Modeling and Simulation of Multiphase/Multicomponent Flows

Modeling and Simulation of
Multiphase/Multicomponent Flows
by
Randy S. Lagumbay
B.S., MSU-IIT, Philippines, 1997
M.S., The University of New South Wales, Australia, 2001
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Department of Mechanical Engineering
2006
This thesis entitled:
Modeling and Simulation of Multiphase/Multicomponent Flows
written by Randy S. Lagumbay
has been approved for the Department of Mechanical Engineering
Prof. Oleg V. Vasilyev
Prof. David Kassoy
Prof. John Daily
Prof. Sedat Biringen
Dr. Jin Wang
Date
The final copy of this thesis has been examined by the signatories, and we find that
both the content and the form meet acceptable presentation standards of scholarly
work in the above mentioned discipline.
iii
Randy S. Lagumbay, (Ph.D., Mechanical Engineering)
Modeling and Simulation of Multiphase/Multicomponent Flows
Thesis directed by Prof. Oleg V. Vasilyev
Most of the existing approaches in dealing with multiphase/multicomponent flows
are limited to either single-phase multicomponent or multiphase single-component mix-
tures. To remove this limitation, a new model for multiphase/multicomponent flows with
an arbitrary number of components in each phase is developed. The proposed model is
based on a homogenized fluid mixture approach. The model is hyperbolic and gives an
accurate value for the mixture speed of sound when compared to experimental data. A nu-
merical method based on the homogenized mixture formulation and extension of Roe and
HLLC scheme is developed. The new approach for modeling and numerical simulation of
multiphase/multicomponent flows is applicable to a variety of flow configurations includ-
ing shock waves, cavitation, single and multiphase turbulent flow, free surface flow, high-
pressure and high-speed flows. In order to validate the proposed approach for a variety of
flow regimes, a novel “idealized” fluid-mixture model is proposed and an exact solution
for the multiphase/multicomponent Riemann problem in one dimension is derived. A
number of existing benchmark problems for single-phase multicomponent flows becomes
a subset of this new model problem. To verify the accuracy of the proposed numerical
method and to demonstrate the physical capability of the proposed model, three classical
benchmark problems (single-phase two-component shock tube, shock wave propagation in
a single-phase two-component fluid, and single-phase shock-bubble interaction) and two
novel benchmark problems for an “idealized” fluid-mixture model (two-phase shock-tube
and two-phase rarefaction problems) are presented. Furthermore, the flexibility of the
new approach is demonstrated by extending the methodology to cavitating and turbulent
flows. Representative cavitation and multiphase/multicomponent subgrid scale models

iv
in the context of direct numerical simulations and large eddy simulations are considered
and numerical simulations demonstrating these capabilities are performed.

Dedication
To my father, mother, brother and sister.

vi
Acknowledgements
This thesis would not have been possible without the support of many people. I
would like to express many thanks with deepest sincerity to my advisor, Professor Oleg
V. Vasilyev, for his tremendous support and guidance, expertise and excellent advice,
patience and understanding throughout this work. I would like to acknowledge the
support and help of Professor Andreas Haselbacher for providing the Rocflu code and
giving valuable comments and suggestions. Thanks to my committee members, Profes-
sor David Kassoy, Professor John Daily, Professor Sedat Biringen and Dr. Jin Wang
who offered guidance and support. I would like to acknowledge the support of Argonne
National Laboratory for sponsoring this work under grants number 2-RP50-P-00005-
00, 3B-00061, 4B-00821, 5F-00462. Also, the support from MSU-IIT are gratefully
acknowledge. I would like to extend my appreciation to my friends and colleagues in
MU-Columbia and CU-Boulder, Dr. Dan Goldstein, Tanarit Sakulyanontvittaya, Dr.
Alexei Vesolainen, Jonathan Regele, Qianlong Liu, Dr. Chetan Malhotra, Dr. Hrishik-
ish Panchawagh and Myongjai Lee for their encouragements and discussions. I am very
grateful and thankful to my father, mother, brother, sister, relatives and numerous
friends who endured this long process with me, always offering support and love. Above
all, I offer this success to the Almighty God for giving me knowledge, strength, courage
and determination.
vii
Contents
Chapter
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Experimental Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Objectives of the Research . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Mathematical Formulation 15
2.1 Homogenized-Mixture Model . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Transport Variables Model . . . . . . . . . . . . . . . . . . . . . 19
2.3 Mixture Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.1 Mixture Pressure using Dalton’s Law . . . . . . . . . . . . . . . . 21
2.3.2 Mixture Pressure using Amagat-Leduc Law . . . . . . . . . . . . 22
2.4 Mixture Speed of Sound . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Idealized Fluid-Mixture Model 30
3.1 Density and Equation of State for Idealized Fluid-Mixture . . . . . . . 30
3.2 Entropy for Idealized Fluid-Mixture . . . . . . . . . . . . . . . . . . . . 32

viii
3.3 Speed of Sound for Idealized Fluid-Mixture . . . . . . . . . . . . . . . . 34
3.4 Riemann Invariants for Idealized Fluid-Mixture . . . . . . . . . . . . . . 34
3.5 Riemann Problem for Idealized Fluid-Mixture . . . . . . . . . . . . . . . 35
3.5.1 Left Shock Wave Analysis . . . . . . . . . . . . . . . . . . . . . . 37
3.5.2 Right Shock Wave Analysis . . . . . . . . . . . . . . . . . . . . . 38
3.5.3 Left Rarefaction Wave Analysis . . . . . . . . . . . . . . . . . . . 39
3.5.4 Right Rarefaction Wave Analysis . . . . . . . . . . . . . . . . . . 40
3.5.5 Complete Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4 Numerical Method for Simulation of Multiphase/Multicomponent Flows 42
4.1 Extended Roe’s Approximate Riemann Solver . . . . . . . . . . . . . . . 42
4.2 Extended HLLC Approximate Riemann Solver . . . . . . . . . . . . . . 49
5 Code Verification and Model Validation 54
5.1 Single-Phase Two-Component Shock-Tube Problem . . . . . . . . . . . 55
5.2 Shock-Wave Propagation in a Single-Phase Two-Component Fluid . . . 55
5.3 Two-Phase Shock-Tube Problem for Idealized Fluid Mixture . . . . . . . 58
5.4 Two-Phase Rarefaction Problem for Idealized Fluid Mixture . . . . . . . 60
5.5 Single-Phase Two-Component Shock-Bubble Interaction . . . . . . . . . 61
6 Multiphase and Multicomponent Free Surface Nozzle Jet Flow 67
6.1 Shock Tube Analysis in an Expanded Duct . . . . . . . . . . . . . . . . 68
6.2 Multiphase/Multicomponent Free Surface Jet Flow . . . . . . . . . . . . 70
6.2.1 Jet Formation and Jet Radial Profile . . . . . . . . . . . . . . . . 74
6.3 Multiphase/Multicomponent Cavitating Nozzle Jet Flow . . . . . . . . . 77
6.3.1 Cavitation Overview . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.3.2 Cavitation Modeling and Simulation . . . . . . . . . . . . . . . . 80
6.3.3 Cavitation in Nozzle Injector . . . . . . . . . . . . . . . . . . . . 84

ix
6.4 The Effect of Fluctuating Inflow Condition to the External Jet Formation 90
7 Multiphase and Multicomponent Turbulent Flows 92
7.1 Turbulent Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.2 Multiphase/Multicomponent Large Eddy Simulation . . . . . . . . . . . 94
7.2.1 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2.2 Subgrid Scale (SGS) Model . . . . . . . . . . . . . . . . . . . . . 97
7.3 Large Eddy Simulation of a Plane Jet . . . . . . . . . . . . . . . . . . . 100
7.3.1 Single Phase LES of a Plane Jet . . . . . . . . . . . . . . . . . . 102
7.3.2 Multiphase/Multicomponent LES of a Plane Jet . . . . . . . . . 109
8 Conclusions and Future Research 119
8.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8.2 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Bibliography 125
x
Tables
Table
5.1 Properties of air and helium used in the simulations of shock-bubble
interaction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2 Comparison of computed velocities with those measured by Haas and
Sturtevant [53] and computed by Quirk and Karni [105]. . . . . . . . . . 65

xi
Figures
Figure
1.1 Complex physics of the jet inside and near the nozzle region. . . . . . . 3
1.2 Cavitation in the vena contracta region. Reference: [114] . . . . . . . . . 5
1.3 Shock wave generation in a gaseous medium due to a high pressure and su-
personic jet flow. The image of shock wave is captured using synchrotron
x-radiography. Reference: [89] . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Speed of sound of a water-air mixture at P = 1 · 105 Pa and T = 298.15K
as a function of the volume fraction of air. Speed of sound predicted by
Eq. (2.50). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
as a function of the volume fraction of air. Comparison of predicted speed
of sound with experimental data of Karplus [72] for frequencies of 1 kHz
(diamonds), 0.5 kHz (squares) and extrapolated to zero frequency (circles). 26
as a function of the mass fraction of air. Speed of sound predicted by Eq.
(2.50) and using Eq. (2.4). . . . . . . . . . . . . . . . . . . . . . . . . . . 27
as a function of the mass fraction of air. Comparison of predicted speed
of sound with experimental data of Karplus [72] for frequencies of 1 kHz
(diamonds), 0.5 kHz (squares) and extrapolated to zero frequency (circles). 27

xii
2.5 Speed of sound of a liquid-gas mixture predicted by Eq. (2.50) at P =
1·105 Pa and T = 298.15K, for a liquid-gas density ratio, ρliquid /ρgas ≈ 100. 28
2.6 Speed of sound of a liquid-gas mixture predicted by Eq. (2.50) at P = 1 ·
105 Pa and T = 298.15K, for a liquid-gas density ratio, ρliquid /ρgas ≈ 1000. 28
2.7 Speed of sound of a liquid-gas-vapor mixture predicted by Eq. (2.50) at
P = 1 · 105 Pa and T = 298.15K. . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Structure of the solution of the Riemann problem for the idealized-mixture
formulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Four possible wave patterns of the solution of the Riemann problem [141]:
(a) left rarefaction, contact, and right shock; (b) left shock, contact, and
right rarefaction; (c) left rarefaction, contact, and right rarefaction; (d)
left shock, contact, and right shock. . . . . . . . . . . . . . . . . . . . . 37
4.1 The Riemann wave diagram used in the Roe scheme. . . . . . . . . . . . 45
4.2 The simplified Riemann wave diagram with two intermediate states used
in the HLLC scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.1 Comparison of numerical (◦) and exact (solid line) solutions for single-
phase two-component shock-tube problem at t ≈ 517 µs. . . . . . . . . . 56
5.2 Comparison of numerical (◦) and exact (solid line) solution for shock-
wave propagation in a single-phase two-component fluid at t ≈ 864 µs. . 57
5.3 Comparison of numerical (◦) and exact (solid line) solution for two-phase
shock-tube problem for an idealized fluid mixture at 240 µs. . . . . . . . 59
5.4 Comparison of numerical (◦) and exact (solid line) solution for two-phase
rarefaction problem for an idealized fluid mixture at 540 µs. . . . . . . . 60
5.5 Schematic diagram of the computational domain for the shock-bubble
interaction problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
xiii
5.6 Numerically generated Schlieren images of the shock-bubble interaction
computations on the fine grid at various times. . . . . . . . . . . . . . . 63
5.7 Numerically generated Schlieren images of the shock-bubble interaction
computations at approximately 674 µs for different grid spacings. . . . . 64
5.8 Definition of locations at which velocities are measured. VS - incident
shock, VT - transmitted shock, Vui - upstream edge of bubble, and Vdi -
downstream edge of bubble. . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.1 Plot of (a) mass fraction of liquid, (b) x-velocity component, (c) mix-
ture pressure, and (d) mixture temperature, after 5.7689 × 10−07 seconds
through an expanded duct, showing the normal traveling shock. . . . . . 69
6.2 Plot of: (top left) mixture density showing an oblique shock wave in a
gaseous medium, (top right) contact discontinuities of the mixture density
across the interface in the centerline of the jet axis, (bottom left) mass
fraction of the liquid, and (bottom right) contact discontinuities of the
mass fraction of the liquid across the interface in the centerline of the jet
axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3 Plot of: (top left) mixture pressure showing an oblique shock wave in
a gaseous medium, (top right) mixture pressure along the centerline of
the jet axis, (bottom left) mixture temperature showing an oblique shock
wave in a gaseous medium, and (bottom right) mixture temperature along
the centerline of the jet axis. . . . . . . . . . . . . . . . . . . . . . . . . 73
6.4 Plot of: (top left) contact discontinuities of the x-velocity component of
the mixture across the interface in the centerline of the jet axis, (top
right) velocity profile of the jet at 1.8 mm from the exit of the nozzle,
and (bottom) mixture x-velocity component showing an oblique shock
wave in a gaseous medium. . . . . . . . . . . . . . . . . . . . . . . . . . 74

xiv
6.5 Plot of the mass fraction of the liquid. The leading edge of the jet is
about 3.0 mm from the nozzle exit. . . . . . . . . . . . . . . . . . . . . . 75
6.6 Jet radial profile across the body of the jet at a distant of about 3.0 mm
from the nozzle exit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.7 Comparison of the results between numerical simulation (bottom) and
experimental observation of MacPhee et al. [89](top). . . . . . . . . . . 76
6.8 Plot of the mass fraction of the vapor after 2.000e-04 sec. Inflow condi-
m
tion: P = 1,161,000 Pa; u = 10 s2 ; domain length = 6 mm; orifice over
at
nozzle entrance ratio: ai = 0.3 . . . . . . . . . . . . . . . . . . . . . . . 82
6.9 Plot of the mass fraction of the vapor after 2.202e-04 sec. Inflow condi-
m
tion: P = 1,161,000 Pa; u = 10 s2 ; domain length = 6 mm; orifice over
at
nozzle entrance ratio: ai = 0.3 . . . . . . . . . . . . . . . . . . . . . . . 82
6.10 Plot of the mass fraction of vapor after 2.502e-04 sec. Inflow condition:
m
P = 1,161,000 Pa; u = 10 s2 ; domain length = 6 mm; orifice over nozzle
at
entrance ratio: ai = 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.11 Plot of the mass fraction of vapor after 3.000e-04 sec. Inflow condition:
m
P = 1,161,000 Pa; u = 10 s2
; domain length = 6 mm; orifice over nozzle
at
entrance ratio: ai = 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.12 Plot of the pressure after 3.000e-04 sec. Inflow condition: p = 1,161,000
m
Pa; u = 10 s2
; domain length = 6 mm; orifice over nozzle entrance ratio:
at
ai = 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.13 Plot of the initial condition of the system. The nozzle region is filled with
liquid and the chamber downstream from the exit of the nozzle is filled
with gas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.14 Plot of the computational mesh and boundary conditions of the two-
dimensional plane model nozzle injector. . . . . . . . . . . . . . . . . . . 85

xv
6.15 Cavitation inside the nozzle injector. Experimental results taken from
Roosen et al. [114]. Injection pressure, Pinlet = 80 bar; exit pressure,
Pexit = 11 bar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.16 Cavitation inside the nozzle injector. Numerical simulation by Yuan et
al. [155]. Injection pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar. 87
6.17 Results of the numerical simulation for a cavitating nozzle injector using
the multiphase mixture formulation with a cavitation model. Injection
pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar. . . . . . . . . . 87
6.18 Contour plot of the vapor (left) and gas (right) volume fraction inside
and close to the exit of the nozzle injector. Constant steady injection
pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar. . . . . . . . . . . 88
6.19 Contour plot of the liquid volume fraction (left) and velocity profile
(right) inside and close to the exit of the nozzle injector. Constant steady
injection pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar. . . . . . 89
6.20 Plot of the unsteady external jet formation (left) and distribution of vapor
volume fraction (right) inside and close to the exit of the nozzle injector.
Periodic unsteady injection pressure, Pinlet = 80 ± 10 bar; frequency,
f = 3.725 mhz; exit pressure, Pexit = 11 bar. . . . . . . . . . . . . . . . 91
7.1 Computational grid on a x - y section. . . . . . . . . . . . . . . . . . . . 101
7.2 Zoom view near the inlet region. . . . . . . . . . . . . . . . . . . . . . . 101
7.3 Inflow boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.4 Profiles of the longitudinal velocity at different streamwise locations. . . 104
7.5 Profiles of the streamwise turbulence intensity at different locations. . . 104
7.6 Contours of the magnitude of vorticity on an xy-plane at different time.
(a) t=5.0E-05sec., (b) 1.04E-04sec., (c) 1.4E-04sec., (d) 1.88E-04sec. . . 105

xvi
7.7 Snapshots of contours of the (a)magnitude of vorticity on an xy-plane,
(b) velocity, (c) density and (d) pressure at time, t=2.05E-04sec. . . . . 106
7.10 Comparison of a snapshot of turbulent jet. (a) extended LES approach
for multiphase/multicomponent flows, (b) LES approach from [6]. . . . . 109
7.11 Snapshots of multiphase jet at time, t=5.0E-04 sec.(a) mass fraction of
liquid, (b) mass fraction of gas, (c) mass fraction of vapor. . . . . . . . . 111
liquid, (b) vorticity, (c) density, (d) energy, (e), pressure and (f) velocity
vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Chapter 1
Introduction
This thesis focuses on the development of analytical and computational approach
for modeling and numerical simulation of multiphase (liquid or gas) and multicompo-
nent (several species of the same phase) flows, which is applicable to a variety of flow
configurations including shock waves, cavitation, single and multiphase turbulent flows,
free surface flow, high-pressure and high-speed flows. The approach is demonstrated by
solving many problems with different flow configurations.
1.1 Motivation
Multiphase and multicomponent flows are common in many engineering applica-
tions. Relevant examples are fuel sprays in combustion process, liquid-jet machining
of materials, and steam generation and condensation in nuclear reactors. The physical
mechanisms underlying multiphase and multicomponent flows as well as the interplay of
these mechanisms is very complex. In multiphase and multicomponent flows the phases
and/or components can assume a large number of complicated configurations; small-
scale interactions between the phases can have a profound impact on macroscopic flow
properties [54]. The modeling and numerical simulation of multiphase and multicompo-
nent flows poses far greater challenges than that of single-phase and single-component
flows. These challenges are due to interfaces between phases and large or discontinuous
property variations across interfaces between phases and/or components.

2
High-pressure and supersonic multiphase and multicomponent jet flow is one of
the most challenging problems in multiphase flow due to the complexity of the dynamics
of the jet. For example, the presence of cavitation and gas entrapment inside the
nozzle orifice can greatly affect the development and formation of the external jet.
Experimental evidence demonstrates that cavitation within the nozzle influences the
characteristics of the nozzle exit spray [60, 128]. The jet exits the nozzle orifice as
a bubbly fluid, which carries some of the trapped gas and cavitation bubbles. The
jet properties behave as a multiphase mixture of liquid, gas and vapor. Due to a high
injection pressure and a small nozzle orifice, the multiphase and multicomponent jet flow
often exits the nozzle orifice at supersonic condition with respect to the surrounding gas;
hence, generates an oblique shock wave in the gaseous medium [89]. Figure 1.1 shows a
sketch of the complex physics of a supersonic multiphase and multicomponent jet flow
inside and near the nozzle region.
Most of the current research on high-pressure supersonic multiphase and multi-
component jet flows through a gaseous medium is experimental, and the evolution of
the jet hydrodynamics is not completely and clearly revealed. The understanding of
the flow pattern of the jet is still rather limited. The characteristic properties of the
jet depend on many factors such as whether flow is laminar or turbulent, subsonic or
supersonic, and single phase or multiphase/multicomponent. The jet could be a sin-
gle phase or multiphase/multicomponet fluid. A typical multiphase/multicomponet jet
consists of bubbles that are carried in the flow stream. These bubbles could be the va-
por generated in the nozzle orifice due to cavitation and/or could be the gas entrapped
inside the nozzle. In addition, jet velocity could be larger than the speed of sound in
the gaseous medium resulting in generation of shock waves to the surrounding gas en-
vironment, and producing compression waves inside the jet in an opposite direction of
the jet velocity. Another phenomenon is the jet entrained near the surrounding media
towards the direction of the flow, which creates a secondary flow of the surrounding
3
Figure 1.1: Complex physics of the jet inside and near the nozzle region.
medium. Further downstream, the surrounding medium near the main jet flow diffuses
due to the viscous diffusion of the momentum.
1.2 Experimental Study
Experimental investigation in this area has been rare and detailed quantitative
data is very limited due to the difficulties involved resulting from the small size of the
injector nozzle that ranges from 100 µm - 1 mm, high injection pressure that may reach
upwards to 2000 bar in a very short time period, the high speed flow, which may reach
supersonic velocity [155], and the cost and difficulty of manufacturing experimental
devices. The aforementioned time frame, on the order of 1 - 10 µs [155], presents the
greatest challenge to modeling the internal flow of a nozzle and to capturing the process
of cavitation formation for investigation.
Most experimental studies have been performed on large scale transparent models
in order to visualize the cavitation structure. The first cavitation water tunnel was built
by Sir Charles Parsons in 1895 after identifying the reason why the ship propellers failed
to produce the design thrust. However, Parsons experienced with unsuccessful results of
the trials as reported by Burrill [21]. In 1910, Parsons constructed a large water tunnel,
4
which could test 12 in. diameter propellers under cavitating conditions. During the
period of World War II, a few other tunnels for propeller research with the refinements
and improvements over Parson’s tunnel were built in Europe and the USA. By the
same period, the general interest in cavitation phenomena was extended (1) to tests in
venturi meters and other conduits with restrictions, (2) to circuits for tests of single-
hydrofoil sections and (3) to variable-pressure circuits for hydraulic pump and turbine
investigations. Knapps et al. [75] have presented and discussed some special equipment
developed to produce cavitation under controlled conditions in a specified locations,
where observations can be made.
After the end of World War II, the interest in cavitation phenomena increased
and many of the research has focused on the effect of cavitation development in in-
jection nozzles to the jet formation and spray atomization. In 1959, Bergwerk [16]
performed an experimental study of the flow pattern in diesel nozzle spray holes. It was
observed during the experiment that at high injection rates, the fluid pressure in the
vena contracta region inside the nozzle decreased below the saturation vapor pressure
of the fluid, which causes a vapor formation in that region [114]. Similar observations
have been reported in the experiments described in Refs. [7, 84, 99, 114]. Figure 1.2
shows cavitation in the vena contracta region. Although this cavitation can improve
jet break-up [7, 109] and can enhance jet atomization [134], it increases the hydraulic
resistance of the nozzle and produces noise and vibrations that can cause damage to
the working surfaces [75]. The experimental research in nozzle cavitation continues to
progress. Hiroyasu et al. [63] have studied the internal flow in a nozzle injector and the
effect of cavitation in the break up length of a liquid spray. Arcoumanis et al. [9] have
investigated the development of cavitation in a vertical multi-hole diesel injector. Sote-
riou et al. [128] have studied the effect of cavitation and hydraulic flip on atomization
in direct injection diesel sprays, while He et al. [60] and Ruiz et al. [115] have looked
into the effect of cavitation and turbulence on high-speed atomization. The spectral
5
Figure 1.2: Cavitation in the vena contracta region. Reference: [114]
characteristics of turbulent flow in a diesel fuel nozzle injector were studied by Knox et
al. [77]. The application of refractive index matching to study the internal flow in a
diesel nozzle injector were conducted by Arcoumanis et al. [10]. The use of laser light
illumination to visualize the development of cavitation in the orifice of a diesel injector
were performed by Soteriou et al. [129]. Ganippa et al. [48] have used a transparent
scaled-up single-hole diesel nozzle to study the cavitation phenomenon.
Further experimental studies of cavitation have shown that the characteristics of
the flow from a large scale transparent models differs from an actual scale [8]. The
difference is caused by the flow scaling effects associated with imperfections in micro-
and macro- geometry of the flow and the nature of the flow. Because of this flow scaling
effects, many researchers [8, 12, 23, 25, 69, 83, 117, 134] have used an actual scale in
conducting an experiments in order to determine the correct characteristics of the flow.
Recent experimental studies have focused on the spray cone angles for a cavitating
nozzle flow [25, 152], cavitation oscillation frequencies [23], and the length of the cavity
[117]. Payri et al. [100] examined the influence of cavitation on the internal flow and the
6
macroscopic behavior of spray in diesel injection nozzles, and concluded that cavitation
leads to incremental changes of the spray cone angle and increases the outlet velocity.
So far, most of the experimental research mentioned in the previous paragraph
has focused on the development of cavitation inside the nozzle injector and the effect
of cavitation to the jet formation and spray atomization. However, none of them have
paid attention close to the exit of the nozzle region. Over the past few years, despite of
the substantial advances in laser diagnostics [26, 62, 156], the region close to the nozzle
still has not yielded desired quantitative information. A high-pressure fuel spray has
never been recognized as supersonic under typical fuel injection conditions [97, 124].
Recently, MacPhee et al. [89] have used a synchrotron x-radiography and a fast x-ray
detector to record the time evolution of the transient fuel sprays from a high-pressure
injector. In their experiment, the propagation of the spray-induced shock waves in a
gaseous medium were captured and the complex nature of the spray hydrodynamics
were revealed. They have found out that under injection conditions similar to those in
operating engines, the fuel jets can exceed supersonic speeds and result an oblique shock
wave in the gaseous medium, see Figure 1.3. However, the effect of this shock wave to
the atomization of the fuel and the combustion processes is currently not known.
1.3 Numerical Study
In the experimental studies, the understanding of the complex nature of the high-
pressure supersonic multiphase jet flow is limited. However, numerical and theoretical
investigations complement each other and provide detailed quantitative explanations of
the complex structure of the jet from single phase flow to a more complex multiphase
flows. Numerical modeling and simulation has become the best of the alternative tools
to supplement experimental study, and provide very promising results in improving the
understanding of the complex nature of the jet flow. With the advancement of computer
technology and improvement of numerical technique, the majority of current research

7
Figure 1.3: Shock wave generation in a gaseous medium due to a high pressure and su-
personic jet flow. The image of shock wave is captured using synchrotron x-radiography.
Reference: [89]
have focused on numerical modeling and simulation.
Bunnell et al. [20] have performed a three-dimensional simulation of unstable
cavitating flows in injector passages and have demonstrated the effect of cavitation on
both the mean and unsteady components of the orifice discharge coefficient. Similarly,
Dirke et al. [33] have carried out three-dimensional simulations of cavitating flows
in diesel injectors and have shown the distribution of cavitation zones. Yuan et al.
have demonstrated the strong interaction of a cavitating nozzle flow with the external
jet formation [155] and the effect of injection pressure fluctuations on the cavitation
processes in injection nozzles [154]. A numerical study of cavitating flow through various
nozzle geometries was performed by Schmidt et al. [122] and the results show that the
upstream geometry has a small influence on the nozzle flow. A numerical simulation of
three-phase (air, fuel liquid, and fuel vapor) flow in a high pressure swirl injectors was
presented by Alajbegovic et al. [5]. A formation of a thin conical fuel sheet with an air
core and cavitation in the pressure depression initiated in the air core was predicted in
8
the simulation.
The flow inside the nozzle is composed of a mixture of liquid, vapor, and gas, and
exits the nozzle through a gaseous medium. The numerical modeling and simulation of
a multiphase and multicomponent flow is a very daunting task as compared to a single
phase flow. In a multiphase and multicomponent flow the phases and/or components will
assume a large number of complicated configurations, in which small-scale interactions
between the phases and/or components can have profound effects on the macroscopic
properties of the flow [54]. The most common problem encountered in multiphase and
multicomponent flow modeling is the fluid interface, where a large density variation
exists. This causes flow phenomena to become more complex because of additional
non-linearity introduced by the indeterminacy of such surfaces. In addition, the fluid
interface may be unstable, changing the flow configuration of the problem [103].
Several numerical simulations of multiphase and multicomponent flow have been
conducted over the past few years with different modeling approach and numerical
techniques. It has been shown that each model and numerical method has limits to
their respective capabilities. To this date, there has been a great need of improving the
technique of modeling multiphase and multicomponent flow and of a numerical method
to perfectly capture the complex nature of the flow.
Dumont et al. [37] and Shin [125] have used the homogeneous equilibrium model
to solve two-phase (liquid and vapor) cavitating flows. Their models differ on the
equation of state of the mixture and the numerical scheme. Taking a different approach,
Alajbegovic et al. [5] and Tatschl et al. [136] have used the two-fluid formulation for
multiphase flows [35] to simulate three phase cavitating flows. The conservative form of
the Favre-averaged Navier-Stokes equations have been utilized by Senocak et al. [123]
to study cavitating flows through convergent-divergent nozzles. A direct calculation
of cavitating flows by the Space-Time Conservation Element and Solution Element
(CE/SE) method is reported by Qin et al. [104]. The CE/SE method is applicable
9
for flows at wide range of Mach numbers and suitable for time accurate simulations.
Vortman et al. [149] have proposed a new approach based on postulating Gibbs free
energy for the phase mixture. The two-phase flow is treated numerically by combining
the rate equation with a volume of fluid approach.
Singhal et al. [126] have developed a full cavitation model that accounts for
the formation and transport of vapor bubble, turbulent fluctuations of pressure and
velocity, and the magnitude of non-condensible gases that are dissolved in the operating
liquid. A reduced form of Rayleigh-Plesset equation for bubble dynamics has been used
to derive the phase change rate equations. The model has assumed an isothermal flow
process, and decoupling the cavitation module from heat transfer and radiation modules.
The model has been implemented in an advanced, commercial, general-purpose CFD
code called CFD-ACE+ by CFD Research Corporation (CFDRC). Taking a similar
approach, Delale et al. [32] have used a continuum bubbly liquid flow model with
bubble nucleation, and nonlinear bubble dynamics described by the classical Rayleigh-
Plesset equation in quasi one dimensional steady state cavitating nozzle flows. Another
approach using multi-dimensional modeling of multiphase flow to examine the multi-
physics of high-speed nozzle and jet flows were presented by Murphy et al. [96].
In addition, Ahuja et al. [4] have formulated a multiphase model for low speed
gas/liquid mixtures by reducing the compressible system of equations to an acoustically
accurate form for multi-fluid mixtures. Also, Hosangadi et al. [65] have developed a
generalized numerical framework for transient and multiphase problems that involve
a combination of gas, bulk liquid, and a dense dispersed phase. The model has been
improved by allowing flexibility of specifying variables for thermodynamic properties
and specifying physical equations of state for mixture constituents [3].

10
1.4 Objectives of the Research
The main objective of this thesis is to develop a general approach for modeling and
numerical simulation of multiphase and multicomponent, compressible and incompresi-
ble flows, which is applicable to a variety of flow configurations including shock waves,
cavitation, single and multi-phase turbulent flows. The approach is flexible and can have
an arbitrary number of components in each phase using either Euler or Navier-Stokes
formulation. It must be emphasized that the modeling and computational approach
developed in this thesis is flexible enough to incorporate a variety of different cavitation
and/or turbulence models. To illustrate these capabilities numerical simulation using
representative models are performed.
1.5 Research Methodology
Two approaches are commonly used for the simulation of multiphase and multi-
component flows. In the first approach, each phase and/or component is considered to
occupy a distinct volume and the interfaces between the phases and/or components are
tracked explicitly, see, e.g., [17, 36, 42, 50, 87, 118, 135]. In the second approach, the
phases and/or components are spatially averaged to lead to a homogeneous mixture and
are considered to occupy the same volume. The phases and/or components need not be
in equilibrium, i.e. their mechanical and thermodynamical properties may differ. The
advantage of the homogenized-mixture approach compared to the interface-tracking ap-
proach is that it solves only one set of equations for the mass, momentum, and energy
of the mixture, supplemented by equations for the mass or volume fraction of the mix-
ture constituents [94]. In this research, the homogenized-mixture approach is used and
mechanical and thermodynamical equilibrium is assumed, i.e. pressure, temperature,
and velocity are identical for all the phases and components.
Multiple approaches of dealing with multiphase or multicomponent mixtures ex-

11
ist. However, most of them are limited to either single phase multicomponent fluids
[1, 70, 79, 80] or multiphase single component mixtures [13, 22, 133]. To remove this
limitation, a new model for multiphase and multicomponent flows with an arbitrary
number of components in each phase is developed. The model is hyperbolic, allowing
the construction of upwind methods for the computation of convective fluxes. Further-
more, the model is acoustically and thermodynamically consistent, which means that
the model gives an accurate value for the mixture speed of sound.
To perform simulations which demonstrate the capabilities of the new multiphase
and multicomponent model, a numerical method is developed, based on a finite-volume
framework [57, 58, 59]. The Roe’s approximate Riemann solver [112] and the modified
Harten, Lax and van Leer scheme (HLLC) [15] are extended to multiphase and mul-
ticomponent flows and used to capture shock waves and contact discontinuities. The
numerical method is verified by applying it to a number of test problems. The problems
were chosen to highlight the flexibility and robustness of the new approach and cover
the following flow regimes:
1. single-phase single-component fluid;
2. single-phase multicomponent fluid;
3. multiphase single component fluid;
4. multiphase multicomponent fluid.
It should be noted that a number of benchmark problems for cases 1-3 are available
[1, 2, 13, 22, 71, 120, 150]. However, no test problem for case 4 that has an exact
closed-form solution for arbitrary initial conditions and arbitrary numbers of phases and
components has been reported until now. To address this deficiency, a novel “idealized”
fluid mixture model is developed, which allows the derivation of an exact solution for
the multiphase and multicomponent Riemann problem in one dimension. A number of

12
existing benchmark problems for single-phase multicomponent flows become a subset of
this new problem.
1.6 Organization of the Thesis
In the thesis body, additional short reviews of the research areas related to the
particular applications are given in an appropriate places. The rest of this thesis is or-
ganized as follows. Chapter 2 introduces the mathematical formulation for multiphase
and multicomponent flow. The governing multiphase and multicomponent equations
in conservative form and the mathematical model for the mixture variables are pre-
sented. The mathematical model for multiphase and multicomponent flow is based
on the homogenized-mixture approach. The governing equations of the homogenized-
mixture model for the simulation of multiphase and multicomponent flows are described.
The determination of the speed of sound of the mixture as well as the validation of the
mixture speed of sound is presented. Also, the calculation of the mixture equation of
state using two different approaches such as Dalton’s Law and Amagat’s Law is pre-
sented.
The governing equations for the multiphase and multicomponent flows cannot
be solved exactly for arbitrary mixtures. In Chapter 3, a novel idealized fluid-mixture
model is developed. This new model allows derivation of exact solution for one dimen-
sional multiphase/multicomponent Riemann problem. A number of existing benchmark
problems for single-phase multicomponent flows becomes a subset of this new model
problem. In addition, the expressions of the equation of state and entropy for an ide-
alized fluid-mixture are derived. Also, the Riemann invariants and eigenvectors of the
idealized fluid mixture are presented for one-dimensional flows.
The development of the numerical method is presented in Chapter 4. Two ad-
vanced numerical schemes based on Roe‘s approximate Riemann solver [112] and the
modified Harten, Lax and van Leer scheme (HLLC) [15] are extended to multiphase
13
and multicomponent flows and used to capture shock waves and contact discontinu-
ities. The spatial and temporal discretization as described in [57, 58, 59] is adopted,
except that the equations being solved represent the mixture of liquid, gas, and vapor
with two additional conservation equations for the mass fractions of the gas and vapor.
The spatial discretization is based on the unstructured finite-volume framework and
the classical four-stage Runge-Kutta method in low-storage formulation is used for the
temporal discretization.
In Chapter 5, the new developed solver for simulation of multiphase and multi-
component flows is verified and validated by solving some benchmark problems includ-
ing the novel benchmark problems for the idealized fluid-mixture model. The problems
considered are:
1. Single-phase two-component shock-tube problem;
2. Shock-wave propagation in a single-phase two-component fluid;
3. Two-phase single-component shock-tube problem for idealized-fluid mixture;
4. Two-phase single-component rarefaction problem for idealized fluid mixture;
5. Single-phase two-component shock-bubble interaction.
Chapter 6 demonstrates the flexibility of the new developed general approach
to incorporate a variety of cavitation model for modeling and numerical simulation
of multiphase and multicomponent flows. A representative cavitation model proposed
by Hosangadi et al. [64] is integrated into the framework. The approach with the
incorporated cavitation model is tested to perform simulation of multiphase and mul-
ticomponent cavitating nozzle jet flow through a gaseous medium. The results of the
numerical simulation, which captures the cavitation process in the nozzle orifice are
presented.
14
Chapter 7 extends the new developed general approach to modeling and sim-
ulation of turbulent multiphase and multicomponent flows. The extended large eddy
simulation (LES) methodology for multiphase and multicomponent flows is incorporated
into the general framework. The capability of the extended LES for multiphase and mul-
ticomponent flows is examined by conducting a numerical simulation of turbulent plane
jet in both single phase and multiphase/multicomponent flow regime.
Finally, conclusions and possible future extensions of the work are outlined in
Chapter 8.
Chapter 2
Mathematical Formulation
The mathematical model for multiphase and multicomponent flow is based on the
homogenized-mixture approach. The phases and/or components are spatially averaged
and are considered to occupy the same volume. The advantage of the homogenized-
mixture approach compared to the other approach, i.e. interface-tracking approach,
is that it solves only one set of equations for mass, momentum and energy of the
mixture, supplemented by equations for the mass or volume fraction of the mixture
constituents [94]. However, there are challenges associated with the use of this approach
like the mathematical closure of the system that is acoustically and thermodynamically
consistent.
The governing equations of the homogenized-mixture model for the simulation of
multiphase and multicomponent flows are described in the following. The mixture is
assumed to be in thermodynamic and mechanical equilibrium so that the temperature
T , pressure P , and velocity components u, v and w of the phases and components are
identical. In the description to follow, the mixture is assumed to consist of two phases,
namely liquid and gas, and the gas phase is assumed to consist of two components,
namely a generic gas and a vapor. These are denoted by the subscripts l, g, and v for
liquid, gas, and vapor, respectively. It should be noted, however, that the homogenized-
mixture model can be extended in a straightforward fashion to an arbitrary number
of phases and components. Variables without subscripts are applicable to the mixture
16
only. The subscript i is used to denote a specific component.
2.1 Homogenized-Mixture Model
The homogenized-mixture model is based on the notion that quantities associated
with a given phase and/or component are averaged to give the corresponding mixture
quantity. Accordingly, quantities per unit volume are averaged by their respective vol-
ume fraction φi . For example, the mixture density is given by
X
ρm = ρi φi , (2.1)
i=l,g,v
where the volume fractions satisfy the constraint
X
φi = 1. (2.2)
i=l,g,v
Conversely, quantities per unit mass are averaged by their respective mass fractions Yi .
For example, the specific heat at constant volume of the mixture is given by
X
cvm = cvi Yi . (2.3)
i=l,g,v
The volume and mass fractions are related through
ρi φi = ρm Yi . (2.4)
2.2 Governing Equations
The Navier-Stokes equations governing the evolution of mass, momentum, energy,
and composition of the mixture can be written in vector form as
∂Q ∂E ∂F ∂G ∂Ev ∂Fv ∂Gv

+ + + = + + + S, (2.5)
∂t ∂x ∂y ∂z ∂x ∂y ∂z
where Q is the vector of the conserved variables and E, F and G are the convective
flux vectors given by

17
       
 ρm   ρm u   ρm v   ρm w 
       
       
 ρm u   ρm u2 + P   ρm vu   ρm wu 
       
       
 ρm v   ρm uv   ρm v 2 + P   ρm wv 
       
       
       
Q =  ρm w  , E =  ρm uw  , F =  ρm vw  , G =  ρm w2 + P  .
       
       
       
 ρm emT   (ρm emT + P )u   (ρm emT + P )v   (ρm emT + P )w 
       
       
 ρg φg   ρg φg u   ρg φg v   ρg φg w 
       
       
ρv φv ρv φv u ρv φv v ρv φv w
(2.6)
where u, v and w are the x-, y- and z−components of the velocity vector of the mixture,
respectively, and the total energy, emT , is defined as
1 2
emT = cvm T + u + v 2 + w2 . (2.7)
2
The constitutive equations of the liquid, gas, and vapor are assumed to take the form
ρi = ρi (P, T ). (2.8)
The mathematical model derived in this thesis is general and can be used for arbitrary
forms of the equation of state for each phase. However, in the present study the gas
and vapor are assumed to obey the ideal-gas laws
P
ρg = , (2.9)
Rg T
P
ρv = , (2.10)
Rv T
while the liquid is assumed to be a linear dependent of pressure and temperature

2
1 βl
ρl = ρo + 2 (P − Po ) − (T − To ), (2.11)
Cl Cl
where ρo , Po and To are the reference density, pressure and temperature of the liquid,
respectively. Cl and βl are the isothermal speed of sound and compressibility of the
liquid, respectively.
18
The vectors Ev , Fv and Gv on the right hand side of Eq. (2.5) are the viscous
flux vectors written as
Ev = [0, τxx , τyx , τzx , Θx , Ψgx , Ψvx ]T , (2.12)
Fv = [0, τxy , τyy , τzy , Θy , Ψgy , Ψvy ]T , (2.13)
Gv = [0, τxz , τyz , τzz , Θz , Ψgz , Ψvz ]T . (2.14)
The total stress, τij , are defined as
2 ∂vk
τij = 2µm Sij − µm δij , (2.15)
3 ∂xk
where µm is the viscosity of the mixture and Sij = 12 (∂e

ui /∂xi + ∂e
ui /∂xi ) is the strain
rate tensor. The work of the total stress and heat conduction in the fluid mixture are
defined as
∂ Te
Θx = u e xz + e
eτxx + veτxy + wτ km , (2.16)
∂x
e
Θy = u
eτyx + veτyy + wτ em ∂ T ,
e yz + k (2.17)
∂y
∂ Te
Θz = u e zz + e
eτzx + veτzy + wτ km , (2.18)
∂z
where, km is the thermal conductivity of the mixture. Furthermore, the scalar fluxes of
the gas and vapor are given by
∂ρm Yg
Ψgi = ϕ
e , (2.19)
∂xi
∂ρ Yv
Ψvi e m ,
= ϕ (2.20)
∂xi
where ϕ is the diffusion coefficient, which is considered to be zero, unless otherwise
specified.
In addition, the vector S on the the right hand side of Eq. (2.5) represents the
source terms written as
S = [0, 0, 0, 0, 0, 0, Sv ]T , (2.21)
19
where Sv is the cavitation source term. Sv is zero until Section 6.3.2, where cavitation
model is introduced. In general, source terms are non-zero in the case when there is
phase transformation, chemical reactions, and/or body force. In this thesis, the source
terms are limited only to mass transfer between liquid and vapor phases, which is a
characteristic of cavitation process.
On the other hand, Eq. (2.5) can be simplified into Euler equations in the case of
inviscid flow, with the assumptions of no viscosity effect, no heat transfer, and no body
force. Thus, Eq. (2.5) can be reduced to
∂Q ∂E ∂F ∂G
+ + + = S. (2.22)
∂t ∂x ∂y ∂z
The governing Euler equations for the mixture given in Eq. (2.22) are used in Chapters
5 and 6 for the numerical simulation of multiphase and multicomponent flows, while
the governing Navier-Stokes equations given in Eq. (2.5) are considered in Chapter 7
for the numerical simulation of turbulent multiphase and multicomponent flows.
2.2.1 Transport Variables Model
The viscosity µm and thermal conductivity km of the mixtures are modeled by
averaging the viscosity and thermal conductivity of the mixture constituents. These are
given by
X
µm = µi Yi , (2.23)
i=l,g,v
X
km = ki Yi . (2.24)
i=l,g,v
The viscosity for each constituents are modeled in the following. For the gas and
vapor, the viscosity can be approximated using the Power law and Sutherland law
[151], respectively given by

n
T
µi=g,v ≈ µo , (2.25)
To
3/2
T To + Sth
µi=g,v ≈ µo , (2.26)
To T + Sth
20
where µo is a reference viscosity at a reference absolute temperature To . The constants
n and Sth are fit to the data. For example, for air, n = 0.7, Sth ≈ 110.4 K, with To = 273
K and µo = 1.71 × 10−5 kg/(m · s). For the liquid, the viscosity is approximated using
the empirical function [151],

2
µl To To
ln ≈a+b +c . (2.27)
µo T T
where, for water with To = 273.16 K, µo = 0.001792 kg/(m · s), the suggested values
for the constants are a = −1.94, b = 4.80, and c = 6.74.
The thermal conductivity for each constituents are calculated using the equation
of Prandtl number,
cpi µi
ki=l,g,v ≈ . (2.28)
Pr
Some complex models of the properties of liquids and gases can be found in [108].
2.3 Mixture Equation of State
The equation of state of the mixture is very important in describing the physical
and thermodynamical properties of the mixture of fluids. However, the evaluation of
the properties of mixture is difficult, because any one property depends not only on
two independent properties like pressure and temperature, but also on a specification of
the composition of the mixture such as the mass or volume fraction of each component.
On the other hand, the equation of state of the mixture provides a mathematical re-
lationship between numerous state functions associated with matter, such as pressure,
temperature, mass or volume fraction, speed of sound, specific heat, and many others.
The mathematical expressions of the equation of state for multiphase (liquid or
gas) and multicomponent (several instances of the same phase) mixture can be obtained
using the principle of either Dalton’s Law or Amagat-Leduc Law. In the evaluation,
each phase and/or component behaves as a compressible fluid. The derivation of the
21
equation of state is difficult and depends on the constitutive equation for each phase
and/or component.
2.3.1 Mixture Pressure using Dalton’s Law
In Dalton’s law, the total pressure exerted by a mixture of gases is the sum of the
component pressures Pi , each measured exclusively at the temperature T and volume
V of the mixture. Hence, Dalton’s law when applied to multicomponent mixture of gas
and vapor can be written in the form

X
Pgv = Pi = Pg + Pv , (2.29)
i=g,v
Vgv = Vg = Vv , (2.30)
where Pgv is the total pressure and Vgv is the volume of the multicomponent mixture of
gas and vapor. Applying Eqs. (2.9) and (2.10) to Eqs. (2.29) and (2.30), supplemented
with the mass balance of the gas and vapor mixture, yields
Pgv = ρgv Rgv T, (2.31)
where
ρg φg + ρv φv
ρgv = ,
φgv
φgv = φg = φv ,
ρg φg Rg + ρv φv Rv
Rgv = .
ρg φg + ρv φv
Now, let us consider the case for a mixture of liquid, gas and vapor. The pressure
exerted by the multicomponent mixture of gas and vapor, Pgv , and the pressure exerted
by the liquid, Pl , in the overall mixture are assumed in equilibrium. Thus,
P = Pl = Pgv . (2.32)
From Eq. (2.31), the following is obtained
P
φgv = ρg φg + ρv φv , (2.33)
Rgv T
22
using the relation, φl + φgv = 1, yields
P
(1 − φl ) = ρg φg + ρv φv . (2.34)
Rgv T
Simplification gives
P (ρl − ρl φl ) = (ρg φg + ρv φv .) ρl Rgv T. (2.35)
To obtain a close form solution of Eq. (2.35), the constitutive equation of state of the
liquid must be defined. Introducing Eq. (2.11) to Eq. (2.35), yields the exact solution
of the mixture equation of state

p
ψ1 − ψ2 + (ψ2 − ψ1 )2 + 4ψ2 ψ3
P = , (2.36)
2
where
2 !
βl Po
ψ1 = Cl2 ρo − (T − To ) − 2 − ρl φl ,
Cl Cl
ψ2 = (ρv φv Rv + ρg φg Rg )T,
2 !
2 βl Po
ψ3 = Cl ρo − (T − To ) − 2 .
Cl Cl
2.3.2 Mixture Pressure using Amagat-Leduc Law
In Amagat-Leduc law, the volume of a gas mixture is equal to the sum of the
volumes of all constituents at the same temperature and pressure as the mixture. Hence,
Amagat-Leduc law when applied to multicomponent mixture of gas and vapor can be
written in the form
Pgv = Pg = Pv , (2.37)
X
Vgv = Vi = Vg + Vv , (2.38)
i=g,v
where Pgv is the total pressure and Vgv is the volume of the multicomponent mixture of
gas and vapor. Applying Eqs. (2.9) and (2.10) to Eqs. (2.37) and (2.38), supplemented
with the mass balance of the gas and vapor mixture, yields
Pgv = ρgv Rgv T, (2.39)

23
where
ρg φg + ρv φv
ρgv = ,
φgv
φgv = φg + φv ,
ρg φg Rg + ρv φv Rv
Rgv = .
ρg φg + ρv φv
Again, let us consider the case for a mixture of liquid, gas and vapor. The pressure
exerted by the multicomponent mixture of gas and vapor, Pgv , and the pressure exerted
by the liquid, Pl , in the overall mixture are assumed in equilibrium. Thus,
P = Pl = Pgv . (2.40)
From Eq. (2.39), the following is obtained
P
φgv = ρg φg + ρv φv , (2.41)
Rgv T
using the relation, φl + φgv = 1, yields
P
(1 − φl ) = ρg φg + ρv φv . (2.42)
Rgv T
Simplification gives
P (ρl − ρl φl ) = (ρg φg + ρv φv .) ρl Rgv T, (2.43)
which is the same with Eq. (2.35). In order to obtain a close form solution of Eq. (2.43),
the constitutive equation of state of the liquid must be defined. Introducing Eq. (2.11)
to Eq. (2.43), yields the exact solution of the mixture equation of state, which is the
same with Eq. (4.17).
2.4 Mixture Speed of Sound
The acoustic differential of Eq. (2.8) is given by

2
1 βi
dρi = 2 dP + dT, (2.44)
Ci Ci
24
1/2
where Ci = (∂P/∂ρi )1/2 and βi = (∂P/∂T )i are the isothermal speed of sound and
compressibility of the ith component, respectively. Using Eqs. (2.1) and (3.13), the
mixture-density differential can be written as

!
1 1
dρm = (ρv − ρl )dφv + (ρg − ρl )dφg + 2 − 2 dP, (2.45)
Cφ Cφβ
P P
where 1/Cφ2 = 2
i φi /Ci
2 =
and 1/Cφβ i φi βi2 /Ci2 .
The mixture speed of sound can be obtained easily by transforming Eq. (2.5)
from conservative variables to primitive variables form
∂Qv ∂E ∂F ∂G ∂Ev ∂Fv ∂Gv

Γ + + + = + + + S, (2.46)
∂t ∂x ∂y ∂z ∂x ∂y ∂z
where Qv is the vector of the primitive variables
Qv = [P, u, v, w, T, φg , φv ]T (2.47)
and Γ is the transformation matrix defined by

dQi ∂Qi
Γ= = (2.48)
dvj i,j=1,...,7 ∂Qvj
The elements of the matrix are computed, for example; ∂Q5 /∂Qv1 = ∂(ρm emT )/∂P =
el φl /Cl2 + eg φg /Cg2 + ev φv /Cv2 . After careful evaluation of all the elements in the matrix,
the following is obtained

 
1
 Cφ2
0 0 0 − C12 ρg − ρl ρv − ρl 
φβ
 
 u 
 ρm 0 0 − Cu2 (ρg − ρl )u (ρv − ρl )u 
 Cφ2 φβ 
 
 v 
 Cφ2
0 ρm 0 − Cv2 (ρg − ρl )v (ρv − ρl )v 
 φβ 
 
Γ= 
 w
Cφ2
0 0 ρm − Cw2 (ρg − ρl )w (ρv − ρl )w 
 .
 φβ 
 
 e l φl
+
e g φg
+ e v φv
ρm u ρm v ρm w ρl φl cvl + βT cvl φl T − 12 ( u
2 +v 2 +w 2
) ρg eg − ρl el ρv ev − ρl el 
 Cl2 Cg2 Cv2 2
Cφβ 
 
 βg2 φg 
 φg
0 0 0 − ρg 0 
 Cg2 Cg2 
 
2
φv
Cv2 0 0 0 − βCv φ2v 0 ρv
v
The eigenvalues of the transformation matrix give the speed of sound. The result-
ing eigenvalues, λi of the system are all real and have an exact analytical expressions.
These are given by
λi = (V · n̂ − Cm , V · n̂, V · n̂, V · n̂, V · n̂, V · n̂, V · n̂ + Cm ), (2.49)

25
where V = (u, v, w) is the velocity vector, n̂ = (nx , ny , nz ) is the unit normal vector,
and Cm is the speed of sound of the mixture given by

X φi βi 2
ρm cvm + P
ρi Ci
2 i
Cm = X φi 1 . (2.50)
2
ρm cvm
ρi Ci2
i
Note that in the case of multicomponent gases, Eq. (2.50) becomes
Cm = (γm Rm T )1/2 , (2.51)

P P P
where γm = cpm /cvm = i Yi cpi / i Yi cvi , Rm = i Yi Ri , and Ri is the gas constant of
the ith gas component. Furthermore, it should be noted that due to the assumption of
thermodynamic and mechanical equilibrium, the speed of sound predicted by Eq. (2.50)
is applicable only to disturbances whose frequency tends to zero.
The speed of sound of a mixture of water and air at sea-level conditions predicted
by Eq. (2.50) is plotted in Figure 2.1 as a function of the volume fraction of air. Note
that over a wide range of volume fractions, the speed of sound of the mixture is much
lower than the speed of sound of either medium [98]. The presence of gas in a liquid
dramatically reduces the speed of sound in the liquid [14, 72, 90, 93]. The speed of
sound in water is about 1480 m/s and about 340 m/s in air, but in an air-water mixture
lies to about 20 m/s. A 1% by volume of air in water drops the speed of sound of the
liquid by 95% to 100 m/s. This dramatic phenomenon occurs due to the compressibility
effect of the gas in the liquid.
Similarly, the speed of sound of a mixture of water and air at sea-level conditions
predicted by Eq. (2.50) and using Eq. (2.4) is plotted in Figure 2.3 as a function of the
mass fraction of air. The speed of sound is approximately 1350 m/s for mass fraction
of air at 10−8 , and decreases to 25 m/s for mass fraction of air at 10−3 . As shown in
Figure 2.4, there is a definite dependence of speed of sound on frequency, but in general
the measured data by Karplus [72] confirm the theoretical prediction of a large decrease
of the value of speed of sound with admixture of gas into liquid.

26
1400 PURE LIQUID
1200
Speed of sound (m/s)

1000
800
600
400
PURE AIR
200
0
0 0.2 0.4 0.6 0.8 1
Volume fraction of air

Figure 2.1: Speed of sound of a water-air mixture at P = 1 · 105 Pa and T = 298.15K
as a function of the volume fraction of air. Speed of sound predicted by Eq. (2.50).
100
90
80
Speed of sound (m/s)
70
60
50
40
30
20
10
0
0 0.2 0.4 0.6 0.8 1
Volume fraction of air

as a function of the volume fraction of air. Comparison of predicted speed of sound
with experimental data of Karplus [72] for frequencies of 1 kHz (diamonds), 0.5 kHz
(squares) and extrapolated to zero frequency (circles).
27
1400
1200
Speed of Sound (m/s)

1000
800
600
400
200
0 −8 −6 −4 −2 0
10 10 10 10 10
Mass fraction, Air

as a function of the mass fraction of air. Speed of sound predicted by Eq. (2.50) and
using Eq. (2.4).
140
120
Speed of Sound (m/s)
100
80
60
40
20
0 −5 −4 −3 −2 −1
10 10 10 10 10
Mass fraction, Air

as a function of the mass fraction of air. Comparison of predicted speed of sound with
experimental data of Karplus [72] for frequencies of 1 kHz (diamonds), 0.5 kHz (squares)
and extrapolated to zero frequency (circles).
28
1200
Speed of Sound of the Mixture

1000
LIQUID
800
LIQUID DENSITY : 100 kg/m3
GAS DENSITY : 1.189 kg/m3
600
400
GAS
200
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Volume fraction of Fluid
Figure 2.5: Speed of sound of a liquid-gas mixture predicted by Eq. (2.50) at P = 1 · 105
Pa and T = 298.15K, for a liquid-gas density ratio, ρliquid /ρgas ≈ 100.
1200
LIQUID
Speed of Sound of a Mixture
1000
800
LIQUID DENSITY : 1000 kg/m3
GAS DENSITY : 1.189 kg/m3
600
400
GAS
200
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Volume fraction of Fluid
Figure 2.6: Speed of sound of a liquid-gas mixture predicted by Eq. (2.50) at P = 1 · 105
Pa and T = 298.15K, for a liquid-gas density ratio, ρliquid /ρgas ≈ 1000.
29
Figure 2.7: Speed of sound of a liquid-gas-vapor mixture predicted by Eq. (2.50) at

P = 1 · 105 Pa and T = 298.15K.
Figures 2.2 and 2.4 show a good agreement between the value predicted by Eq.
(2.50) and the experimental data of Karplus [72]. The predicted speed of sound was also
compared to Refs. [3, 4, 35, 74, 143] and, although not plotted, is in good agreement.
Moreover, the speed of sound of the mixture varies also with the variation of the
density ratio of the liquid and gas. A high density ratio of liquid and gas will have a
much lower speed of sound as compared to a low density ratio. For a density ratio of
100, the speed of sound of the mixture is about 100 m/s, see Figure 2.5, however, for
a density ratio of 1000, the speed of sound of the mixture drops off rapidly to about
25 m/s, see Figure 2.6. In addition, for a mixture of liquid, gas, and vapor at sea-level
conditions, the speed of sound is shown in Figure 2.7.

Chapter 3
Idealized Fluid-Mixture Model
The governing equations for the multiphase and multicomponent flows cannot be
solved analytically for arbitrary mixtures. The equation of state for each component
must be carefully chosen if an exact solution is to be found. In this section, a novel
idealized fluid-mixture model is developed, and exact solutions for one dimensional mul-
tiphase and multicomponent Riemann problem are derived. In addition, the Riemann
invariants and eigenvectors of the idealized fluid mixture are presented.
3.1 Density and Equation of State for Idealized Fluid-Mixture
To obtain a closed-form solution of the governing equations, the mixture entropy
is assumed to be a function of pressure, temperature and mass fractions of the species,
i.e. sm = sm (P, T, Yi ). Differentiation gives
∂sm ∂sm X ∂sm

dsm = dP + dT + dYi , (3.1)
∂P ∂T ∂Yi
i=l,g,v
and because dsm is assumed to be an exact differential, then it follows that
∂ 2 sm ∂ 2 sm
= , (3.2)
∂P ∂T ∂T ∂P
∂ 2 sm ∂ 2 sm
= , (3.3)
∂Yi ∂T ∂T ∂Yi
∂ 2 sm ∂ 2 sm
= . (3.4)
∂Yi ∂P ∂P ∂Yi
31
Using the T -ds equation applied to the mixture gives

dT P 1 1 X
dsm = cvm + d + Li dYi , (3.5)
T T ρm T
i=l,g,v
where Li is the latent heat of phase change and is assumed to be a function of pressure
and temperature, i.e. Li = Li (P, T ).
From Eq. (2.4) the following relation is obtained
1 X Yi
= , (3.6)
ρm ρi
i=l,g,v
and differentiation gives

X dYi
1 dρi
d = − Yi 2 . (3.7)
ρm ρi ρi
i=l,g,v
Substituting the differential form of Eq. (2.8) into Eq. (3.7) results
X dYi
1 ∂ 1 ∂ 1
d = + Yi dP + Yi dT . (3.8)
ρm ρi ∂P ρi ∂T ρi
i=l,g,v
Subsequent substitution of Eqs. (2.3) and (3.8) into Eq. (3.5) yields
X cvi Yi P ∂

1

P ∂

1

P Li

dsm = + Yi dT + Yi dP + + dYi .
T T ∂T ρi T ∂P ρi T ρi T
i=l,g,v
(3.9)
Applying exact differential properties (3.1)-(3.4) to Eq. (3.9) results in the fol-
lowing constraints
X Yi ∂ρi P ∂ρi
+ = 0, (3.10)
ρ2 ∂T T ∂P
i=l,g,v i

∂ Li cvi P
= + 2 , (3.11)
∂T T T T ρi

∂ Li 1
=− . (3.12)
∂P T T ρi
In order for Eq. (3.10) to be satisfied for arbitrary mass fractions, the expression
inside of the brackets must be equal to zero for each phase/component. Consequently,
the constitutive equation for each phase/component must be a function of the ratio of
pressure and temperature, i.e.

P
ρi = ρi . (3.13)
T
32
Thus, to obtain an analytical solution for the mixture entropy, the density of each
component must be a function of the ratio of pressure and temperature. Because the
gas and vapor are assumed to follow the ideal-gas laws, see Eqs. (2.9) and (2.10), they
automatically satisfy Eq. (3.13). For the liquid, the following relation is proposed
P
ρl = ρo + α , (3.14)
T
where α = To /Cl2o , To is the reference temperature, Clo is the reference speed of sound,
and ρo is the reference density of the liquid. A liquid obeying Eq. (3.14) is called an
idealized liquid in this thesis. Equation (3.14) can be regarded as a model for a liquid
described by Eq. (2.8). To see this, note that the linearized model given in Eq. (2.11)
can be approximated by Eq. (3.14), provided that the temperature variations are small.
Combining Eqs. (2.9), (2.10), (3.14) and (3.6), the mixture density can be written
as
z z
ρm = = , (3.15)
zYl /ρl + Rg Yg + Rv Yv −1
zYl (ρo + αz) + Rg Yg + Rv Yv
where z = P/T . The mixture defined by Eq. (3.15) is called an idealized fluid mixture
because it is derived from the idealized liquid defined above and an ideal gas and vapor.
Note that Eq. (3.15) can also be interpreted as the equation of state of the mixture.
3.2 Entropy for Idealized Fluid-Mixture
Integrating Eqs. (3.11) and (3.12) from some reference state (P r , T r ) the following
equation for the latent heat of phase change is obtained
Li T Lr
= cvi ln r + Fi (z) + ir , (3.16)
T T T
where
Z z
1
Fi (z) = − dζ. (3.17)
zr ρi (ζ)
Integrating Eq. (3.9) results in the following expression for the entropy of the
33
mixture
X T P Yi

P Lri

sm = cvi Yi ln r + + Fi + r Yi . (3.18)
T T ρi T T
i=l,g,v
The expression (3.18) can be further simplified by using Eqs. (2.3) and (3.6)
T P X P Lr
sm = cvm ln r + + Fi + ir Yi . (3.19)
T ρm T T T
i=l,g,v
Substituting equations of state (2.9), (2.10), (3.14) for the gas-vapor-liquid mix-
ture into Eq. (3.19) yields

z Y ρ X Lr
T Yl Yl ρl l 0 i
sm = cvm ln r + a1 + − ln − a1 ln − + Yi , (3.20)
T α α ρrl zr α ρl Tr
i=l,g,v
where a1 = Rg Yg + Rv Yv .
Evaluating Eq. (3.20) at two different states and subtracting one from another,
one obtains
2 h z i2
T 2 Yl 2 1 ρl
[sm ]21 = cvm ln r + a1 + − Yl ln − a1 ln
T 1 α 1 α ρrl 1 zr 1
2 X Lr
ρ0 Yl
− + i
[Yi ]21 , (3.21)
α ρl 1 Tr
i=l,g,v
where square brackets denote the following operation [(·)]21 = (·)2 − (·)1 .
If no mass transfer between the phases is present, i.e. the mass fractions are
assumed constant, Yi1 = Yi2 = Yi , Eq. (3.21) reduces to

T2 Yl ρl2 z2 ρo Yl 1 1
sm2 − sm1 = cvm ln − ln − a1 ln − − , (3.22)
T1 α ρl1 z1 α ρl2 ρl1
For isentropic changes of state, Eq. (3.22) leads to

a1 Yl
T2 z2 cvm ρl2 αcvm ρo Yl 1 1
= exp − . (3.23)
T1 z1 ρl1 αcvm ρl2 ρl1
For the case of a pure gas, Eq. (3.23) reduces to

(γg −1)/γg
T2 P2
= , (3.24)
T1 P1
where γg = 1 + Rg /cvg . For the case of a pure liquid Eq. (3.23) reduces to
1
T2 ρl2 αcvm ρo 1 1
= exp − . (3.25)
T1 ρl1 αcvm ρl2 ρl1
34
3.3 Speed of Sound for Idealized Fluid-Mixture
The speed of sound of the idealized fluid mixture can be obtained by applying
Eqs. (2.9)-(3.14) to Eq. (2.50), giving

φl φv φg
ρm cvm + P β2 + +
2 ρl Cl2 ρv Cv2 ρg Cg2
Cm = , (3.26)
2
φl φv φg
ρm cvm + +
ρl Cl2 ρv Cv2 ρg Cg2
where Cl = (T /α)1/2 , Cg = (Rg T )1/2 , and Cv = (Rv T )1/2 are the isothermal speeds of
sound in the liquid, gas, and vapor, respectively, and β = βl = βg = βv = (P/T )1/2
is the compressibility. Note that the compressibilities are identical for the idealized
mixture. Simplification yields

2 + P β2
ρm cvm Cφ/ρ
2
Cm = , (3.27)
ρ2m cvm
where
1 φl φv φg
2 = 2 + 2
+ . (3.28)
Cφ/ρ ρl Cl ρv Cv ρg Cg2
Equations (3.27) and (2.50) give practically identical results if the sound speeds of the
idealized and non-idealized liquid are matched at the reference temperature To .
3.4 Riemann Invariants for Idealized Fluid-Mixture
For one-dimensional problems, the eigenvalues are given by
λ = (u − Cm , u, u, u, u + Cm )T , (3.29)
and the corresponding set of eigenvectors are

35
 
1 1
 2
− 0 0 0 
 ρm Cm Cm 
 P 1 
 − 2 2 0 0 0 
 ρm Cm cvm T T 
 ! 
 φg βg2 P 
 
Π= − ρg Cg2 − ρm Cm
2
+ 0 0 1 0 . (3.30)
 ρg ρm Cg2 Cm
2 ρm cvm 
 
 φv βv2 P 
 − 2 2
1 
 ρ ρ C 2 C 2 ρv Cv − ρm Cm + ρm cvm
0 0 0 
 v m v m 
 1 1 
2
0 0 0
ρm Cm Cm
The Riemann invariants are computed from the relation dΥ = Π dK, where K =
[P, u, T, φg , φv ]T ,
 
dP du
 2
− 
 ρm Cm Cm 
 P dP dT 
 
 − 2 2 + 
 ρm Cm cvm T T ! 
 2P 
 φg β g 
dΥ = 
 − ρg ρm C 2 C 2 ρg Cg2 − ρm Cm 2
+ dP + dφg .
 (3.31)
 g m ρm cvm 
 
 φv βv2 P 
 − 2
ρv Cv − ρm Cm +2
dP + dφv 
 ρv ρm Cv2 Cm2 ρm cvm 
 
 dP du 
2
+
ρm Cm Cm
3.5 Riemann Problem for Idealized Fluid-Mixture
The Riemann problem is characterized by uniform initial conditions except for a
discontinuity at x = 0 on an infinite one-dimensional domain. The lack of an intrinsic
length or time scale means that the solution to the Riemann problem is self-similar.
The solution of the Riemann problem for scalar conservation laws, linear hyperbolic
systems of equations, and the single phase Euler equations can be derived, see, e.g.,
[78] and [141]. For this reason, the Riemann problem is often used to verify numerical
methods. In this section, the solution of the Riemann problem for the idealized mixture
is presented.
36
t
u, u, u
u − Cm u + Cm
* *
KL KR
Q Q
L R
x
0
Figure 3.1: Structure of the solution of the Riemann problem for the idealized-mixture
formulation.
Let us consider the initial conditions



 QL if x < 0,
Q(x, 0) = (3.32)

 QR if x ≥ 0,
where Q = [ρm , ρm u, ρm emT , ρm Yg , ρm Yv ]T . The solution of the Riemann problem
involves expansion waves, shock waves, and a contact discontinuity. The structure
of the solution is shown in Figure 3.1. The four regions with constant solutions are
separated by five wave families. The challenge in finding the solution of the Riemann
problem lies in determining the unknown states Q∗L and Q∗R to the left and right of the
contact discontinuity, see Figure 3.1. These regions are referred to as the left and right
star regions, respectively. The corresponding unknown primitive variables are
∗ T
K ∗L = [PL∗ , u∗L , TL∗ , YgL
∗
, YvL ] , (3.33)
∗ T
K ∗R = [PR∗ , u∗R , TR∗ , YgR
∗
, YvR ] . (3.34)
There are four possible wave patterns in the solution of the Riemann problem
as shown in Figure 3.2, see, e.g, Toro [141]. These wave patterns are considered in
constructing the exact solution. The eigenstructure of the mixture formulation reveals
37
t t
11111111
00000000 011011111111
00000000
00000000
11111111 101011111111
00000000
00000000
11111111
00000000
11111111 00000000
11111111
0000000000000000
1111111111111111
000000000000000
111111111111111
0000000000000
1111111111111 00000000
11111111 00000000
11111111
101111111111111111
0000000000000000
00000000
11111111
0000000000000000
1111111111111111
000000000000000
111111111111111
0000000000000
1111111111111
0000000000000000
1111111111111111 000000000000000000
111111111111111111
00000000
11111111 0000000000000000000
1111111111111111111 0
1
00000000000000000
11111111111111111
0000000000000000
1111111111111111
00000000
11111111
000000000000000
111111111111111
0000000000000
1111111111111
0000000000000000
1111111111111111 000000000000000000
111111111111111111
00000000
11111111 1111111111111111111 0
1
000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
00000000000000000001111111111111111111
0000000000000000000
00000000
11111111
000000000000000
111111111111111
0000000000000
1111111111111
000000000000000000
111111111111111111
0000000000000000
1111111111111111
000000000000000
111111111111111 000000000000000000
111111111111111111
00000000
11111111 0000000000000000000
1111111111111111111 0
1
000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
0000000000000000000
1111111111111111111
00000000
11111111
0000000000000
1111111111111
000000000000000000
111111111111111111
0000000000000000
1111111111111111
000000000000000
111111111111111 000000000000000000
111111111111111111
00000000
11111111 0000000000000000000
1111111111111111111 0
1
000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
0000000000000000000
1111111111111111111
0
1 00000000
11111111
0000000000000
1111111111111
000000000000000000
111111111111111111
0000000000000000
1111111111111111
000000000000000
111111111111111
1111111111111 000000000000000000
111111111111111111
0000000000000011111111
00000000 1111111111111111111 000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
00000000000000000001111111111111111111
0000000000000000000
10011111111
00000000
000000000000000000
111111111111111111 x
000000000000000000
111111111111111111 x
(a) (b)
t 0110t
1111111
0000000
0000000
1111111 10
0000000
1111111
0000000
1111111 1010
0000000
1111111
0000000000000000
1111111111111111 0000000
1111111 0000000
111111110111111111111111111
0000000000000000
1111111111111111 0000000
1111111
000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000
1111111 0000000000000000001111111
1111111111111111110000000
0000000
1111111 000000000000000000
10111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
000000000000000000
111111111111111111 000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000
1111111
000000000000000000
111111111111111111 000000000000000000
111111111111111111
000000000000000000
1111111111111111110000000
1111111 000000000000000000
10111111111111111111
000000000000000000
00000000000000000
11111111111111111
0000000000000000
1111111111111111
000000000000000000
111111111111111111
000000000000000000
111111111111111111 00000000000000000
11111111111111111
0000000000000000000
1111111111111111111
0000000
1111111
000000000000000000
111111111111111111
00000000000000000
11111111111111111 000000000000000000
1111111111111111110000000
1111111 10111111111111111111
000000000000000000
00000000000000000
11111111111111111
0000000000000000
1111111111111111
000000000000000000
111111111111111111
000000000000000000
111111111111111111 0000000000000000000
1111111111111111111
0000000
1111111
000000000000000000
111111111111111111
00000000000000000
11111111111111111 0000000
1111111
0000000000000000001111111
111111111111111111 10111111111111111111
000000000000000000
10111111111111111111
00000000000000000
11111111111111111
0000000000000000
1111111111111111
000000000000000000
111111111111111111
000000000000000000
111111111111111111 0000000000000000000
1111111111111111111 0000000000000000000000000
0000000000000000001111111
11111111111111111 0000000
000000000000000000
111111111111111111
00000000000000000
11111111111111111
0000000000000000000
1111111111111111111 x 111111111111111111 000000000000000000
100 x
(c) (d)
Figure 3.2: Four possible wave patterns of the solution of the Riemann problem [141]: (a)
left rarefaction, contact, and right shock; (b) left shock, contact, and right rarefaction;
(c) left rarefaction, contact, and right rarefaction; (d) left shock, contact, and right
shock.
that the pressure P ∗ and velocity u∗ are constant across the contact discontinuity, while
other thermodynamic variables such as ρ∗m and T ∗ are discontinuous. The unknown
variables K∗L and K∗R are connected by the condition that the pressure P ∗ and velocity
u∗ are constant across the contact discontinuity. In the following, detailed analyses
of the conditions across the left shock wave and left rarefaction wave, denoted by a
subscript L, are presented. The conditions across the right shock and right rarefaction
wave can be obtained by replacing the subscript L by R.
3.5.1 Left Shock Wave Analysis
The left wave is assumed to be a shock wave moving with speed SL , see Figure
3.2(b) and (d). The pre-shock variables are PL , uL , TL , YgL , and YvL . The post-shock
38
variables are PL∗ , u∗L , TL∗ , YgL

∗ , and Y ∗ . The Rankine-Hugoniot conditions are applied
vL
across the left shock, leading to
ρmL uL − ρ∗mL u∗L = SL (ρmL − ρ∗mL ), (3.35)
(ρmL u2L + PL ) − (ρ∗mL u∗2 ∗ ∗ ∗

L + PL ) = SL (ρmL uL − ρmL uL ), (3.36)
(ρmL emL uL + PL uL ) − (ρ∗mL e∗mL u∗L + PL∗ u∗L ) = SL (ρmL emL − ρ∗mL e∗mL ), (3.37)
ρmL YgL uL − ρ∗mL YgL uL = SL (ρmL YgL − ρ∗mL YgL

∗ ∗ ∗
), (3.38)
ρmL YvL uL − ρ∗mL YvL uL = SL (ρmL YvL − ρ∗mL YvL

∗ ∗ ∗
). (3.39)
Equations (3.35)-(3.39) can be solved to give
u∗L = uL − fL (PL∗ , ρ∗mL , QL ), (3.40)
gL (PL∗ , ρ∗mL , TL∗ , QL ) = 0, (3.41)
∗
YiL = YiL , (3.42)
c∗vmL = cvmL , (3.43)
where
1
(PL∗ − PL )(ρ∗mL − ρmL ) 2
fL (PL∗ , ρ∗mL , QL ) = , (3.44)
ρmL ρ∗mL
and

1 1 1
gL (PL∗ , ρ∗mL , TL∗ , QL ) = cvmL TL − c∗vmL TL∗ − (PL + PL∗ ) − . (3.45)
2 ρ∗mL ρmL
3.5.2 Right Shock Wave Analysis
The analysis is analogous to that for the left shock wave. By replacing the sub-
script L by R, the following conditions are obtained,
u∗R = uR − fR (PR∗ , ρ∗mR , QR ), (3.46)
gR (PR∗ , ρ∗mR , TR∗ , QR ) = 0, (3.47)
YiR
∗
= YiR , (3.48)
c∗vmR = cvmR , (3.49)

39
where
1
(PR∗ − PR )(ρ∗mR − ρmR ) 2
fR (PR∗ , ρ∗mR , QR ) = , (3.50)
ρmR ρ∗mR
and

1 1 1
gR (PR∗ , ρ∗mR , TR∗ , QR ) = cvmR TR − c∗vmR TR∗ − (PR + PR∗ ) − . (3.51)
2 ρ∗mR ρmR
3.5.3 Left Rarefaction Wave Analysis
Let us assume that the left wave is a rarefaction wave, see Figure 3.2(a) and (c).
Then the unknown state K∗L is connected to the known left state QL using the isentropic
relation given by Eq. (3.23) and the generalized Riemann invariants for the left wave.
The Riemann invariant across the left rarefaction is given by
dP du
dΥ1 = 0 = 2
− , (3.52)
ρm Cm Cm
from Eq. (3.31). Integration across the left rarefaction yields
u∗L = uL + fL (PL∗ , ρ∗mL , QL ), (3.53)
where
Z ∗
dP
fL (PL∗ , ρ∗mL , QL ) = . (3.54)
L ρm Cm
The integral in Eq. (3.54) is evaluated using adaptive Simpson quadrature [88] due to
the complexity of the integrand.
Similarly, from Eq. (3.21) the following is obtain for isentropic conditions the
non-linear algebraic equation
∗ ∗ ∗
T L Yl L 1 ρl L
gL (PL∗ , TL∗ , QL ) = cvm ln r + a1 + − Yl ln
T L α L α ρrl L
h z i ∗ ∗ X Lr h i ∗L
L ρ0 Yl L i
− a1 ln − + Yi = 0,(3.55)
zr L α ρl L Tr L
i=l,g,v
h i ∗
L
where square brackets denote the following operation (·) = (·)∗L − (·)L .
L
40
3.5.4 Right Rarefaction Wave Analysis
The analysis is analogous to that for the left rarefaction wave, except that the
Riemann invariant across the right rarefaction is given by
dP du
dΥ5 = 0 = 2
+ , (3.56)
ρm Cm Cm
from Eq. (3.31). By replacing the subscript L by R, the following conditions are ob-
tained,
u∗R = uR − fR (PR∗ , ρ∗mR , QR ), (3.57)
where
Z ∗
dP
fR (PR∗ , ρ∗mR , QR ) = , (3.58)
R ρm Cm
and
∗ ∗ ∗
T R Yl R 1 ρl R
gR (PR∗ , TR∗ , QR ) = cvm ln r + a1 + − Yl ln r
T R α R α ρl R
h z i ∗ ∗ X Lr h i ∗R
R ρ0 Yl R i
− a1 ln − + Yi = 0,(3.59)
zr R α ρl R Tr R
i=l,g,v
h i ∗
R
where square brackets denote the following operation (·) = (·)∗R − (·)R .
R
3.5.5 Complete Solution
Now the conditions for all four possible wave patterns, shown in Figure 3.2, can
be determined. The unknown states K∗L and K∗R can be computed by utilizing the
condition that the pressure and velocity are constant across the contact discontinuity,
i.e.
PL∗ = PR∗ = P ∗ , (3.60)
and
u∗L = u∗R = u∗ . (3.61)

41
By eliminating u∗ from Eqs. (3.40) or (3.53) and (3.46) or (3.57), a single non-linear
algebraic equation is obtained,
fL (P ∗ , ρ∗mL , QL ) + fR (P ∗ , ρ∗mR , QR ) + uR − uL = 0. (3.62)
Note that Eq. (3.62) has three unknowns, namely, P ∗ , ρ∗mL , and ρ∗mR . To close
the problem, Eqs. (3.41) and (3.55) for the state to the left of the contact discontinuity
and Eqs. (3.47) and (3.59) for the state to the right of the contact discontinuity are
used, in addition to Eq. (3.15). Therefore, the following four cases are considered.
1. If P ∗ > PL then a shock wave is traveling to the left and the function fL is given
by Eq. (3.44) supplemented by Eqs. (3.41) and (3.15) for the left star region.
2. If P ∗ ≤ PL then a rarefaction wave is moving to the left and the function fL
is given by Eq. (3.54) supplemented by Eqs. (3.55) and (3.15) for the left star
region.
3. If P ∗ > PR then a shock wave is traveling to the right and the function fR is
given by Eq. (3.50) supplemented by Eqs. (3.47) and (3.15) for the right star
region.
4. If P ∗ ≤ PL then a rarefaction wave is moving to the right and the function fR
is given by Eq. (3.58) supplemented by Eqs. (3.59) and (3.15) for the right star
region.
The speed of the contact discontinuity can be determined from
1 1
u∗ = (uL + uR ) + [fR (P ∗ , ρ∗mR , QR ) − fL (P ∗ , ρ∗mL , QL )] . (3.63)
2 2
Chapter 4
Numerical Method for Simulation of Multiphase/Multicomponent

Flows
This chapter presents the development of the numerical method for modeling
and numerical simulation of multiphase and multicomponent flows. The numerical
method is used to solve the governing equations of multiphase and multicomponent
flows, which are systems of non-linear hyperbolic partial differential equations. Two
advanced numerical scheme such as Roe’s approximate Riemann solver [112] and the
modified Harten, Lax and van Leer scheme (HLLC) [15] are extended to multiphase and
multicomponent flows and used to capture shock waves and contact discontinuities.
4.1 Extended Roe’s Approximate Riemann Solver
The Roe’s approximate Riemann solver [112] is widely known in finding an ap-
proximate solution of the Riemann problem. The method has been applied to a variety
of physical problems. For example, Toro [140] has used the method to solve reactive two-
phase flow problems via a phase-splitting procedure, and Sainaulieu [116] has extended
the method to a class of two-phase flow problems without phase-splitting.
In this thesis, the Roe’s approximate Riemann solver is utilized and extended to
the mixture of liquid, gas and vapor. The extended Roe scheme is used to compute the
convective fluxes of the mixture. The flux at the cell interface is calculated by evaluating
the change in flux associated with each wave strength component characterized by the
43
eigenvalues of the Jacobian matrix and the right characteristic vector. The convective
fluxes of the gas and vapor are computed as suggested by [79] to ensure positivity. The
face states required by the flux computation are computed from a simplified WENO
scheme [58].
To demonstrate the Roe approach in calculating the fluxes at the cell interface,
a one dimensional time dependent Euler equations for the mixture of liquid, gas and
vapor is presented for simplicity. However, in the actual solver a three dimensional time
dependent Euler equations for the mixture is solved.
Let us consider a Riemann problem for one dimensional time dependent Euler
equations for the mixture of liquid, gas and vapor given by
∂Q ∂F
+ = 0, (4.1)
∂t ∂x


 QL if x < 0,
Q(x, 0) = (4.2)

 QR if x ≥ 0,
where the vectors of conserved variables and fluxes are

   
 ρm   ρm u 
   
   2 
 ρm u   ρm u + P 
   
   
Q=
 ρm emT  ,
 F=
 (ρm emT + P )u 
 (4.3)
   
   
 ρg φg   ρg φg u 
   
   
ρv φv ρv φv u
Roe linearized Eq. (4.1) by introducing the Jacobian matrix A(Q) = ∂F/∂Q.
Using chain rule, Eq. (4.1) can be written as
∂Q ∂Q
+ A(Q) =0 (4.4)
∂t ∂x
Roe then replaced the Jacobian matrix A(Q) by a constant Jacobian matrix Ã =
Ã(QL , QR ), which is a function of the left and right states. The original non-linear
44
system of equations (4.1) are replaced by a linearized system of equations with constant
coefficients
∂Q ∂Q
+ Ã =0 (4.5)
∂t ∂x
Now, the original Riemann problem Eq. (4.1) is replaced by the approximate Riemann
problem
∂Q ∂Q
+ Ã = 0, (4.6)
∂t ∂x


 QL if x < 0,
Q(x, 0) = (4.7)

 QR if x ≥ 0,
which is then solved exactly. The initial condition of the approximate Riemann problem
is the same from the original Riemann problem. The elements on the Roe Jacobian
matrix Ã are the Roe average quantities. Furthermore, the Roe Jacobian matrix Ã
must satisfy the following properties [141]:
1. Hyperbolicity of the system of equations: The linearized system of equations
must preserve the mathematical character of the original non-linear system of
equations. The Roe Jacobian matrix Ã must have real eigenvalues and complete
set of linearly independent right eigenvectors.
2. The Roe Jacobian matrix Ã must be consistent with the exact Jacobian matrix
A:
Ã(Q, Q) = A(Q) (4.8)
3. Conservation across discontinuities:
F(QR ) − F(QL ) = Ã(QR − QL ) (4.9)
The Riemann wave diagram used in the Roe scheme is shown in Figure 4.1. The
contact wave, whose speed is denoted by SM , separates the left, UL , and right, UR ,
45
t
SL SM SR
UL UR
x
xL xR
Figure 4.1: The Riemann wave diagram used in the Roe scheme.
states. The minimum and maximum velocities of the acoustic waves are denoted by SL
and SR , respectively. In Roe’s scheme, the approximate Riemann solution is defined as



 UL if SL > 0,
U= (4.10)

 UR if SR < 0,
where U is the vector of the conserved variables defined similar to Q in Eq. (4.3). The
corresponding interface flux, F, is defined as



 FL if SL > 0,
F= (4.11)

 FR if SR < 0.
The flux at the cell interface, x/t = 0 is given by

n
1 1X
F [Q(0)] = [F(QR ) + F(QL )] − r̃i |λ̃i |dṽi (4.12)
2 2
i=1
where r̃i are the right characteristic vectors, λ̃i are the eigenvalues of the system and
dṽi are the wave strengths or Riemann invariants.
In the following, the algorithm in calculating the fluxes at the cell interface using
the Roe’s approximate Riemann solver are outlined.

46
1. Given the conserved variables, Q = [ρm , ρm u, ρm emT , ρm Yg , ρm Yv ]T , for the
left (L) and right (R) states, the primitive variables, K = [P, u, T, Yg , Yv ]T ,
for the left (L) and right (R) states are calculated in the following:
ρm u
u = , (4.13)
ρm
ρm emT − 12 ρm u2
T = , (4.14)
ρm cvm
ρm Yg
Yg = , (4.15)
ρm
ρm Yv
Yv = , (4.16)
ρm
and
p
ψ1 − ψ2 + (ψ2 − ψ1 )2 + 4ψ2 ψ3
P = , (4.17)
2
where
2 !
βl Po
ψ1 = Cl2 ρo − (T − To ) − 2 − ρm Yl ,
Cl Cl
ψ2 = (ρm Yv Rv + ρm Yg Rg )T,
2 !
2 βl Po
ψ3 = Cl ρo − (T − To ) − 2 ,
Cl Cl
2. Compute the Roe-average quantities of the mixture as follows:
ρ̃m = Rρ ρmL , (4.18)

uL + uR Rρ
ũ = , (4.19)
1 + Rρ
PL + PR Rρ
P̃ = , (4.20)
1 + Rρ
TL + TR Rρ
T̃ = , (4.21)
1 + Rρ
YiL + YiR Rρ
Ỹi = , (4.22)
1 + Rρ
X
c̃vm = Ỹi cvi . (4.23)
i
p
where Rρ = ρmR /ρmL .
47
For the mixture speed of sound, the Roe-average is computed from Eq. (2.50)
!2
X ρ̃m Ỹi β̃i
ρ̃m c̃vm + P̃
i
ρ̃2i C̃i
2
C̃m = , (4.24)
X ρ̃m Ỹi 1
2
ρ̃m c̃vm
i
ρ̃2i C̃i2
The remaining variables appearing in Eq. (4.24) are computed as ρ̃i = ρi (P̃ , T̃ ),
β̃i = βi (P̃ , T̃ ), and C̃i = Ci (P̃ , ρ˜i ).
3. Compute the Roe-average wave speeds of the mixture. The Roe-average wave
speeds are the eigenvalues of the systems from Eq. (2.49) in one dimension.
λ1 = ũ − C̃m , (4.25)
λ2 = ũ, (4.26)
λ3 = ũ, (4.27)
λ4 = ũ, (4.28)
λ5 = ũ + C̃m . (4.29)
4. Compute the wave strengths of the mixture. The wave strengths are computed
using the relation ∆v = Π ∆K, where K are the primitive variables, and Π are
the eigenvectors of the system. The resulting wave strengths are
1 1
∆v1 = ∆P − ∆u, (4.30)
2
ρ̃m C̃m C̃m
P̃ 1
∆v2 = − ∆P + ∆T , (4.31)
2 ρ̃ c̃
ρ̃m C̃m m vm T̃ T̃
" #
φ̃g β̃g2 P̃
∆v3 = − ρ̃g C̃g2 − ρ̃m C̃m
2
+ ∆P + ∆φg , (4.32)
2 2
ρ̃g C̃g ρ̃m C̃m ρ̃m c̃vm
" #
φ̃v β̃v2 P̃
∆v4 = − ρ̃v C̃v2 − ρ̃m C̃m
2
+ ∆P + ∆φv , (4.33)
2 2
ρ̃v C̃v ρ̃m C̃m ρ̃m c̃vm
1 1
∆v5 = ∆P + ∆u . (4.34)
ρ̃m C̃m2 C̃m
where ∆P = PR − PL , ∆T = TR − TL , ∆u = uR − uL , ∆φg = φgR − φgL and
∆φv = φvR − φvL .

48
5. Compute the right characteristic vectors. The right characteristic vectors are
as follows (note: all the quantities are in Roe-average) :
   
β2
 −( Cφ2 )T  ρg − ρl
 φ   
   
2
 −( βφ2 )T u   
 Cφ   (ρg − ρl )u 
   
 2   
r1 =  ρl φl cvl T + βcvl T 2 φl − 1 ( βφ2 )T u2 , r2 = 
 ρg eg − ρl el 
,
 2 Cφ   
   
 βg2 φg T   
 − C2   ρg 
 g   
 
βv2 φv T 0
− C2
v
(4.35)
     
 ρv − ρl   Π1   Π1 
     
     
 (ρv − ρl )u   Π1 u − ρm2Cm   Π1 u + ρm2Cm 
     
     
r3 = 
 ρv ev − ρl el 
, r4 = 
 Π2 − 2
ρm Cm u ,
 r5 = 
 Π2 + 2
ρm Cm u ,

     
   ρg φg   ρg φg 
 0     
   2   2 
     
ρv φv ρv φv
ρv 2 2
(4.36)
where
β2
h i
ρm C 2 (ρg −ρl )φg βg2 P
Π1 = 2Cφ2
− 21 ( Cφ2 ) ρmPcvm + 2ρg Cg2 ρg Cg2 − ρm Cm
2 +
ρm cvm +
φ
h i
(ρv −ρl )φv βv2 P
2ρv Cv2
2 +
ρv Cv2 − ρm Cm ρm cvm , (4.37)
h i
2
ρm Cm e l φl e v φv e g φg P β2
Π2 = 2 Cl2
+ Cv2 Cg2
+ 2ρm cvm ρl φl cvl + βT cvl φl T − 21 ( Cφ2 )u2 +
φ
h i
(ρg eg −ρl el )φg βg2 P
2ρg Cg2
ρg Cg2 − ρm Cm
2 +
ρm cvm +
h i
(ρv ev −ρl el )φv βv2 P
2ρv Cv2
ρv Cv2 − ρm Cm
2 +
ρm cvm , (4.38)
49
and
1
el = cvl T + u2 , (4.39)
2
1
ev = cvv T + u2 , (4.40)
2
1 2
eg = cvg T + u . (4.41)
2
6. Finally, calculate the flux at x/t = 0

n
1 1X
F [Q(0)] = [F(QR ) + F(QL )] − r̃i |λ̃i |dṽi (4.42)
2 2
i=1
1
where 2 [F(QR ) + F(QL )] is the arithmetic average of the convective fluxes and
1 Pn
2 i=1 r̃i |λ̃i |dvi is the numerical dissipation, where n = 5.
Note that in the Roe’s approximate Riemann solver, all three waves have zero spread,
which means that the method cannot capture the finite spread of the expansion fan.
4.2 Extended HLLC Approximate Riemann Solver
Approximately two years after Roe’s [112] breakthrough in developing an approx-
imate Riemann solver, Harten, Lax, and van Leer [55] have devised a new approximate
Riemann solver denoted as HLL scheme. In their scheme, a single state approximate
Riemann solver can be constructed in a way that the entropy condition will be automat-
ically satisfied and will yield an exact resolution of isolated shocks. A modification of
the scheme was designed by Toro et al. [142], where a two-state approximate Riemann
solver can be devised so that both shocks and contact waves will be resolve exactly. The
modified scheme is called HLLC (C stands for Contact) Riemann solver. The scheme
has been applied to many physical problems such as supersonic two dimensional, time
dependent Euler equations [139] and two dimensional, time dependent shallow water
equations [45]. Further improvement of the scheme was presented by Batten et al. [15],
which show that with proper choice of acoustic and contact wave velocities, the HLLC
50
t
SL SM SR
* *
UL UR
FL FR
UL UR
x
xL xR
Figure 4.2: The simplified Riemann wave diagram with two intermediate states used in
the HLLC scheme.
scheme will yield an exact resolution of isolated shock and contact waves. The resulting
scheme is positively conservative, which forces the numerical method to preserve ini-
tially positive pressures and densities. For an in-depth description of the HLLC method,
see [15], and [141].
In this thesis, the HLLC approximate Riemann solver of [15] is extended and is
used to compute the convective fluxes of the mixture. The convective fluxes of the gas
and vapor components are computed as suggested by [79] to ensure positivity. The face
states required by the flux computation are computed from a simplified WENO scheme
[58]. In the following, an outline of the extended HLLC approximate Riemann solver of
[15] for the mixture of liquid, gas, and vapor is presented.
The simplified Riemann wave diagram with two intermediate states is illustrated
in Figure 4.2. The two averaged intermediate states, U∗L and U∗R , are separated by a
contact wave, whose speed is denoted by SM . The acoustic waves have the minimum
and maximum velocities, SL and SR respectively, of all waves exist in the exact solution.
51
The flow is supersonic from left to right if SL > 0, and the upwind flux is defined from
F = F (UL ), where UL is the left states. Conversely, the flow is supersonic from right
to left if SR < 0, and the flux is defined from F = F (UR ), where UR is the right states.
Harten, Lax, and van Leer [56] defined the two-state approximate Riemann solution,
UHLLC as 

 UL if SL > 0,






 U∗ if SL ≤ 0 < SM ,
L
UHLLC = (4.43)



 U∗R if SM ≤ 0 ≤ SR ,




 U
R if SR < 0,
where U is the vector of the conserved variables defined similar to Q in Eq. (2.6). The
corresponding interface flux, FHLLC , is defined as



 FL if SL > 0,




 ∗

 F
L if SL ≤ 0 < SM ,
FHLLC = (4.44)



 FR if SM ≤ 0 ≤ SR ,
∗




 F
R if SR < 0.
The two intermediate states U∗L and U∗R are determined following the approach
suggested by Toro et al. [142], which assumed constant velocity particle between the
acoustic waves. Applying the Rankine-Hugoniot conditions across the SL waves and
simplifying yields
F∗L = FL + SL (U∗L − UL ), (4.45)
SL U∗L − F∗L = SL UL − FL . (4.46)
Similarly, for the SR waves yields
F∗R = FR + SR (U∗R − UR ), (4.47)
SR U∗R − F∗R = SR UR − FR . (4.48)
To compute U∗L and F∗L , Batten et al. [15] made a specific assumption such that
SM = qL∗ = qR
∗
= q∗ , (4.49)
52
where q ∗ is the average directed velocity between the two acoustic waves, and
ρR qR (SR − qR ) − ρL qL (SL − qL ) + PL − PR
SM = . (4.50)
ρR (SR − qR ) − ρL (SL − qL )
The wave speeds SL and SR are computed from
SL = min[qL − CmL , q − C m ], (4.51)
SR = max[qR + CmR , q + C m ], (4.52)
where qL , qR , CmL , and CmR are the face-normal velocities and the speeds of sound
of the mixture at the left and right state, respectively, and q̃ is the Roe-averaged [112]
face-normal velocity.
The primary difficulty in extending the HLLC method to multiphase and multi-
component flows is due to the treatment of the speed of sound of the mixture. In the
HLLC method of [15] for single-phase flow, C̃m is computed from the the constant ratio
of specific heats and the Roe-averaged total enthalpy and velocities. For the multiphase
and multicomponent mixtures considered in this thesis, this is not possible because the
speed of sound given by Eq. (2.50) cannot be related to the total enthalpy in a straight-
forward fashion. Instead, the following is proposed to compute C̃m from Eq. (2.50)
as !2
X ρ̃m Ỹi β̃i
ρ̃m c̃vm + P̃
i
ρ̃2i C̃i
2
C̃m = , (4.53)
X ρ̃m Ỹi 1
ρ̃2m c̃vm
i
ρ̃2i C̃i2
where, using the definition,
p
Rρ = ρmR /ρmL , (4.54)
the Roe-averaged of the mixture density is defined as
ρ̃m = Rρ ρmL , (4.55)
and the Roe-averaged mass fractions is given by,
YiL + YiR Rρ
Ỹi = . (4.56)
1 + Rρ
53
Moreover, the Roe-averaged of the specific heat at constant volume of the mixture is
defined as
X
c̃vm = Ỹi cvi . (4.57)
i
In addition, for lack of a better approach, the mixture pressure and temperature is
defined as,
PL + PR Rρ
P̃ = , (4.58)
1 + Rρ
and
TL + TR Rρ
T̃ = , (4.59)
1 + Rρ
respectively. The remaining variables appearing in Eq. (4.53) are computed as ρ̃i =
ρi (P̃ , T̃ ), β̃i = βi (P̃ , T̃ ), and C̃i = Ci (P̃ , ρ˜i ).

Chapter 5
Code Verification and Model Validation
The new developed solver for simulation of multiphase and multicomponent flows
is verified and validated by solving some benchmark problems including the novel bench-
mark problems for the idealized fluid-mixture model. The problems considered are:
1. Single-phase two-component shock-tube problem: Tests the accuracy with which
shock waves and contact discontinuities are captured.
2. Shock-wave propagation in a single-phase two-component fluid: Tests the accu-
racy of computing the shock wave refraction at a component interface.
3. Two-phase single-component shock-tube problem for idealized-fluid mixture:
Tests the accuracy with which two-phase flows are solved if all solution variables
are discontinuous.
4. Two-phase single-component rarefaction problem for idealized fluid mixture:
Tests the accuracy for low-density flows.
5. Single-phase two-component shock-bubble interaction: Demonstrates the abil-
ity to solve the interaction of a shock-wave and a material interface in two
dimensions.
For the first four problems, the accuracy is assessed by comparing the numerical solu-
tions to the appropriate exact solutions. For the fifth problem, the accuracy is evaluated
55
by comparing the results to the experiments of Haas and Sturtevant [53] and the simu-
lations of Quirk and Karni [105].
5.1 Single-Phase Two-Component Shock-Tube Problem
The initial conditions correspond to two different ideal gases [70, 1, 79, 71, 138],


 QL if x < 0.5,
Q(x, 0) = (5.1)

 QR if x ≥ 0.5,
where Q = [ρm , u, P, γm , Y1 , Y2 ]T and
QL = [ 1.000, 0.0, 1.0 · 105 , 1.4, 1.0, 0.0 ]T ,

(5.2)
QR = [ 0.125, 0.0, 1.0 · 104 , 1.2, 0.0, 1.0 ]T .
The primary difficulty is the capturing of the contact discontinuity without oscillations.
The computational domain 0 ≤ x ≤ 1 is discretized uniformly with 1000 cells. Excellent
agreement between the numerical and exact solutions is obtained as shown in Figure
5.1. In particular, the results do not exhibit oscillations like the results presented in
[71].
5.2 Shock-Wave Propagation in a Single-Phase Two-Component

Fluid
The problem suggested by [71] is considered, in which a shock wave refracts at
a gas interface, leading to a transmitted and a reflected shock wave. The transmitted
shock wave may travel faster or slower than the incident shock wave depending on the
sound speeds of the respective gases. The reflected wave is either a shock wave or
a rarefaction wave depending on the ratio of the acoustic impedances [29, 61]. The
interface is set into motion by the shock wave.
The initial conditions correspond to a weak shock wave with a Mach number
56
(a) Density. (b) Pressure.
(c) Velocity. (d) Temperature.
(e) Mach number. (f) Speed of sound.
Figure 5.1: Comparison of numerical (◦) and exact (solid line) solutions for single-phase
two-component shock-tube problem at t ≈ 517 µs. Number of cells = 1000.
57
(c) Mass fraction of helium. (d) Temperature.
(e) Mach number. (f) Speed of sound.
Figure 5.2: Comparison of numerical (◦) and exact (solid line) solution for shock-wave
propagation in a single-phase two-component fluid at t ≈ 864 µs. Number of cells =
1000.
58
Ms = 1.1952 in air propagating toward a region occupied by helium,





 QA1 if 0.0 ≤ x < 0.25,


Q(x, 0) = QA2 if 0.25 ≤ x < 0.5, (5.3)





 QHe if 0.5 ≤ x, ≤ 1.0
where Q = [ρm , u, P, γm , Y1 , Y2 ]T and
QA1 = [ 1.7017, 98.956, 1.5 · 105 , 1.40, 1.0, 0.0 ]T ,
QA2 = [ 1.2763, 0.000, 1.0 · 105 , 1.40, 0.0, 1.0 ]T , (5.4)
QHe = [ 0.1760, 0.000, 1.0 · 105 , 1.67, 0.0, 1.0 ]T .

The vectors QA1 , QA2 , and QHe correspond to the post-shock variables in air, pre-
shock variables in air, and pre-shock variables in helium, respectively. For these initial
conditions, the transmitted shock wave is very weak and the reflected wave is a slender
rarefaction wave. The computational domain 0 ≤ x ≤ 1 is discretized uniformly with
1000 cells.
Figure 5.2 shows the comparison of the numerical and exact solution at t ≈ 864 µs.
As for the first test problem, the results exhibit no oscillations like those presented by
[71]. The transmitted shock wave travels faster than the incident shock wave since the
acoustic speed in helium is greater than the acoustic speed in air.
5.3 Two-Phase Shock-Tube Problem for Idealized Fluid Mixture
The two-phase shock-tube problem of [121] is considered. The driver section
contains a liquid at high pressure and the driven section contains a gas at low pressure.
The initial conditions are



 QL if x < 0.7,
Q(x, 0) = (5.5)

 QR if x ≥ 0.7,
where Q = [ρm , u, P, Y1 , Y2 ]T and
QL = [ 1500.0, 0.0, 1.12 · 109 , 1.0, 0.0 ]T ,

(5.6)
QR = [ 50.0, 0.0, 1.00 · 105 , 0.0, 1.0 ]T .
59
(c) Mass fraction of gas. (d) Velocity.
Figure 5.3: Comparison of numerical (◦) and exact (solid line) solution for two-phase
shock-tube problem for an idealized fluid mixture at 240 µs. Number of cells = 1000.
Note that the problem is very stiff. The density and pressure differ by ratios of 30 and
about 104 across the discontinuity, respectively. The computational domain 0 ≤ x ≤ 1
is discretized uniformly with 1000 cells.
Figure 5.3 shows the comparison of the numerical and exact solution at 240 µs.
Small oscillations in the density and velocity between the contact discontinuity and
shock wave are visible due to the proximity of the contact discontinuity and the shock
wave. This was also observed by [121]. However, the solution shows no oscillations at
the tail of the rarefaction wave.

60
(c) Volume fraction. (d) Velocity.
Figure 5.4: Comparison of numerical (◦) and exact (solid line) solution for two-phase
rarefaction problem for an idealized fluid mixture at 540 µs. Number of cells = 1000.
5.4 Two-Phase Rarefaction Problem for Idealized Fluid Mixture
A two-phase rarefaction problem is considered. The solution consists of two sym-
metric rarefaction waves and a trivial stationary contact discontinuity. The initial con-
ditions are 

 QL if x < 0.5,
Q(x, 0) = (5.7)

 QR if x ≥ 0.5,
where Q = [ρl , φl , ρg , φg , u, P ]T and
QL = [ 1000.0, 0.1, 1.2342, 0.9, −200.0, 9.7 · 104 ]T ,

(5.8)
QR = [ 1000.0, 0.1, 1.2342, 0.9, 200.0, 9.7 · 104 ]T .
61
Component γ R cv
Air 1.400 287.0 720.0
He 1.648 1578.0 244.0
Table 5.1: Properties of air and helium used in the simulations of shock-bubble interac-
tion. Note that the helium is contaminated with 28% air and its properties have been
adjusted accordingly.
The computational domain 0 ≤ x ≤ 1 is discretized uniformly with 1000 cells. Figure 5.4
shows the comparison of the numerical and exact solution at t = 540 µs. No problems
are encountered despite the low pressure and density.
5.5 Single-Phase Two-Component Shock-Bubble Interaction
Haas and Sturtevant [53] presented experiments of a weak planar shock wave with
Ms = 1.22 in air interacting with a cylindrical bubble filled with helium. The shock wave
is transmitted through the bubble and sets it into motion. The results presented focus
only on the early stages of interaction and include a comparison with the experiments
of Haas and Sturtevant and the computations of Quirk and Karni [105].
Both air and helium are assumed to be perfect gases with the properties listed in
Table 5.5. As indicated by Haas and Sturtevant, the helium bubble is contaminated with
air about 28% by mass [53]. The initial flow field is determined from the standard shock
relations given the strength of the incident shock wave and considering the density and
pressure of the quiescent flow ahead of the shock to be 1 kg/m3 and 105 Pa. The bubble
is assumed to be in thermodynamic and mechanical equilibrium with its surroundings.
Therefore, its initial density is given by ρHe = ρAir RAir /RHe , where RAir and RHe are
the gas constants for air and helium, respectively.
The computational domain is shown in Figure 5.5. Only the upper half is actually
computed because the flow is symmetric about the shock-tube axis. Quadrilateral grids
with uniform resolutions of h = ∆x = ∆y of 224 · 10−6 m, 112 · 10−6 m, and 56 · 10−6

62
0.0445
B C
Incident Shock
y[m]
A Cylindrical D
0.0000
Bubble
0.050
−0.0445
0.000 0.050 0.085 0.300

x[m]
Figure 5.5: Schematic diagram of the computational domain (not drawn to scale).
m are employed. The smallest grid spacing is comparable to that used on the finest
refinement level in the adaptive-grid computations of Quirk and Karni. (Of course, this
is not to advocate using grids with uniform spacing for purposes other than focussed
verification and validation studies.) Referring to Figure 5.5, solid-wall and symmetry
conditions are applied to BC and AD, respectively. Along AB, inflow conditions are
specified using the conditions behind the incident shock wave. An outflow condition
is applied along CD. Figure 5.6 shows a sequence of numerically generated Schlieren
images, illustrating the interaction of the shock wave with the bubble as computed on
the finest grid. The computation reproduces all the features of the interaction, and is
in good agreement with the experiments by Haas and Sturtevant and the simulations
by Quirk and Karni.
Figure 5.6(a) shows the helium bubble at t ≈ 32 µs, after it is hit by the incident
shock wave. A curved refracted shock is generated inside the bubble. Since the helium
has a higher speed of sound than the surrounding air, the refracted shock wave travels
faster than the incident shock wave. Also, a weak expansion wave is reflected outside
the bubble. The refracted shock wave eventually emerges from the bubble to become
the transmitted wave, see Figure 5.6(b). The incident shock diffracts at approximately
63
(a) 32 µs.
(b) 89 µs.
(c) 201 µs.
(d) 400 µs.
(e) 674 µs.
Figure 5.6: Numerically generated Schlieren images of the shock-bubble interaction

computations on the fine grid at various times.
64
(a) Fine grid.
(b) Medium grid.
(c) Coarse grid.
Figure 5.7: Numerically generated Schlieren images of the shock-bubble interaction

computations at approximately 674 µs for different grid spacings.
201 µs as shown in Figure 5.6(c). The bubble is deformed into a kidney shape and
spreads laterally, see Figure 5.6(d). The deformation is caused by vorticity generated
at the edge of the bubble due to the passage of the shock, which induces a jet of air
along the axis of symmetry [105]. Figure 5.6(e) shows the formation of the bubble into
a distinct vortical structure.
The accuracy of the numerical solutions are evaluated by conducting a conver-
gence study. Figure 5.7 shows the comparison of the results on the coarse, medium,
65
Figure 5.8: Definition of locations at which velocities are measured. VS - incident shock,
VT - transmitted shock, Vui - upstream edge of bubble, and Vdi - downstream edge of
bubble.
VS EVS VT EVT Vui EVui Vdi EVdi

Experiment [53] 410 − 393 − 170 − 145 −
Computation [105] 422 +3.0 377 −4.1 178 +4.7 146 +0.7
Coarse grid 426 +3.9 375 −4.6 172 +1.2 136 −6.2
Medium grid 426 +3.9 380 +3.3 168 −1.2 137 −5.5
Fine grid 423 +3.2 391 −0.5 169 −0.6 140 −3.4
Table 5.2: Comparison of computed velocities with those measured by Haas and Sturte-
vant [53] and computed by Quirk and Karni [105]. Percentage errors with respect to
the measurements by Haas and Sturtevant are also shown. The notation is defined in
Figure 5.8.
and fine grids at approximately 674 µs. The location of the bubble is approximately
identical for the three grids. As expected, the vortical structures are more pronounced
on the fine grid.
To validate the numerical solutions, the computed velocities of some prominent
flow features as defined in Figure 5.8 are compared to the measurements of Haas and
Sturtevant [53] the computations of Quirk and Karni[105]. The velocities are determined
66
from a linear least-squares curve fit of positions obtained from digitizing a sequence of
numerically generated Schlieren images. Table 5.5 lists the velocities and the percentage
errors of the computed results relative to the experimental values. The result are in
good agreement with the measured values and thus demonstrate the accuracy of the
approach.
Chapter 6
Multiphase and Multicomponent Free Surface Nozzle Jet Flow
This chapter illustrates the capabilities of the new developed analytical and com-
putational approach for modeling and numerical simulation of multiphase and multi-
component flows to model high pressure and supersonic multiphase and multicomponent
free surface nozzle jet flow. All numerical simulations performed in this Chapter are
based on the governing Euler equations for the mixtures.
The results of the simulations including shock wave generation in the gaseous
medium and cavitation in the nozzle injector are presented. The effects of cavitation to
the jet flow rate and jet flow distribution are investigated. Furthermore, the effects of
fluctuating inflow condition to the external jet formation are studied. In the simulation,
the extended Roe and HLLC approximate Riemann solver for multiphase and multi-
component flow are used. First, the two schemes are tested to solve a high speed flow
problems in an expanded duct. The purpose of this scenario is to test the capabilities
of the scheme to handle problems with sudden expansion. This could help in choosing
the appropriate scheme for simulating a supersonic multiphase and multicomponent free
surface jet flow. Second, with the appropriate scheme, simulation of high pressure and
supersonic multiphase and multicomponent free surface nozzle jet flow is conducted. A
shock wave generation in a gaseous medium is captured. Third, cavitation model is
introduced into the governing equation of multiphase and multicomponent flow. The
purpose of the model is to capture cavitation in the nozzle orifice and to investigate its
68
effect on the jet flow rate. Finally, simulation of a supersonic multiphase and multi-
component nozzle jet flow with a fluctuating inflow condition is performed in order to
study the instability and formation of the external jet.
6.1 Shock Tube Analysis in an Expanded Duct
A shock tube analysis in an expanded duct is performed to test the capabilities
of the two numerical schemes before simulation of free surface nozzle jet flow, which
requires considerably larger resources. There are four scenarios studied. First, a high
speed flow gas enters into an expanded duct that is filled with gas (gas into gas scenario).
Second, a high speed flow liquid enters into an expanded duct that is filled with liquid
(liquid into liquid scenario). Third, a high speed flow liquid enters into an expanded duct
that is filled with gas (liquid into gas scenario). Finally, a high speed flow of liquid/gas
enters into an expanded duct that is filled with gas (liquid/gas into gas scenario).
In all four scenarios, the two schemes (extended Roe and HLLC approximate
Rieman solver) were able to solve the problems like capturing the traveling normal
shock in the duct. However, when the shock reaches the region where there is an abrupt
change (sudden expansion) of the cross sectional area of the flow, the extended Roe
approximate Riemann solver failed, in particular to the third and fourth scenario. It
was observed during the simulation that the extended Roe approximate Riemann solver
had difficulty in dealing with large contact discontinuities and capturing the finite spread
of the expansion fan as compared to the extended HLLC approximate Riemann solver.
The mathematical formulation for multiphase and multicomponent flow was suc-
cessfully implemented in the incompressible single phase limit, compressible single phase
limit and incompressible-compressible two-phase limit. For the compressible and in-
compressible single phase limit, like in the first scenario and second scenario, the two
schemes (extended Roe and HLLC approximate Rieman solver) have no problem in
69
(a) (b)
(c) (d)
Figure 6.1: Plot of (a) mass fraction of liquid, (b) x-velocity component, (c) mixture
pressure, and (d) mixture temperature, after 5.7689 × 10−07 seconds through an ex-
panded duct, showing the normal traveling shock. Liquid density, ρl = 100.0 kg/m3 ;
gas density, ρg = 1.189 kg/m3 ; inflow condition, inflow jet velocity is twice the speed of
sound of the gas (Mach = 2.0 with respect to the gas).
dealing with sudden expansion of the cross sectional area of the flow. However, for the
incompressible-compressible two-phase limit, like in the third and fourth scenario, the
extended Roe approximate Riemann solver has encountered a problem just after the
expansion of the duct. This is due to the large contact discontinuities in the density
and velocity fields in both normal and tangential directions. Also, a sudden jump of
the momentum occurs when the fluid exits just after the expansion of the duct. Unlike
70
the extended Roe solver, the extended HLLC solver has no problem in dealing with this
region.
For an illustration, a selected third scenario (liquid into gas) is presented. In this
case, the duct is initially filled with stationary gas of density ρg = 1.189 kg/m3 . Then
a liquid of density ρl = 100 kg/m3 enters the duct at supersonic condition (Mach=2.0)
with respect to the gas. A total of 54,962 computational cells is used in the simulation.
Figure 6.1 shows the plot of the mixture density, mass fraction of the liquid, mixture
pressure and mixture x-velocity component, respectively. A normal traveling shock is
captured and travels with an accurate velocity. At the contact, the pressure is high due
to the momentum of the liquid that hits the stationary gas. The results are obtained
after t = 5.7689 × 10−07 seconds, where at this time the liquid travels a distance of
approximately 0.4 mm.
6.2 Multiphase/Multicomponent Free Surface Jet Flow
A simulation of a high pressure and supersonic multiphase and multicomponent
free surface nozzle jet flow is performed. The purpose of this simulation is to support
the experimental results of MacPhee et al. [89] and to provide more quantitative and
qualitative information of the structure of the jet close to the nozzle region. The result
of the latter experiment shows an oblique shock wave in a gaseous medium due to a
high injection pressure and high speed flow of the jet.
In the simulation, a nozzle with an orifice diameter of 178 µm (the same diameter
used by MacPhee et al. [89]) is connected to a chamber. Then a jet of fluid enters
the nozzle and exits the nozzle orifice through a chamber generating an oblique shock
wave. The intensity of the shock wave depends on the initial and inflow condition. The
injection pressure ranges from 50 MPa to 135 Mpa. The chamber initial condition is at
1 atm. and 30o C. A total of 137,400 computational cells is used in the simulation.
There are four cases considered in the simulation to test the physical capabilities
71
of the multiphase and multicomponent mixture model. First, a jet of gas is injected into
the chamber that is filled with gas (gas into gas case). Second, a jet of liquid is injected
into the chamber that is filled with liquid (liquid into liquid case). Third, a jet of liquid
is injected into the chamber that is filled with gas (liquid into gas case). Finally, a jet of
liquid/gas is injected into the chamber that is filled with gas (liquid/gas into gas case).
The multiphase and multicomponent mixture model was successfully implemented
in the incompressible single phase limit (liquid into liquid case), compressible single
phase limit (gas into gas case) and incompressible-compressible two-phase limit (liquid
into gas and liquid/gas into gas case) at subsonic and supersonic inflow conditions. In
all four scenarios, the extended HLLC approximate Riemann solver is used since the
extended Roe approximate Riemann solver has difficulty in handling problems where
there is a sudden expansion of the cross sectional area of the flow. The extended HLLC
solver is capable of solving multiphase and multicomponent flow problems with large
density and flow field variations.
Results of the simulation for test case four (liquid/gas into gas) are shown in
Figures 6.2 to 6.2. The inflow conditions are homogeneous mixture of 80% liquid, 10%
gas and 10% vapor by volume, respectively. It is assumed in the simulation that the jet
carries 10% of gas by volume due to gas entrapment and 10% of vapor by volume due
to cavitation in the nozzle. The corresponding density of the liquid, gas and vapor are
ρl = 1000.0 kg/m3 , ρg = 1.189 kg/m3 and ρv = 0.722 kg/m3 , respectively. The velocity
of the jet at the inflow is assumed supersonic with a Mach number of 2 with respect to
the gas inside the chamber. The results presented are at t ≈ 5µs.
Figure 6.2 shows the contour plot of the mixture density and mass fraction of
the liquid with the corresponding plot of the contact discontinuities across the interface
in the centerline of the jet axis, respectively. An oblique shock wave in the gaseous
medium is generated due to a high speed flow of the jet. The leading part of the jet
looks like a tail of a “needle” and in this region the mass is highly concentrated. The
72
Figure 6.2: Plot of: (top left) mixture density showing an oblique shock wave in a
gaseous medium, (top right) contact discontinuities of the mixture density across the
interface in the centerline of the jet axis, (bottom left) mass fraction of the liquid,
and (bottom right) contact discontinuities of the mass fraction of the liquid across the
interface in the centerline of the jet axis. Liquid density, ρl = 1000.0 kg/m3 ; gas density,
ρg = 1.189 kg/m3 ; vapor density, ρv = 0.772 kg/m3 . Inflow condition, Mach = 2.0 with
respect to the gas inside the chamber.
jet travels about 3.2 mm after t ≈ 5µs.
The contour plot of the pressure and temperature of the mixture with the corre-
sponding plot of the magnitude along the centerline of the jet axis is shown in Figure
6.2 respectively. A pressure and temperature wave is captured in the gaseous medium.
At the leading edge of the jet, the pressure and temperature increases tremendously due
to the strong impact and high momentum of the jet that hits the stationary gas.
The x-velocity component and the corresponding velocity profile across the lead-
73
Figure 6.3: Plot of: (top left) mixture pressure showing an oblique shock wave in
a gaseous medium, (top right) mixture pressure along the centerline of the jet axis,
(bottom left) mixture temperature showing an oblique shock wave in a gaseous medium,
and (bottom right) mixture temperature along the centerline of the jet axis. Liquid
density, ρl = 1000.0 kg/m3 ; gas density, ρg = 1.189 kg/m3 ; vapor density, ρv = 0.772
kg/m3 . Inflow condition, Mach = 2.0 with respect to the gas inside the chamber.
ing edge of the jet in the radial direction is plotted in Figure 6.2. The velocity of the
jet expands in the radial direction as it travels further downstream with a decreasing
magnitude. The jet entrained the near surrounding gas towards the direction of the
flow, which results to a recirculation of the gas on the side of the jet near the exit of
the nozzle region.

74
Figure 6.4: Plot of: (top left) contact discontinuities of the x-velocity component of the
mixture across the interface in the centerline of the jet axis, (top right) velocity profile of
the jet at 1.8 mm from the exit of the nozzle, and (bottom) mixture x-velocity component
showing an oblique shock wave in a gaseous medium. Liquid density, ρl = 1000.0 kg/m3 ;
gas density, ρg = 1.189 kg/m3 ; vapor density, ρv = 0.772 kg/m3 . Inflow condition, Mach
= 2.0 with respect to the gas inside the chamber.
6.2.1 Jet Formation and Jet Radial Profile
Figure 6.5 shows the jet formation. The leading part of the jet expands (slightly)
radially and forms like a tail of a “needle”. A bulk of fluid immediately followed the
leading edge of the jet, which is caused by the accumulation of the vapor and gas as the
jet impacted on the ambient gas at the leading edge. Note that the jet is composed of a
liquid, gas and vapor. The liquid at the leading edge of the jet moves at different rate
than those of the average body of the jet. Without the resistance to the surrounding
75
Figure 6.5: Plot of the mass fraction of the liquid. The leading edge of the jet is about
3.0 mm from the nozzle exit.
Figure 6.6: Jet radial profile across the body of the jet at a distant of about 3.0 mm
from the nozzle exit.
76
Figure 6.7: Comparison of the results between numerical simulation (bottom) and ex-
perimental observation of MacPhee et al. [89](top). (Left) Jet formation. (Right) Jet
radial profile across the leading part of the jet.
gas, the jet body can move faster than the leading edge and with speeds close to the
trailing edge. This phenomenon was observed by MacPhee et al. [89].
The jet radial profile across the body of the jet at a distant of 3.0 mm from the
exit of the nozzle is shown in Figure 6.6. The results of the numerical simulation is in
good agreement with the experimental results of MacPhee et al. [89]. A comparison
of jet formation and jet radial profile between numerical simulation and experimental
observation is shown in Figure 6.2.

77
6.3 Multiphase/Multicomponent Cavitating Nozzle Jet Flow
6.3.1 Cavitation Overview
Cavitation is a serious physical hydrodynamic phenomenon, which is common
problem to many fluid related devices such as in pumps, turbines, control valves and
nozzles. It can cause serious wear and tear and possibly damages the devices. This
happens when the vapor collapses after evaporation due to the decreased in the flow
velocity and to the increased in pressure. Cavitation may occur when the local static
pressure in a fluid drops below the saturation vapor pressure of the liquid at the actual
temperature for sufficient duration of time by static or dynamic means. Cavitation is
a dynamic phenomena since it is concerned with the growth and collapse of cavities.
There are two stages of the cavitation process as described in [75]. The first stage is
called “incipient stage”, where cavitation is just barely detectable. Here, the discernible
bubbles of incipient cavitation is small and a limited zone of cavitation. The second stage
is called “developed stage”, where evaporation rates increased and cavitation grows due
to the changes in conditions such as pressure, velocity and temperature. The occurrence
of the inception and development of cavitation depend on the condition of the liquid,
including the presence of contaminations, either solid or gaseous, and on the pressure
field in the cavitation zone. A good reference for a general study of cavitation can be
found in [75].
6.3.1.1 Types of Cavitation
1. Traveling Cavitation
It is composed of individual transient cavities or bubbles, which form in the
liquid and move with the liquid as they expand, shrink, and then collapse.
These traveling transient bubbles may occur at the low pressure region along a
solid boundary or in the liquid interior either at the cores of the moving vortices
78
or in the high turbulence region in a turbulent shear field.
2. Fixed Cavitation
A cavitation that develops after inception, in which the liquid flow detaches from
the rigid boundary of an immersed body or a flow passage to form a pocket or
cavity attached to the boundary.
3. Vortex Cavitation
The cavities are found in the cores of vortices that form in high shear zone.
Here, the cavitation may appear as traveling or fixed cavities. This is one of the
earliest observed cavitation types, since it often occurs on the tip of the blade’s
propeller. This may also occur on the boundary surfaces of a submerged jets
where there is sufficiently high shear rate. The life of vortex cavity is longer
than the traveling cavity due to the angular momentum of the liquid.
4. Vibratory Cavitation
Cavitation occurs and collapse due to a continuous series of high-amplitude,
high-frequency pressure pulsations in the liquid. These pressure pulsations are
generated by a submerged surface that vibrates normal to its face and sets up
pressure waves in the liquid.
6.3.1.2 Effects of Cavitation
Knapps et al. [75] classified the effects of cavitation into three general categories.
First, cavitation can modify the hydrodynamics of the flow of the fluid. The flow pattern
can be modified and the dynamic interaction between the liquid and its boundaries can
be altered as the cavity volume displaces the liquid. The effect of cavitation tends
to limit or lessen the force that can be applied to the liquid by the surface. Second,
cavitation can cause damage on the solid-boundary surface of the flow. It can cause
wear and tear of the surface of any solid boundaries including metals either, hard or
79
soft, brittle or ductile, chemically active or chemically inert. However, nonmetallic solid
boundaries like rubber, plastic, glass, quarts and concrete are susceptible to cavitation
damage. Third, cavitation can induced noise and vibration when bubbles or cavities
collapse.
6.3.1.3 Experimental Studies of Cavitation
The main difficulties of understanding the physical nature of cavitation is due
to the fact that cavitation is a high-speed phenomenon. This high-speed phenomenon
limits to investigate the process of inception, growth and collapse in both flowing and
non-flowing environments. The main problems of cavitation are (1) how to provide a
suitable means for producing a controlled cavitation, (2) how to detect and locate the
cavitation and (3) how to “arrest” the motion. Knapps et al. [75] presented and dis-
cussed some special equipment that is developed to produced cavitation under controlled
conditions in a specified locations, where observations can be made.
Knapps et al. [75] categorized and explained several methods for detecting the
presence of cavitation, either direct or indirect observation. For indirect observations,
cavitation can be detected (1) by determining the effect of cavitation on the performance
of a piece of equipment, (2) by measuring the effect of cavitation on the distribution
of pressure over the boundary at which cavitation occurs, (3) by sensing the noise
emitted by cavitation and (4) by allowing cavitation to scatter laser-beam light into
a photocell [39]. For direct observations, cavitation can be detected by visual and
photographic means. This photographic technique has the possibility for detailed study
of the hydrodynamic phenomenon either at inception or for advanced stages of cavity
development.
The study of cavitation grew rapidly when it was realized that cavitation effect
can cause a tremendous damage to many fluid related devices. Since then, there have
been many methods developed to detect and locate cavitation. Richards et al. [111]
80
conducted a cavitation experiment using a water shock tube. They used a pressure
transducer to monitors the pressure changes and used schlieren photography techniques
to study cavitation bubbles. Escaler et al. [41] conducted an experimental investigation
to evaluate the detection of cavitation in actual hydraulic turbines. Their methodology
is based on the analysis of structural vibrations, acoustic emissions and hydrodynamic
pressures measured in the machine. A flow visualization of cavitation with particle and
bubble using the Particle Image Velocimetry (PIV) image processing method is per-
formed by Kato et al. [73]. Another experiment that uses an ultra high speed video
camera with the maximum frame speed of 106 fps to observe the detailed aspects of three
cavitation patterns such as separated vortex cavitation in a convergent-divergent chan-
nel, Karman-vortex like cavitation in the wake flow of a circular cylinder and vibratory
cavitation in an ultrasonic vibratory apparatus is conducted by Sato et al. [119].
6.3.2 Cavitation Modeling and Simulation
Most of the experimental methods in detecting and locating cavitation are in
macroscopic scale and are limited to give detailed picture of the hydrodynamic phe-
nomenon. To perform a cavitation study by experimental method, it requires some
costly equipments or devices. However, with the advent of the computer and devel-
opment of the numerical technique, a numerical study of cavitation became the best
alternative to supplement the experimental results.
The new developed general framework for simulation of multiphase and multicom-
ponent flow is capable of incorporating a variety of cavitation models. To demonstrate
this, a cavitation model by Hosangadi et al. [64] is considered. The cavitation source
term is given by
Sv = Kl ρl φl + Kv ρv φv (6.1)
81
where Kl and Kv are defined as

 
 0 P < Pv 
Kv =  h i , (6.2)
1 Q∞ P −Pv
τv L∞ 1
ρ Q2
P > Pv
2 ∞ ∞
 
 0 P > Pv 
Kl =  h i . (6.3)
1 Q∞ P −Pv
τf L∞ 1
ρ Q2
P < Pv
2 ∞ ∞
and τf and τv are the rate for vapor formation and liquid reconversion, respectively, and
can be obtained experimentally. The cavitation number is given by:

" #
P − Pv
C= 1 2
. (6.4)
2 ρ∞ Q∞
where, Pv is the saturation pressure of the vapor at a given temperature.
Some results of unsteady cavitation in a nozzle is shown in Figure 6.3.2 to Figure
6.3.2. Cavitation starts to develop at the entrance of the nozzle orifice as shown in
Figure 6.3.2. This is an example of an “incipient stage” as described by Knapps et al.
[75]. This is followed by a “developed stage”, where evaporation rates increased and
cavitation grows due to the changes in conditions of pressure, velocity and temperature,
see Figs. 6.3.2 to 6.3.2. In Figs. 6.3.2 to 6.3.2, the cavities or bubbles move with
the liquid as they expand, shrink, and then collapse. This type of cavitation is called
traveling cavitation. Another type of cavitation that can be observed in this process is
the fixed cavitation, see Figs. 6.3.2 and 6.3.2, where a fixed cavity or bubble is attached
to the wall of the nozzle orifice. This can cause the fluid flow to detaches from the wall
of the nozzle orifice. Furthermore, Figure 6.3.2 shows the contour plot of the pressure.
A pressure drop in the nozzle orifice can be observed. Due to a sudden change of the
cross sectional area of the flow, the fluid flow accelerates in this region that can cause
a pressure drop, which can be below the saturation pressure of the vapor.
82
Figure 6.8: Plot of the mass fraction of the vapor after 2.000e-04 sec. Inflow condition:
P = 1,161,000 Pa; u = 10 sm2 ; domain length = 6 mm; orifice over nozzle entrance ratio:
at
ai = 0.3; mesh size = 8,366.
Figure 6.9: Plot of the mass fraction of the vapor after 2.202e-04 sec. Inflow condition:
P = 1,161,000 Pa; u = 10 sm2 ; domain length = 6 mm; orifice over nozzle entrance ratio:
at
ai = 0.3; mesh size = 8,366.
83
Figure 6.10: Plot of the mass fraction of vapor after 2.502e-04 sec. Inflow condition: P
= 1,161,000 Pa; u = 10 sm2 ; domain length = 6 mm; orifice over nozzle entrance ratio:
at
ai = 0.3; mesh size = 8,366.
Figure 6.11: Plot of the mass fraction of vapor after 3.000e-04 sec. Inflow condition: P
= 1,161,000 Pa; u = 10 sm2 ; domain length = 6 mm; orifice over nozzle entrance ratio:
at
ai = 0.3; mesh size = 8,366.
84
(a) .
Figure 6.12: Plot of the pressure after 3.000e-04 sec. Inflow condition: P = 1,161,000
Pa; u = 10 sm2 ; domain length = 6 mm; orifice over nozzle entrance ratio: aati = 0.3;
mesh size = 8,366.
6.3.3 Cavitation in Nozzle Injector
Cavitation will normally occur inside the nozzle injector in many modern fuel
injection systems primarily due to high injection pressure. Other important parameters
that additionally affect cavitation are the orifice inlet curvature, injection angle, and
nozzle aspect ratio. A numerical simulation is carried out for a high injection pressure to
confirm cavitation inside a nozzle injector. The two-dimensional plane experimental test
case of Roosen et al. [114] is used in the simulation. The same test case used by Yuan et
al. [155] to validate their numerical simulation. The dimension of the nozzle hole is 1 mm
× 0.28 mm × 0.2 mm (length × height × width). Figure 6.13 shows the computational
domain of the nozzle injector and the treatment of the boundary conditions. To reduce
the computational time, a symmetric boundary condition is imposed along the nozzle
85
Water Air
2 mm 1 mm 5 mm
Figure 6.13: Plot of the initial condition of the system. The nozzle region is filled with
liquid and the chamber downstream from the exit of the nozzle is filled with gas.
Nozzle Orifice
( 0.28 mm )
Pinlet Pexit
Air
Water
1 mm
2 mm 1 mm 5 mm
Figure 6.14: Plot of the computational mesh and boundary conditions of the two-
dimensional plane model nozzle injector.
86
Figure 6.15: Cavitation inside the nozzle injector. Experimental results taken from
Roosen et al. [114]. Injection pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar.
axial axis, and solves only the lower half of the computational domain. A total of 7140
computational cells are used in the numerical computation. The system is initialized
with liquid in the nozzle region and gas in the exit region. Figure 6.14 shows the initial
condition of the system. Note the sharp interface between the liquid and gas just after
the nozzle exit.
There are two scenarios performed in the simulation to compare with the experi-
mental results of Roosen et al. [114], and the numerical simulation results of Yuan et al.
[155]. The first scenario considers an injection pressure of 80 bar and an ambient con-
dition of 21 bar, while the second scenario considers an injection pressure of 80 bar and
an ambient condition of 11 bar. The injection pressure is assigned a value of the inlet
pressure for simplicity and is assumed to be a constant steady inlet pressure throughout
the calculation. In the next section, the injection pressure is fluctuated and the effect
to the external jet formation is observed.
In both scenarios, the results of the simulation, see Figure 6.17, were in substantial
87
Figure 6.16: Cavitation inside the nozzle injector. Numerical simulation by Yuan et al.
[155]. Injection pressure, Pinlet = 80 bar; exit pressure, Pexit = 11 bar.
Figure 6.17: Results of the numerical simulation for a cavitating nozzle injector using the
multiphase mixture formulation with a cavitation model. Injection pressure, Pinlet = 80
bar; exit pressure, Pexit = 11 bar.
agreement with the experimental results of Roosen et al. [114], see Figure 6.15, and the
numerical simulation results of Yuan et al. [155], see Figure 6.16, respectively. Figure
6.18 shows the contour plot of the vapor and gas volume fraction, inside and close to
the exit of the nozzle injector at different time levels. It was observed that the time
scale of the internal cavitation dynamics is on the order of 10 µs, which confirms the
assertion of Yuan et al. [155]. It was also noticed that the bubble cavities start to
develop near the inlet corner of the nozzle orifice, and extend further downstream to
the exit of the nozzle. The maximum length of cavity is observed after steady flow is
achieved. Maximum length of cavity is reached after 5 × 10−5 seconds. As can be seen
88
Figure 6.18: Contour plot of the vapor (left) and gas (right) volume fraction inside and
close to the exit of the nozzle injector. Constant steady injection pressure, Pinlet = 80
bar; exit pressure, Pexit = 11 bar.
in Figure 6.18, for the gas volume fraction, a presence of gas inside and near the exit
of the nozzle is observed. This could be a re-entrant jet of gas that starts to penetrate
the cavity and causes it to collapse. However, the re-entrant jet is not strong enough
to penetrate inside the nozzle further upstream. This might be due to the steadiness of
the flow and constant steady injection pressure.
Figure 6.19 shows the contour plot of the liquid volume fraction and velocity,
inside and close to the exit of the nozzle injector at different time levels. It is clear that
cavitation causes a reduction of the cross sectional area of the liquid jet flow as can be
seen on the left of Figure 6.19. This means that cavitation chokes the flow of liquid and
89
Figure 6.19: Contour plot of the liquid volume fraction (left) and velocity profile (right)
inside and close to the exit of the nozzle injector. Constant steady injection pressure,
Pinlet = 80 bar; exit pressure, Pexit = 11 bar.
reduces the discharge significantly. The coefficient of discharge within the orifice for a
1.2 cavitation number was computed to be equal to a value of 0.75, which is about 7
% higher as compared to Nurick’s [99] correlation and Singhal et al. [126] simulation.
Also, shown to the right of Figure 6.19 is how cavitation separates the flow inside the
nozzle injector and affects the velocity profile of the jet flow.
90
6.4 The Effect of Fluctuating Inflow Condition to the External Jet

Formation
It has been illustrated in the previous section that cavitation occurs inside the
nozzle injector and has significant effect on the liquid flow discharge and velocity profile
of the flow. In the actual injection process, the injection pressure and exit pressure
fluctuate at high frequencies. To understand the effect of injection pressure fluctuation
to the cavitation process and the external jet formation, a periodic rectangular inflow
pressure condition is implemented. Figure 6.20 shows the results of the simulation at
different time levels for a fluctuating inflow pressure condition. The volume fraction of
liquid that relates to the amount of liquid discharge and the formation of the external
jet is shown on the left of Figure 6.20, while the volume fraction of vapor inside and
close to the exit of the nozzle injector is depicted on the right of Figure 6.20.
The cavitation process in the actual injection system becomes more complex
due to fluctuating inflow boundary condition. There is significant interaction of the
bubble cavities and gas from the nozzle exit. It can be seen on the right of Figure
6.20 that the cavitation process is more complicated. A re-entrant jet of gases from
the downstream chamber causes the cavities to collapse and separate. The separation
of the bubble cavities creates a bubble cloud downstream of the cavity. The bubble
clouds will interact with a large rotating vortex of gas external to the nozzle as it exits
from the nozzle injector. Qin et al. [104] further postulated that the external pressure
propagates into the orifice once the re-entrant jet reaches the nozzle orifice inlet and
occupies the entire upper part of the orifice, causing the flow in the nozzle to revert to
a non-cavitating mode. This phenomenon is called hydraulic flip, which was observed
by Bergwerk et al. [16].
The discharge of the nozzle is strongly dependent on the cavitation process and
the magnitude of the bubble cavities. As can be seen on the left of Figure 6.20 , the
91
Figure 6.20: Plot of the unsteady external jet formation (left) and distribution of vapor
volume fraction (right) inside and close to the exit of the nozzle injector. Periodic
unsteady injection pressure, Pinlet = 80 ± 10 bar; frequency, f = 3.725 mhz; exit
pressure, Pexit = 11 bar.
magnitude of discharge fluctuations is related to the fluctuation of the inflow boundary
condition. It should be noticed that the cavitation process for a fluctuating inflow
condition intensifies the unsteadiness of the external jet formation. The instability of
the jet formation and the fluctuating amount of discharge greatly affect the jet break-up
and the atomization process further downstream .

Chapter 7
Multiphase and Multicomponent Turbulent Flows
This chapter extends the multiphase and multicomponent approach to turbulent
flows. Multiphase and multicomponent turbulent flows are widely present in various
engineering, environmental and geophysical applications. In most cases, flow configura-
tions in these areas are very complex. Due to the complexity of the flow, modeling of
turbulent flows of multiphase and multicomponent fluids is very difficult. The develop-
ment of a turbulence model is the most challenging task due to a high density contrast
of the fluids.
Turbulence develops whenever incipient instabilities, which are not dissipated
quickly enough by the stabilizing forces acting in the fluid motion, became able to
extract continuously energy from the flow [24]. Turbulence can be characterized by the
Reynolds number and affected by the density variations of the fluid flow. When the
density is no longer constant, new additional driving and stabilizing forces introduces
new instability mechanisms. The basic mechanism from which turbulence can originate
is dominated by inertia effects. At low Reynolds number, density variations influences
the unsteady turbulent flow through gravity, inertia, and viscosity [24]. While, at high
Reynolds number, the pressure gradient through the pressure-density coupling gains
significant influence on the unsteadiness of turbulent flow. Some study of variable-
density effects in low-speed regime can be found in [46, 47].

93
7.1 Turbulent Modeling
Current methods for modeling and simulation of turbulent flows can be grouped
into three general categories: Direct Numerical Simulation (DNS), Reynolds Averaged
Navier-Stokes (RANS), and Large Eddy Simulation (LES). The following briefly de-
scribed the three approaches:
1. Direct Numerical Simulation (DNS) is an approach that solves the Navier-Stokes
equations directly, which implies that the numerical simulation employs com-
putational mesh capable of resolving turbulent structures all the way to Kol-
mogorov scale.
2. Reynolds Averaged Navier-Stokes (RANS) approach is derived by averaging
Navier-Stokes equations over time, resulting either in the set of steady equations
for mean quantities in the case of steady RANS or a set of evolution equations
for slow varying averaged variables in the case of unsteady RANS.
3. Large Eddy Simulation (LES) is an approach, where the formal scale separation
is obtained by means of a low-pass filtering operation applied to the Navier-
Stokes equations. The filtered Navier-Stokes equations are closed by modeling
the subgrid-scale (SGS) stresses that account for the effect of the unresolved
small-scale motions [51].
The advantage of DNS approach is no modeling needed to close the governing Navier-
Stokes equations, but the approach is limited to feasible small Reynolds number flows
due to computational cost. On the other hand, RANS approach is computationally af-
fordable and is widely used in many industrial engineering applications. However, there
are some disadvantages of RANS approach such as modeling all the scales, obtaining
only the mean flow characteristics, and unable to accurately predict spatio-temporal
characteristics of turbulent flows in complex geometries and in transitional regimes.

94
With respect to LES, the approach is more accurate compared to RANS and able to
predict spatio-temporal turbulent flow characteristics. The LES approach is physically
and mathematically rigorous, and less empirical in nature. Although the LES method is
still computationally intensive compared to RANS, but, with the recent breakthrough
in computer technology, LES of flows of engineering interest has become computation-
ally feasible. As a result, LES methodology is gaining momentum and is starting to be
widely used in industry. The LES approach has been successfully applied to a variety of
turbulent flows [19, 38, 68, 102, 113, 131, 132, 148]. On the other hand, most of the cur-
rent LES implementations are limited to either single phase and single component flows
[11, 18, 31, 76, 92, 107] or single phase and multicomponent flows [24, 27, 95, 110, 147].
In this thesis, the LES methodology is extended to multiphase and multicomponent
flows.
7.2 Multiphase/Multicomponent Large Eddy Simulation
In LES, the large scale or the resolve scale flow field is directly calculated while the
effects of small scale called subgrid scales (SGS) are modeled. The resolved variables
and their governing equations are defined by explicitly applying the spatial low pass
filter H to each term in the governing equations. This can be done by decomposing
each field quantity Q in the flow domain Ω into large and small scale components,
′
Q = Q̄ + Q , (7.1)
where the filtered quantity Q̄ is defined as

Z
Q(x) = H(x − ζ, ∆)Q(ζ)dζ, (7.2)
Ω
′
and Q is the SGS part that accounts for small scales not resolved by the filter width
∆. Q is the vector of the conserved variables.
In compressible flows, the Favre-filtering [43, 44] is commonly used to avoid the in-
troduction of SGS terms in the equation of mass conservation. A Favre-filtered variable

95
is defined as
e = ρQ .
Q (7.3)
ρ̄
In this thesis, the Favre-filtering is utilized for multiphase and multicomponent flows
not only to avoid the presence of SGS terms in the continuity equation but also to
account the large density fluctuation. Thus, the resolved variables (denoted by tilde)
such as velocity, temperature and transport scalar fields in the mixture are written in
terms of Favre-filtered quantities [130]. These are given by
Ve = ρV , (7.4)
ρ̄
ρT
Te = , (7.5)
ρ̄
ρYv
Yev = , (7.6)
ρ̄
ρYg
Yeg = . (7.7)
ρ̄
7.2.1 Governing Equations
The governing three-dimensional time-dependent Navier-Stokes equations for the
mixture given by Eq. 2.5 are extended to a full three-dimensional time-dependent
LES equations. Applying the Favre-filtering operation to Navier-Stokes equations, the
multiphase/multicomponent LES governing equations for the mixture are written as
e
∂Q e
∂E e
∂F e
∂G ∂Eev ∂Fev ∂Gev
+ + + = + + e
+ S, (7.8)
∂t ∂x ∂y ∂z ∂x ∂y ∂z
e F
where Q is the vector of the resolved conservative variables and E, e and G
e are the
convective flux vectors given by Eq. 2.6, whose quantities are Favre-filtered. The vectors
e v, F
E e v and G
e v on the right hand side of Eq. (7.8) are the viscous flux vectors, which
include the SGS terms and whose quantities are Favre-filtered. The viscous flux vectors
96
are written as
e v = [0, τxx , τyx , τzx , Θx , Ψgx , Ψvx ]T ,

E (7.9)
e v = [0, τxy , τyy , τzy , Θy , Ψgy , Ψvy ]T ,

F (7.10)
e v = [0, τxz , τyz , τzz , Θz , Ψgz , Ψvz ]T .

G (7.11)
The total stress, τij , are defined as
2 ∂e
vk
µm Seij − µ
τij = 2e em δij − τijSGS , (7.12)
3 ∂xk
em is the resolved viscosity of the mixture, Seij = 12 (∂e

where µ ui /∂xi + ∂e
ui /∂xi ) is the
resolved strain rate tensor, and τijSGS is the SGS stress. The work of the total stress
and heat conduction in the fluid mixture are defined as
∂ Te
Θx = u e xz + e
eτxx + veτxy + wτ km − ExSGS , (7.13)
∂x
∂ Te
Θy = u e yz + e
eτyx + veτyy + wτ km − EySGS , (7.14)
∂y
∂ Te
Θz = u e zz + e
eτzx + veτzy + wτ km − EzSGS , (7.15)
∂z
where, e
km is the resolved thermal conductivity of the mixture and EiSGS is the SGS
energy flux. Furthermore, the scalar fluxes of the gas and vapor are defined as
∂ρm Yg
Ψgi = ϕ
e − ΨSGS
gi , (7.16)
∂xi
∂ρ Yv
Ψvi e m − ΨSGS
= ϕ vi , (7.17)
∂xi
where ϕ
e is the resolved diffusion coefficient, which is considered to be zero, unless
otherwise specified. ΨSGS

gi and ΨSGS
vi are the SGS scalar flux of the gas and vapor,
e on the right hand side of Eq. (7.8) represents the
respectively. In addition, the vector S
filtered source terms written as
e = [0, 0, 0, 0, 0, 0, S̃v ]T ,
S (7.18)
where S̃v is the filtered cavitation source term.

97
7.2.2 Subgrid Scale (SGS) Model
The SGS terms that appears in the momentum, energy and scalar equation due
to the low-pass filtering of the governing equations are not known and must be modeled.
The most commonly used subgrid scale model in LES is the Smagorinsky eddy-viscosity
model [127], which has been widely used for incompressible flows. Piomelli et al. [102]
have extended the model by using a scale function to avoid the model’s excessive damp-
ing of SGS instabilities. An extension of the model to compressible isotropic turbulence
flow was developed by Speziale et al. [130] and Erlebacher et al. [40]. To avoid the ad
hoc treatment of the model constants, a dynamic subgrid-scale eddy viscosity model for
incompressible flows was introduced by Germano et al. [49]. Furthermore, Moin et al.
[95] have extended the dynamic model for compressible flows and scalar transport. So
far, most of the subgrid scale models are limited to either incompressible or compressible
turbulent flows. In this thesis, the extension of the LES to multiphase/multicomponet
flows is proposed. In the following, the SGS model for momentum, energy and scalar
flux for the mixture are outlined.
7.2.2.1 SGS Model for the Momentum Equation
The subgrid momentum flux, τijSGS , represents the effect of the unresolved subgrid-
scales in the momentum equation. This is defined as
τijSGS = ρ̄m (ug

i uj − ũi ũj ). (7.19)
The term ug
i uj on the right hand side of Eq. 7.19 is not available, hence Eq. 7.19 cannot
be solved. However, instead of solving the right hand side of Eq. 7.19, it is modeled
using the eddy-viscosity model. Here, the proposed subgrid-scale eddy viscosity model
of Yoshizawa [153] and Moin et al. [95] is utilized for modeling the SGS momentum flux
of the mixture. The SGS model for the momentum flux is defined as
e Seij + 2 ρm CI ∆2 |S|
τijSGS = −ρm CS ∆2 |S| e 2 δij (7.20)
3
98
e = (2Seij Seij )1/2 and

where static coefficients CS = 0.0512, CI = 0.09 were used and |S|
e 2 = Se2 /2. Note that the two model coefficients can be determined dynamically
|S| ij
rather than input a priori [49, 92, 95].
7.2.2.2 SGS Model for the Energy Equation
The subgrid energy flux, EiSGS , represents the effect of the unresolved subgrid-
scales in the energy equation. This comprised the SGS heat flux, EiSGSHF ; SGS turbu-
lent dissipation, EiSGST D ; and SGS viscous dissipation, EiSGSV D :
EiSGS = EiSGSHF + EiSGST D + EiSGSV D . (7.21)
The SGS heat flux is defined as
EiSGSHF = ρm (ug
i T − ũi T̃ ), (7.22)
and is modeled using an eddy-diffusivity model similar to Refs. [92, 95], except that it
is utilized for the mixture
νT ∂ T̃
EiSGSHF = −ρ̄m c̃pm . (7.23)
P rT ∂xi
e is the turbulent eddy viscosity of Eq. 7.20. Simplifying Eq.

The term νT = CS ∆2 |S|
7.23 gives
e ∂ T̃
CS ∆2 |S|
EiSGSHF = −ρ̄m c̃pm . (7.24)
P rT ∂xi
where CS is the eddy-viscosity coefficient of Eq. 7.20 that can be either defined statically
or computed dynamically [49, 92, 95]. Similarly, the turbulent Prandtl number, P rT
can be either assigned or computed dynamically [92, 95]. For simplification, a new
parameter, CT = CS /P rT is introduced, which can also be either defined statically or
computed dynamically [28, 38, 49, 95, 101]. Finally, the SGS model for heat flux is
reduced to
e ∂ T̃ ,
EiSGSHF = −ρ̄m c̃pm CT ∆2 |S| (7.25)
∂xi
99
where CT = 0.03125 is used as in the case of Moin et al. [95] experiment.
The SGS turbulent dissipation is defined as
EiSGST D = ρm (uig
uj uj − ũi ug
j uj ), (7.26)
and is modeled based on the proposed model by Knight et al. [76]
EiSGST D = ũj τijSGS . (7.27)
The effect of SGS turbulent dissipation is included in Eq. 7.13. A dynamic version
of the SGS turbulent dissipation has been developed by Martin et al. [92]. The effect
of SGS viscous dissipation in the energy equation is minimal and is about 5% of the
divergence of SGS heat flux [92]. Since the contribution of SGS viscous dissipation to
the total energy is small, EiSGSV D ≈ 0, the effect of this is neglected in this thesis.
7.2.2.3 SGS Model for the Scalar Equation
The subgrid scalar flux, ΨSGS

gi and ΨSGS
vi , represent the effect of the unresolved
subgrid-scales in the gas and vapor species equation, respectively. These are defined as
ΨSGS
gi = ρ̄m (ug
i Ygi − ũi Ỹgi ), (7.28)
ΨSGS
vi = ρ̄m (ug
i Yvi − ũi Ỹvi ). (7.29)
These terms can be modeled either using a variety of models, e.g. [95, 101] or the
dynamic structure model of Chumakov et al. [28]. An eddy viscosity type approach
is considered in this thesis and is utilized for the SGS scalar flux of the gas and vapor
species equation. The SGS scalar fluxes are modeled as
e
ΨSGS e ∂ Yg ,
= −ρm CY g ∆2 |S| (7.30)
gi
∂xi
e
ΨSGS
vi
e ∂ Yv ,
= −ρm CY v ∆2 |S| (7.31)
∂xi
where, in general, CY g 6= CY v are the eddy-viscosity coefficients, which can either be
assigned statically or computed dynamically [49, 92, 95]. However, note that in order
100
the continuity equation for the mixture to be automatically satisfied, the eddy viscosity
coefficients must be equal, CY g = CY v .
7.3 Large Eddy Simulation of a Plane Jet
To demonstrate the capabilities of the extended LES approach for modeling and
simulation of multiphase and multicomponent turbulent flows, LES of a plane jet is
performed. Numerical simulations of a plane jet are very common in single phase flow
[18, 31, 66, 67, 81, 82, 86, 91, 107], but not in multiphase and multicomponent flows.
In this section, numerical simulations of multiphase and multicomponent jet flows are
presented. In the simulation, the computational domain of the jet that is considered
is 40 mm x 30 mm x 4 mm with the corresponding computational grid of 120 x 160
x 16. The computational mesh is shown in Figure 7.1 with a zoom view near the jet
inlet shown in Figure 7.2. Note that this grid was chosen based on the information
provided by [81, 86]. The center of the jet inlet, having a width d = 1 mm , is located
at x = 0 mm and y = 0 mm. A uniform grid is used in the z direction with spacing
hz = 0.25. In the x direction, the grid is relatively fine between 0 mm and 2 mm with
spacing hx = 0.1 in order to have adequate numerical resolution of the inflow region,
and then stretched. In the y direction, the grid is relatively fine in the ±0.5 mm in
order to resolve the initial shear layer and then stretched on both sides of the inlet.
The inflow boundary condition, see Figure 7.3, is adopted from [67], which is a
top-hat mean velocity profile with a co-flow velocity of Uc = 0.1Uj , where Uj is the jet
velocity. A three-dimensional disturbance is applied to the streamwise and cross-flow
velocity at a single frequency and single spanwise wavenumber. For the outflow and
upper and lower sidewall boundaries, a non-reflecting boundary condition is used.

101
0.0150
0.0100
0.0050
y [m]
0.0000
-0.0050
-0.0100
-0.0150
0.000 0.010 0.020 0.030 0.040
x [m]
Figure 7.1: Computational grid on a x - y section.
0.0050
y [m]
0.0000 d
-0.0050
Figure 7.2: Zoom view near the inlet region.

102
Uc
Outflow
Inflow
d Uj
Uc
Figure 7.3: Inflow boundary condition.
7.3.1 Single Phase LES of a Plane Jet
A single phase compressible plane jet with Mach number, Mj = 0.9 is considered.
The jet Reynolds number is Red = ρ∆Ūj d/µ ≈ 2000, where ∆Ūj is the velocity difference
between the two streams at the inflow, d is the slot width, and ρ and µ are the density
and viscosity of the jet, respectively. The properties of the gas that are used in the
kg N −s
simulations are ρ = 1.18 m3
and µ = 0.00089 m2
. The jet is strong since the co-flow
velocity is weak compared to the jet centerline velocity, which results in a strong shear
layer at the jet edges. The ratio of the jet slot width to the inflow momentum thickness
is 20. Also, the velocity ratio between the low-speed and high-speed streams is 0.1.
103
7.3.1.1 Computational Results
The profiles of the longitudinal velocity is shown in Figure 7.4. Similarity coordi-
nates are used with the transverse y direction normalized by the jet half-width and the
jet velocity difference with respect to the small co-flow normalized by its value at the
centerline. The velocity is obtained by taking the average velocity along the span-wise
direction at different stream-wise location, and then computing the average velocity at
different time intervals. As can be seen in Figure 7.4, the longitudinal mean veloc-
ity profile near the inflow is almost flat and then downstream, for x > 20, develops
rapidly to self-preserving profiles. In addition, figure 7.5 shows the profiles of stream-
wise fluctuation intensity at various locations. The profiles are nearly self-similar. The
fluctuations observed in the figure are due to insufficient sampling sizes. The profiles of
the longitudinal velocity and turbulence intensity are in good agreement although not
plotted, with the experiment and simulation [81].
The early evolution of the jet, plotted with an iso-surface contour of the magnitude
of vorticity, is shown in Figure 7.6. The development of the shear layer instabilities,
which causes the generation of the Kelvin Helmholtz rollers is observed. The inception
of flow instabilities and the amplification of fluctuating vorticity tends to further confine
the vorticity. The disturbance of a vortex sheet continues to grow as the jet propagates in
the stream-wise direction. The development of the jet is characterized by the dynamics
of large scale vortex rings, and by strong vortex interactions that leads to a more
disorganized flow regime [52]. The formation of the jet depends on the generation of
vorticity near the jet exit and the jet convection and diffusion. Furthermore, the jet
regime can be characterized by the Reynolds number. At moderately high Reynolds
numbers, laminar jets are not stable, even in the absence of density variation and
thermal effects, and rapidly develop into fully three-dimensional turbulent flows [137].
The vortex rollers continues to develop as the jet propagates downstream. The
104
Figure 7.4: Profiles of the longitudinal velocity at different streamwise locations.
Figure 7.5: Profiles of the streamwise turbulence intensity at different locations.

105
(a) (b)
(c) (d)
Figure 7.6: Contours of the magnitude of vorticity on an xy-plane at different time. (a)
t=5.0E-05sec., (b) 1.04E-04sec., (c) 1.4E-04sec., (d) 1.88E-04sec.
disturbance of the vortex rollers intensifies the fluctuation of the jet downstream. Figure
7.7 shows the evolution of the jet at time, t = 2.05E − 04 sec. The contour plot of the
magnitude of vorticity as shown in Figure 7.7 (a) illustrates the development of the
Kelvin-Helmholtz instability. The jet structure behaves as sinusoidal wave. This is
typical behavior for small density variations, where the most dominant unstable mode
is the sinusoidal mode. The velocity of the jet spreads downstream with a decreases
in momentum, and at the centerline the velocity oscillates in a sinusoidal wave. The
106
(a) (b)
(c) (d)
Figure 7.7: Snapshots of contours of the (a)magnitude of vorticity on an xy-plane, (b)

velocity, (c) density and (d) pressure at time, t=2.05E-04sec.
contour plot of the magnitude of velocity is shown in Figure 7.7 (b). The jet diffuses
downstream and the magnitude of density decreases as can be seen in Figure 7.7 (c).
Series of recirculation zones on both sides of the jet due to the effect of vorticity and
the entrainment of the surrounding medium are observed. This is visible in the contour
plot of pressure field, see Figure 7.7 (d).

107
(a) (b)
(c) (d)

The merging of the two unstable shear layers, which was observed in Figure 7.7
initiates the development of turbulent jet. This is illustrated in Figure 7.8. When the
two unstable shear layers merge, the jet breakups and the flow separates. The contour
plot of the magnitude of vorticity as shown in Figure 7.8 (a) illustrates the breakups of
the jet and the separations of the flow. The momentum of the jet decreases downstream
and the velocity of the jet spreads non-symmetrically. This can be observed in Figure
108
(a) (b)
(c) (d)

7.8 (b), which plots the contour of the jet velocity. In addition, the jet diffuses and the
magnitude of density continues to decrease due to the interaction of vortices as shown
in Figure 7.8 (c). Similarly, the magnitude of pressure field decreases due to an increase
of recirculations as observed in Figure 7.8 (d).
The turbulent structure of the jet including the formation of jet breakups and
separation of flows varies at different time. Figure 7.9 shows the evolution of turbulent
109
(a) (b)
Figure 7.10: Comparison of a snapshot of turbulent jet. (a) extended LES approach for
multiphase/multicomponent flows, (b) LES approach from [6].
jet at time, t = 2.80E − 04 sec. The formation of the jet breakups as shown in Figure
7.9 (a) differs from Figure 7.8 (a). The jet continues to loss momentum downstream as
can be seen in Figure 7.8 (b). It can also be noticed in Figure 7.8 (c) the decrease of
magnitude of density. Likewise, in Figure 7.8 (d), the decrease of magnitude of pressure
field is observed.
A qualitative comparison of the result of numerical simulation for turbulent jet
between the extended LES approach for multiphase/multicomponent flows and the LES
approach from [6] is shown in Figures 7.10 (a) and (b), respectively. A good agreement
of the results showing the structure of a turbulent jet is observed between the two
approach.
7.3.2 Multiphase/Multicomponent LES of a Plane Jet
A multiphase/multicomponent plane jet with Mach number, Mj = 0.9 with re-
spect to the gas is now considered. The jet is assumed to be a mixture of 80% liquid,
10% gas and 10% vapor by mass fraction. The jet exit through a gaseous medium. The
110
jet Reynolds number is Red = ρm ∆Ūj d/µm ≈ 2000, where ∆Ūj is the velocity differ-
ence between the two streams at the inflow, d is the slot width, and ρm and µm are the
mixture density and mixture viscosity of the jet, respectively. The fluid properties used
in the simulations are:
1. liquid (water) with density, ρl = 1000 kg/m3 and viscosity, µl = 0.00089 Pa-s.
2. gas1 (air) with density, ρg = 1.18 kg/m3 and viscosity, µg = 0.0000173 Pa-s.
3. gas2 (vapor) with density, ρv = 0.83 kg/m3 and viscosity, µv = 0.000013 Pa-s.
The velocity ratio between the low-speed and high-speed streams at the inflow is 0.1.
The ratio of the jet slot width to the inflow momentum thickness is 20.
7.3.2.1 Computational Results
Results of numerical simulation for multiphase and multicomponent jet flows
demonstrate the capabilities of the extended LES approach for multiphase and multi-
component flows. The computational results confirm physically observed phenomena,
which shows that the extended LES approach works, at least qualitatively. Further test
will need to be done to confirm the model behavior quantitatively. In addition, the fully
dynamic models will need to be used.
Figure 7.11 shows the snapshots of the mass fraction of (a) liquid, (b) gas and
(c) vapor, at time t = 5.0E − 04 sec. The formation of the jet and the value of the
mass fraction are consistent for each phase/component. A mushroom-shape structure
is observed at the leading part of the jet. The jet spreads radially entraining the
surrounding gas. The entrainment of the surrounding gas is controlled by the speed of
the jet. The entrainment rate of the jet defines the rate of propagation of the interface
between rotational and irrotational fluid [137]. The corresponding contour plot of the
magnitude of vorticity, density, energy, pressure and velocity with vectors are shown in
Figures 7.11 (b-f), respectively. The body of the jet at the leading part becomes thin
111
(a) (b)
(c)
Figure 7.11: Snapshots of multiphase jet at time, t=5.0E-04 sec.(a) mass fraction of
liquid, (b) mass fraction of gas, (c) mass fraction of vapor.
as can be observed in Figures 7.11 (c) and (d). The thinning maybe be caused by the
kinematic requirements, or viscous extrusion [85]. The vorticity, see Figure 7.11 (b), at
the leading part of the jet develops recirculation of flows on both sides of the leading
part of the jet as can be seen in Figures 7.11 (e) and (f). In addition, the magnitude of
pressure and velocity in this region decreases.
The jet expands as it continues to propagate and the structure of the mushroom-
112
(a) (b)
(c) (d)
(e) (f)
liquid, (b) vorticity, (c) density, (d) energy, (e), pressure and (f) velocity vectors.
113
(a) (b)
(c) (d)
(e) (f)
114
shape at the leading part of the jet starts to deform. Figure 7.13 plots the contour of the
mass fraction of liquid (a), magnitude of vorticity (b), density (c), energy (d), pressure
(e) and velocity with vectors (d) at time t = 7.7E − 04 sec. The body of the jet at
the leading part remains thin, see Figure 7.13 (a). The surface area of the jet reduces
due to axisymmetric deformation. The decrease of magnitude of density and energy is
observed in Figures 7.13 (c) and (d). The axial momentum of the jet decreases along
the jet axis. The magnitude of pressure, see Figure 7.13 (d), at the leading part of the
jet decreases also due to recirculation of flow as can be seen in Figure 7.13 (e).
In Figure 7.14, the contour of the mass fraction of liquid (a), magnitude of vor-
ticity (b), density (c), energy (d), pressure (e) and velocity with vectors (d) at time
t = 1.0E − 03 sec. is plotted. The leading part of the jet diffuses and expands in axial
direction. The mushroom-shape at the leading part is completely deformed and disin-
tegrated as can be observed in Figures 7.14 (a), (c) and (d). Flow recirculation on the
side of the leading part of the jet is still noticeable as shown in Figure 7.14 (f), which
corresponds to a decrease in pressure, see Figure 7.14 (e).
The imbalance between stabilizing and destabilizing forces in fluid motion results
to the instability of the jet [24]. In constant density, homogeneous, single phase flows,
such forces refers to viscous and inertial forces, when neglecting gravity and capillary
effects. The instability mechanisms corresponds to the emergence of various types of
vortex structures. For example, in thin shear layer flows, the basic instability processes
that takes place is the so-called Kelvin-Helmholtz instability, associated with an inflex-
ional velocity profile. On the other hand, in variable density fluid motions, the situation
is different because of the new mechanisms of vorticity generation by density gradients
in presence of pressure gradients.
Figures 7.15 and 7.16 show the instability of the jet at time t = 1.2E − 03
sec. and t = 1.4E − 03 sec., respectively, with the corresponding plot of (a) mass
fraction of liquid, (b) magnitude of vorticity, (c) density, (d) energy, (e) pressure and
115
(a) (b)
(c) (d)
(e) (f)
116
(a) (b)
(c) (d)
(e) (f)
117
(a) (b)
(c) (d)
(e) (f)
118
(f) velocity with vectors. The instability of the jet is characterized by the unstable
varicose mode [30, 34, 85] as can be seen in Figures 7.15 (a) and 7.16 (a). Rayleigh [106]
have studied the hydrodynamic stability of a free liquid jet in air and have observed the
symmetric instability of the jet, which he described as varicose. In addition, the pressure
of the surrounding gas and the shear fluctuations provides energy to the disturbance in
varicose mode instability. The unstable disturbances propagates in both the upstream
and downstream direction accompanied by a pressure fluctuation. Furthermore, the
deformation of the jet depends on several factors including: (1) the non-uniformity of
the velocity of the jet, (2) the interaction of the jet with the surrounding gas, and
(3) the viscosity of the jet. Also, turbulence controls the transfer of axial and radial
momentum, which can modify the shape of the jet.
The transition from instability to turbulence depends on the boundary and initial
conditions applied to the flow field. A reflection of wave from the outer boundary was
observed (results not shown) after continuing the jet to propagate the outer boundary.
The reflection is caused by the use of one-dimensional characteristic outflow boundary
conditions, which are known to cause strong reflection waves in highly nonuniform region
of the flow such as wake. Furthermore, the flow has high density variations and subsonic
flow condition with very low speed of sound of the mixture. Note that the jet was a
mixture of liquid, gas and vapor with a corresponding mixture speed of sound to about
25 m/s. The reflection was not experienced before in Chapter 6 because the simulations
were performed at supersonic condition.

Chapter 8
Conclusions and Future Research
This chapter concludes the thesis with a summary of the computational capa-
bilities of the new developed analytical and computational approach for modeling and
numerical simulation of multiphase and multicomponent flows. Possible extensions of
the work are also outlined in this chapter.
8.1 Concluding Remarks
A general approach for modeling and numerical simulation of multiphase and mul-
ticomponent, compressible and incompresible flows, which can be applied to a variety
of flow configurations including shock waves, cavitation, single and multi-phase turbu-
lent flows has been successfully developed. The approach is flexible and can have an
arbitrary number of components in each phase using either Euler or Navier-Stokes for-
mulation. The modeling and computational approach developed in this thesis is flexible
enough to incorporate a variety of different cavitation and/or turbulence models. The
new developed approach for multiphase and multicomponent flows does not require an
ad hoc closure for the variation of mixture density with regards to the attendant pres-
sure and yields a thermodynamically accurate value for the mixture speed of sound.
A good agreement of the value of mixture speed of sound was observed between the
derived speed of sound based on the proposed mixture model (2.50) and the experi-
mental data of Karplus [72]. The predicted mixture speed of sound was also compared
120
to Refs. [3, 4, 35, 74, 143] and in good agreement. The accuracy of the value of the
speed of sound of the mixture is very important in modeling and numerical simulation
of multiphase and multicomponent flows.
The analytical solution of the governing equations of the mixture is very useful for
validation and verification of the accuracy of the new developed approach, especially if
no available experimental data. However, the governing equations of the multiphase and
multicomponent mixture are not possible to solved analytically for arbitrary mixtures.
The equation of state for each component must be carefully chosen, if an exact solution
is to be found. To obtain an exact solution of the mixture equations, a novel “idealized”
fluid mixture model was developed, which allows the derivation of an exact solution for
the multiphase and multicomponent Riemann problem in one dimension. A number
of existing benchmark problems for single phase and multicomponent flows become a
subset of this new model.
The Roe and HLLC approximate Riemann solver, originally developed for single-
phase single-component fluids, was successfully extended to multiphase and multicom-
ponent fluids and used to capture shock waves and contact discontinuities. The accuracy
of the proposed numerical method and physical model were verified and validated by
solving a number of test problems. Three classical benchmark problems (single-phase
two-component shock tube, shock-wave propagation in a single-phase two-component
fluid, and single-phase shock-bubble interaction) and two novel benchmark problems
for the idealized fluid-mixture model (two-phase shock-tube and two-phase rarefaction
problems) were presented. For all problems allowing an exact solutions, good agreement
between numerical and exact solutions was observed. For the case of single-phase shock-
bubble interaction problem, the numerical results were compared with the experiments
by [53] and the simulations by [105], and exhibited good agreement.
The new developed analytical and computational approach for modeling and sim-
ulation of multiphase and multicomponent flows was successfully applied to simulate

121
high pressure and supersonic multiphase and multicomponent free surface jet flows.
Shock wave generation in a gaseous medium, which was observed by MacPhee et al.
[89] using a synchrotron x-radiography and a fast x-ray detector in their experiments,
was successfully captured in the simulation. The results of the simulation compliment
the experiments and provide useful information for better understanding of the complex
structure of the jet close to the nozzle region.
The flexibility of the new developed general approach to incorporate a variety of
cavitation model was demonstrated by integrating the cavitation model proposed by
Hosangadi et al. [64] into the general framework. The approach with the incorporated
cavitation model was tested to perform simulation of multiphase and multicomponent
cavitating nozzle jet flow through a gaseous medium. The results of the numerical
simulation captures the cavitation process in the nozzle orifice and shows the influence
of cavitation to the external jet formation. It was observed that the time scale of the
internal cavitation dynamics is in the order of 10 µs, which confirms the assertion of
Yuan et al. [155]. It was found out also that the discharge of the nozzle is strongly
dependent on the cavitation process and the magnitude of the bubble cavities. When
the inflow condition is fluctuated, it was noticed that the cavitation separates the flow
inside the nozzle injector and intensifies the unsteadiness of the external jet formation.
Furthermore, It was believed that downstream the nozzle and the reattachment region,
turbulence affect the breakup and coalescence of the collapsing bubble cavities.
The new developed general approach was successfully extended for modeling and
numerical simulation of turbulent multiphase and multicomponent flows. The flexibility
of the approach to incorporate a variety of turbulence model was demonstrated by
incorporating the extended large eddy simulation (LES) methodology for multiphase
and multicomponent flows into the general framework. The capability of the extended
LES for multiphase and multicomponent flows was examined by conducting a numerical
simulation of turbulent plane jet in both single phase and multiphase/multicomponent

122
flow regime. The results of the numerical simulation provide a very useful information
in understanding the behavior of a turbulent plane jet. The computational results
confirm physically observed phenomena including the generation of vortex rollers and
development of jet instability. It was also found out that the deformation of the jet
depends on several factors including: (1) the non-uniformity of the velocity of the jet,
(2) the interaction of the jet with the surrounding gas, and (3) the viscosity of the jet.
Computational Capabilities
The following summarizes the capabilities of the new developed analytical and
computational approach for modeling and numerical simulation of multiphase and mul-
ticomponent flows;
1. Numerical Capabilities:
(a) Finite Volume Method,
(b) Multiphase and Multicomponent Mixture Formulation,
(c) Euler and Navier-Stokes Formulation,
(d) Coupled Incompressible-Compressible Solver,
(e) Unstructured Mesh Formulation,
(f) Serial and Parallel Code,
2. Physical Capabilities:
(a) Steady and Transient Flows,
(b) Laminar and Turbulent Flows,
(c) Steady and Unsteady Cavitation,
(d) Shock Wave Generation,
(e) Subsonic and Supersonic Flows,
(f) Transport of Species (Passive and Active Scalar).

123
8.2 Future Research
The research presented in this thesis could be extended in three main directions:
further development of numerical methodology, model improvements, and technological
applications.
Algorithm development. Current explicit formulation of the method put a
great constraint on time step. Implicit implementation would remove this limitation
and make it more robust and computationally efficient. Another area of improvement
is the use of adaptive mesh refinement strategies. This is especially important in the
regions with large density contrast or contact discontinuity. Having an ability to have
fine mesh along the interphase boundary will drastically decrease the error and improve
the quality of the results. A good candidate for the extension of the algorithm is
an adaptive wavelet collocation method [144, 145, 146]. In addition, the present one-
dimensional characteristic outflow boundary conditions need to be improved by making
use of non-linear multi-dimensional characteristic boundary conditions and/or applying
the sponge layer technique.
Model Development. There are many types of cavitation and turbulence mod-
els available. Each model has its own strengths and weaknesses in simulation of a partic-
ular flow configuration. Thus, it could be possible to incorporate a variety of cavitation
and/or turbulence model into the new developed general approach for modeling and
simulation of multiphase and multicomponent flows. Consequently, perform numerical
simulations of different flow configurations and conduct a qualitative and quantitative
analysis for different types of cavitation and/or turbulence models. With regards to
the turbulence model, the current subgrid-scale model (SGS) implemented in the LES
methodology is of eddy-viscosity type. A dynamic subgrid-scale model of both eddy
and non-eddy viscosity type for multiphase and multicomponent flow needs to be added
in the approach in order to avoid ad hoc treatment of the model coefficients.

124
Applications. The approach needs to be applied to problems of engineering
interest to demonstrate technological applications of the new developed analytical and
computational approach for modeling and numerical simulation of multiphase and mul-
ticomponent flows. Examples of engineering applications for which the new developed
general approach for numerical modeling and simulation of multiphase and multicom-
ponent flows could be advantageously applied are in diesel and gasoline fuel injection
systems, liquid jet-machining, thermal and plasma spray coating, and gas and oil trans-
port. In addition, the approach could also be utilized for biomedical applications such
as blood flow, drug/gene delivery, and shock wave lithotripsy.

Bibliography
[1] R. Abgrall. Generalization of the Roe scheme for the computation of mixture of
perfect gases. Rech. Aerospat., 6:31–43, 1988.
[2] R. Abgrall and S. Karni. Computations of compressible multifluids.
J. Comput. Phys., 169:594–623, 2001.
[3] V. Ahuja and A. Hosangadi. Unstructured simulation of multi-phase mixtures
with compressibility effects. AIAA-2001-2671, pages 1–15, 2001.
[4] V. Ahuja, A. Hosangadi, R. Ungewitter, and S.M. Dash. A hybrid unstructured
mesh solver for multi-fluid mixtures. AIAA-99-3330, pages 1–9, 1999.
[5] A. Alajbegovic, G. Meister, D. Greif, and B. Basara. Three phase cavitating flows
in high pressure swirl injectors. In Fourth International Conference on Multiphase
Flow. ICMF, 2001.
[6] W. Altai and V.H. Chu. Les of the turbulent plane jet. Fourteenth Engineering
Mechanics Conference, 2000.
[7] M. Arai, M. Shimizu, and H. Hiroyasu. Breakup length and spray angle of high
speed jet. Proc. of the 3rdInternational Conference on Liquid Atomization and
Spray Systems, 1985.
[8] C. Arcoumanis, M. Badami, H. Flora, and M. Gavaises. Cavitation in real size
multi-hole diesel injector nozzles. SAE Paper 2000-01-1249, 2000.
[9] C. Arcoumanis, H. Flora, M. Gavaises, N. Kampanis, and R. Horrocks. Investiga-
tion of cavitation in a vertical multi-hole diesel injector. SAE Paper 1999-01-0524,
1999.
[10] C. Arcoumanis, J.M. Nouri, and R.J. Andrews. Application of refractive index
matching to a diesel nozzle internal flow. Proceedings of the ImechE Seminar on
Diesel Fuel Injection System, April 14-15, 1992.
[11] Vreman B., B. Geurts, and H. Kuerten. Subgrid-modeling in les of compressible
flow. Applied Scientic Research, 54:191–203, 1995.
[12] C. Badock, R. Wirth, and C. Tropea. The influence of hydro-grinding on cavitation
inside a diesel injection nozzle and primary break-up under unsteady pressure
conditions. Proceedings of the 15th ILASS-Europe 99, pages 5–7, Toulouse, July
5-7, 1999.
126
[13] M. Baer and J. Nunziato. A two-phase mixture theory for the deflagration to deto-
nation transition in reactive granular materials. Int. J. Multiphase Flow , 12:861–
889, 1986.
[14] F.J. Barclay, T.J. Ledwedge, and G.C. Cornfiled. Some experiments on sonic
velocity in two-phase one-component mixtures and some thoughts on the nature
of two-phase critical flow. Proc. Inst. Mech. Eng., 184(3C):185, 1969.
[15] P. Batten, N. Clarke, C. Lambert, and D. M. Causon. On the choice of wavespeeds

for the HLLC Riemann solver. SIAM Rev., 18, 6:1553–1570, 1997.
[16] W. Bergwerk. Flow pattern in diesel nozzle spray holes. Proc. Inst. Mech. Eng.,
173:655–660, 1959.
[17] S. Biswas, A. Esmaeeli, and G. Tryggvason. Comparison of results from

DNS of bubbly flows with a two-fluid model for two-dimensional laminar flows.
Int. J. Multiphase Flow , 31:1036–1048, 2005.
[18] B.J. Boesrma and S.K Lele. Large eddy simulation of a mach 0.9 turbulent jet.
A99-27941, 1999.
[19] A.C. Buckingham. Large eddy simulation of shockwave passage through turbu-
lence. Engineering Applications of Large Eddy Simulations, 162:73–80, 1993.
[20] R.A. Bunnell and S.D. Heister. Three-dimensional unsteady simulation of cav-
itating flows in injector passages. Journal of Fluids Engineering, 122:791–797,
2000.
[21] L.C. Burrill. Sir charles parsons and cavitation. 1950 Parsons Memorial Lecture,
63,no.8:149–167, 1951.
[22] C. E. Castro and E. F. Toro. A Riemann solver and upwind methods for a
two-phase flow model in non-conservative form. Int. J. Numer. Meth. Fluids, 60,
3:275–307, 2006.
[23] B.W. Chandra and S.H. Collicott. Experimental investigation of cavitation fre-
quency in a slot orifice. ILASS-99, pages 379–384, Indianapolis, IN, 1999.
[24] P. Chassaing, R.A. Antonia, F. Anselmet, L. Joly, and S. Sarkar. Variable Density
Fluid Turbulence. Kluwer Academic Publisher, 2002.
[25] H. Chavez, M. Knapp, and A. Kubitzek. Experimental study of cavitation in the

nozzle hole of diesel injectors using transparent nozzles. SAE Paper No. 950290,
1995.
[26] N. Chihier. Prog. Energy Combust. Sci, 17:211, 1991.
[27] S. Chumakov and C.J. Rutland. Dynamic structure models for scalar flux and
dissipation in large eddy simulation. AIAA Journal, 42,6:1132–1139, 2004.
[28] S.G. Chumakov and C.J. Rutland. Dynamic structure subgrid-scale model for
large eddy simulation. Int. J. Numer. Meth. Fluids, 47:911–923, 2005.
127
[29] R. Courant and K. O. Friedrichs. Supersonic flow and shock waves. Interscience,
New York, 1948.
[30] W.O. Criminale, T.L. Jackson, and R.D. Joslin. Theory and Computation in
Hydrodynamic Stability. Cambridge University Press, 2003.
[31] J.R. Debonis. A large-eddy simulation of a high reynolds number mach 0.9 jet.
AIAA 2004-3025, 2004.
[32] C.F. Delale, K. Okita, and Y. Matsumoto. Steady-state cavitating nozzle flows
with nucleation. Fifth International Symposium on Cavitation, November 1-4,
Osaka, Japan, 2003.
[33] von M. Dirke, A. Krautter, J. Ostertag, M. Mennicken, and C. Badock. Simulation

of cavitating flows in diesel injectors. Oil and Gas Science and Technology, Vol.
54, No. 2:223–226, 1999.
[34] P.G. Drazin and W.H Reid. Hydrodynamic Stability. Cambridge University Press,
1981.
[35] D.A. Drew and S.L. Passman. Theory of multicomponent fluids. 1998.
[36] J. Du, B. Fix, J. Glimm, X. Jia, X. Li, Y. Li, and L. Wu. A simple package for
front tracking. J. Comput. Phys., 213:613–628, 2006.
[37] N. Dumont, O. Simonin, and C. Habchi. Numerical simulation of cavitat-

ing flows in diesel injectors by a homogeneous equilibrium modeling approach.
CAV2001:sessionB6.005, 2001.
[38] N.M. El-Hady, T.A. Zang, and U. Piomelli. Dynamic subgrid-scale modeling for
high-speed transitional boundary layers. Engineering Applications of Large Eddy
Simulations, 162:103–111, 1993.
[39] A.T. Ellis. Some effects on macromolecules on cavitation inception and noise.
Calif. Inst. of Tech. Div. of Engrg. and Appl. Sci. Rep., June, 1967.
[40] G. Erlebacher, M.Y. Hussaini, C.G. Speziale, and T.A. Zang. Toward the large-
eddy simulation of compressible turbulent flows. J. Fluid. Mech., 238:155–185,
1992.
[41] X. Escaler, E. Egusquiza, M. Farhat, Avellan F., and M. Coussirat. Detection

of cavitation in hydraulic turbines. Mechanical Systems and Signal Processing,
2004.
[42] A. Esmaeeli and G. Tryggvason. A front tracking method for computations of

boiling in complex geometries. Int. J. Multiphase Flow , 30:1037–1050, 2004.
[43] A. Favre. Equations des gaz turbulents compressible. i. formes generales. J. Mec.,
4:361–390, 1965.
[44] A. Favre. Equations des gaz turbulents compressible. ii. methode des vitesses
moyennes; methode des vitesses macroscopiques ponderees par la masse volu-
mique. J. Mec., 4:391–421, 1965.
128
[45] L. Fraccarollo and E. F. Toro. Experimental and numerical assessment of the

shallow water model for two dimensional dam-break type problems. Journal of
Hydraulic Research, 33, 1995.
[46] L. Fulachier, R. Borghi, F. Anselmet, and P. Parenthoen. Influence of density

variations on the structure of low-speed turbulent flows: A report on euromech.
J. Fluid Mech., 237,203:577–593, 1989.
[47] L. Fulachier, J.L. Lumley, and F. Anselmet. Iutam symposium on variable density
low-speed turbulent flows. Number 41 in fluid mechanics and its applications,
1997.
[48] L.C. Ganippa, G. Bark, S. Anderson, and J. Chomiak. Comparison of cavitation

phenomenon in transparent scaled-up single-hole diesel nozzles. Proceedings of
the 4th International Symposium on Cavitation, June 20-23, 2001.
[49] M. Germano, U. Piomelli, P. Moin, and W.H. Cabot. A dynamic subgrid-scale

eddy viscosity model. Phys. Fluid A, 3,7:1760–1765, 1991.
[50] J. Glimm, M. J. Graham, J. Grove, X. L. Li, T. M. Smith, D. Tan, F. Tangerman,

and Q. Zhang. Front tracking in two and three dimensions. Comput. Math. App.,
35:1–11, 1998.
[51] D. E. Goldstein and Vasilyev O. V. Stochastic coherent adaptive large eddy

simulation method. Phys. Fluid, 16,7:2497–2513, 2004.
[52] F.F Grinstein, M. Glauser, and W.K. George. Vorticity in jets, chapter iii of fluid
vortices, ed. s. green. pages 65–94, 1995.
[53] J. F. Haas and B. Sturtevant. Interaction of weak shock waves with cylindrical
and spherical gas inhomogeneities. J. Fluid Mech., 181:41–76, 1987.
[54] T. J. Hanratty, T. Theofanous, J.-M. Delhaye, J. Eaton, J. McLaughlin, A. Pros-

peretti, S. Sundaresan, and G. Tryggvason. Workshop findings. International
Journal of Multiphase Flow, 29:1047–1059, 2003.
[55] A. Harten, P. Lax, and B. Van Leer. On upstream differencing and Godunov-type
schemes for hyperbolic conservation law. SIAM Review , 25:35–61, 1983.
[56] A. Harten, P.D. Lax, and B. Van Leer. On upstream differencing and godunov-
type schemes for hyperbolic conservation laws. SIAM Rev., 25:35–61, 1983.
[57] A. Haselbacher. The RocfluMP Book. Center for Simulation of Advanced Rockets,
University of Illinois at Urbana-Champaign, 2003-2006.
[58] A. Haselbacher. A WENO reconstruction method for unstructured grids based

on explicit stencil construction. AIAA-2005-0879 , 2005.
[59] A. Haselbacher. On constrained reconstruction operators. AIAA-2006-1274 , 2006.
[60] L. He and F. Ruiz. Effect of cavitation on flow and turbulence in plain orifices for
high-speed atomization. Atom Sprays, 5:569–84, 1995.
129
[61] L. F. Henderson, P. Colella, and E. G. Puckett. On the refraction of shock waves

at a slow-fast gas interface. J. Fluid Mech., 224:1–27, 1991.
[62] W. Hentschel and K.P. Schindler. Opt. Laser Eng., 25:401, 1996.
[63] H. Hiroyasu, M. Arai, and M. Shimizu. Break-up length of a liquid spray and
internal flow in a nozzle. ICLASS-91, Gaithersburg, Maryland , July, 1991.
[64] A. Hosangadi, V. Ahuja, and S. Arunajatesan. A generalized compressible cavi-

tation model. CAV2001:sessionB4.0043, 2001.
[65] A. Hosangadi, N. Sinha, and S.M. Dash. A unified hyperbolic interface capturing
scheme for gas-liquid flows. AIAA-97-2081, pages 1227–1237, 1997.
[66] Z.W. Hu, C.L. Morfey, and N.D. Sandham. Sound radiation from a subsonic
turbulent plane jet. AIAA 2002-2421, 2002.
[67] Z.W. Hu, C.L. Morfey, and N.D. Sandham. Large eddy simulation of plane jet
sound radiation. AIAA 2003-3166, 2003.
[68] S.A. Jordan and S.A Ragab. On the unsteady and turbulent characteristics of
the three-dimensional shear-driven cavity flow. Engineering Applications of Large
Eddy Simulations, 162:127–146, 1993.
[69] T. Karasawa, M. Tanaka, K. Abe, and T. Kurabayashi. Effect of nozzle configu-

ration on the atomization of a steady spray. Atom. Sprays, 2:411–426, 1992.
[70] S. Karni. Multicomponent flow calculations by a consistent primitive algorithm.

J. Comput. Phys., 112:31–43, 1994.
[71] S. Karni. Hybrid multifluid algorithms. SIAM J. Sci. Comput., 17, 5:1019–1039,
1996.
[72] H. B. Karplus. The velocity of sound in a liquid containing gas bubbles.

Res. Develop. Rep. C00-248 , 1958.
[73] K. Kato, Y. Matudaira, and H Obara. Flow visualization of cavitation with

particle and bubble image processing. Proceedings of ASME FEDSM, July 6-10,
2003.
[74] S. W. Kieffer. Sound speed in liquid-gas mixtures: water-air and water-steam.

J. Geophys. Research, 82, 20:2895–2904, 1977.
[75] R.T. Knapp, J.W. Daily, and F.G Hammitt. Cavitation. McGraw-Hill Book
Company, 1970, 1970.
[76] D. Knight, G. Zhou, N. Okong’o, and V. Shukla. Compressible large eddy simu-
lation using unstructured grids. AIAA Paper 98-0535, 1998.
[77] A.L. Knox-Kelecy and P.C. Farrell. Spectral characteristics of turbulent flow in a
scale model of a diesel fuel injector nozzle. SAE Paper 972942, 1997.
[78] C.B. Laney. Computational gasdynamics. Cambridge University Press, 1998.

130
[79] B. Larrouturou. How to preserve the mass fraction positive when computing
compressible multi-component flows. J. Comput. Phys., 95:59–84, 1991.
[80] B. Larrouturou and L. Fezoui. On the equations of multicomponent perfect or

real gas inviscid flow. Lect. Notes in Math., 1402:69–97, 1989.
[81] C. Le Ribault, S. Sarkar, and S.A. Stanley. Large eddy simulation of a plane jet.
Physics of Fluids, 11,10:3069–3083, 1999.
[82] C. Le Ribault, S. Sarkar, and S.A. Stanley. Large eddy simulation of a plane jet.
AIAA Journal, 39,8:1509–1515, 2001.
[83] Li, Haiyun, and S.H. Collicott. Initial images of cavitation inside high-pressure
atomizers. ILASS-99, pages 409–413, Indianapolis, IN, 1999.
[84] A. Lichtarowicz, R.K. Duggins, and E. Markland. Dsicharge coefficients for in-
compressible non-cavitating flow through long orifices. J. Mech. Eng. Sci., Vol. 7,
No. 2:210–219, 1965.
[85] S.P. Lin. Breakup of Liquid Sheets and Jets. Cambridge University Press, 2003.
[86] Y. Liu and P.G. Tucker. Linear and nonlinear model large-eddy simulations of a
plane jet. 44th AIAA Aerospace Sciences Meeting and Exhibit, 2006.
[87] J. Lu, S. Biswas, and G. Tryggvason. A DNS study of laminar bubbly flows in a
vertical channel. Int. J. Multiphase Flow , 32:643–660, 2006.
[88] J. Lyness. Notes on the adaptive Simpson quadrature routine. J. of the ACM ,
16, 3:483–495, 1969.
[89] A.G. MacPhee, M.W. Tate, C.F. Powell, Y. Yue, M.J. Renzi, A. Ercan,
S. Narayanan, E. Fontes, J. Walther, J. Schaller, S.M. Gruner, and J. Wang.
X-ray imaging of shock waves generated by high pressure fuel sprays. SCIENCE,
295:1261–1263, 2002.
[90] A. Mallock. The damping of sound by frothy fluids. Proc. Roy. Soc. London Ser.
A, 84:391, 1910.
[91] L. di. Mare, M. Klein, W.P. Jones, and J. Janicka. Synthetic turbulence inflow
conditions for large-eddy simulation. Physics of Fluids, 10, 025107, 2006.
[92] M.P. Martin, U. Piomelli, and G.V. Candler. Subgrid-scale models for com-
pressible large-eddy simulations. Theoretical and Computational Fluid Dynamics,
13:361–376, 2000.
[93] D. McWilliam and R.K. Duggins. Speed of sound in bubbly liquids. Proc. Inst.
Mech. Eng., 184(3C):102, 1969.
[94] C.L. Merkle, J.Z. Feng, and P.E.O. Buelow. Computational modeling of the
dynamics of sheet cavitation. Grenoble, 1998.
131
[95] P. Moin, K. Squires, W. Cabot, and S. Lee. A dynamic subgrid-scale model for
compressible turbulence and scalar transport. Phys. Fluid A, 3,11:2746–2757,
1991.
[96] J. Murphy, D. Schmidt, S.P. Wang, and M.L. Corradini. Multi-dimensional mod-
elling of multi-phase flow physics: high-speed nozzle and jet flows - a case study.
Nuclear and Engineering, pages 177–190, 2001.
[97] T. Nakahira, M. Komori, K. Nishida, and K. Tsujimura. Shock W aves, K.

Takayama, Ed., 2:1271–1276, 1992.
[98] V.E. Nakoryakov, B.G. Pokusaev, and I.R Shreiber. Wave propagation in gas
liquid media. CRC Press, 1993.
[99] W. H. Nurick. Orifice cavitation and its effect on spray mixing. ASME J. Fluids
Eng., pages 681–687, 1976.
[100] F. Payri, R. Bermudez, R. Payri, and F.J. Salvador. The influence of cavitation
on the internal flow and the spray characteristics in diesel injection nozzles. Fuel,
83:419–431, 2004.
[101] C. Pierce and P. Moin. A dynamic model for subgrid-scale variance and dissipation
rate of a conserved scalar. Phys. of Fluids, 12:3041–3044, 1998.
[102] U. Piomelli and T.A. Zang. Large eddy simulation of transitional channel flow.
Comput. Phys. Commun., 65:224–230, 1991.
[103] H. Power and C.A. Brebbia. Computational Methods in Multiphase Flow. WIT-
PRESS Southampton, Boston, 2001.
[104] J.R. Qin, S.T. Yu, and M.C. Lai. Direct calculations of cavitating flows by
the method of space-time conservation element and solution element. SAE
1999-01-3554, October 25-28, Toronto, Ontario, Canada, 1999.
[105] J. Quirk and S. Karni. On the dynamics of a shock-bubble interaction.

J. Fluid Mech., 318:129–163, 1996.
[106] L. Rayleigh. On the instability of jets. Proc. London Math. Soc., 10:4–13, 1879.
[107] R.S. Reichert and S. Biringen. Numerical simulation of compressible plane jets.
AIAA, 1997.
[108] R.C. Reid, J.M. Prausnitz, and B.E. Poling. The Properties of Gases and Liquids.
McGraw-Hill, Inc, New York, 1987.
[109] R.D. Reitz. Modeling atomization processes in high-pressure vaporizing sprays.

Atomization and Spray Technology, 3:309–337, 1987.
[110] C.L. Ribault, S. Sarkar, and S.A. Stanley. Large eddy simulation of evolution of
a passive scalar in plane jet. AIAA Journal, 39,8:1509–1516, 2001.
[111] B.E. Richards, D.H. Trevena, and D.H. Edwards. Cavitation experiments using a
water shock tube. J. Phys. D: Appl. Phys., 13:1315–1323, 1980.
132
[112] P. L. Roe. Approximate Riemann solvers, parameter vectors and difference

schemes. J. Comput. Phys., 43:357–372, 1981.
[113] R.S. Rogallo and P. Moin. Numerical simulation of turbulent flows. Annu. Rev.
Fluid Mech., 16:99–137, 1984.
[114] P. Roosen, O. Unruh, and M. Behmann. Untersuchung und modeleirung des tran-
sienten verhaltens von kavitationserscheinungen bei ein- und mehrkomponentigen
kraftstoffen in schnell durchstromten dusen. Report of the Institute for Technical
Termodynamics, aachen, Germany, 1996.
[115] F. Ruiz and L. He. Turbulence under quasi cavitating conditions: a new species.
Atom Sprays, 9:419–29, 1999.
[116] L. Sainaulieu. Finite volume approximation of two-phase fluid flows based on an

approximate roe-type riemann solver. J. Comput. Phys., 121:1–28, 1995.
[117] Sanchez and Collicott. Statistics of cavitation length in a 2-d slot orifice. ILASS-99,
pages 403–408, Indianapolis, IN, 1999.
[118] K. Sankaranarayanan, I. G. Kevrekidis, S. Sundaresan, J. Lu, and G. Tryggvason.

A comparative study of Lattice Boltzmann and front-tracking finite-difference
methods for bubble simulations. Int. J. Multiphase Flow , 29:109–116, 2003.
[119] K. Sato, S. Shimojo, and J. Watanabe. Observation of chain-reaction behavior

at bubble collapse using ultra high speed video camera. Proceedings of ASME
FEDSM, July 6-10, 2003.
[120] R. Saurel and R. Abgrall. A multiphase Godunov method for compressible mul-
tifluid and multiphase flows. J. Comput. Phys., 150:425–467, 1999.
[121] R. Saurel and R. Abgrall. A simple method for compressible multifluid flows.
SIAM J. Sci. Comput., 21, 3:1115–1145, 1999.
[122] D.P. Schmidt, C.J. Rutland, and M.L. Coarradini. A numerical study of cavitating
flow through various nozzle shapes. Transactions of the SAE, 106(3):1644–1673,
1997.
[123] I. Senocak and W. Shyy. Computations of unsteady cavitation with a pressure-

based method. FEDSM2003-45009, 2003.
[124] H.H. Shi, K. Takayama, and O Onodera. JSME Intl. J. Ser. B, 37:509, 1994.
[125] B.R Shin. Numerical analysis of unsteady cavitating flow by a homogeneous

equilibrium model. Fluid Dynmics Conference, A01-31290, 2001.
[126] A. Singhal, M. Athavale, H. Li, and Y. Jiang. Mathematical basis and validation
of the full cavitation model. Journal of Fluids Engineering, 124:617–624, 2002.
[127] Smagorinsky:1963. General circulation experiments with the primitive equations.

i. the basic experiment. Mon. Weather Rev., 91:99–164, 1963.
133
[128] C. Soteriou, R. Andrews, and M. Smith. Direct injection diesel sprays and the
effect of cavitation and hydraulic flip on atomization. SAE Paper 950080, 1995.
[129] C. Soteriou, M. Smith, and R. Andrews. Diesel injector laser light sheet illumi-
nation of the development of cavitation in orifices. Proc ImechE C529/018/98,
1998.
[130] C.G. Speziale, G. Erlebacher, T.A. Zang, and M.Y. Hussaini. The subgrid-scale
modeling of compressible turbulence. Phys. Fluid, 31,4:940–942, 1988.
[131] K.D. Squires. Large eddy simulation of rotating isotropic turbulence. Engineering
Applications of Large Eddy Simulations, 162:65–71, 1993.
[132] G. D. Stefano, O. V. Vasilyev, and D. E. Goldstein. A-priori dynamic test for deter-
ministic/stochastic modeling in large-eddy simulation of turbulent flow. Computer
Physics Communications, 169:210–213, 2005.
[133] B. H. Stewart and B. Wendroff. Two-phase flow: models and methods.

J. Comput. Phys., 56:363–409, 1984.
[134] N. Tamaki, M. Shimizu, K. Nishida, and H. Hiroyasu. Effects of cavitation and

internal flow on atomization of a liquid jet. Atom. Sprays, 8:179–197, 1998.
[135] S. Tanguy and A. Berlemont. Application of a level set method for simulation of
droplet collisions. Int. J. Multiphase Flow , 31:1015–1035, 2005.
[136] R. Tatschl, B. Wiesler, A. Alajbegovic, and C v. Knsberg Sarre. Advanced 3d

fluid dynamic simulation for diesel engines. THIESEL, September 2000.
[137] H. Tennekes and J.L. Lumley. A first course in turbulence. 1972.
[138] V. T. Ton, A. R. Karagozian, B. E. Engquist, and S. J. Osher. Nu-

merical simulation of inviscid detonation waves with finite rate chemistry.
In Proceedings of Combustion Institue, Fall Meeting, Paper 91-101 , 1991.
[139] E. F. Toro and A. Chakraborty. Development of an approximate Riemann solver

for the steady supersonic euler equations. The Aeronautical Jornal, 98, 1994.
[140] E.F. Toro. Riemann-problem based techniques for computing reactive two-phase
flows. Proc. Third. Intern. Confer. on Numerical Combustion, No. 351 in Lecture
Notes in Physics, Antibes, France:472–481, 1989.
[141] E.F. Toro. Riemann Solvers and Numerical Methods for Fluid Dynamics.
Springer-Verlag, (1999).
[142] E.F. Toro, M. Spruce, and W. Speares. Restoration of the contact surface in the
hll-riemann solver. Shock Waves, 4:25–34, 1994.
[143] I. Toumi, A. Kumbaro, and H. Paillere. Approximate Riemann solvers and flux
vector splitting schemes for two-phase flow. 30th Computational Fluid Dynamics,
von Karman Institute for Fluid Dynamics, Lecture Series, 1999.
134
[144] O.V. Vasilyev and Bowman. Second generation wavelet collocation method for
the solution of partial differential equations. J. Comp. Phys., 165:660–693, 2000.
[145] O.V. Vasilyev and N.K.-R. Kevlahan. Hybrid wavelet collocation - brinkman
penalization method for complex geometry flows. Int. J. Numer. Meth. in Fluids,
40:531–538, 2002.
[146] O.V. Vasilyev and S. Paolucci. A dynamically adaptive multilevel wavelet colloca-
tion method for solving partial differential equations in a finite domain. J. Comp.
Phys., 125:498–512, 1996.
[147] M.S. Vazquez, W. Vicente, A. Espinosa, and E. Barrios. Large eddy simulation
of an ammonia jet. American Journal of Applied Science, 2,8:1270–1273, 2005.
[148] P.R. Voke and S. Gao. Large eddy simulation of heat transfer from an impinging
plane jet. Engineering Applications of Large Eddy Simulations, 162:29–35, 1993.
[149] C. Vortman, G.H. Schnerr, and S. Seelecke. Thermodynamic modeling and sim-
ulation of cavitating nozzle flow. International Journal of Heat and Fluid Flow,
24:774–783, 2003.
[150] S. Wang, M. Anderson, J. Oakley, M. Corradini, and R. Bonazza. A thermody-

namically consistent and fully conservative treatment of contact discontinuities
for compressible multicomponent flows. J. Comput. Phys., 195:528–559, 2004.
[151] F.M. White. Fluid Mechanics, 5th. edition. McGraw-Hill, Inc, New York, 2003.
[152] K.J. Wu, C.C. Su, R.L. Steinberg, D.A. Santavicca, and F.V. Bracco. Measure-
ments of the spray angle of atomizing jets. ASME J. Fluids Eng., 105:406–413,
1983.
[153] A. Yoshizawa. Statistical theory for compressible turbulent shear flows, with the
application to subgrid modeling. Phys. Fluid A, 29:2152–2164, 1986.
[154] W. Yuan and G. Schnerr. Cavitation in injection nozzles: effect of injection

pressure fluctuations. CAV2001:session A7.001, 2001.
[155] W. Yuan and G. Schnerr. Numerical simulation of two-phase flow in injection

nozzles: Interaction of cavitation and external jet formation. Journal of Fluids
Engineering, 125:963–969, 2003.
[156] A.J. Yule and A.P. Watkins. Atomization Sprays, page 1441, 1991.

Modeling and Simulation of Multiphase/Multicomponent Flows

Uploaded by

Copyright:

Available Formats

Modeling and Simulation of Multiphase/Multicomponent Flows

Uploaded by

Document Information

Original Title

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

Modeling and Simulation of Multiphase/Multicomponent Flows

Uploaded by

Copyright:

Available Formats

Modeling and Simulation of

B.S., MSU-IIT, Philippines, 1997

M.S., The University of New South Wales, Australia, 2001

A thesis submitted to the

Faculty of the Graduate School of the

University of Colorado in partial fulfillment

of the requirements for the degree of

Department of Mechanical Engineering

Prof. Oleg V. Vasilyev

Prof. David Kassoy

Prof. John Daily

Prof. Sedat Biringen

Dr. Jin Wang

Randy S. Lagumbay, (Ph.D., Mechanical Engineering)

Modeling and Simulation of Multiphase/Multicomponent Flows

Thesis directed by Prof. Oleg V. Vasilyev

Most of the existing approaches in dealing with multiphase/multicomponent flows

are limited to either single-phase multicomponent or multiphase single-component mix-

an arbitrary number of components in each phase is developed. The proposed model is

multiphase/multicomponent flows is applicable to a variety of flow configurations includ-

for the multiphase/multicomponent Riemann problem in one dimension is derived. A

number of existing benchmark problems for single-phase multicomponent flows becomes

benchmark problems (single-phase two-component shock tube, shock wave propagation in

a single-phase two-component fluid, and single-phase shock-bubble interaction) and two

novel benchmark problems for an “idealized” fluid-mixture model (two-phase shock-tube

new approach is demonstrated by extending the methodology to cavitating and turbulent

flows. Representative cavitation and multiphase/multicomponent subgrid scale models

and numerical simulations demonstrating these capabilities are performed.

To my father, mother, brother and sister.

giving valuable comments and suggestions. Thanks to my committee members, Profes-

acknowledge. I would like to extend my appreciation to my friends and colleagues in

MU-Columbia and CU-Boulder, Dr. Dan Goldstein, Tanarit Sakulyanontvittaya, Dr.

1.2 Experimental Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.4 Objectives of the Research . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.6 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.1 Homogenized-Mixture Model . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.1 Transport Variables Model . . . . . . . . . . . . . . . . . . . . . 19

2.3 Mixture Equation of State . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.1 Mixture Pressure using Dalton’s Law . . . . . . . . . . . . . . . . 21

2.3.2 Mixture Pressure using Amagat-Leduc Law . . . . . . . . . . . . 22

2.4 Mixture Speed of Sound . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Idealized Fluid-Mixture Model 30

3.1 Density and Equation of State for Idealized Fluid-Mixture . . . . . . . 30

3.2 Entropy for Idealized Fluid-Mixture . . . . . . . . . . . . . . . . . . . . 32

3.3 Speed of Sound for Idealized Fluid-Mixture . . . . . . . . . . . . . . . . 34

3.4 Riemann Invariants for Idealized Fluid-Mixture . . . . . . . . . . . . . . 34

3.5 Riemann Problem for Idealized Fluid-Mixture . . . . . . . . . . . . . . . 35

3.5.1 Left Shock Wave Analysis . . . . . . . . . . . . . . . . . . . . . . 37

3.5.2 Right Shock Wave Analysis . . . . . . . . . . . . . . . . . . . . . 38

3.5.3 Left Rarefaction Wave Analysis . . . . . . . . . . . . . . . . . . . 39

3.5.4 Right Rarefaction Wave Analysis . . . . . . . . . . . . . . . . . . 40

3.5.5 Complete Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4 Numerical Method for Simulation of Multiphase/Multicomponent Flows 42

4.1 Extended Roe’s Approximate Riemann Solver . . . . . . . . . . . . . . . 42

4.2 Extended HLLC Approximate Riemann Solver . . . . . . . . . . . . . . 49

5 Code Verification and Model Validation 54