
kPa at an equivalent position. When the inlet pressure was reduced to 70 kPa this pressure was 140
kPa.
5.1.5 CFD image features
Pressure is one of the better flow parameter from which to visualise the different processes occurring
in the Scramjet. The figure below shows the flow around the injector section.
Figure 19: Pressure around the injection
Clearly visible are the shock waves (red and yellow) caused initially by the injector leading edge. Not
so clear are the expansion waves (blue/green) originating from the back of the leading edge and rear of

the injector. Behind the injector, recompression shocks are visible after the flow has expanded around
the back of the injector.
The shock and expansion waves can be seen to reflect off the walls and injector of the Scramjet. The
shock and expansion wave become noticeably ‘smeared’ as they travel downstream. The first shock
waves are clearer than those downstream. This is one of the problems with any finite element method
and can only be partially improved by increasing the grid density. After the injector the flow becomes
more complex, with increasing interactions between the reflecting shock and expansion waves.
5.1.6 Overall Results
The figures below show the comparison of CFD and experimental results for each of the 5
configurations. The images below each plot show where on the Scramjet the pressure measurement
were taken. It should be noted that for the configuration li-0f, the grid resolution was double the other
configurations.
Floor Pressures - Long injector, 0mm flat duct (li-0f)

High grid density models
120000
100000
80000
Pressure [Pa]
60000
40000
Experiment - Combustion (9942)
Experiment - Non-combustion (9943)
20000
CFD - Non-combustion
CFD - Combustion - 7 step finite rate
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Figure 20: Floor pressure – long injector, 0mm flat duct

180000
160000 Experiment - Non-combustion (9947)
140000
CFD - Combustion - 7 step finite rate
120000
Pressure [Pa]
100000
80000
60000
40000
20000
0
0 0.1 0.2 0.3 0.4 0.5 0.6
500000

450000
CFD - Instantaneous Combustion (choked)
400000 CFD - Combustion - 7-step finite rate
350000
300000
Pressure [Pa]
250000
200000
150000
100000
50000
0
0 0.1 0.2 0.3 0.4 0.5 0.6

Floor Pressure - Short Injector, 80mm flat duct (si-80f)
160000
Experiment - 9948
140000 CFD - Non-combustion
CFD - Combustion - 7-step finite rate
120000
100000
Pressure [Pa]
80000
60000
40000
20000
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Figure 23: Floor pressure – short injector, 80mm flat duct
Floor Pressure - Short injector, 117mm flat duct (si-117f)
200000
180000 Experiment - Non-combustion (9950)
160000
CFD - Combustion - 7-step finite rate
140000
120000
Pressure [Pa]
100000
80000
60000
40000
20000
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Figure 24: Floor pressure – short injector, 1170mm flat duct

It should be noted that for the configurations where no experimental results are shown for the non-
combustion cases, data was unavailable.
5.1.7 Comparison of CFD and previous experimental results
The most obvious difference between the experimental and CFD results, for all the configurations, is
that the large fluctuations in the experimental results are not present in the CFD results. These
fluctuations are due to reflected shock and expansion waves in the duct. CFD-ACE is not particularly
good at resolving shock waves. The grid often is not of a high enough resolution to pick up the large
changes in fluid properties across them. Increasing the grid resolution does increase their “sharpness”,
(as shown in Figure 14) but at the expense of computational time. Figure 14 also shows that no matter
what grid resolution, the shock waves are “smeared’ as they travel downstream. This cannot be
avoided using CFD-ACE. In reality, shock waves decrease in strength as they are reflected off the
walls, but to a much lesser degree than apparent in CFD-ACE.
All 5 CFD configuration show there is an increase in pressure when comparing the combustion and
non-combustion cases. This is as expected and is due to the temperature rise associated with the
combustion reaction.
The CFD result using instantaneous combustion for the long injector with 87 mm flat duct
configuration shows a very large increase in pressure. This is due to the Scramjet “choking” (choking
occurs when flow in the Scramjet becomes subsonic). The experimental results show that the flow did
not choke, however this has been attributed to transient effects - it takes a finite time before the flow
chokes and the test times were too short to observe the final state of the flow. This was the only case
where the CFD results differed to such a large extent from the experimental.
For the non-combustion cases, CFD results show a good correlation with experimental results, in
terms of both overall trends (decreasing pressure as the floor diverges) and magnitude. This can be
partially attributed to the modification of the inlet pressure (see section 5.1.4).
The CFD combustion models show significantly less correlation with experimental results. Overall the
pressures are lower than the experiments would indicate. There are two likely reasons for this: grid
resolution and inlet conditions.
In all of the experimental combustion results, there is an initial sharp rise in the pressure in the duct. In
Figure 20 this can be seen approximately half-way along the thrust surface, in the remaining
configurations it can be seen toward the beginning of the thrust surface. This increase is due to the
pressure rise associated with combustion. It is thought that the combustion is initiated by the flow
passing through a shock wave. The shock wave increases the pressure and temperature above the
ignition conditions and combustion occurs rapidly. The ignition temperature is between 1100 – 1200
K. Changes over this range have a significant effect on the reaction rate. For the medium grid
resolution used in the above configurations, the pressure and temperature rise associated with the
shock waves was not ‘sharp’ enough (as seen in Figure 14) to reach the ignition temperature and so the
flow did not fully ignite. The difference in pressure between the non-combustion and combustion CFD
cases does show that combustion is occurring, however by doubling the grid resolution, as was the
case for long injector, 0mm flat duct configuration (Figure 20), the full pressure rise associated with
combustion can be seen. The pressure for the combusting mixture is still however lower than

experiments suggest. If the grid resolution was increased for the remaining configurations it is thought
that the larger pressure rise would be seen. Due to time constraints these models were not generated.
The other possible explanation for the difference in combustion results could lie in the inlet conditions
input to CFD. Since combustion is highly sensitive to temperature in the range 1000 – 1200 K, a slight
change in the inlet temperature could have a significant effect on combustion in the duct. Inlet
pressure was changed from the original values in order to reproduce the non-combustion results,
which could suggest that other parameters may not be correct. Uncertainty was also associated with
the inlet conditions, since for Doolan’s experiments the flow passed through a diffuser (a set of
converging plates) to slow the flow before it entered the Scramjet. The pressure and Mach number
were verified after the diffuser, but there was more uncertainty with the Scramjet inlet conditions than
if no diffuser was used (as was the case for this project).
5.1.8 Summary
The lessons learnt in this phase were very important for the subsequent CFD modelling. Turbulent and
laminar flows were investigated to see which was most appropriate; several different combustion
models were tested and the most appropriate chosen; and problem with low grid resolution identified.
The models used in the next phase of the project where based on the combustion and non-combustion
models shown above. The most important modifications to these were a doubling of the grid resolution
and more accurate predictions for the inlet conditions.
5.2 COMPARISON OF CFD AND EXPERIMENTAL RESULTS
In this section results obtained from CFD and experiments perform in this project are compared. For
conciseness, results from four representative configurations are shown below. Pressure measurements
for all configurations modelled and tested can be found in Appendix C.
Combustion Non-Combustion
300 300
250 250
200 200
Pressure [kPa]
150 150
100 100
50 50
Experiment Experiment
CFD CFD
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60
Distance from front of Scramjet [m] Distance from front of Scramjet [m]
Figure 25: Scramjet pressure – 200 mm, 1.75° thrust surface

300 300
250 250
200 200
Pressure [kPa]
150 150
100 100
50 50
CFD CFD
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60
Figure 26: Scramjet pressure – 150 mm flat floor, 7° thrust surface
300 300
250 250
200 200
Pressure [kPa]
150 150
100 100
50 50
CFD CFD
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60
Figure 27: Scramjet pressure – 100 mm flat floor, 1.75° thrust surface

300 300
250 250
200 200
Pressure [kPa]
150 150
100 100
50 50
CFD CFD
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.00 0.10 0.20 0.30 0.40 0.50 0.60
Figure 28: Scramjet Pressure – 0 mm flat duct, 7° thrust surface
The CFD results above show a much better correlation with experimental results than for comparisons
with Matthew Doolan’s results. As was mentioned previously this was most likely due to the increased
grid resolution and better estimate for the inlet conditions.
The above results show a much better prediction of the shock and expansion waves. All the major
pressure increases and decreases are predicted and the variations in pressures associated with the
shock and expansion waves are much greater in magnitude than in the previous results. The overall
magnitudes and trends associated with combustion, non-combustion and expansion around the thrust
surface are also well predicted. In CFD results, the large step down in pressure occurs over the initial
change in geometry around the thrust surface.
The most obvious difference between results is the position of the shock waves. For the combustion
cases CFD-ACE consistently predicted the shock wave downstream (to the right) of experimental
results. For the non-combustion cases CFD-ACE consistently predicted the shock wave upstream (to
the left) of experiments. Overall the positioning of the shock waves was better predicted in the CFD
combustion models than the non-combustion models. The logical explanation for these differences is
that the shock and expansion wave angles are wrongly predicted. There could be several reasons for
this:
1. The Mach number of the free stream is not right. As seen in section 2.5.1, the angle of an oblique
shock generated from a change in direction of the flow is dependant on the Mach number. An
increase in Mach number will produce a decrease in the shock angle and thus a change in the
position of the shock waves. This explanation would not however explain why there are
differences in position between the combustion and non-combustion models.
2. The other possibility is that the conditions in the hydrogen fuel stream are incorrect. Shock and
expansion waves are reflected and transmitted at varying angles as they pass through the hydrogen
stream. The angle that they are reflected and transmitted is dependant on the Mach numbers of
both the free stream air and the hydrogen stream. The combustion process would change the Mach
number in the hydrogen stream.

3. The third possibility is that for the non-combustion cases the nitrogen and air results cannot be
compared (Nitrogen was used in the free stream in experiments while air was used in CFD). Using
Nitrogen, which has a fractionally lower molecular weight, in the CFD models would have the
effect of increasing the shock wave angle and shifting the shock waves further to the left, however
it may have some other effect that could not be forseen.
Regardless of these differences, the above results show CFD predicts the flow in the Scramjet
surprisingly well. Shock and expansion waves are predicted accurately in magnitude (but not so well
in position) and overall pressures and trends are well predicted.

CHAPTER 6 - RESULTS ANALYSIS AND DISCUSSION
6. RESULTS ANALYSIS AND DISCUSSION
6.1 COMPARISON OF EXPERIMENTAL AND CFD THRUST CALCULATIONS
Combustion Non-Combustion Combustion Non-Combustion

Experiment CFD Experiment CFD Experiment CFD Experiment CFD
300 140
250 120
100
200
Thrust [N]
Thrust [N]
80
150
60
100
40
50
20
0 0
0 50 100 150 200 250 0 50 100 150 200 250
Flat floor length [mm] Flat floor length [mm]
(a) (b)
Figure 29: Comparison of CFD and experimental thrust calculations: (a) 7° thrust surface, (b) 1.75° thrust
surface
Figure 29 shows the comparison between thrust determined from CFD and from experiments for two
thrust surface angles. It shows that for the combustion cases, CFD thrust calculations are higher than
the experimental calculations while for the non-combustion cases the CFD results are consistently
lower than experiments. This is consistent with the comparison of pressure measurements (see
Appendix C).
The remaining analyses in this chapter are based on the CFD results.
6.2 EFFECT OF THRUST SURFACE ANGLE
Chapter Six – Results Analysis and Discussion 39

0
200 mm
-50 150 mm
100 mm
50 mm
Total Thrust [N]
-100
-150
-200
-250
-300
0 1 2 3 4 5 6 7 8
Thrust Surface angle [deg]
Figure 30: Effect of thrust surface angle on thrust generated (combustion)
Figure 30 shows the total thrust in the Scramjet as a function of the thrust surface angle. These values
are plotted for all four of the flat duct lengths. The thrust generated is actually negative since the shear
forces generated on the walls of the Scramjet are greater than the thrust developed by the thrust
surface. This is not surprising due to the low equivalence ratio used (0.5).
Clearly shown in Figure 30 is the trend of increasing thrust as the thrust surface angle is increased.
This is because as the angle is increased the component of the pressure force in the x direction is
increased. The maximum thrust is generated for the 7 degree case, however the limit to this increase is
undetermined. In hindsight larger thrust surface angles should have been investigated. The angles used
where based on the 3.5 degree angle used in previous experiments by Doolan and proposed by
researchers at the Japanese National Aerospace Laboratories (NAL). There was however some
confusion about this angle which resulted in the incorrect assumption that it would be appropriate to
investigate angles around this value for use as the thrust surface angle.
A single configuration with 15° thrust surface and 100 mm flat duct length was investigated after the
above results were found. The results of this are shown in Figure 31.

50
0
0 2 4 6 8 10 12 14 16
-50
Total Thrust [N]
-100
-150
-200
-250 200 mm
100 mm
-300
Thrust Surface angle [deg]
Figure 31: Thrust generated by 15° thrust surface
This shows that a 15° thrust surface generates positive 30 N of thrust. It transpired that previous
investigations have occurred for thrust surface angles around this value. Further investigation of thrust
surface angles above and below this value would be required to determine the maximum thrust
obtainable
6.3 EFFECT OF FLAT DUCT LENGTH
0
7
-50 5.25
3.5
1.75
Total Thrust [N]
-100
-150
-200
-250
-300
0 50 100 150 200 250
Flat duct length [mm]
Figure 32: Effect of flat duct length on thrust generated (combustion)
Figure 32 shows that varying the flat duct length has much less effect than varying the thrust surface
angle. It also shows that the thrust surface length (which is dependant on the length of the flat duct

length since the total length of the Scramjet is not varied) does not have as large an effect on the thrust
generated as the thrust surface angle. For the 7° thrust surface maximum thrust is produced for a flat
duct length of 100 mm. This is shown in more detail in Figure 33.
-20 7
Total Thrust [N]

-40
-60
-80
-100
-120
0 50 100 150 200 250
Figure 33: Variation of flat duct length for 7° thrust surface angle
The decreasing thrust at lower flat duct lengths is due to a lack of combustion in the duct. When the
duct is expanded too close to the injector, the expansion reduces the pressure and temperature before
the fuel has time to ignite. This can be seen in Figure 34 below.
Figure 34: H2O mass fraction – 50 mm flat duct, 7° thrust surface
Figure 34 shows that the bottom side of the fuel stream does not combust until near the end of the
duct. Figure 35 shows the water mass fraction for the next longest flat floor configuration with both
sides of the fuel stream combusting.
Figure 35: H2O mass fraction – 100 mm flat duct, 7° thrust surface
In this configuration the flat duct is sufficiently long to allow the ignition delay to occur before the
pressure and temperature is decreased by the expansion. For the zero flat duct configuration virtually
no combustion occurs (This can also be seen in Appendix C where the pressures from the combustion
and non-combustion cases are almost the same).

Expanding the flow too early causes the temperature to decrease below the ignition temperature
(between 1100 and 1200 K). Figure 36 shows the temperature over the range 1000 – 1200 K in the
region immediately after the expansion, for the 50 mm flat duct case. This highlights the subtle but
important temperature differences between the upper and lower sections of the duct. The upper
red/purple section just after the expansion (caused by a reflected shock wave) causes the upper
boundary of the fuel stream to ignite and combust. In the lower section, the expansion causes the
temperature to decrease below the ignition temperature and so the lower boundary of the fuel stream
does not combust.
Figure 36: Temperature around the expansion region (50 mm flat duct, 7° thrust surface)
For configurations with longer flat duct lengths, the temperature also drops below the ignition
temperature after the expansion, however the combustion process can sustains itself since the length is
such that it has already begun to combust further upstream.
These results highlight the important relationship between ignition delay time (or ignition delay
length) and flat duct length. They show that for combustion to occur in the duct the thrust surface must
not begin before ignition of the fuel-air mixture. The length of duct required for combustion, using the
free stream conditions investigated in this project, is 100 mm. For different free stream conditions this
value would be most dependant on free stream Mach number and temperature. This length could also
be reduced when using other injectors that increase the mixing efficiency in the duct12.
As the thrust surface angle is decreased the flat duct length has less effect on the combustion in the
duct. This can be seen for the 1.75° case in Figure 32 where thrust produced increases until 0 mm of
flat floor. In this case the thrust surface begins before the mixture begins to combust, however the
expansion is not large enough to cause the temperature to decrease below the ignition value. For the
3.5° and 5.25° cases the optimum flat duct length is undefined. The optimum flat duct length may vary
for the thrust surface cases between 1.75° and 7° depending on the effect of the expansion process,
however thrust surface angles above 7° would all have an optimum value of 100 mm (since regardless
of the angle, combustion begins before the thrust surface begins).
6.4 EFFECT OF THRUST SURFACE LENGTH

350
Incremental Thrust/T.S. Length

7
300
5.25
250 3.5
1.75
200
[N/m]
150
100
50
-50
0 0.1 0.2 0.3 0.4 0.5
Thrust Surface Length [m]
Figure 37: Incremental thrust / Thrust surface length
Figure 37 shows the effect of thrust surface length on the incremental thrust generated by the
Scramjet. The incremental thrust is the thrust generated in combustion minus the thrust generated in
non-combustion. Lower thrust surface lengths correspond to the long flat duct lengths. It shows that,
per unit length, the shorter thrust surface is more efficient at generating thrust, particularly for the
higher thrust surface angles. It should be stressed however that this represents the increase in thrust of
combustion over non-combustion. If only the combustion cases are considered – which is what we are
ultimately concerned with – the configurations corresponding to the 100 mm flat duct lengths generate
the most combustion (per unit length of thrust surface).
The relationship between thrust surface length and unit incremental thrust can be explained by the fact
the shorter the thrust surface length, the longer the flat duct length and therefore the higher the
pressure present at the beginning of the thrust surface. It is also useful, however, to look at where on
the thrust surface the majority of the thrust is generated. By considering the pressure in the duct
(shown in Figure 38 for the 100 and 200 mm flat duct lengths) it can be seen that a significant portion
of the incremental thrust is due to the large difference in pressure of the first compression wave (peak)
after the thrust surface begins. This pressure difference soon decreases. The advantage of having a
longer thrust surface becomes less, since the pressure difference toward the end of the duct is
relatively low. This means that the majority of the thrust is generated at the beginning of the thrust
surface and although increasing the length of the thrust surface increases the thrust, the thrust obtained
from the extra length is diminishing.

300 300
Combustion Combustion
Non-Combustion
250 250 Non-Combustion
Difference
Difference
Pressure [kPa]
200 200
150 150
100 100
50 50
0 0
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(a) (b)
Figure 38: Scramjet floor pressure for 7° thrust surface: (a) 200 mm flat duct, (b) 100 mm flat duct
6.5 EFFECT OF COMBUSTION
70
7
60 5.25
Incremental Thrust [N]
50 3.5
40 1.75
30
20
10
0
-10
0 50 100 150 200 250
Figure 39: Incremental thrust
Figure 39 shows the effect of increasing flat duct length on incremental thrust. This shows that when
considering the effect of combustion over non-combustion, the optimum flat duct length is not 100
mm. For the 7° thrust surface case the maximum incremental thrust is greater than 200 mm. The large
decrease in thrust associated with short flat duct lengths can also be seen clearly. For the 3.5° and
5.25° thrust surface cases the results do not show the maximum. The above graph shows there is a
different relationship between incremental thrust and flat duct length than for total thrust and flat duct
length. For maximum incremental thrust the optimum flat duct length varies depending on the thrust
surface angle.
The above figure shows the effect of isolating the difference between combustion and non-combustion
cases. As mentioned previously, since the ultimate aim is to generate the maximum total thrust for the
combustion case this leads to the conclusion that the maximum thrust is generated for 100 mm of flat
duct and a thrust surface angle greater than 7°.

SUMMARY AND CONCLUSIONS
7. SUMMARY AND CONCLUSIONS
During the course of this project the Computational Fluid Dynamics package CFD-ACE was used to
model the flow inside the ANU’s experimental Scramjet engine.
The project was broken down into three stages: verification of CFD-ACE with previous work;
optimisation of the Scramjet geometry using CFD-ACE; and experimental verification of the optimum
configuration. This involved developing numerous CFD models of Scramjet configurations and
performing experiments in the T3 shock tunnel.
To verify the operation of CFD-ACE several investigations were performed to develop an accurate
CFD model of the flow. Turbulent and laminar flow, grid resolution, inlet conditions and a number of
combustion models were all investigated. It was found that for the models developed pressure trends
along the floor of the Scramjet were well predicted for non-combusting flow, however for the
associated combusting flows pressure was somewhat under-predicted. Several explanations could
account for this, including uncertainties of the model inlet conditions and low grid resolution. It was
found that the low grid resolutions used for models in this stage of the project did not accurately
resolve the shock waves in the flow. As a result the peak pressures and temperatures in the duct were
not obtained and resulted in under-prediction of the combustion process.
CFD models were developed for configurations investigated in the current project with a higher grid
resolution and a more accurate calculation of the inlet conditions. CFD results showed a much better
correlation with the pressure measurements determined in the experimental stage of this project. Both
the position and magnitude of shock and expansion waves and the general trends in the Scramjet duct
were surprisingly well predicted. For the combustion models CFD-ACE consistently predicted
shockwave slightly upstream of experimental results while the non-combustion models predicted the
shockwaves to be downstream. Several suggestions were made for these differences, including
incorrect free stream Mach number or hydrogen fuel inlet conditions, and the invalid assumption of
comparing nitrogen and air free streams as equivalent.
Thrust generated by the Scramjet was calculated using CFD-ACE and included all pressure and shear
forces generated on the walls of the Scramjet. For all configurations initially investigated the thrust
generated was negative, ie. drag was developed.
Thrust was found to increase as thrust surface angle increased from 1.75° to 7° but no limit to this
increase was found. This was consistent for all the flat duct lengths. The maximum thrust developed
was shown to be greater than 7°. An additional configuration with 15° thrust surface angle was also
investigated. This developed a positive thrust significantly greater than the lower angled configuration.
More CFD modelling of thrust surface angles above and below this value would be required to
determine the optimal angle.
The optimum flat duct length was found to be 100 mm. Lengths below this value were found to hinder
the combustion process by expanding the flow before the combustion process began. For the smaller
flat duct cases the temperature was reduced below the ignition temperature and resulted in little or no
combustion. For the 50 mm flat duct case combustion occurred on the top side of the fuel stream but
Summary and Conclusions 46

not on the bottom. For the 0 mm flat duct case, ignition did not occur at all and resulted in very little to
no combustion.
The effect of thrust surface length was investigated by looking at the incremental thrust developed per
unit length. This highlighted the fact that per unit length the short thrust surfaces produced more thrust
than the longer length. This is partially attributed to the larger pressure in the duct at the beginning of
the expansion region (due to the longer length of duct before it) but is also due to a decaying pressure
differential that does not significantly increase as the thrust surface length increases. Put simply, the
majority of the incremental thrust is generated near the start of the thrust surface.
Overall, CFD-ACE was found to be a useful and accurate tool for modelling the supersonic,
combusting flow in a Scramjet engine. While the initial development of a CFD model was time
consuming, once an accurate model was determined, modifications of flow conditions and geometric
parameters could be easily and quickly made. It is interesting to note that obtaining experimental
pressure measurements was a quicker process than CFD modelling in terms of both overall time taken
and the time to investigate each configuration. However, CFD generates a much larger number of flow
parameters than can be experimentally determined and is significantly less expensive, in terms of both
personnel and equipment, than performing experiments in the shock tunnel. Also, once a model has
been developed and verified subsequent modelling is significantly quicker and easier than
experiments.
Hopefully this project provides a basis for subsequent CFD modelling of the Scramjet engine at the
ANU.
Summary and Conclusions 47

FURTHER WORK
8. FURTHER WORK
The are a number of areas from this project in which further work can be carried out. The first of these
is to investigate larger thrust surface angles. In this project angles of below 7° were investigated,
however it transpired that these value were to small. Investigation above and below 15° would be
required to find the optimum value. If experiments were to be performed using the Scramjet in the
shock tunnel, a number of modifications would be required so that the larger angles could be
accommodated.
Once both thrust surface angle and flat duct length are investigated changes to free stream and injector
flow conditions could be made to determined the effect of Mach number and equivalence ratio in the
duct. This could also include investigation of other flow conditions, including combustion efficiency.
This project looked at a two dimensional model of the Scramjet. The logical extension to this would be
to use a three dimensional model. For a three dimensional model, the more complex injectors currently
being studied at the ANU could be included in the model. These increase the mixing efficiency of the
fuel-air mixture. Also, the shock waves produce on the side walls could be captured. The disadvantage
with 3 dimension models is that they take significantly more computational time than two dimensional
models. Since run-time for the models in this project were up to 2 days (for the 7-step combustion
models), three dimensional models would be impractical unless a CFD package with shock-capturing
features (such as a self generating grid that increases grid resolution near shockwaves) could be used.
Further Work 48
REFERENCES
1
Erdos, J.I., Scramjet testing in Shock-Heated Tunels, ISSW21, Paper 9030, 1997
2
Heiser, W. H., Pratt, D. T., Hypersonic Airbreathing Propulsion, AIAA, Washignton DC, 1994, pp
22-26
3
Kerrebrock, J. L. Aircraft Engines and Gas Turbines, second edition, MIT Press, London, 1992
4
O’Byrne, S, (1997) Examination of Transient Mixing and Combustor Geometry in a Scramjet,
Masters Thesis,
5
Curran, E.T., Heiser, W.H., Pratt, D.T., Fluid Phenomena in Scramjet Combustion Systems, Annual
Review of Fluid Mechanics, Vol. 28, 1996
6
Fox, J.S., O’Byrne, S.O., Gaston, M.J., Houwing, A.F.P., PLIF and Shadowgraph Imaging of
Supersonic Mixing and Combustion, 1997
7
Doolan, M.C., (1997) Optimising the Combustor Geometry in a Supersonic Combustor Ramjet,
Honours thesis, Engineering Department, Australian National University
8
Rabbath, P.A., (1994) Investigation of supersonic combustion in a Scramjet by using flow
visualisation and pressure measurements, Honours thesis, Engineering Department, Australian
National University
9
Fox, J.S., (1998) PhD thesis, to be submitted
10
Gaston, M.J., Mudford, N.R., Houwing, A.F.P., A Comparison of Two Hypermixing Fuel Injectors
in a Supersonic Combustor, AIAA Journal of Propulsion and Power, 1997
11
CFD Research Corporation, CFD-ACE, Graphical User Interface Manual, Version 2.2, May 1997
12
Gaston, M.J., (1998) PhD thesis, to be submitted
13
CFD Research Corporation, CFD-GEOM , May 1997
14
CFD Research Corporation, CFD-VIEW, 3-D Computer Graphics and Animation Software, Version
3.6, May 1997
15
Brescianini, C & Morgan, R, Modelling of a Scramjet Flow Using Various Turbulence Models, 11th
Australasian Fluid Mechanics Conference, December 1992
16
CFD Research Corporation, CFD-ACE Theory Manual, September 1993
17
Rogers, R.C., Chinitz, W., Using a Global Hydrogen-Air Combustion Model in Turbulent Reacting
Flow Calculations, AIAA Journal, Vol. 21 No. 4
18
Anderson, John D., Modern Compressible Flow, 2nd edition, McGraw-Hill, New York, 1990
19
Gaston,M.J. (1998) PhD thesis to be submitted
49

Thesis Hyslop

Uploaded by

Copyright:

Available Formats

Thesis Hyslop

Uploaded by

Document Information

Original Description:

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

Thesis Hyslop

Uploaded by

Copyright:

Available Formats

DEPARTMENT OF ENGINEERING

AUSTRALIAN NATIONAL UNIVERSITY

CFD MODELLING OF SUPERSONIC COMBUSTION

Supervisors: Dr Frank Houwing

Finally, thankyou to my parents for the years of encouragement and support.

5. RESULTS AND PRELIMINARY ANALYSIS 27

6. RESULTS ANALYSIS AND DISCUSSION 39

7. SUMMARY AND CONCLUSIONS 46

APPENDIX A – CFD-ACE GRIDS 51

APPENDIX B – LIST OF SHOCK TUNNEL SHOTS 52

APPENDIX C – SCRAMJET MODIFICATION DRAWINGS 53

APPENDIX D – ADDITIONAL CFD AND EXPERIMENTAL RESULTS 62

1.1 SCRAMJET ENGINES

Chapter One – Introduction 1

Figure 1: Generic Scramjet engine

Chapter One – Introduction 2

1.2 ANU’S SCRAMJET

Chapter One – Introduction 3

2.1 ANU’S SCRAMJET

Chapter Two – Background 4

Figure 3: Inner sections of ANU’s Scramjet

2.2 THE T3 FREE-PISTON SHOCK TUNNEL

Figure 4: T3 shock tunnel schematic (taken from ALDiR website)

Chapter Two – Background 5

To produce supersonic flow in the tunnel the following events occur:

1. A high pressure is built up in the high pressure reservoir

2.3 PREVIOUS WORK

Mach number, M = 2.5

Chapter Two – Background 6

Chapter Two – Background 7

2.4.1 Solution Methodology and Governing Equations

Chapter Two – Background 8

2.4.2 Turbulence Models

where µlaminar is the laminar viscosity and ρ the density.

2.4.3 Reaction Models

Chapter Two – Background 9

The equilibrium model

The multi-step finite rate model

k = ApT n e ( − Ea /RT ) (2.5)

where: k is the reaction rate coefficient

Ap, Ea/R and n are determine experimentally for a particular reaction.

2-step finite rate reaction

No. Reaction Ap n Ea/R

7-step finite rate reaction

Chapter Two – Background 10

2.5 SUPERSONIC FLOW THEORY

expansion wave normal shock wave

M2<M1 M3>M2 M4<M3<1

Figure 6: Shock and expansion waves

2.5.1 Oblique shockwave relation

Chapter Two – Background 11

Figure 7: Oblique shockwaves

2.5.2 Expansion waves

2.5.3 Normal shockwaves

Chapter Two – Background 12

2.6 EQUIVALENCE RATIO

Chapter Two – Background 13

1. Verification of CFD-ACE with previous work

3.2 VERIFICATION OF CFD-ACE WITH PREVIOUS WORK