computational fluid dynamic modelling of zinc slag fuming ... · computational fluid dynamic...
TRANSCRIPT
Computational Fluid Dynamic Modelling of
Zinc Slag Fuming Process
A Thesis Presented for the Degree of Doctor of Philosophy
By
Md. Nazmul Huda
Faculty of Engineering and Industrial Sciences
Swinburne University of Technology
Melbourne, Australia
2012
i
To My To My To My To My Beloved ParentsBeloved ParentsBeloved ParentsBeloved Parents
ii
Declaration
The candidate hereby declares that the work in this thesis, presented for the degree of
Doctor of Philosophy submitted to the Faculty of Engineering and Industrial Sciences,
Swinburne University of Technology; is that of the candidate alone and has not been
submitted previously, in whole or in part, in respect of any other academic award and
has not been published in any form by any other person except where due reference is
given, and has been carried out during the period from January 2008 to November
2011 under the supervision of A/Prof. Jamal Naser and Prof. Geoffrey Brooks from
Swinburne University of Technology, Melbourne, Australia and Prof. Markus Reuter
and Mr. Robert Matusewicz from Outotec Pty. Ltd., Melbourne, Australia.
---------------------------------------
Md. Nazmul Huda
Certification
This is to certify that the above statement made by the candidate is correct to the best
of our knowledge.
A/Prof. Jamal Naser
Prof. Geoffrey Brooks
iii
Abstract
Slag fuming is a reductive treatment process for molten zinciferous slags for
extracting zinc in the form of metal vapour by injecting or adding a reductant source
such as pulverized coal or lump coal and natural gas. The process has been operative
since 1930’s for recovering zinc from lead blast furnace slag. Though slag fuming is a
well-established process and has been industrially operative for over eighty years,
there is only limited understanding of the process kinetics and fluid flow behaviour
inside the slag fuming furnace. The purpose of this study is detailed fluid dynamic
analysis including combustion behaviour, gas-liquid momentum interaction,
generation of splashing due to gas injection process in slag fuming furnace, analysis of
fuming behaviour at different locations of the furnace by using computational fluid
dynamic (CFD) modelling technique.
This PhD thesis focuses on the Computational Fluid Dynamic (CFD) modelling study
of the zinc slag fuming process. In the first stage of the present research, a
Computational Fluid Dynamic (CFD) modelling study of the Top Submerged Lance
(TSL) gas injection process was carried out in a laboratory scale isothermal air-water
model. The multiphase flow simulation, based on Euler-Euler approach, elucidated the
effect of swirl and non-swirl flow inside the bath. The effects of lance submergence
level and air flow rate were also investigated in that phase. The simulation results for
velocity fields and generation of turbulence in the bath were validated against existing
experimental data from the previous water model experimental study of Morsi et al.
[1].
In the next stage of the research, a Computational Fluid Dynamic (CFD) model of the
pilot plant scale top submerged lance slag fuming furnace was developed to
investigate details of fluid flow, combustion behaviour, reaction kinetics and heat
transfer in the furnace. The model integrates submerged CH4 combustion at the lance
tip and chemical reactions with the heat, mass and momentum interfacial interaction
between the phases present in the system. Commercial CFD package AVL Fire 2009.2
(AVL, Graz, Austria) coupled with a number of user defined subroutines in
FORTRAN programming language were used to develop the model. The model is
based on 3-D Eulerian multiphase flow approach and it predicted the velocity and
iv
temperature field of the molten slag bath, generated turbulence, vortex and plume
shape at the lance tip. The model also predicted the mass fractions of slag and gaseous
components inside the furnace. The model was validated by adopting the macro – step
validation approach by using the zinc fuming rate against the pilot plant scale
experimental study on top submerged lance zinc fuming process carried out by
Waladan et al. [2].
Finally, the developed CFD model for TSL furnace was extended for submerged coal
combustion instead of CH4 combustion and applied to a conventional tuyere blown
slag fuming furnace. The model considered a thin slice of a conventional tuyere blown
slag fuming furnace consisting two opposing set of tuyere. The model was developed
in Eulerian multiphase flow approach by employing 3D hybrid unstructured
orthographic grid system. The aim was to investigate details of fluid flow, submerged
coal combustion dynamics, coal utilization behaviour, jet penetration behaviour, bath
interaction conditions and generation of turbulence in the bath. The model was
developed by coupling the CFD with the kinetics equations developed by Richards et
al. for a zinc slag fuming furnace. The model predicted the velocity, temperature field
of the molten slag bath, generated turbulence and vortex, coal utilization behaviour
from the slag bath. The jet penetration depth at the tuyere tip was validated against the
experimental study carried out by Hoefele and Brimacombe [3].
v
Acknowledgements
I would like to express my heartiest gratitude to a number of great persons who
contributed to my PhD journey in innumerable ways - academically, professionally
and psychologically.
First and foremost, I offer my sincerest gratitude to my supervisor Dr Jamal Naser, for
his contributions of time, ideas, and arrangement for funding to make my Ph.D.
experience productive and stimulating. In particular, I appreciate his invaluable and
enterprising guidance throughout the execution of this project. He has supported me
throughout this challenging journey with his patience and knowledge, whilst allowing
me the room to work in my own way. In addition, he was always accessible and
willing to help his students with their research. From the beginning, he inspired me to
successfully complete my degree. He provided me with appropriate direction,
technical support and became more of a mentor and friend, than a supervisor.
I gratefully thank my second supervisor Professor Geoffrey Brooks for his constant
encouragement, valuable suggestions throughout the project and patience in reading
and correcting the thesis within a short period of time. I feel myself truly honoured to
be a part of this research group and to have the opportunity to work with my
supervisors. One simply could not wish for better or friendlier supervisors. I mastered
the research techniques and have been well trained by my supervisors from the initial
to the final level of my degree.
I have been truly inspired by my industrial supervisors Professor Markus Reuter and
Mr. Robert Matusewicz, throughout this successful journey. I gratefully acknowledge
their constant support, valuable discussions and contributions on the pilot plant data
used in this study.
I would also like to take the opportunity to express my gratitude to my beloved wife,
Maimuna Musarrat, for her inspiring support towards the end of this tiring journey.
She came in my life as the greatest blessing and has been inspiring me to finish it on
vi
time. She provided me with immense mental support and accelerated my strengths
during the writing of this thesis.
From my childhood, I have been fortunate enough to have the most beloved siblings in
the world. I wish to express my sincerest gratitude to all of my siblings for their
continual support from the very beginning of my study life till now. No one could
wish to have better siblings than I have.
In my daily work I have been blessed with a friendly and cheerful group of fellow
students. Thanks to members of the High Temperature Processing Group of the
Faculty of Engineering and Industrial Science for their ongoing support and good
humour.
I wish to thank all of my friends outside the department for bearing with me and their
understanding. I offer my regards and blessings to all of my friends here in
Melbourne, my housemates and colleagues. My special thanks to Mustafa Tareq,
Adity Tareq, Abdul Aziz and Shovin Aziz, who have always treated me as a member
of their family in Australia. They gave me strength to stand for who I am and provided
me with mental support whenever I was in need of it.
I'm sure I've forgotten someone. I assure you that this is a shortcoming on my part and
not on yours. I beg you to forgive me for my oversight.
Finally I gratefully acknowledge the financial and other support received for this
research from the Faculty of Engineering and Industrial Science, Swinburne
University of Technology.
Above all, this thesis is dedicated to my beloved parents, Md. Ishaque and
Meherunnesa Begum. They both are the true inspirations and love of my life. They
sacrificed everything for me to let me grow and achieve every success of my life.
Their unconditional love and continual support in all that I have done till now are the
keys to all my achievements.
vii
Table of Contents
Declaration .................................................................................................................... ii
Certification ................................................................................................................... ii
Abstract ......................................................................................................................... iii
Acknowledgements ......................................................................................................... v
Table of Contents ....................................................................................................... vii
Lists of Figures ............................................................................................................ 12
Lists of Tables .............................................................................................................. 18
Nomenclature............................................................................................................... 19
Chapter 1 ..................................................................................................................... 22
1 Introduction ......................................................................................................... 23
1.1 Primary Research Theme .............................................................................. 24
1.2 Research Overview ....................................................................................... 25
1.3 Contributions of this research ....................................................................... 26
1.4 Publications from present research ............................................................... 28
1.5 Thesis Structure ............................................................................................. 29
Chapter 2 ..................................................................................................................... 31
2 Process Overview of TSL Technology and Zinc Extraction ........................... 32
2.1 Gas Injection into molten system .................................................................. 32
2.2 Development of TSL Technology ................................................................. 32
2.3 Operating Principle of TSL Technology ....................................................... 33
2.4 History of Zinc .............................................................................................. 36
2.5 Physical and chemical properties of zinc ...................................................... 39
2.6 Uses of zinc ................................................................................................... 40
2.7 Extraction Methods ....................................................................................... 45
2.7.1 Primary zinc production .......................................................................... 45
viii
2.7.1.1 Hydrometallurgical/ Electrolytic route ............................................ 48
2.7.1.2 Pyrometallurgical route.................................................................... 50
2.7.2 Slag Fuming/ Secondary zinc production ............................................... 54
2.7.2.1 Conventional Slag Fuming Operation ............................................. 56
2.7.2.2 Application of TSL Technology in Zinc Processing ....................... 58
CHAPTER 3 ................................................................................................................ 66
3 Literature Review................................................................................................ 67
3.1 Cold Model Investigations ............................................................................ 67
3.1.1 Experimental studies ............................................................................... 68
3.1.2 Swirl and Non-swirl Investigation .......................................................... 70
3.1.3 Formation of Bubbles and Splashing ...................................................... 73
3.1.4 Bath mixing characteristics ..................................................................... 79
3.1.5 Numerical Investigation .......................................................................... 80
3.2 Review of Zinc Slag Fuming Process ........................................................... 83
3.2.1 Slag fuming by conventional tuyere blown process ............................... 84
3.2.2 Slag fuming by TSL process ................................................................... 87
3.2.3 Other studies based on zinc extraction .................................................... 90
3.3 Research Objectives ...................................................................................... 92
Chapter 4 ..................................................................................................................... 94
4 Modelling Techniques and Model Features ...................................................... 95
4.1 CFD Modelling ............................................................................................. 95
4.1.1 Finite Volume Method ............................................................................ 97
4.2 Multiphase Flow Modelling .......................................................................... 98
4.2.1 Approaches to Multiphase Modelling ..................................................... 98
4.2.1.1 The Euler-Lagrange Approach ........................................................ 98
4.2.1.2 The Euler-Euler Approach ............................................................... 99
4.2.1.2.1 Homogeneous Model ................................................................ 100
4.2.1.2.2 Multi-fluid Model ...................................................................... 100
4.2.1.2.3 VOF Model ................................................................................ 100
4.3 Model Geometry and Computational Methodology ................................... 101
ix
4.3.1 Air water Model .................................................................................... 102
4.3.1.1 Model Features .............................................................................. 103
4.3.1.2 Governing Equations ..................................................................... 104
4.3.1.2.1 Continuity .................................................................................. 104
4.3.1.2.2 Momentum conservation ........................................................... 104
4.3.1.2.3 Interfacial Momentum Exchange .............................................. 106
4.3.1.3 Boundary conditions ...................................................................... 107
4.3.1.3.1 Inlet ............................................................................................ 107
4.3.1.3.2 Outlet ......................................................................................... 108
4.3.1.3.3 Wall ........................................................................................... 109
4.3.1.4 Initial Conditions and Fluid Properties .......................................... 110
4.3.2 Zinc Fuming TSL Model ...................................................................... 111
4.3.2.1 Slag Composition ........................................................................... 112
4.3.2.2 Model Features .............................................................................. 114
4.3.2.3 Governing Equations ..................................................................... 115
4.3.2.3.1 Enthalpy conservation ............................................................... 115
4.3.2.3.2 Interfacial Energy Exchange ..................................................... 116
4.3.2.3.3 Combustion Modelling .............................................................. 117
4.3.2.3.4 Chemical Reactions in the Slag Bath ........................................ 118
4.3.2.3.5 Interfacial Mass Exchange ........................................................ 120
4.3.2.4 Boundary conditions ...................................................................... 120
4.3.2.4.1 Inlet ............................................................................................ 120
4.3.2.4.2 Outlet ......................................................................................... 121
4.3.2.4.3 Wall ........................................................................................... 121
4.3.2.5 Initial Conditions and Fluid Properties .......................................... 122
4.3.3 Conventional Tuyere blow model ......................................................... 124
4.3.3.1 Model Geometry and computational mesh .................................... 124
4.3.3.2 Slag Composition ........................................................................... 127
4.3.3.3 Model Features .............................................................................. 127
4.3.3.4 Governing equations ...................................................................... 128
4.3.3.4.1 Coal combustion ........................................................................ 128
4.3.3.4.2 Devolatilization ......................................................................... 130
4.3.3.4.3 Gas Phase Combustion .............................................................. 131
4.3.3.4.4 Char oxidation ........................................................................... 132
x
4.3.3.4.5 Chemical Reactions in the Slag Bath ........................................ 133
4.3.3.5 Boundary conditions ...................................................................... 134
4.3.3.5.1 Inlet ............................................................................................ 134
4.3.3.5.2 Outlet ......................................................................................... 134
4.3.3.5.3 Wall ........................................................................................... 134
4.3.3.5.4 Symmetry .................................................................................. 135
4.3.3.6 Initial Conditions and Fluid Properties .......................................... 135
Chapter 5 ................................................................................................................... 137
5 Cold Flow CFD Model of the TSL Gas Injection Process ............................. 138
5.1 Test of Grid Independence .......................................................................... 138
5.2 Results and Discussion ................................................................................ 141
5.2.1 Effect of Swirl Intensity ........................................................................ 141
5.2.2 Effect of Submergence Level ................................................................ 148
5.2.3 Effect of Air Flow Rate ......................................................................... 150
5.2.4 Mixing in the Liquid Bath ..................................................................... 152
5.2.5 Turbulence Mixing ................................................................................ 152
5.2.6 Mean Convective Mixing ...................................................................... 156
5.2.7 Effect of Density ................................................................................... 158
5.2.8 Effect of Viscosity................................................................................. 160
Chapter 6 ................................................................................................................... 161
6 Numerical Investigation of Zinc Fuming Bath in TSL Furnace ................... 162
6.1 Test of Grid Independency .......................................................................... 162
6.2 Results and Discussion ................................................................................ 164
6.2.1 Bath Behaviour...................................................................................... 165
6.2.2 Combustion Behaviour ......................................................................... 171
6.2.3 Zinc Fuming Behaviour ........................................................................ 175
6.2.4 Effect of Lance submergence level ....................................................... 180
Chapter 7 ................................................................................................................... 186
7 CFD Modelling of Conventional Zinc Fuming Furnace ................................ 187
xi
7.1 Test of Grid Independency .......................................................................... 187
7.2 Results and Discussion ................................................................................ 189
7.2.1 Bath behaviour ...................................................................................... 190
7.2.2 Jet Penetration ....................................................................................... 196
7.2.3 Bath zones ............................................................................................. 200
7.2.4 Fuming Behaviour ................................................................................. 201
Chapter 8 ................................................................................................................... 205
8 Conclusions and Recommendations ................................................................ 206
8.1 Conclusions ................................................................................................. 206
8.2 Recommendations for further work ............................................................ 208
References: ................................................................................................................. 210
12
Lists of Figures
Figure 1-1: Schematic diagram of combustion zone and molten metal bath of a TSL
furnace ........................................................................................................................... 24
Figure 2-1: A cutaway schematic diagram of the typical Outotec Ausmelt TSL furnace
[12] ................................................................................................................................ 34
Figure 2-2: Schematic representation of Indian method for producing zinc [23] ......... 36
Figure 2-3: William Champion’s Zinc smelting furnace [23] ...................................... 37
Figure 2-4: Zinc demand by First use in 2005 estimates [26]....................................... 40
Figure 2-5: Zinc demand by End use in 2003 estimates [26] ....................................... 41
Figure 2-6: Zinc consumption in the World during 1960-2005 (kilo tonne) [26] ........ 41
Figure 2-7: Schematic diagram of Multiple Hearth Roaster (Image taken from
www.energytek.com.tw) ............................................................................................... 46
Figure 2-8: Schematic diagram of a fluidised bed roaster [26] ..................................... 47
Figure 2-9: Schematic diagram of an electrolytic cell .................................................. 50
Figure 2-10: Schematic illustration of the slag fuming process .................................... 55
Figure 2-11: Schematic of fuming furnace cross section [18] ...................................... 56
Figure 2-12: A schematic diagram of a rectangular tuyere blow conventional slag
fuming operation (Image taken from US Patent by Quarm [22]) ................................ 57
Figure 2-13: Korea Zinc’s integrated flow sheet using TSL technology to recover zinc
and lead from various slags and residues created during primary zinc and lead
concentrate processing [Hoang et al. [39]] ................................................................... 59
Figure 2-14: Schematic diagram of the zinc fumer at Onsan, Korea [Choi et al. [33]] 59
Figure 2-15: Flow circuit of the zinc fuming plant at Onsan, Korea. (Floyd et al. [25])61
Figure 2-16: Flow circuit of ISP Slag Fumer at Hachinohe, Japan. (Floyd et al. [37]] 62
13
Figure 2-17: Flow circuit of the commercialised TSL zinc technology to recover zinc
from primary leach residues (Image taken from Hoang et al. [12]).............................. 64
Figure 2-18: Conceptual flow circuit of the TSL Direct Zinc Smelting technology
(Image taken from Hoang et al. [12])............................................................................ 65
Figure 3-1: Schematic of the submerged plasma process for the high temperature
fuming of zinc from zinc containing residues [Image taken from Verscheure et al.] ... 90
Figure 3-2: Schematic diagram of the Enviroplas pilot plant (Image taken from Latif
[104]) ............................................................................................................................. 92
Figure 4-1: Schematic diagram of the air-water model .............................................. 102
Figure 4-2: Mid Plane cross sectional view of generated grid for CFD analysis ....... 103
Figure 4-3: Velocity vectors (m/s) for swirl air injection at the lance tip (Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =57.5o) ........................................................................................... 108
Figure 4-4: Schematic diagram of the modelled furnace for Outotec TSL pilot plant111
Figure 4-5: Generated grid of the modelled pilot plant scale TSL furnace for CFD
analysis ........................................................................................................................ 112
Figure 4-6: Simplified phase relationships for the reduction step in an Outotec TSL
furnace for the components FeOx, ZnO, CaO and SiO2 generated by FACT Sage
[101] for the given temperature, partial pressure as well as the lime content. ............ 113
Figure 4-7: Schematic view of the modelled thin slice rectangular tuyere blow furnace
(Isometric Layout) ....................................................................................................... 125
Figure 4-8: (a) Generated surface mesh, (b) Volume mesh for CFD analysis (course
grid) ............................................................................................................................. 126
Figure 4-9: Coal combustion process flow chart ........................................................ 130
Figure 5-1: Mean tangential velocity (m/s) distribution for different grid
configurations .............................................................................................................. 140
14
Figure 5-2: Axial velocity (w) distribution (m/s) for the water model simulation ..... 143
Figure 5-3: Axial velocity (w) distribution (m/s) from experimental results of Morsi et
al. [1] ........................................................................................................................... 145
Figure 5-4: Tangential velocity (v) distribution (m/s) for the water model simulation146
Figure 5-5: Tangential velocity (v) distribution (m/s) from experimental results of
Morsi et al. [1] ............................................................................................................. 147
Figure 5-6: Mean tangential velocity comparison between swirl and non-swirl flow
from the simulation results and comparison with water model experiment of Morsi et
al. [1] (z/Z= 0.92, H/L= 2/3, Q=2.67 x 10-3
m3/s) ....................................................... 148
Figure 5-7: Average volume fraction of water at 68.0=Z
z for different submergence
level for the water model simulation (Q=2.67 x 10-3
m3/s, Ф = 57.5
o) ....................... 149
Figure 5-8: Relation between vertical penetration distance for annulus air injection
(va
L ) and modified Froude number (m
Fr ) as derived from the water modelling
simulation results (H/L=1/3, Ф = 0o) .......................................................................... 151
Figure 5-9: Turbulent Kinetic energy (k) distribution (m2/s2)- (a) Q=2.67 x 10-3
m3/s,
H/L=2/3, Ф = 0o, t = 60 sec, (b) Q=2.67 x 10
-3 m
3/s, H/L=2/3, Ф = 57.5
o, t = 60 sec 153
Figure 5-10: Turbulent kinetic energy (k) distribution (m2/s
2) from experimental
results of Morsi et al. [1] ((a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 0
o, (b) Q=2.67 x 10
-3
m3/s, H/L=2/3, Ф =57.5
o) ............................................................................................ 154
Figure 5-11: Volume fraction for water after 180 seconds (Q=2.67 x 10-3
m3/s,
H/L=2/3, Ф =57.5o) ..................................................................................................... 155
Figure 5-12: Velocity vectors for liquid phase (m/s) after 180 seconds
(Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =57.5
o) ................................................................... 156
Figure 5-13: Volume exchange effectiveness along radial direction from present water
model simulation (Q=2.67 x 10-3
m3/s, H/L=1/3) ....................................................... 157
15
Figure 5-14: Contours for volume exchange effectiveness from present water model
simulation .................................................................................................................... 158
Figure 5-15: Average volume fraction at 40mm height (z/Z=0.66) above the liquid
bath (Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o) .......................................................... 159
Figure 5-16: Average volume fraction at 60mm height (z/Z=0.625) above the liquid
bath .............................................................................................................................. 159
Figure 6-1: Mean tangential velocity distributions for different grid configurations (Qa
= 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5) ................................................................ 163
Figure 6-2: Volume fraction distribution for molten slag phase along the vertical cross
section in X-Z plane at four different time step of the transient simulation (Qa = 0.05
kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 13th
second, (b) 13.5th
second, (c) 20.5th
second and (d) 32nd
second. ........................................................................................ 167
Figure 6-3: Volume fraction distribution for molten slag phase along the cross section
in X-Y plane (top view of the modelled furnace) at different time steps of the transient
simulation (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 21.5th
second, (b) 22nd
second .......................................................................................................................... 168
Figure 6-4: Iso-contour plot of the molten slag phase at different time steps of the
transient simulation (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 25th
second,
(b) 26.5th
second .......................................................................................................... 169
Figure 6-5: Streamlines distribution of the molten slag phase showing slag phase
movement inside the furnace at different time steps of the transient simulation (Qa =
0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 16.5th
second, (b) 27th
second............. 170
Figure 6-6: Velocity vectors inside the furnace in the molten slag bath: (Qa = 0.05
kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5) ........................................................................... 171
16
Figure 6-7: Species mass fraction (kg/kg) distribution for (a) CH4, (b) CO2 (c) CO and
(d) O2 ........................................................................................................................... 173
Figure 6-8: Temperature distribution for molten slag phase only (b) Species mass
fraction (kg/kg) distribution for O2 (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5) . 174
Figure 6-9: Fumed zinc distribution inside the furnace (a) at initial stage (after 1
second) (b) after 30 seconds (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5) ............ 175
Figure 6-10: ZnO distribution in the slag bath at different depths below the lance tip
along a line in radial direction from the lance (Qa = 0.05 kg/s, Qf = 0.0035 kg/s,
LH ′′ = 1/5) ................................................................................................................ 177
Figure 6-11: CO mass fraction (kg/kg) distribution along the radial direction at
different depths below the lance tip (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5) 178
Figure 6-12: Zinc fuming rate from some published experimental work ................... 178
Figure 6-13: Fuming rate from present simulation results (Qa = 0.05 kg/s, Qf = 0.0035
kg/s, LH ′′ = 1/5) and comparison with experimental data (Curve regenerated from
CZF5 of Figure 11 from Waladan et al. [2]) ............................................................... 179
Figure 6-14 (a): Zinc fuming rate and amount of splash at different heights above the
bath (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3) ................................................. 181
Figure 6-15: Comparison of zinc fuming rate and amount of splash at h/H = 0.09
above the bath for three submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s) ....... 184
Figure 6-16: Comparison of zinc fuming rate and amount of splash at h/H = 0.73
above the bath for three submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s) ....... 184
Figure 6-17: Overall zinc fuming rate comparison for three different lance
submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s) .............................................. 185
Figure 7-1: Mean upward velocity, W (m/s) distribution of slag phase for three
different grids along axial direction (Uo = 67.8 m/s, d/D = 0.5, h/H = 0.02) ............. 189
17
Figure 7-2: Slag phase volume fraction distribution (a) mid-plane cut at X-Z plane (b)
Iso-contour plot of slag phase ..................................................................................... 192
Figure 7-3: (a) Diagram showing injection phenomena in zinc fuming, (b) schematic
representation of the sequence of reactions in the bath (both the figures are taken from
Richards et al.[21]) ...................................................................................................... 193
Figure 7-4: (a) Injected coal mass at the tuyere tip (b) Average coal mass along the
axial direction from the tuyere tip(Uo = 86 m/s, d/D = 0.5, h/H = 0.02).................... 195
Figure 7-5: Turbulent kinetic energy distribution inside the modelled furnace (Uo = 86
m/s) .............................................................................................................................. 195
Figure 7-6: Gas phase volume fraction at tuyere 1 tip showing jet penetration length
(lp) and jet expansion angle ( Θ ) for Uo = 86 m/s ...................................................... 197
Figure 7-7 (a): Comparison of the CFD results of tuyere jet penetration length (lp) with
correlation provided by Hoefele and Brimacombe [3] from experimental work........ 198
Figure 7-8: Velocity vectors of the slag phase inside the molten slag bath (Uo = 86
m/s) .............................................................................................................................. 201
Figure 7-9: ZnO distribution in the slag bath along the axial direction from the tuyere
tip at three different heights from the bottom wall (Uo = 86 m/s) .............................. 202
Figure 7-10: Schematic illustration of the tuyere tip jet and coal behaviour as observed
from the present simulation ......................................................................................... 202
18
Lists of Tables
Table 2-1: Physical and chemical properties of zinc [24] ............................................. 39
Table 2-2: Typical operating results of the zinc fumer at Onsan, Korea. [Floyd et al.
[37]] ............................................................................................................................... 60
Table 2-3: Typical operating conditions of the Hachinohe Smelter, Japan [Floyd et al.
[37]] ............................................................................................................................... 63
Table 4-1: Slag composition for TSL Zinc Fuming Model ........................................ 113
Table 4-2: Fluid and thermal properties of molten slag phase for TSL model ........... 123
Table 4-3: Injection conditions (TSL model).………………………………………119
Table 4-4: Comparisons of the simulation and plant data…..………………………121
Table 4-5: Initial Slag Composition for Tuyere blow model……………………….123
Table 4-6: Fluid and thermal properties of molten slag phase ……………………...131
Table 5-1: Overview of computational grids .............................................................. 139
Table 5-2: Overview of the simulation and experimental conditions ......................... 140
Table 5-3: Simulation conditions and corresponding figures ..................................... 141
Table 6-1: Overview of computational grids .............................................................. 163
Table 6-2: Injection conditions (CFD and Experimental) .......................................... 164
Table 7-1: Overview of computational grids .............................................................. 188
19
Nomenclature
cA pre-exponential factor
vA pre-exponential factor
DC drag coefficient
D cylinder diameter
Do outlet diameter of the cylinder
bD bubble diameter
mkD
, diffusion coefficient (m
2/s)
od outer diameter of the lance
id inner diameter of the lance
cE activation energy constant
vE activation energy constant
mFr modified Froude number
f body force vector
g gravitational body force (m/s2)
H lance submergence level for air-water model
H ′ lance submergence level for TSL pilot plant model
k turbulent kinetic energy
vK rate constant
gask total number of chemical species
L liquid level in the cylinder for air-water model
L′ liquid level in the cylinder for TSL pilot plant model
20
l vertical depth from the lance tip for TSL pilot plant model
pl jet penetration length at the tuyere tip
vaL vertical penetration distance for air jet injected through annulus
N number of phases
P local pressure
AP atmospheric pressure
Q air flow rate through lance for air-water model
aQ air flow rate through lance for TSL pilot plant model
fQ fuel flow rate through lance for TSL pilot plant model
R universal gas constant (J K-1
mol-1
)
bRe bubble Reynolds number
r radial distance from the centre point
pR radius of the coal particle
kS species source term
tSc turbulent Schmidt number
pT temperature of the coal particle
gT gas temperature
t
kT phase k Reynolds stress
V released volatiles
v velocity vector
fV ultimate volatile content
X radial coordinate
Y tangential coordinate
21
Z axial coordinate
Greek letters
µ molecular viscosity (N·s/m2)
tµ turbulent viscosity
SIt ,µ shear induced turbulent viscosity
BIt ,µ bubble induced turbulent viscosity
α volume fraction
ρ density (kg/m3)
τ shear stress
δ kronecker delta function
ε dissipation rate
22
Chapter 1
23
1 Introduction
Pyrometallurgy is one of the fundamental branches of extractive metallurgy. It is an
ancient technology which has defined significant stages of human development [4].
Smelting and refining operations are the major process of any pyrometallurgical route,
which is achieved through gas injection process. For the purpose of refining and
smelting ferrous and non-ferrous materials, gas injection process has been successfully
in operation since the 1800s. Based on the mode of injection, gas injection process can
be classified into following four categories:
Top submerged injection (TSL/Sirosmelt Furnace)
Top non-submerged injection (BOF steelmaking, Hismelt, Mitsubishi Smelter)
Bottom blow using tuyeres or nozzles (Bessemer, QSL furnace)
Side blow (horizontal or inclined) using tuyeres or nozzles (Pierce Smith
converter, Noranda reactor)
The major purpose of gas injection process is to create agitation in the molten metal
bath to accelerate mixing and promote refining reactions. Guthrie [5-9] described the
detail hydrodynamics of gas-stirred melts and fluid flows in metallurgy. Submerged
gas injection is known to be an effective way of stirring the bath for mixing, refining
and smelting due to the rigorous agitation achieved. The idea of submerged injection
was first proposed by Sir Henry Bessemer in 1860 for his bottom blown Bessemer
steelmaking process. Since then submerged gas injection has played an important role
in metals refining processes.
Top submerged lance (TSL) technology (also known as Sirosmelt) has been in
successful operated and established world wide as an effective smelting technology.
This unique Australian technology was developed by Dr. John Floyd [10] at the
Commonwealth Scientific and Industrial Research Organisation (CSIRO) in the
1970s. Since it’s invention, TSL technology has been used to both ferrous and non-
ferrous metallurgical process industries.
24
1.1 Primary Research Theme
This thesis is primarily focused on computational fluid dynamic (CFD) investigation
of the zinc slag fuming behaviour in a top submerged lance (TSL) zinc fuming
furnace. The primary objective was to predict lance tip combustion and bath
interaction conditions in order to achieve the maximum combustion efficiency and
control over the desired operating conditions in a TSL furnace. Though TSL
technology is successfully operating around the world, a complete understanding of
the non-isothermal process in three dimensional flow field structure including
submerged combustion and chemical reactions still needs thorough investigation.
Process optimization through real plant observation is difficult and gaining insight into
such a complex industrial system through the application of high temperature
experimentation is also hard to achieve. The process involves a very high temperature
robust combusting environment with interfacial mass, momentum and energy transfer
and complex chemical reactions among the species of the phases.
Figure 1-1: Schematic diagram of combustion zone and molten metal bath of a TSL
furnace
25
With the advancement of high performance computing facilities, computational fluid
dynamic (CFD) modelling technique has evolved as a powerful tool for the
researchers working in the metallurgical field. CFD can predict flows ranging from
simple single phase flows to complex multiphase flows in high temperature
combusting environment associated with metallurgical process industries. A number
of studies are reported in the open literature employing CFD tool for investigating
smelting and refining operations [11-18]. Successful and efficient development of a
CFD model can predict the fluid flow behaviour, combustion behaviour, generation of
turbulence and splashing and other fluid dynamic parameters inside the furnace.
Hence, CFD was chosen as the modelling tool for a comprehensive analysis of the
zinc fuming TSL furnace.
1.2 Research Overview
The current research consists of three major stages, starting with cold flow air-water
CFD model, followed by the CFD model of the non-isothermal pilot plant scale zinc
slag fuming furnace and finally extension of the developed model into the
conventional tuyere blow furnace configuration. All the models were developed by
different versions of AVL FIRE (AVL, Graz, Austria). The second and the third stage
work incorporate a number of User Defined Subroutines in FORTRAN programming
language. Details of the developed models and modelling techniques are described in
Chapter 4. The three stages are summarized as follows:
In the first step, a cold flow isothermal air-water model of a laboratory scale
top submerged lance gas injection process was developed. A comprehensive
analysis of fluid dynamic behaviour of the TSL gas injection process was
carried out by CFD technique in 3D hybrid unstructured grid system. This
model was successfully validated against an experimental study of Morsi et al.
[1].
In stage two, a pilot plant scale CFD model of zinc fuming TSL furnace was
developed. The model integrates submerged CH4 combustion at the lance tip,
interfacial mass, momentum and energy exchange and chemical reactions in
26
the slag bath. Despite some limitations, the zinc fuming rate predicted by the
model agreed reasonably well with the experimental data of Waladan et al. [2].
Finally, the developed code was further extended to be applied to a
conventional tuyere blow zinc slag fuming furnace to check the applicability of
the developed code in other furnace configurations. The submerged
combustion code for CH4 was extended for coal combustion. The model
incorporates submerged multiphase coal combustion with devolatilization, char
oxidation and char chemical reaction. Details of the kinetics of the zinc slag
fuming process for a conventional zinc fuming furnace were described by
Richards et al. [19-21]. The chemical reactions in the developed CFD model
were based on the kinetic theory of Richards et al. [19-21]. A detailed fluid
dynamic analysis, tuyere tip combustion behaviour, coal utilization behaviour
and slag fuming behaviour at different furnace locations of the process were
analysed in the developed model.
1.3 Contributions of this research
The air-water simulation results from the first stage of the research showed that 2/3
lance submergence level provides better mixing and high liquid velocities for
generation of turbulence inside the water bath. However, it is also responsible for
generating more splash in the bath compared to 1/3 submergence level. An approach
generally used by Heating, Ventilation and Air Conditioning (HVAC) system
simulation was applied in that stage to predict the convective mixing phenomena. The
simulation results for the air-water system showed that, mean convective mixing for
swirl flow is more than twice than that of non-swirl in close proximity to the lance. A
semi-empirical equation was proposed from the results of present simulation to
measure the vertical penetration distance of the air jet injected through the annulus of
the lance in the cylindrical vessel of the model, which can be expressed as,
( ) 4745.0275.0 miova FrddL −= 1.1
27
After the successful validation of the air-water model in the first stage, the model was
extended to measure the degree of splash generation at certain heights above the free
surface for different liquid density, as an exploratory step towards the development of
a more complex pilot plant scale combusting system.
Investigation of the pilot plant scale zinc fuming TSL furnace predicted that the % of
ZnO in the slag bath decreases linearly with time and is broadly consistent with the
experimental data. The model results further predicted that the rate of ZnO reduction is
controlled by the diffusion of ZnO from the bulk slag to slag-gas interface and rate of
gas-carbon reaction. Three different lance submergence levels were considered to
study the effect of lance submergence level on splash generation and overall fuming
rate. The overall fuming rate for 1/3 lance submergence level were found to be 0.25 to
0.4 wt%/min, which is around 1.3 times higher than 1/5 lance submergence level.
In the third stage, tuyere jet penetration length ( pl ) was compared with the equation
provided by Hoefele and Brimacombe [3] from isothermal experimental work
(equation 1-2) and found 2.26 times higher, which is due to coal combustion and gas
expansion at high temperature.
( ) ( ) 35.046.07.10 lgFr
o
P Nd
lρρ′=
1.2
The jet expansion angle measured for the slag system studied is 85o for the specific
inlet conditions during the simulation time studied. Highest coal penetration distance
was found to be l/L = 0.2, where l is the distance from the tuyere tip along the centre
line and L is the total length (2.44m) of the modelled furnace. The model also
predicted that 10% of the injected coal bypasses the tuyere gas stream un-combusted
and carried to the free surface by the tuyere gas stream, which contributes to zinc
oxide reduction near the free surface.
28
1.4 Publications from present research
Following papers have resulted from the present research:
Journals:
1) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: CFD
Modelling of Swirl and Non-swirl gas Injections Into Liquid Baths Using Top
Submerged Lances. Metallurgical and Materials Transactions B, 2010, vol
41(1), pp. 35-50
2) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: CFD
Modelling of Zinc Fuming Process in Top Submerged Lance Smelting
Furnace. Metallurgical and Materials Transactions B, 2012. vol 43(1): p. 39-
55.
3) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: CFD
Investigation of Submerged Combustion Behaviour in a Tuyere Blown Slag
Fuming Furnace. Published online in Metallurgical and Materials Transactions
B, 2012, DOI: 10.1007/s11663-012-9686-7
Conference Publications and Presentations
1) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “CFD
Modelling of Gas Injections in Top Submerged Lance Smelting” in TMS
Annual General Meeting. San Francisco, California, USA, 15-19 February,
2009, pp. 95-102.
2) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “CFD
Modelling of Top Submerged Lance Gas Injection”, in High Temperature
Processing Symposium, Melbourne, Australia, 9 February, 2009
3) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “Combustion
Modelling of Top Submerged Lance furnace by using CFD tool”, in High
Temperature Processing Symposium, Melbourne, Australia, 8-9 February,
2010, ISBN 978-0-9806708-0-6
4) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “A
Computational Fluid Dynamic Modelling study of slag fuming in Top
Submerged Lance Smelting Furnace” in World Congress on Engineering.
London, UK, 30th
June- 2nd
July, 2010.
5) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “Applications
of CFD Modelling in Metallurgical Process Industries”, in 15th
Biennial
Computational Techniques and Applications Conference (CTAC 2010),
Sydney, Australia, 28th
November – 1st December, 2010.
29
6) N. Huda, J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz: “Applications
of CFD Modelling in Smelting Industries – Some Recent Developments”,
in High Temperature Processing Symposium, Melbourne, Australia, 7-8
February, 2011, ISBN 978-0-9806708-2-0.
7) N.Huda, J. Naser, G. Brooks: CFD Modelling of Combustion Behaviour in
Slag Fuming Furnaces. TMS Annual General Meeting (International Smelting
Technology Symposium) Orlando, Florida, USA, 11-15 March, 2012, pp. 251-
258
1.5 Thesis Structure
A detail description of the overview of TSL technology including its development and
operation is presented in Chapter 2. Chapter 2 also contains process overview of zinc
extraction which highlights the history of zinc, its uses and extraction methods
including hydro and pyro-metallurgical circuits of zinc extraction. Then a review of
the previous studies based on gas injection process, TSL model studies and zinc
fuming kinetics are presented in Chapter 3. Chapter 3 begins with the current state of
knowledge on gas injection process and cold model experimental studies on gas
injection, followed by top submerged lance investigations – including modelling and
experimental work. At the end of Chapter 3, a comprehensive literature survey based
on zinc slag fuming process is also cited.
Chapter 4 begins by describing the methods of numerical investigation and CFD
modelling approach for current research followed by procedure of solving the
governing equations for fluid motion, combustion, interfacial phenomena and
chemical reactions. The detailed description and dimensions of the developed models,
methodologies and boundary conditions used in this research program are discussed in
the last portion of chapter 4.
A detailed hydrodynamic analysis of the top submerged lance gas injection process is
discussed in Chapter 5. In this chapter, effect of swirl and non-swirl flow, gas
injection rate, lance submergence level, density and viscosity is discussed for air-water
system. Chapter 6 demonstrates the results obtained from the developed CFD model
of the pilot plant scale zinc fuming TSL furnace. Chapter 7 recapitulates the zinc
fuming behaviour for a different furnace configuration – i.e. side injection through
30
tuyeres. Experimental results and validation of the developed models is outlined in
each of the corresponding chapters. Both Chapters 6 and 7 discusses the results of the
developed CFD model for zinc fuming process and examines the process variables
that influence the zinc fuming rate. Finally conclusions from the study are drawn and
recommendations for future work are suggested in Chapter 8.
31
Chapter 2
32
2 Process Overview of TSL Technology and Zinc
Extraction
This Chapter begins with an overview of Top Submerged Lance (TSL) furnace, its
history and development, followed by its basic working principle. Then, the chapter
will focus on the history and extraction of zinc both from its ores and from
metallurgical wastes or slags, different uses of zinc and available extraction methods.
At the end of this chapter, application of TSL technology to slag fuming process is
discussed.
2.1 Gas Injection into molten system
Gas injection in pyrometallurgy plays significant roles through creating mixing,
promoting interfacial reactions, producing an interface between phases, promoting
foaming and supplying energy into the system by combustion. Gas injection can also
be used as a medium to inject solids into the molten bath (i.e. combustion and
reductant coal). Different types of gas injection process commonly used in
pyrometallurgy have been discussed in Chapter 1 (Section 1). This Chapter will focus
on the development, operation and working principle of the TSL technology, as the
primary research theme is based on top submerged lancing (TSL) system.
2.2 Development of TSL Technology
The Top Submerged Lance (TSL) technology was developed by CSIRO scientist Dr.
John Floyd and his team during the 1970s. Initially it was called High temperature
submerged combustion, then SIROSMELT. It was then commercialized by the then
Ausmelt Limited (Now known as Outotec Ausmelt) and Xstrata Technology formerly
known as Isasmelt (Mount Isa Mines). The technology has been successfully
implemented in over 35 furnaces worldwide in 23 locations in 14 countries by Outotec
Ausmelt for the production of tin, copper, nickel, lead, platinum group metals, zinc
and aluminium [4].
33
2.3 Operating Principle of TSL Technology
The top submerged lance (TSL) furnace is a long cylindrical furnace with the lance
positioned centrally. It is a high temperature bath smelting process that uses the
vertically suspended lance to inject process air, oxygen and fuel to molten slag bath
for submerged combustion to supply energy and to increase the stirring to promote
reactions in the molten metal bath. Hence, acute process phenomena like primary
combustion, energy transfer, feed material dissolution, slag-metal reaction, take place
in the severely agitated slag layer. The addition of post combustion air with the lance
shroud system is used to maximize the recovery of energy available in the system. The
furnaces are tightly sealed and operate under a slightly negative pressure to eliminate
fugitive emissions to the environment. Feed material is supplied through a sealed port
in the top of the furnace. The Outotec TSL furnace treats a wide range of feed
materials [10]:
copper, nickel, lead, tin and polymetallic concentrates
copper and lead secondaries
zinc bearing residues
various waste materials and ferrous feedstock
The lance is suitable for operation under oxidizing, neutral, or reducing conditions to
provide a better control of the slag chemistry at the lance tip and gas-rise region of the
slag bath. The slag used for nonferrous processing applications, generally are solutions
of the oxides of iron, calcium, silicon, and aluminum. The composition is controlled
primarily to remove the unwanted components in the feed to the furnace, with fluxing
employed to provide the required viscosity at the chosen temperature of operation [4].
The vertically positioned lance consists of a series of concentric steel pipes through
which air, oxygen and fuel are delivered to the molten bath. The lance submergence
causes a high rate of heat and mass transfer between the molten bath and the slag
layer. The depth of submergence can also be varied which provides a better control of
the furnace to operate in different furnace condition. There is also a flexibility to use
different types of fuel. Another important factor that needs to be mentioned here is that
the lances are non consumable. During operation, the outer portion of the lance that
34
will be submerged is coated with a solidified slag layer before lowering into the slag
bath and then is operated with a submergence of about 100 to 500 mm in a bath of 800
to 2000 mm depth, depending on the requirements of the application [4]. The lance tip
operates submerged under the slag with lance material protected by the frozen slag
layer. This frozen slag layer is induced by the cooling effect of a swirled gas flow in
an outer annulus of the lance [22].
Figure 2-1 shows a diagram of the Ausmelt TSL furnace:
Figure 2-1: A cutaway schematic diagram of the typical Outotec Ausmelt TSL furnace
[12]
35
The reactor can either be used as primary smelters or for slag or waste treatment in
production of zinc, tin, copper, nickel, lead, platinum-group metals, and aluminum.
The furnace possesses five main reaction regions described by Floyd [4],
The combustion region at the lance tip.
The gas-rise region above the lance tip, where gases generated and any solid
un-reacted feed material at the lance tip further react with each other and the
surrounding and entrained slag.
The splash-cascade region, where ejected liquid slag ejected falls back into the
slag bath.
The post combustion region, where air and/or oxygen is injected into the
splash region of the gas space above the bath.
The bath region significantly beneath the lance-tip level, where slag and metal
phases are relatively quiescent compared with the violently agitated top region
of the bath.
There are other features and facilities of TSL furnace that are in common with other
furnace systems—flue off take, various ports, tap holes or tapping weirs, refractory or
cooled containment systems, etc. Variable level of submergence of the lance in the
bath allows the operator to control the degree of turbulence in the bath and the extent
of splashing of the slag cascade above the bath.
Details of the zinc extraction process by available TSL technique, future possible TSL
methodology and other available technologies from zinc concentrates, residues and
slags are described in the later part of this chapter.
36
2.4 History of Zinc
Zinc, also known as spelter, with atomic number 30, is the first element of group 12 in
periodic table. The word “zinc” is derived from the Persian word ‘sing’ meaning
stone. Before discovering zinc in its metallic form, its ores were used for making brass
and zinc compounds were used for healing wounds and sore eyes. In 13th
Century,
manufacture of zinc oxide was described by Marco Polo in Persia to serve for medical
purposes. Zinc is believed to be first discovered in India in the metallic form by 1374,
which was the eighth known metal at that time [23].
The Hindus described how the new “tin-like” metal was made by indirectly heating
calamine with organic matter in a covered crucible fitted with a condenser in the book
- Rasaratnassamuchchaya in fourteenth century [24]. Zinc vapor was evolved and
the vapor was air cooled in the condenser located below the refractory crucible (as
shown in Figure 2-2).
Figure 2-2: Schematic representation of Indian method for producing zinc [23]
The Chinese learnt about the manufacturing process in the 17th century. From India,
zinc manufacture moved to China where it developed as an industry to supply the
needs of brass manufacture. The Europeans came to know about the existence of the
metal much later, around the end of the 16th century. The first zinc smelter developed
in Europe was in 1743, at Bristol, in the UK. In that smelter, a vertical retort procedure
for zinc extraction was adopted, proposed by an Englishman William Champion
(1709-1789) [24] (as shown in Figure 2-3). Based on that procedure, in 1743 a zinc
37
smelter was established at Bristol in the United Kingdom. A charge of calamine and
carbon was sealed into a clay crucible having a hole in the bottom. This was luted onto
an iron tube extending below the crucible furnace into a cool chamber below. The
closed end of the iron tube sat in a tub of water and it was here that the metallic zinc
was collected (shown in Figure 2-3). The distillation took a total of about 70 hours to
yield 400 kg of metal from all 6 crucibles positioned in the furnace. An annual
production rate of 200 tons was suggested for the works at that time [24].
Figure 2-3: William Champion’s Zinc smelting furnace [23]
In 1758, William’s brother, John, patented the calcination of zinc sulfide to oxide for
use in the retort process, thereby laying the foundation for the commercial zinc
practice which continued well into the twentieth century. The English zinc industry
was concentrated in Bristol and Swansea [24].
A major breakthrough in the technological development of zinc extraction was made
by a German scientist Johann Ruberg (1751-1807), who built the first zinc smelting
works in Wessola in Upper Silesia in 1798 incorporating a horizontal retort process.
Improved fuel efficiency and convenient charging and discharging capabilities are
some of the advantages of the horizontal retort process. The initial raw material used
was zinc galmei (calamine), a by-product of lead and silver production. Later, it
became possible to produce zinc directly from smithsonite, an easily smelted ore. This
was shortly followed by the use of zinc blend, which had first to be converted into the
38
oxide by roasting. After this development, other smelting works were soon erected in
Silesia near the deposits, in the areas around Liège in Belgium, in Aachen, in the
Rhineland and Ruhr regions in Germany. The first Belgian plant was built by Jean-
Jacques Daniel Dony (1759-1819) in 1805 and also used horizontal retorts but of
slightly different design. Zinc production in the United States started in 1850 using the
Belgium process and soon became the largest in the world. In 1907, world production
was 737,500 tons of which the USA contributed 31%, Germany 28%, Belgium 21%,
United Kingdom 8%, and all other countries 12% [24].
Sheet production had begun soon after finding out the exceptional resistive capacity of
zinc towards atmospheric corrosion. The possibility of rolling zinc at 100-150°C was
discovered as early as 1805 and the first rolling mill was built in Belgium in 1812.
More such mills were built in Silesia from 1821 onwards. The oldest anticorrosion
process - hot-dip galvanizing, was introduced in 1836 in France. In the United States,
the rich ore deposits led to rapid growth in zinc production in 1840, so that by 1907,
Germany, which had for long been the world’s leading producer of zinc, was left
behind [24].
39
2.5 Physical and chemical properties of zinc
Zinc is a moderately reactive lustrous bluish-white metal placed in group IIB of the
periodic table. It is nonmagnetic in nature and its common oxidation state is +2.
Metallic zinc is brittle and crystalline at ordinary temperature but at 100 °C to 210 °C
(212 °F to 410 °F), it becomes ductile and malleable and can easily be beaten into
various shapes [25]. It creates fumes of zinc oxide while burning in air with a bright
bluish-green flame. It is a barely reactive metal that will combine with oxygen and
other non-metal and also reacts with dilute acids and alkalis. It releases Hydrogen
while reacting with dilute acids. It is the 27th
most commonly found element in earth’s
crust and it is fully recyclable [23]. It can be recycled indefinitely without losing its
physical or chemical properties.
Some general physical and atomic properties of zinc are mentioned in the following
Table 2-1:
Table 2-1: Physical and chemical properties of zinc [24]
Phase Solid
Atomic number 30
Isotopes 10
Density 7.14 gm. cm-3
Melting point 419.530 C (692.53
0 F)
Boiling point 9070 C (1180
0 F)
Heat of fusion 7.32 kJ mol-1
Heat of vaporization 123.6 kJ mol-1
Specific heat capacity (at 250 C) 25.470 J mol
-1 K
-1
Crystal structure Hexagonal
Oxidation state +2, +1 and 0
Magnetic ordering Diamagnetic
Electrical resistivity (20 °C) 59.0 nΩ·m
Thermal conductivity (300 K) 116 W·m−1
·K−1
Thermal expansion (25 °C) 30.2 µm·m−1
·K−1
40
2.6 Uses of zinc
Over the centuries, zinc as a metal has gone through substantial changes both in the
extraction methods and in the way that it is used. The broad categories of end use
remain much the same, i.e. coatings to protect iron and steel, building/ construction,
automotive/transport, household appliance, fittings, toys, zinc alloy castings, sheet for
building applications and a range of chemical applications. Figure 2-4 and Figure 2-5
are two statistical graphs showing the zinc demand by primary use and by auxiliary
use. Primary uses of zinc after production includes corrosion protection for steel
(galvanizing, zinc thermal spraying, electroplating, zinc rich paints), die casting and
gravity casting, brass, aluminium alloys, magnesium alloys, batteries, rolled zinc
sheets and zinc oxide, zinc stearate and other zinc compounds for chemical and
pharmaceutical industries.
Figure 2-4: Zinc demand by First use in 2005 estimates [26]
41
Figure 2-5: Zinc demand by End use in 2003 estimates [26]
Figure 2-6: Zinc consumption in the World during 1960-2005 (kilo tonne) [26]
Figure 2-6 shows the World demand for zinc increased from 3000 to 11000 kilo
tonne/year during 1960 to 2005. Some of the major uses of zinc are summarized
below:
42
Protecting Steel
Protecting steel from corrosion by metallurgically bonding to steel is zinc’s major
contribution. Zinc increases the durability of steel by protecting it against corrosion.
Zinc does so by two ways – by providing a physical barrier and cathodic protection.
When iron and zinc are both exposed to corrosive medium, they constitute an
electrolytic cell where layer of coated zinc act as cathode, and zinc is attacked
preferentially since its reduction potential of -0.76 is lower than that of iron, which is -
0.41 [27]. The process of metallurgical bonding between steel and zinc is known as
hot dip galvanizing. Some noticeable benefits of zinc coated steel are long service life,
low maintenance cost and minimal service interruption.
Other Coatings
Other than galvanizing, there are some more methods of zinc coating to protect
fabricated steel or sheets of steel such as electroplating (also known as electro
galvanizing), flame sprayed coatings, sherardising, mechanical plating and using zinc
rich paints. The use of zinc in paint industry is also increased in last couple of years.
Improved coatings have encouraged the development of new applications. For
example, galvanized steel with a fine surface finish is used to produce the parts of car
bodies that are vulnerable to corrosion. Such applications have been cited as a major
factor in the market for zinc [26].
Human Health
Zinc is an essential element for human health. For proper functioning of the immune
system, digestion, reproduction, taste and smell and many other natural process,
adequate daily intake of Zinc is essential. Zinc deficiency is now recognized as one of
the important risks to human health and is one of the leading causes of illness and
disease in low- income countries [25].
In Alloy making
In alloy making industry (like brass), zinc plays a noteworthy contribution. Other than
brass, there are series of alloys for coatings have been produced recently. Some of
43
those are Galvalume and Galfan. Galvalume consists of about 55% aluminium and
45% zinc with a small amount of silicon. It is being used extensively around the world
as it has better atmospheric corrosion resistance than pure zinc. Galfan is a zinc and
5% aluminium alloy containing small amounts of rare earth metals which has a
substantial and growing niche market in which its properties are valuable. Its
corrosion resistance is better than that of zinc and it retains some cathodic protection
capability [26].
Zinc castings
Another growth area is that of zinc castings which is based on new alloys and new
technology. Zinc casting now has a family of alloys - the specifier can choose the
alloy and casting process most suited to their product. Zinc alloy castings are unique,
particularly when produced by the pressure die casting process. They can be made to
extremely close tolerances, with excellent surface finish, have a range of useful
mechanical properties (especially ductility) and can receive a wide range of applied
finishes. As a result, zinc castings find a range of applications from automobiles to zip
fasteners. The development of new alloys and dramatic improvements in process
control enabled zinc castings to hold their own in many areas, particularly where
strength and applied finishes are required [26].
In medicine industry
Zinc is also used in medicine industry as it is responsible for the function of about 60
enzymes in human body. Besides oral medicine, zinc is also used in making ointments
for different dermatological treatment [26].
Energy conservation
Zinc is also a source of energy. Zinc batteries have been used for many decades. Zinc
gives a good combination of physical and electrochemical properties. Zinc is a good
reducing agent and it can produce high cell voltage. It has an excellent power density
and is one of the most stable metals in aqueous electrolyte solutions [26].
Zinc in Space:
44
United States National Aeronautics and Space Administration (NASA) scientists were
looking for a coating that could withstand the extreme temperature of space travel.
Then they discovered zinc-oxide as a fruitful solution. Researchers developed a zinc
based coating capable of withstanding thermal cycling between 180o
C and -180o
C
and the bombardment of ultraviolet exposure. Zinc oxide coating is now routinely
used in protecting components of spacecraft [25].
Besides all the uses mentioned above, zinc oxide is used in the manufacture of paints,
rubber products, cosmetics, pharmaceuticals, floor coverings, plastics, printing inks,
soaps, textiles, electrical equipments etc. Zinc has also the phosphorescent property
which makes it possible to glow in the dark. By virtue of this property, it is also a key
ingredient in making X-ray, TV screens [26].
45
2.7 Extraction Methods
Zinc can either be extracted from mined ores or metallurgical waste (discarded slags)
containing significant amount of zinc. Based on the extraction methods, zinc
production can be broadly categorised into two types, namely
Primary zinc production
Slag fuming/ Secondary zinc production
2.7.1 Primary zinc production
Primary zinc production refers to extracting zinc from ores. Ores containing zinc are
widespread geologically and geographically. Zinc ores typically may contain 3 to 11%
zinc in association with cadmium, lead, copper, gold, silver, iron as well as some other
minor elements [25]. Rarely is the ore, as mined, rich enough to be used directly by
pyro or hydrometallurgical operation. After extracting the ore from mines, it needs to
be concentrated before it goes for further processing. The concentration process of
zinc ores is accomplished at or near the mine by crushing, grinding, and floatation
process. Concentrated zinc ores usually contain 55% zinc.
Zinc is found primarily in earth’s crust as zinc sulphide (ZnS) or zinc blende. Some
ores are found as [25]
Calamine (ZnCO3) – 67% Zn
Hemimorphite or (Zn4Si2O7(OH)2.H2O). – 54.2% Zn
Zincite (ZnO)
Willemite (Zn2SiO4) – 58.5% Zn
Once concentrated, the zinc ores is transferred to smelters for the production of zinc or
zinc oxide. Reduction of zinc sulphide concentrates to metallic zinc is accomplished
through either electrolytic deposition from a sulfate solution or by distillation in retorts
or furnaces. Both of these methods begin with the elimination of most of the sulphur
in the concentrate through a roasting process. Roasting is high temperature oxidation
of zinc sulphide concentrates usually carried out at 700~800o C with air blow through
46
which it converts to an impure zinc oxide known as calcine. This is a solid-gas
reaction and the reaction is exothermic which increases the temperature up to 1000o C.
At a temperature of around 950°C, oxidisation of the zinc, iron and sulphur occurs.
The sulphur is collected as SO2 and is used to make sulphuric acid (H2SO4) - a
commercial by-product. Roaster types include multiple-hearth, suspension, or
fluidized bed. The following reactions occur during roasting process,
2
SOZnO22
O3ZnS2 +→+ 2.1
3SO2
2O
2SO2 →+ 2.2
In a multiple-hearth roaster, the concentrate drops through a series of 9 or more
hearths stacked inside a brick-lined cylindrical column (see Figure 2.7). As the feed
concentrate drops through the furnace, it is first dried by the hot gases passing through
the hearths and then oxidized to produce calcine. It usually operates at atmospheric
pressure and at about 690°C (1300°F) temperature. Operating time depends upon the
composition of concentrate and the amount of the sulphur removal required. Multiple
hearth roasters have the capability of producing a high-purity calcine.
Figure 2-7: Schematic diagram of Multiple Hearth Roaster (Image taken from
www.energytek.com.tw)
47
Suspension roasters are also un-pressurized, but operate at a relatively higher
temperature of about 980°C (1800°F) than the multiple hearth roasters. In this type of
roaster, the concentrates are blown into a combustion chamber very similar to that of a
pulverized coal furnace. The roaster consists of a refractory-lined cylindrical steel
shell, with a large combustion space at the top and 2 to 4 hearths in the lower portion,
similar to those of a multiple hearth furnace.
Finely ground sulphide concentrates are suspended and oxidized in a feedstock bed
supported on an air column in a fluidised bed roaster. Both in the suspension roaster
and in the fluidised bed roaster, the reaction rates for desulphurization are more rapid
than in the older multiple-hearth processes. Fluidized-bed roasters operate under a
pressure slightly lower than atmospheric and at temperatures averaging 1000°C
(1800°F). In the fluidized-bed process, no additional fuel is required after ignition has
been achieved. The major advantages of this roaster are greater throughput capacities
and greater sulphur removal capabilities.
Figure 2-8: Schematic diagram of a fluidised bed roaster [26]
Roasted calcine is the raw material for either the electrolytic process or the
pyrometallurgical process to get the final product. Thus, the final product of the zinc
companies, which is the slab zinc, follows two different routes:
48
Hydrometallurgical route, and
Pyrometallurgical route
2.7.1.1 HYDROMETALLURGICAL/ ELECTROLYTIC ROUTE
De-sulphurized calcine from the roaster follows three basic steps in the electrolytic
processing, namely:
a. Leaching
b. Purification
c. Electrolysis
c. Leaching
Leaching is a very common technique in extractive metallurgy to make soluble salts of
metal. Leaching for roasted calcine is carried out by two steps known as double
leaching. In the first step, the calcine, after being reduced to powder form, are leached
with neutral or slightly acidic solution with the liquid passing counter-current to the
flow of calcine, which produce zinc sulphate from a portion of calcine. In this process,
only a portion of the zinc oxide enters into solution.
ZnO + H2SO4 = ZnSO4 + H2O 2.3
The remaining calcine is then leached in a strong acidic solution which dissolves the
remainder of the zinc oxide, along with metallic impurities, such as arsenic, antimony,
cobalt, germanium, nickel, and thallium. Insoluble zinc ferrite, formed during
concentrate roasting by the reaction of iron with zinc, remains in the leach residue,
along with lead and silver. Insoluble zinc ferrite, formed during concentrate roasting
by the reaction of iron with zinc, remains in the leach residue, along with lead and
silver. Lead and silver are then typically shipped to a lead smelter for recovery, while
the zinc is extracted from the zinc ferrite to increase recovery efficiency [28].
b. Purification
The zinc sulphate solution, before proceeding to the electrowinning process, should
follow the purification steps to maximise zinc removal during electrolysis. Presence of
49
impurities in the electrolytic solution may enhance the hydrogen production instead of
zinc metal. The purification process is usually carried out in a large agitated tank to
remove the metallic impurities which would otherwise intercede with zinc deposition
process during electrolysis. After purification, concentrations of these impurities are
limited to less than 0.05 milligram per liter (4 x 10-7 pounds per gallon). A number of
reagents are added in a sequence of steps to enforce this job. The solution is
neutralised and the precipitates i.e. metallic co-products together with any unwanted
contaminants, are removed by filtration. The process takes place at temperatures
ranging from 40 to 85°C (104 to 185°F), and pressures ranging from atmospheric to
240 kilopascals (kPa) (2.4 atmospheres) [28].
c. Electrolysis
The refined solution is then electrolyzed in a cell fitted with a lead anode and an
aluminum cathode. The process is also known as electro-winning. The process works
by passing an electric current through the solution in a series of cells. This causes the
zinc to deposits on the cathodes (aluminum sheets) and oxygen to form at the anodes.
Sulfuric acid is also formed in the process and reused in the leaching process. The
cathode is immersed in the solution for 24 – 48 hours during which time the zinc is
deposited on the aluminum cathode from which it is subsequently stripped [25].
Figure 2.9 shows a schematic diagram of electrolytic cell. Electrolytic zinc smelters
contain as many as several hundred cells. A portion of the electrical energy is
converted into heat, which increases the temperature of the electrolyte. Electrolytic
cells operate at temperature ranges from 30 to 35 °C (86 to 95 °F) and at atmospheric
pressure. A portion of the electrolyte is continuously circulated through the cooling
towers both to cool and concentrate the electrolyte through evaporation of water. The
cooled and concentrated electrolyte is then recycled to the cells. This process accounts
for approximately one-third of all the energy usage when smelting zinc [25].
50
Figure 2-9: Schematic diagram of an electrolytic cell
There are two common processes for electrowinning the metal: the low current density
process, and the Tainton high current density process. The former uses a 10% sulfuric
acid solution as the electrolyte, with current density of 270–325 amperes per square
meter. The latter uses 22–28% sulfuric acid solution as the electrolyte with a current
density of about 1,000 amperes per square meter. The latter gives better purity and has
higher production capacity per volume of electrolyte, but has the disadvantage of
running hotter and being more corrosive to the vessel in which it is done. In either of
the electrolytic processes, each metric ton of zinc production expends about
3,900 kW·h (14 GJ) of electric power [25].
The final process takes place in the foundry where the zinc takes its final form, either
as a metal with a purity of 99.95% or as an alloy together with copper, aluminium or
magnesium. The zinc is cast into various forms (normally ingots or plates) with
weights from 9 kg to 4 tons.
2.7.1.2 PYROMETALLURGICAL ROUTE
High reaction rate resulting in high production rate of zinc and ability to accommodate
a wide variety of zinc bearing materials as feed are the advantageous factors of
pyrometallurgical route over the electrolytic process. On the contrary, downgraded
purity in final product (around 98%) is the major drawback of this route. Basic
working principle of this route is reducing the calcine by carbon in retort. There are
also several pyrometallurgical processes that reduce zinc oxide using carbon, then
distil the metallic zinc from the resulting mix in an atmosphere of carbon monoxide.
The four types of commercial pyrometallurgical processes are the Belgian-type
51
horizontal retort process, the New Jersey Zinc continuous vertical-retort process, the
blast furnace process and the St. Joseph Minerals Corporation's (electrothermic)
process [25]. The reduction process gives the zinc as vapor from which zinc metal is
obtained by condensing. A portion of this metal is used in alloying purposes and from
the other portion pure zinc is obtained through redistilation. Basic reaction occurring
in carbon retort is,
The electrothermic distillation retort process was developed in USA by the St. Joe
Minerals Corporation in 1930 [25]. Electrothermic processing of calcine begins with a
downdraft sintering operation, in which grate pallets are joined to form a continuous
conveyor system. Combustion air is drawn down through the conveyor, and impurities
such as lead, cadmium, and halides in the sinter feed are driven off and collected in a
bag filter. The product sinter typically includes 48% zinc, 8% iron, 5% aluminium, 4%
silicon, 2.5% calcium, and smaller quantities of magnesium, lead, and other metals.
Product sinter and, possibly, secondary zinc materials are charged with coke to an
electric retort furnace. The charge moves downward from a rotary feeder in the
furnace top into a refractory-lined vertical cylinder. Paired graphite electrodes
protrude from the top and bottom of this cylinder, producing a current flow. The coke
serves to provide electrical resistance, producing heat and generating the carbon
monoxide required for the reduction process. Temperatures of 1400°C (2600°F) are
attained, immediately vaporizing zinc oxides according to the following reaction [28]:
2CO)vapor(ZnCOZnO +→+ 2.4
The zinc vapour and carbon dioxide pass to a vacuum condenser, where zinc is
recovered by bubbling through a molten zinc bath. Over 95 percent of the zinc vapour
leaving the retort is condensed to liquid zinc. The carbon dioxide is regenerated with
carbon, and the carbon monoxide is recycled back to the retort furnace [28].
The blast furnace process was developed by the Imperial Smelting Corporation at
Avonmouth, England. The process starts by charging solid sinter and heated coke into
the top of the blast furnace. Preheated air at 190 to 1,050 °C (370 to 1,920 °F) is
blown into the bottom of the furnace. Zinc vapour and sulphides leave through the top
52
and enter the condenser. Slag and lead collect at the bottom of the furnace and are
tapped off regularly. The zinc is scrubbed from the vapour in the condenser via liquid
lead. The liquid zinc is separated from the lead in the cooling circuit [29].
The New Jersey Zinc process begins by roasting concentrates that are mixed with coal
and briquetted in two stages. The briquettes are then heated in an autogenous coker at
700 °C (1,292 °F) and then charged into the retort. There are three reasons to briquette
the calcine: to ensure free downward movement of the charge; to permit heat transfer
across a practical size cross-section; to allow adequate porosity for the passage of
reduced zinc vapor to the top of the retort. The reduced zinc vapor that is collected at
the top of the retort is then condensed to a liquid [29].
53
A generalised process diagram illustrating primary zinc production is shown in the
following flowchart:
Figure 2.10: Generalised flow diagram of primary zinc production showing two
different routes (Hydrometallurgy and Pyrometallurgy)
Alloying
purposes
Purifying
additives
Crushing, Grinding
and Floatation
Zinc
Ore
Zinc
concentrate
Calcine
ZnO (<1%S)
Leaching
Purification
Electrolysis
Cathode zinc
Melting and
casting
Zinc slab
Sintering
Reducing by
carbon in Retort
Casting
Impure
Zinc
Sulfuric
acid
Sulfuric
acid
Roasting
Acid
plant SO2
Dust and
fume
Redistillation
Pure Zinc
Coke
Silica
54
2.7.2 Slag Fuming/ Secondary zinc production
Slag is a pervasive by product of any pyrometallurgical process, also often known as
metallurgical waste. Three premier sources of slag in the context of industrial practice
are, firstly the gangue material introduced with the concentrate of ore, secondly from
the fluxes deliberately added during the smelting operation, and thirdly by oxidation
of the melt, be it metal or matte [30]. In general, slag is a mixture of molten oxides
which may also contain sulphides and halides in smaller quantity [30].
Significant metal values have been confined as metal oxides in the slag produced by
the smelting industries. Zinc in the slag from the lead-zinc blast furnace or lead blast
furnace, however if its content is lower than 10% or so, is not always recovered [31].
Recovery of these metal values before the slag are finally discarded has got
noteworthy interest of metal producers. For example, the primary production of zinc
and lead produces slag or residues, which contain significant amounts of zinc in the
form of ZnO. The zinc content in the slag depends largely on the type of concentrate
and residue materials, method of extraction and equipment used, which can be
recovered from the slag in its molten state in the form of fume by using a suitable
reducing agent. The process is generally known as slag fuming. Slag fuming is an
important secondary unit operation which is in the extraction of non-ferrous metals
and has been used since the 1930s to recover zinc from lead blast furnace slag [32]. It
is mostly a batch process, in which a reducing mixture of air and pulverized coal or
any other reducing agent is injected into the molten slag, however, Korea Zinc also
fume in continuously operating furnaces [33]. The coal-air mixture reduces the zinc
oxide from the slag to metallic zinc vapour.
Hence, slag fuming can be mentioned as one of the important mode of ‘slag cleaning’.
In metallurgical processing, the term ‘slag cleaning’ refers to the process of recovering
valuable metals from the slag phase. There are two other ways of slag cleaning. One is
‘settling’ by which contained metals or metal sulphides rise or sink to form a distinct
layer from which the slag can be separated [30]. The other one is slag milling and
floatation. This involves grinding the cooled slag and separating metallic and sulphide
particles by a standard rougher-cleaner type circuit [30]. But the most imperious of all
55
technique is the slag fuming process of slag cleaning i.e. recovering the valuable
metals.
The earliest reported experimental work on zinc fuming was conducted in Australia by
Sulphide Corporation at Cockle Creek between 1906 and 1920 [21]. The process has
been operative since 1930’s for recovering zinc from lead blast furnace slag. The
process operates between 1423 and 1573 K. The overall reactions occurring in the
bath are,
( ) ( ) ( )ggslag COZnCZnO +→+ 2.5
( ) COCOC coal 22 →+ 2.6
( ) ( ) ( ) ( )g2ggslag COZnCOZnO +→+ 2.7
Figure 2-10: Schematic illustration of the slag fuming process
The overall chemical reaction in the bath is thought to be controlled by the supply of
carbon to the slag-gas interface [19-21]. The main reaction (2.7) is endothermic and
combustion of fuel in the bath supplies the necessary heat. The vaporized zinc
oxidizes when it comes in contact with the air above the zinc bath. The zinc oxide
fume is subsequently collected in the bag house. Figure 2.11 shows a schematic
diagram of the zinc slag fuming process in the case of a top submerged lance (TSL)
56
smelting furnace. Details of the zinc slag fuming process from kinetic and fluid
dynamic perspective will be discussed in the later part of this thesis.
2.7.2.1 CONVENTIONAL SLAG FUMING OPERATION
Conventional slag fuming process on a rectangular furnace is a primeval process of
zinc fuming from slags. Commercial development of the process was made by the
Anaconda Copper Mining Co. and Consolidated Mining and Smelting Co. in the
1920’s [21]. The process usually carried out in a rectangular water jacketed furnace on
batch basis as shown on the schematic cross sectional view in Figure 2.12.
Approximately 50 tonnes of charge (either molten slag or solid slag) is fed to the
furnace at the beginning of each cycle. Two opposing sets of submerged tuyeres are
employed to inject air and pulverized coal into the molten slag bath to carry out
combustion and reducing reactions. The system usually operated within the
temperature ranges of 1150 to 1300o C (1423 to 1573 K).
Figure 2-11: Schematic of fuming furnace cross section [18]
Within the bath, the reducing agent (pulverized coal – 80% -200 mesh) reduces the
dissolved zinc oxide to metallic zinc vapour in the form of fume which is subsequently
captured in a bag house. The overall reactions occurring in the bath are mentioned in
equations 2.5-2.7. Each fuming cycle also termed as “fuming period”, usually operates
57
for 150 minutes, with additional 30 minutes for charging and tapping. Zinc extraction
takes place only during the fuming period. An integrated slag fuming operating unit is
portrayed in the schematic diagram of Figure 2.13.
Figure 2-12: A schematic diagram of a rectangular tuyere blow conventional slag
fuming operation (Image taken from US Patent by Quarm [22])
As described by Quarm [34], Figure 2.13 shows a schematic diagram of 8 feet x 21
feet slag fuming furnace. The furnace body was built up with water jacketed steel
plates 21, the inner surface of which had a frozen slag layer. The furnace was
connected to a waste heat boiler and then to the bag house by a flue 12. Air supply
comes through line 51, while coal fluidized by a minor part of the air is fed through
line 15 and nozzle 16 inside of each of tuyeres 14. The tuyeres are placed on both of
the opposing side of furnace wall and connected to a shut off valve 19, so that the
malfunctioning tuyeres can be removed for servicing. Through the opening 20, feed is
being fed to the furnace and also to add air to oxidize the fumed zinc into zinc oxide.
58
A tap hole 23 at the bottom of the furnace facilitates the removal of dezinced slag
through a launder 24 into the settling tank 26. Matte droplets are suspended from the
discarded slag in the settling tank 26 which is removed through another tap hole 27 by
the launder 28 into the matte ladle 29. The discarded slag is removed similarly by
another arrangement of launder 31 and slag ladle 32 arrangements.
2.7.2.2 APPLICATION OF TSL TECHNOLOGY IN ZINC PROCESSING
TSL technology is very well established technique for smelting industry operating
successfully around the world by implementing state of the art technologies for
extracting both ferrous and non-ferrous metals. Commercial TSL plants commissioned
around the world for both ferrous and non-ferrous metal processing are reported by
several authors [4, 35-38]. Zinc extraction by TSL furnace has started with pilot plant
studies in 1980s and has progressed to the treatment of close to 800,000 tonnes a year
of zinc bearing feeds in the form of residues and slags [39]. During smelting of zinc-
lead concentrates in Imperial Smelting Process (ISP) and in lead sinter plants such as
in Kivcet or QSL process, significant amount of zinc bearing dust is produced every
year. Because of the volatilization property, zinc, lead and silver can be easily
recovered by fuming process from these discarded slags. Outotec TSL technology has
been successfully adopted for zinc extraction from QSL slag at Onsan Zinc Refinery,
Korea [33, 37] and from ISF slag at Mitsui Mining and Smelting, Hachinohe, Japan
[37]. For processing of lead and zinc bearing intermediate industrial products
(residues, slags), twenty one (21) Outotec Ausmelt TSL furnaces are now in operation,
under design or under construction in South Korea, Japan, India and Australia [40].
Among all, Korea Zinc is the most comprehensive evidence to date as to the success
of TSL technology, comprising an integrated flow sheet of seven interdependent
projects utilising twelve TSL furnaces (10 for specific recovery of zinc) [39]. In 2008,
annual production of zinc by Korea Zinc was 445,000 tonnes, pointing it to a major
producer of zinc in the world [41]. Recently, another two stage zinc extraction process
has been proposed by Hoang et al. [39] from zinc concentrates designated as Direct
Zinc Smelting (DZS).
The zinc fuming plant in Korea Zinc incorporating TSL furnaces was designed to treat
120,000 tpa (tons per year) residue and has commenced its operation in 1995 [33].
59
Floyd et al. [37] describes the detailed working stages and flow diagrams of the zinc
fuming process at Onsan as shown in the Figure 2.14. Choi et al. [33] describes the
detailed working principal, plant description, material handling system, gas handling
system and development phase of the Korea Zinc plant at Onsan. As reported by
Floyd et al. [37], the Korea QSL lead smelter in Onsan, Korea, produces a slag of
typical composition of 20 – 22% Fe, 19 – 22% SiO2, 13 – 16% CaO, 13 – 15% Zn and
<5% Pb.
Figure 2-13: Korea Zinc’s integrated flow sheet using TSL technology to recover zinc
and lead from various slags and residues created during primary zinc and lead
concentrate processing [Hoang et al. [39]]
Figure 2-14: Schematic diagram of the zinc fumer at Onsan, Korea [Choi et al. [33]]
60
Figure 2-13 show the flow sheet of zinc fuming plant and Figure 2-14 show the
schematic diagram of the zinc fumer at Onsan, Korea. The process utilizes two stage
fuming process and uses coal as fuel and reductant and has facilities for oxygen
enrichment to 40% O2 in the lance combustion system [37]. As shown in the Figure
2-14, the first furnace is the smelting furnace and the second furnace is the cleaning
furnace. The cleaning furnace operates at extremely reducing conditions [33].
Magnesite chrome brick is used as the refractory in the furnaces. The cleaning furnace
is designed to have a siphon for continuous tapping of slag and a bottom tap – hole to
extract metal batch wise [33]. The typical operating results of the Korea Zinc, as
reported by Floyd et al. [37], are mentioned below:
Table 2-2: Typical operating results of the zinc fumer at Onsan, Korea. [Floyd et al.
[37]]
Operating Parameters Solid slag feed Liquid slag feed
Smelting rate (tph) 6 12
Temperature (o C) 1250 – 1300
o C 1250 – 1300
o C
Zinc fume % Zn 50 – 55 40 – 45
% Pb 20 – 30 25 – 35
Slag product % Zn 5 – 7 6 – 8
% Pb 0.3 – 0.5 < 1.0
The zinc content in the final slag was 3.5% and the recovery of zinc in the fume oxide
was about 82% [33]. Choi et al. [33] further reported that the furnaces needs to be
operated at higher than 1300o C to lower the final zinc content in the slag less that 1%.
61
The flow circuit of the zinc fuming plant at Onsan, Korea, is shown in the Figure 2-15.
Figure 2-15: Flow circuit of the zinc fuming plant at Onsan, Korea. (Floyd et al. [25])
Lead Concentrates, recycle
streams and fluxes, O2, Air and
coal as fuel and reductant
QSL Furnace Fume to
recycle Lead
Bullion
TSL Furnace
QSL
Slag
Air, O2 and
Coal as fuel
and reductant
Slag granulated
and sent to
cement plant
Waste Heat
Boiler
Bag House
ID Fan
Stack
Fume to Leaching
Zinc
solution to
Electrowin
Pb Residue
to QSL
Furnace
62
Figure 2-16: Flow circuit of ISP Slag Fumer at Hachinohe, Japan. (Floyd et al. [37]]
Sinter, Fume Briquettes, Coke,
Hot Air
ISP Furnace Zinc from
splash
condenser
Lead Bullion
Electric
Furnace
ISP
Slag
Bullion
TSL
Fumer
ID Fan
Stack
Slag granulated and
stockpiled or Used
Speiss
Slag
Evaporative
Cooler
Bag House
63
The slag fuming plant at Hachinohe Smelter, Japan, using TSL furnaces was
commissioned in 1993 for zinc recovery from Imperial Smelting Furnace (ISF) slag.
At that plant, the slag is tapped continuously from the ISP through the fore hearth into
an electric furnace and via launder into the fuming furnace. The slag flows
continuously through the fuming furnace where reducing conditions are supplied by
sub stoichiometric combustion of heavy fuel oil with air and with additions of
reductants such as coke breeze [37]. The flow circuit of the ISP slag fuming circuit at
Hachinohe is shown in Figure 2-16.
As reported by Floyd et al. [37], the typical slag composition of the Hachinohe
Smelter is 37% Fe, 20% SiO2, 14% CaO, 6 – 8% Zn and 0.5 – 1.0 % Pb. The fume is
collected in a baghouse and is returned to the sinter plant as feed material. The final
zinc content in the fumer slag is 3.0%. The typical operating conditions of the
Hachinohe Smelter, as reported by Floyd et al. [37], are mentioned below:
Table 2-3: Typical operating conditions of the Hachinohe Smelter, Japan [Floyd et al.
[37]]
Slag feed rate (tph) 10 – 12
Temperature (o C) 1300 – 1350
o C
Zinc fume % Zn 60
% Pb 12
Slag product % Zn 2 – 3
% Pb 0.1 – 0.3
Moreover, to maximize zinc recovery from primary concentrates, retrofitted TSL
furnaces are installed in a number of plants to extract zinc and other valuable elements
like In and Ge from primary leach residues [39]. A flow sheet of the commercialised
TSL zinc technology to recover zinc from primary leach residues is shown on Figure
2-17.
64
Figure 2-17: Flow circuit of the commercialised TSL zinc technology to recover zinc
from primary leach residues (Image taken from Hoang et al. [12])
More recently Hoang et al [39, 42] proposed a process for direct smelting of zinc from
zinc sulphide concentrates where the roasting process is eliminated by the TSL
smelting process [42]. The technology of Direct Zinc Smelting (DZS) process
involves treating of lower grade unclean concentrates in a combination of
pyrometallurgical recovery of zinc as ZnO fume which further follows the leach and
electrowinning circuit while utilising sulphide sulphur as fuel. Flow circuit of the
conceptual DZS model is presented in Figure 2-18. In the first stage of the two stage
TSL application, sulphide sulphur from the zinc concentrates will be used as energy
carrier and fume 60 – 65% of the zinc content. Molten slag from stage 1 will be used
in stage 2 for further recovery (99%) of zinc to produce final discard benign slag.
65
Figure 2-18: Conceptual flow circuit of the TSL Direct Zinc Smelting technology
(Image taken from Hoang et al. [12])
From the aforementioned discussions, the need for a more fundamental and clear
understanding of the submerged combustion process and detail hydrodynamic
characteristics in addition to chemical kinetics has evolved as obvious. As the primary
focus of this research is the submerged combustion dynamics and detail kinetics of the
zinc slag fuming process, it will be dealt with in the later part of this thesis
extensively.
66
CHAPTER 3
67
3 Literature Review
Since the invention of TSL technology in 1970s numerous studies had been carried
out for fundamental understanding of the process. The aim was to improve the system,
to make it an environmentally friendly, efficient and optimum process in terms of fuel
usage, product quality and waste material. This chapter starts with a comprehensive
literature survey on various experimental and numerical studies carried out on cold
flow top submerged lance gas injection process. An extensive literature survey on top
submerged lance zinc slag fuming process and conventional zinc slag fuming process
is also presented, followed by the research objective of this thesis.
3.1 Cold Model Investigations
Cold model investigations provide the fundamental understanding of the fluid flow
behaviour and other hydrodynamic parameters related to the process. Process
optimization can be achieved through proper manipulation of some hydrodynamic
parameters by cold model investigation. It is treated as the beginning step of the
research that leads to the real plant scale investigation. For gas injection systems in
metallurgical process industries, cold model investigation provides the bath interaction
characteristics due to the injection process, which is impossible to know in real
furnace scenario. Thus cold flow models can play a significant role in understanding
the basic working principle of any process that deals with high temperature robust
combusting environment. Major limitations of the cold model studies are it deals with
isothermal condition, hence unable to provide information on heat transfer phenomena
and combustion behaviour inside the furnace. In addition, if chemical interactions
among the species within the bath are of significant interest, then cold model studies
are not a good solution. Nevertheless, cold flow models are the gateway for research
in any high temperature process.
Cold model studies carried out by the previous researchers are based on both
experimental and numerical techniques. Significant research interest were mostly
based on bath mixing characteristics, formation of bubbles and splashing, effect of
different hydro-dynamic parameters (i. e. lance submergence level, flow rate, lance
68
diameter, lance position, angle of injection). In the subsequent sections, these topics
will be discussed further.
3.1.1 Experimental studies
As discussed in the Introduction (Section 1), gas injection in metallurgical processing
has been in operation since 1800’s. There had been numerous cold modelling studies
on different mode of gas injection systems.
Mazumdar and Guthrie [7] carried out experimental and numerical modelling study on
a 0.3 scale cold flow water model of a 150 ton steelmaking ladle. The experimental
work dealt with some geometrical change of the ladle, like with and without tapered
side walls, and with and without surface baffles around the rising plume. The
numerical modelling study was based on a generalized two dimensional, steady state
computational scheme for predicting flows generated by fully submerged and partially
submerged axi-symmetric gas injection lances.
The authors developed an equation of average plume velocity considering the case of
axi-symmetric gas injection into a cylindrical vessel, based on the previous work of
Sahai and Guthrie [8].
3/1
4/13/131
pR
LQkU β′′= 3.1
Where Up is the average plume velocity, Q is the flow rate, R is ladle radius, L is the
liquid depth and β is a fraction such that 0< β<1.
Morsi et al. [1] developed a similar type of cold flow air-water model of a 150 ton
steel-making ladle as developed by Mazumdar and Guthrie [7]. Morsi et al. [1] used
Laser Doppler Anemometer (LDA) technique to measure the velocity fields inside the
liquid bath. The authors investigated the effect of swirl and non-swirl gas injections
into liquid baths using submerged vertical lances. They reported that swirl gas
injection and two-third lance submergence level promoted better mixing in the bath.
They also examined the applicability of the concept of isotropic turbulence inside the
bath and concluded that it may hold outside the plume region.
69
Similar technique was also used by Taniguchi et al. [43] in a water model experiment
to measure the velocity fields. Taniguchi et al. [43] also measured volumetric
coefficient and free surface fluctuation in a water vessel with 0.145 m radius and 0.2
m height by injecting Nitrogen instead of air. Iguchi et al. [44-51] also carried out
extensive research on cold flow model of both top submerged injection and bottom
injection. Koh and Taylor [52] developed another cold flow air-water model to
measure liquid splashes at the bath surface for different flow rate and different lance
geometries. Jet penetration and liquid splashes in submerged gas injection were also
studied by Igwe et al. [53] through cold flow air-water model. Koria and Singh [54]
carried out another experimental investigation where the influence of different lance
lengths and diameter were studied experimentally on the upstream and downstream
flow properties of gas.
In another experimental study, Morsi et al. [55] investigated the flow field
characteristics within an elliptical liquid bath by using laser diagnostics technique and
high speed photography. The model was designed to reveal the flow behaviour of a
conceptual AusIron furnace design which consists of an elliptical Perspex vessel and
two vertically supported lances. The authors reported that higher flow rate created a
strong recirculation zone at the bottom of the bath. The authors also investigated the
effect of level of submergence for two vertically supported lances and concluded that
lower level of submergence caused a rapid spread of the gas jet at the top section of
the bath and higher submergence level improved the agitation in the bath.
Nilmani and Conochie [56] investigated the effect of viscosity in a cold flow model
top submerged lance experimental rig. The authors also found that the velocity of rise
of gas in the sirosmelt furnace is much higher than in the water model. In their cold
flow experimental study the authors used two gas: air and helium and three liquids:
water, glycerol/water with viscosity of 56 centipoise and glycerol/water with viscosity
of 200 centipoise. The authors reported that increasing liquid viscosity reduces gas
dispersion and with a less dense gas a greater volume flow is required to maintain the
same injection characteristics. Neven et al [57] also used Helium instead of air as the
injected gas into the water in another cold model study. The effect of viscosity of
liquid were also examined through cold flow experimentation by Iguchi et al [47],
where air was injected into a bath containing aqueous glycerol solution through a
70
single hole nozzle installed at the centre of the bottom of an acrylic cylindrical vessel
by means of a compressor. They examined the effects of the viscosity of liquids on the
bubble dispersion and reported that with increasing the viscosity of liquid, bubbles
tended to flock together around the centreline of the vessel and rise after the foregoing
bubble.
3.1.2 Swirl and Non-swirl Investigation
Top submerged lance technology incorporates helical swirler in the annulus area of the
lances to provide a swirling motion into the injected air. Thus, the injected gas jet
through the submerged lances can have either swirling or non-swirling effect. There
have been numerous studies regarding investigation of swirl and non-swirl air
injection into the liquid bath. Nilmani and conochie [56] did some experimental work
on swirl flow investigation and reported that swirler improves the radial dispersion of
gas bubbles, produces finer bubbles and minimises bath slopping and splashing.
Later in the year 1987-88, Dave et al. [58] investigated thoroughly the effect of
constant and variable pitch swirled insert through some experimental work. The
authors investigate the flow characteristics of both fixed and variable pitch inserts in
the case of sirosmelt lance. According to the authors, the swirl flow has the following
benefits over the non-swirl flow:
1. For the same mass flow, swirl flow can provide a higher velocity as compared
to the axial flow.
2. Because of the swirl inserts the fluid has to travel a greater distance through
the annulus which in turn increase the heat transfer rate through the lance to
the outer hot surface and helps to create the protective slag layer outside the
lance in a shorter time.
3. The rotating swirl flow creates a centrifugal force field which has a favourable
convection effect into the molten bath.
The authors subdivided flow behaviour across a given swirler into three regions:
1. Entrance region (flow development)
71
2. Fully developed flow region (constant velocity profile)
3. Decaying flow region
The authors also reported that fully developed flow loss consists of momentum and
frictional pressure loss. They developed two equations for momentum pressure loss
and for frictional pressure loss. The momentum pressure loss equation was given as,
−+
−=∆
c
cbw
iec
g
mA
RR
PA
mP
44
2
3
2
22
111 π
ρρ 3.2
and the frictional pressure loss equation was given as,
θ
ρ
cos
2 2
,
h
whgh
md
VLfP =∆ 3.3
Where helical flow friction factor, hf , can be related to the fully developed helical
Reynolds number and the maximum helical velocity which occurs at the tube radius
wR , can be expressed as,
( )[ ]
z
w
wh VP
RPV
21
22
,
2π+= 3.4
where, average axial velocity,
gc
g
zA
mV
ρ= 3.5
cA represents the flow cross-sectional area of the annulus with an n start helical vane
insert and is given by,
( ) ∫−−=w
cb
R
R
cbwc drntRRA θπ sec22 3.6
Thus by using equation (3.3) and (3.4), the flow losses for fully developed helical flow
in the insert region of a multistart gas injection lance can be estimated for any given
gas flow rate.
Solnordal and Gray [22] also did some experimental work to measure pressure losses
in the lance. The authors investigated swirl decay characteristics, heat transfer
coefficient, pressure losses, heat transfer per unit pumping power and heat transfer
72
mechanism of an operating top submerged lance through experimental work. The
authors reported that helical vane swirler can increase heat transfer coefficient by a
factor of 2.35 times over axial flow values. But swirlers incorporate a great pressure
loss as compared to plain lance and pressure loss due to poor design of the swirler
entrance region contribute up to 80% of the total pressure loss. The authors suggested
using a series of short variable pitch swirlers within the lance, with each swirler
entrance aligned with the bulk flow direction to reduce entrance pressure loss.
Neven et al. [57] reported that the presence of swirler does not have any significant
effect on the bubbling frequency. Shinichiro et al. [59] did some experimental work to
investigate the effect of swirler on formation of fine bubbles. The authors further
proposed an equation to measure the bubble diameter in the bath, which can be written
as,
( ) ( ) 21
31
212
1
=
ρσR
wB 3.7
where w is the tangential velocity, R is the radius of container, σ is surface tension of
molten steel. The authors reported that,
1. An increase in the centrifugal force induced by imparting a swirling motion in
the liquid accelerates in creating fine bubbles.
2. The penetrated volume efficiency of bubbles increases with increasing the
tangential velocity and injected air flow rate.
3. The diameter of bubbles decreases with increasing the tangential velocity.
4. The diameter of bubbles increases with increasing injecting air flow rate.
Iguchi et al. [45] also investigated bubbling phenomena in a cylindrical bath with
centric bottom gas injection by inducing swirl motion. The authors classified swirl
motion as two types depending on the bath depth. First kind of swirl motion is formed
when the bath depth is less or nearly equal to the bath diameter which was caused by
internal forced oscillation due to quasi-periodical bubble formation and in the second
kind bath depth is equal or greater than twice the bath diameter which was formed by
instability of large scale ring vortex enclosing the bubbling jet. The authors also
proposed some empirical correlations for the initiation of swirl motion and swirl
73
period. Later, Iguchi et al. [60] carried out further experiments to investigate the effect
of first kind of swirl motion on the liquid flow characteristics, bubble characteristics,
mass transfer from a solid body immersed in the bath and mixing time of the bath by
using high speed video camera, electro-resistivity probe and Laser Doppler
Velocimeter (LDV). The authors concluded that swirl motion enhanced the mass
transfer coefficient and reduced the mixing time significantly. The authors also
reported that the erosion of vessel wall was also enhanced by swirl motion.
Furthermore, Iguchi et al. [44] investigated the effect of swirl motion at reduced
pressure on the bath surface where several factors such as starting time, period,
amplitude and damping time of swirl motion occurring in a water bath were
experimentally investigated.
Ihira et al. [61] experimentally investigated the effect of multi lance configuration on
the swirl motion and mixing time inside the bath. The authors reported that there are
two types of unsteady motion in the deep water wave regime: swirl and reciprocating
motion.
3.1.3 Formation of Bubbles and Splashing
Formation of bubbles and splashing in high temperature process is an unavoidable
phenomena and key concern of the researchers. Many researchers and operators have
claimed that formation of splash and bubbles is a limiting factor in many gas injected
process. Irons and Guthrie [6] described in detail the formation of bubble in molten
metal bath. Liow et al. [62, 63] investigated thoroughly the dynamics of formation of
splash and discussed the macro scale and micro scale splash formation mechanism and
their effects on metallurgical processes. Liow et al. [63] further reported a number of
sources for formation of splash, which includes:
a. The impingement of drops on liquids
b. Slopping of bath liquid
c. Breakup of gas bubbles at the liquid-gas interface
d. Shearing of liquids by gases, and
e. Impact of packets of fluids on solids
74
The authors [63] studied the splash formation by lances on an industrial scale by
injecting air through the lance into a layer of molten slag in the flash smelting furnace
at the Kalgoorlie Nickel Smelter (KNS). The study shows that the cumulative weight
of the splash collected showed an exponential decrease with an increase in height from
the slag-air interface. They proposed a correlation for the amount of splash collected,
W, at a given height h,
khCeW
−= 3.8
Where, C is the total splash at the free surface, and k is the splash decay constant with
distance.
They also estimated roughly the amount of bubble volume formed at the lance exit by
using following equation,
6.02.1378.1 −= gQV 3.9
Liow et al. [63] reported two different splash formation mechanism from the KNS trial
study. The first mechanism was the formation of the slag sheet around the cavity
formed by the gas injected into the bath. The second mechanism was the Kelvin-
Helmholtz instability mechanism which is responsible for producing finer splash
particles. This mechanism occurs during the break-up of drops by the shearing action
of the gas inside the cavity. The criteria for stability is given by We.Fr<4 where,
( )
( )( )glgl
gl LUWe
ρρρρ
ρρ
+−=
22
22
3.10
gL
UFr
2
= 3.11
where, U is the relative gas to liquid velocity, L characteristic length.
Bubbling frequency and injection dynamics were investigated by Neven et al. [57] by
both cold flow water model and hot combustion testing. The authors validated the
Davidson and Schüler [64] formula for bubbling frequency and bubble volume that is
derived from the force equilibrium between inertial and buoyancy force. Neven et al.
[57] considered that the other two effect in bubbling phenomena - viscosity and
density becomes less important at high flow rates. Force due to inertia results from a
75
fact that a growing gas bubble at the lance tip has to accelerate an amount of liquid
surrounding the bubble. The inertia force corresponds to,
dt
dsV
dt
dF slagI
= ρ..
16
11 3.12
and the buoyancy force,
( )airslagB gVF ρρ −= .. 3.13
where,
ρ = density (kg/m3)
V = volume of the gas bubble (m3)
s = distance between lance tip and centre of the bubble (m)
g = gravitational acceleration (m/s2)
The force equilibrium results in a quite simple formula that gives the bubbling
frequency, freq, as a function of gas flow rate, G:
freq = 2.84 5
1−
G 3.14
Formation and rise of a bubble stream in viscous liquid was investigated by Snabre
and Magnifotcham [65], where they developed a semi-empirical model based on force
balance around the spherical bubble at the instant just previous to detachment,
σFFFFFF idpgb ++=++ 3.15
where,
Buoyancy force, ( )gVF gb ρρ −=
Gas momentum force, 22
4ggag WdF ρ
π= , with
2
4
a
gd
QW
π=
Pressure force, ( )PPdF gap −= 2
4
π
Drag force, *2
2 .4
.2
1dd C
dWF
πρ=
76
Inertial force, γρρ
ρα VF
g
i
+=
Surface tension force, σπσ adF =
Where gW is the gas velocity through the tube, W the average velocity of bubble
expansion, γ the average bubble acceleration, *
dC the average drag coefficient, σ the
surface tension of the liquid, gP the gas pressure in the bubble and P the average liquid
pressure.
Koh and Taylor [52] studied the splashing at the free surface of the sirosmelt bath in
an air-water model using an electrical conductivity method. They measured the
average splash height and total splash volume at various air flow rate for a range of
lance geometries. The authors investigated three modes of gas injection system:
bottom injection, injection through plain lance and injection through swirled lance and
reported that the splash height and volume produced by the top submerged swirled gas
injection system are the smallest. The authors also reported that splash height and
volume are significantly decreased when the swirler is located at the tip of the lance.
The similar findings were also reported by Igwe et al. [53], where they did similar
type of cold flow nitrogen-water experiments to measure the jet penetration and liquid
splashes in submerged gas injection systems. They studied the effect of lance design,
nozzle dimensions, gas driving pressure and liquid densities on jet penetration, bubble
dispersion and liquid splashes. The authors reported that penetration of the jet is a
function of the term, mFr , which is a modification of the jet Froude number,
( )og
g
mdg
vFr
ρρ
ρ
−=
1
2
3.16
Where 1ρ is the liquid phase density, gρ is the density of gas, g is the gravitational
constant, v is gas flow velocity and do is the orifice diameter. The higher is the
modified Froude number, the greater is the horizontal component of the jet
penetration.
Iguchi et al. [51] measured the vertical migration distance of the bubbles from the
lance exit using electro resistivity probe and developed an empirical equation to
77
calculate the vertical migration distance of the air jet for top submerged lance gas
injection which also depends on air flow rate. The empirical equation proposed by
Iguchi et al. [51] for vertical migration distance of the injected air into the liquid bath
can be expressed as,
31
1.4 mnv FrdL = , 2< Frm < 6x103
3.17
where Lv is the vertical migration distance of the injected air, dn is the nozzle inner
diameter at the exit and Frm is the modified Froude number which can be expressed as,
5
2
nL
gg
mgd
QFr
ρ
ρ= 3.18
where gρ is the density of gas, Lρ is the density of liquid, Qg is the gas flow rate and
g is the acceleration due to gravity. In this equation dn is used as the inner diameter of
the nozzle.
Iguchi et al. [48] investigated the mechanism of heat transfer between bubbles and
liquid and the effect of heat transfer on the formation of bubbles and the rising
characteristics of bubbles, using air and helium in a water model. In real conditions,
gas is blown into the molten bath at a temperature much lower than the liquid
temperature and hence heat transfer takes place between bubbles and liquids. So, in
the experimental investigation of by Iguchi et al. [48], air and helium were cooled to -
110o
C and blown into water to investigate the effect of temperature difference and
heat transfer on the formation of bubbles and the rising characteristics of the bubbling
jets. The authors concluded that bubbles tended to spread in the radial direction rather
than upward and gas hold-up for a cold gas injection became relatively large compared
with the ambient gas injections. Komarov and Sano [66] carried out similar type of
experiments to reveal bubble behaviour and heat transfer in preheated gas injection
into liquid baths. In that study, preheated gas (N2, He, Ar-He mixture) were injected
into the bath of volatile (water) and non-volatile (ethylene glycol, methyl carbitol and
ethylene glycol-glycerine mixture) liquid. The results showed that the size of the
rising bubbles decreases as the surface tension decreases, so the bubble diameter
decreases when preheated gas is injected.
78
Kumagai and Iguchi [67] carried out further experimental study of instability
phenomena due to formation of splash at the bath surface in top lance injection. The
authors reported that instability phenomena of the bubbling jet appear when the gas
flow rate is higher than a certain critical value. The phenomena are strongly dependent
on the penetration depth of the bubbling jet into the bath. Iguchi et al. [68] studied the
behaviour of an air-water bubbling jet subjected to Coanda effect. In the case of
eccentric lance gas injection, jet or bubble plume moving near a wall is pulled towards
the wall, attaches to it and finally moves along it. This effect is known as the Coanda
effect, which is caused due to the pressure difference along an eccentrically
submerged lance. In that study, Iguchi et al. [68] measured bubble frequency, gas
holdup, mean bubble rising velocity and mean bubble cord length by using two-needle
electro-resistivity probe. The results showed that, these quantities were hardly
influenced by the side wall of the vessel and the vertical bubbling jet is free from the
Coanda effect.
Iguchi with some other co-workers [46] developed another electro-resistivity probe
being able to continuously measure bubble characteristics in a molten iron bath
agitated by gas injection for two hours. The authors measured axial and radial
distribution of gas holdup, bubble frequency, mean bubble rising velocity and mean
bubble diameter in a molten iron bath at a temperature of 1250oC and compared the
measured values with the previous cold model experimental data and empirical
equations proposed by the same author. Mean bubble rising velocity and bubble
diameter are also measured by Diaz et al. [69]. The authors used a rotary L-shaped
lance to study bubble behaviour and absorption rate of gas injected into liquid bath.
Bubble diameter and bubble rising velocity were measured by using high speed video
recording system and absorption rate of injected CO2 gas in aqueous NaOH solution
was measured by using a pH meter. The authors reported that the lance rotation leads
to decrease in average bubble diameter and the rising velocity as compared with the
stationary lance and also the gas absorption rate for rotary lances is larger than that for
the stationary lance.
79
3.1.4 Bath mixing characteristics
Investigation on bath mixing characteristics has also drawn significant attention of the
researchers for different gas injection system. Uniform and rigorous bath mixing
accelerates the chemical reactions within the bath. There had been numerous studies
on the bath mixing characteristics, mostly based on cold model investigation.
Rankin et al. [70] carried out extensive research based on experimental and numerical
studies to reveal the bath mixing characteristics for top submerged lance injection
systems. The authors used PHOENICS computer code to predict the velocity field and
turbulence mixing inside the bath and carried out some pilot plant trials and air-water
model experiments to validate the PHOENICS simulation results.
Singh and Ghosh [71] conducted experimental studies to determine the mixing time
and mass transfer rates between slag and metals in an LBE (Lance Bubbling
Equilibrium) model by means of conductivity measurement and chemical
decolourization methods. The authors reported that mixing time decreases as the
number of porous plugs increases and volumetric mass transfer coefficient for LBE
method is higher than that for top blowing and bottom blowing process.
Investigation on bath mixing by coaxial gas injection through a nozzle fitted at the
bottom of the bath is carried out by Krishna and Mehrotra [72]. The authors described
the re-circulatory flow pattern generation mechanism in such baths and the effect of
process variables on the two phase plume shape and dimensions. Experimental
techniques of characterising the mixing of liquids in these baths by determining
velocity fields or by estimating mixing parameters such as mixing time, degree of
mixing, etc. was also described by the authors. They also proposed a correlation
between mixing time and various operating variables, based on cold model studies.
Diaz et al. [73] used L-shaped lances to reveal mixing characteristics of gas injection
for rotary lances. The authors measured the mixing time by means of the electrical
conductivity method and used KCL as the tracer which was added to the liquid bath
through a funnel. The authors concluded that mixing time is a function of a number of
variables which are bath depth, gas flow rate, lance rotation speed and elbow length
and proposed the following empirical correlation for mixing time , mixt ,
80
∞mixt 21.216.039.039.044.0 DHLRQ −−−− 3.19
where D is the vessel diameter, H bath depth, L elbow length, R rotation speed, Q gas
flow rate.
Later, Krishnakumar et al. [74] also calculated mixing time by using electrical
conductivity measurement technique, where the authors investigated the mixing
phenomena in a VOD (Vacuum Oxygen Decarburisation) ladle and the effect of
different placements of bottom nozzle with and without a top jet. The authors defined
two degrees of mixing (t95 and t99) for their water model experiments. They finally
concluded that mixing time tend to decrease considerably as the position of the bottom
nozzle is moved away from the centre to the mid-radius irrespective of whether a top
jet is present or not.
3.1.5 Numerical Investigation
Numerical technique is also a powerful tool to investigate different fluid dynamic
characteristics in gas injection processes. Though not much study on numerical
investigation of top submerged lance gas injection system are reported in the open
literature, there are significant research findings on the other types of gas injection
processes based on numerical studies.
Schwarz and Koh [75] first developed a numerical model of bath mixing by swirled
gas injection. They used finite domain computer code PHOENICS to model both flow
within the lance, with and without swirlers. The authors reported that swirl flow
within a lance showed that a recirculating pattern is set up, leading to an increase in
the axial velocity at the lance wall by a factor of about three over the mean flow
velocity which enhanced the heat transfer between the lance wall and the gas. The
simulation results were validated against cold flow air-water model experimental data.
In their numerical investigation, Schwarz and Koh [75] used Prandtl mixing length
model for modelling of turbulence and suggested that a more sophisticated turbulence
and turbulence diffusion models must be used to predict void fraction distribution
accurately. Rankin et al. [70] also used the similar type of PHOENICS code to
simulate the bath mixing process. Liovic et al.[76, 77] developed a numerical model
for simulating transient behaviour of multi-fluid problems. They investigated the gas
81
injection process for top submerged lance as a 2D axisymmetric problem by using
CFD. In their simulation, to get a complete cross-sectional view, the results at the
centreline were mirrored. However, Liovic et al.[76] also reported that, 2D
axisymmetric volume tracking was unable to facilitate the simulation of fully 3D
interfacial phenomena. Due to the complex nature of the flow structure involved in the
gas injection system, detailed understanding of the hydrodynamics of the gas injection
system still needs thorough investigation.
Solnordal and Gray [78] developed a mathematical model of heat flow to an operating
sirosmelt lance which predicts lance wall and air temperatures and thickness of the
slag layer on the lance. The model was used to determine both the heat transfer
coefficient between the vessel contents and the lance and the thermal conductivity of
the slag layer by measuring the distribution of wall temperature and slag thickness on
an operating sirosmelt lance. Turkoglu and Farouk [79] numerically analysed the time
evolution of the flow and temperature fields in industrial scale gas-injected molten
iron baths by using Eulerian approach. The authors used a constant effective viscosity
for the gas phase turbulence and predicted the turbulence in the liquid phase using two
equation ε−k turbulence model.
Shin et al. [80] simulated turbulent combustion phenomena in the gas phase occurring
at the post combustion stage in a bath type smelting reduction process with the help of
mathematical model which combines the Simple Chemical Reaction System (SCRS)
combustion model with instantaneous reaction up to the chemical equilibrium state
allowing the concentration fluctuation and the ε−k turbulence model. The authors
investigated the effects of the injecting method on the oxidation. In their study, the
efficiency of the post combustion operation was evaluated by the post combustion
ratio (PCR) at the outlet of the furnace, where,
∑ ∑
∑+
=coco
co
mm
mPCR
ρρ
ρ
2
2 x 100 (%) 3.20
The authors reported that Post Combustion Ratio (PCR) was elevated with the
decrease of the inlet velocity of the injecting gas and the increase of the lance height
from the iron bath. The results further predicted that oxygen should be injected with
82
the low inlet velocity at the large distance from the iron bath surface in order to attain
the efficient post combustion operation.
Zhu et al. [81] carried out similar type of numerical investigation for the three
dimensional turbulent fluid flow and mixing characteristics in gas-stirred ladles. The
authors investigated the effects of gas flow rate, positions of nozzle and tracer and
inclined wall on the flow pattern, and mixing phenomena. They concluded that
eccentric blowing reduces the mixing time and the mixing time is sensitive to the
alloy/tracer adding position especially for centre blowing. For the ladle with inclined
wall, it has shorter mixing time compared with the cylindrical one. Zhu et al. [81] used
the following experimental correlations proposed by Castillejos and Brimacombe [82]
to model the bubble plume zone by considering some assumptions:
−=
4.2
2/max max
7.0expαα
α
x
x 3.21
where,
α = Volume fraction of gas in the plume zone
maxα = Volume fraction of gas at plume centre line
2/max, αxx = Radial coordinate across the plume; half-value radius
The assumptions are,
1. The fluid flow in gas-stirred systems was driven by buoyancy of bubbles, and
the inertial force of bubbles was neglected.
2. The domain is occupied by a single phase fluid with spatially variable density.
3. The top surface is flat, and no tangential stresses are present.
4. The formation of the plume at the first stage of blowing was neglected and it was
taken as stable during the blowing.
Bernard et al. [83] developed another numerical model where the authors proposed a
simple approximation for the buoyant force in a bubble plume. By assuming a uniform
radius and slip velocity for the entire bubble column, the authors derived an
expression for the vertical acceleration of liquid in the column which is directly
proportional to the injected gas flow-rate and inversely proportional to the depth and
83
velocity. This bubble induced acceleration has been implemented with a ε−k
turbulence model in the MAC3D single-phase CFD code. But this model is not
applicable for stratified liquids in which there is a sharp density gradient between two
weakly stratified layers. Milelli et al. [84] also developed another numerical model for
bubble plume using CFD which focuses on the effect of turbulence in promoting bath
mixing and the application of Large Eddy Simulation (LES) methodology to capture
bubble-induced turbulence effects. In that study, the flow considered is also buoyancy-
driven rather than shear induced.
Morsi et al. [85] carried out a numerical investigation of the top submerged gas
injection system by using CFX 4 which is based on Eulerian – Eulerian two phase
model and drag force, lift force and turbulence dispersion force were taken into
account for the interface between gas and liquid. Morsi et al. [85] assumed same
pressure in both phases on the same computational cell and a constant gas velocity at
the exit of the nozzle which is treated as a mass source term with a corresponding
source term in the axial momentum equation at the nozzle exit.
Fluid flow, bubble distribution and gas-liquid mass transfer in a water model vessel
with gas injection were also analysed numerically by Taniguchi et al. [43]. They
investigated the effect of turbulence on metallurgical reactions in ladles. In that study
the authors applied a mathematical model composed of ε−k model and bubble-
dispersion model. Flow pattern, gas hold-up distribution, velocities, k and ε
distributions were calculated and these results were validated with previous
experimental data obtained by LDV. Another CFD modelling of bubbles, droplets and
particles in metals reduction and refining was developed by Cross et al. [86]. Alessio
[87] also investigated numerically the bubbles dispersion patterns in a gas-stirred
ladle.
3.2 Review of Zinc Slag Fuming Process
Smelting industries often produce slag, containing significant amount metal oxides,
which are often discarded as waste. Significant metal values can be recovered from the
slag by proper treatment. Zinc slag fuming is such a process, where a reductant source
(usually coal) is injected in the molten slag bath to reduce dissolved ZnO from the
84
bath into metallic zinc vapour. The process can be carried out in either conventional
tuyere blown furnace or top submerged lance (TSL) smelting furnace. Commercial
development of the process using tuyeres was made by the Anaconda Coper Mining
Co. and Consolidated Mining and Smelting Co. in the 1920’s [21]. The potentiality of
extracting lead and zinc from different source other than the primary ores was
discussed by Ward [88]. Ward [88] discussed the economic values of lead and zinc in
the world market and emphasized for a more suitable environmentally friendly
extraction method for those two valuable metals.
In this section, a comprehensive literature survey on the zinc slag fuming process will
be discussed. As the zinc slag fuming can be carried out by both top submerged lance
smelting furnace and conventional tuyere blown furnace, a detailed literature survey
was carried out on each of the process. Details of the process overview on the zinc
extraction and slag fuming process by employing both top submerged lance and
conventional tuyere blown furnace was discussed on chapter 2 in section 2.7.2.
3.2.1 Slag fuming by conventional tuyere blown process
Though conventional slag fuming has been commercially operative since 1920’s, only
a few studies related to slag fuming kinetics and details of fluid dynamic behaviour
inside the slag fuming furnace has been found in open literature. The earliest studies
were performed in 1950s by Bell et al. [89] and Kellogg [90, 91] and in 1960’s by
Quarm [34, 92, 93] to investigate details of slag fuming behaviour. Later on, Richards
et al. [19-21, 30, 94, 95] carried out extensive research on slag fuming process.
Investigation on slag fuming kinetics carried out by Richards et al. [19-21] comprised
of accumulating several industrial data and mathematical modelling. The industrial
study consisted primarily of slag sampling through five different fuming cycles [21].
From that industrial study, the authors reported that, in general, zinc elimination curve
is linear with time and a portion of the injected coal entrains in the slag [21]. From the
analysis of the tuyere back-pressure fluctuations and movie photographs of the tuyere
tip, they reported that coal-air mixture enters the slag in the form of discrete bubbles.
They divided the fuming furnace into two reaction zones: Reduction zone and
Oxidation zone. In the Reduction zone, the coal entrained in the slag reduces ZnO and
85
Fe3O4 which is responsible for fuming. In the Oxidation zone, the remaining coal in
the tuyere gas stream combusts.
Based on the data obtained in industrial studies, Richards and Brimacombe [20]
developed a mathematical model of zinc slag fuming based on the kinetic concept of
the two reaction zone from the industrial studies. The two zones and the water-
jacketed furnace wall had been linked by overall heat and mass balances. The authors
reported that the model had shown consistent results over five different industrial
fuming operations among the eleven industrial fuming cycles tested. They also
reported that about 33% of the injected coal was entrained in the slag, 55% combusted
in the tuyere gas column and 12% bypassed the bath completely. Later on, Cockcroft
et al.[96] improved that mathematical model of zinc slag fuming, based on rate
phenomena such as mass transfer, chemical kinetics and heat transfer and also
including the behaviour of lead in the bath. That improved model predicted that
fraction of coal entrained is greater with increasing nominal tuyere exit velocity and
that oxygen utilization increases as the bath depth is raised. That model also predicted
that the settling and collection of metallic lead will bear importantly on the removal of
lead from high lead slags produced by the QSL and flash smelting processes.
Finally, Richards and Brimacombe [19] elucidated the rate limiting steps of the
fuming process and predicted the influence of process variables on fuming by using
the mathematical model [20]. The model predicts that fuming efficiency reaches
maximum with increasing residence time of coal particles in the slag. The level of
ferric iron in the slag is an important variable affecting the fuming kinetics. The level
of ferric iron in the slag depends on ferrous iron oxidation rate, melting/freezing of
slag at the water-cooled jacket and ferric iron reduction by coal entrained in the slag.
They further reported that, at the beginning for very short span of time, Boudouard
Reaction controls the zinc reduction kinetics. However, after that it is been controlled
by diffusion of ferric iron to the interface between the secondary bubbles containing
the coal and the slag. They also reported that increase in coal entrainment increases the
fuming efficiency. Coal entrainment can be increased by injecting coal in high-
pressure. Cockcroft et al. [97] reported from another study, that high pressure coal
injection increased coal entrainment about 25%, as a result fuming rates were
increased substantially, to between 70 and 90 pct, depending on the charge mix.
86
Richards [94] commented that coal entrainment is controlled by injection conditions
whereas bath temperature is a function of coal combustion and ferrous oxide
oxidation.
Another mathematical modelling study of a DC electric furnace for zinc recovery from
lead blast furnace slag has been carried out by Chang et al. [98]. In that work the
authors predicted the momentum and heat transfer from the arc to the slag by algebraic
equations and considered a two dimensional steady state model for electrical current
flow, fluid flow and mass transfer. Kellogg [90] developed the first computer model
for slag fuming process. In that model, Kellogg assumes stepwise equilibrium during
each micro-step (0.1 minute in a 90 minute period). By using Kellogg’s computer
model, Grant [99] derived the thermodynamic properties of slags from the slag fuming
plant data.
Scholey et al. [100] investigated the heat transfer phenomena in water cooled zinc-
fuming furnace jackets for conventional rectangular furnace with submerged tuyeres.
From the industrial measurement, Scholey et al. [100] reported the presence of large
thermal transients or temperature “spikes” in the region immediately above the tuyeres
and commented from a mathematical analysis that these temperature spikes are
associated with sudden slag falloff due to charging and tapping of the furnace and
agitation on the bath surface due to gas injection effects. The authors concluded that
the temperature spikes can be reduced by increasing the number of anchoring fins
which will reduce the crack formation and propagation along the furnace wall.
Other than the batch fuming process, some studies on continuous fuming process was
carried out by Richards [95], Haralampiev and Popov [101]. Haralampiev and Popov
[101] carried out a physical investigation of the continuous fuming process by using
water and 72 weight % glycerine solution. They suggested some changes in the
dimensions of the furnace to facilitate the continuous fuming process. Richards [95]
reported that fuming efficiency (Zn/coal) in continuous fuming was predicted to be
lower than the batch process for equivalent overall slag flow rate. Further industrial
studies have followed the studies by Richards et al. [19-21, 94] and Cockcroft et al.
[96, 97]. Lehner and Lindgren [102] reported the detail of fluid flow behaviour by an
industrial study at Boliden’s zinc fuming plant. They carried out a series of
experiments to investigate the mixing behaviour in the slag by adding radioactive
87
gold. They also analysed the fuming behaviour and reported that zinc fuming process
is controlled dominantly by the gas phase.
3.2.2 Slag fuming by TSL process
Slag fuming by using top submerged lance (TSL) technology has been successfully
operating around the world. Zinc slag fuming by TSL technology has started with pilot
plant studies in 1980’s and has progressed to the treatment of close to 800,000 tonnes
a year of zinc bearing feeds in the form of residues and slag [39]. There are a few
studies in the open literature regarding slag fuming behaviour by using TSL
technology, based on both laboratory scale work and pilot plant scale studies.
Suzuki et al. [31] investigated the factors effecting the zinc fuming kinetics such as
gas or slag composition, gas blow rate, slag temperature, viscosity and surface tension.
They carried out laboratory scale experimental work by blowing N2 gas or gas
mixtures of CO, H2 and CH4 with CO2 using a top lance into molten slag containing 6-
12 pct of zinc at 1200-1300o
C. In their research, Suzuki et al. [31] found that
formation of bubbles in the molten slag bath has a great influence on the zinc fuming
kinetics. They reported some important parameters that remarkably effect the
formation of bubbles which are surface tension, viscosity of the liquid. The number of
bubbles became fewer as the values of these physical properties were larger and
therefore each bubble size became larger. Formation of the number of bubbles also
depends on the gas blow rate. The number of bubbles increased with increasing gas
blow rate up to a limiting value, but after exceeding this value it decreased rapidly.
The authors also reported that the forming state of bubbles differed with the kind of
blowing gases, like H2, CO and N2. The authors developed an equation of the zinc
fuming rate from their experimental results, which can be expressed as:
28 2
12/
2aKntn
N−=∆∆ 3.22
where, n∆ is the change of ZnO in slag in the unit of g-mols, a is the activity of ZnO
and 1
K is the equilibrium constant which can be expressed as:
appKozn
/. 2/12
1= 3.23
88
znp and
2op are the equilibrium partial pressures of Zn and O2 respectively. The
activity of ZnO may be expressed as:
)/(.. ∑+==i
nnnNa γγ 3.24
where, N represents the mol fraction of ZnO in the slag, γ is the activity coefficient of
ZnO, n is the g-mols of ZnO in the slag and ∑ in is the summation of g-mols of the
components except ZnO.
Denholm et al. [103] carried out the first experimental work on top submerged lance
investigation after its invention and some pilot plant trials. The experiment was done
in CSIRO and emphasised on copper matte converting and zinc slag fuming. That
study was based on four fundamental investigations to have a clear understanding of
working principle and possible implementation of the TSL technology in industrial
zinc fuming applications. The four studies were:
1. Tracer studies to reveal mixing conditions in the bath by swirl and non swirl
air injection.
2. Sampling and analyses of the slag to establish the rate of reduction reactions
during addition of carbonaceous reductants to copper and zinc slag.
3. Gas pressure fluctuations measured on the lance air supply for a range of
lances and operating conditions.
4. Temperature measurements within the lance.
The study emphasised one of the most important aspect of the present research which
is the investigation of zinc slag fuming by both batch and continuous process. The
authors reported that significantly higher carbon utilization rate was achieved when
zinc was fumed in a continuous process than compared to the batch process.
Waladan et al. [2, 104, 105] reported several studies based on the experimental works
of zinc slag fuming process by using top submerged lance. From a pilot plant scale
study Waladan et al. [2] discussed the detail of zinc fuming behaviour, effect of coal
addition rate and coal particle size on the fuming rate. They also discussed the
comparison between conventional tuyere blown process and TSL process of zinc
fuming. Later on, Waladan and Nilmani [105] carried out a crucible scale
89
experimental work to investigate the effect of different injection parameters like lance
diameter, gas flow rate, slag bath temperature and different reducing agent on fuming
rate. From their work, the authors reported that high momentum coal injection, a thin
wall lance, intense bath agitation and increasing temperature, all improved fuming
rates. They further reported that there was substantial increase in the fuming rate when
they changed their injection lance from an internal lance diameter of 4 mm to 2 mm.
Reddy et al. [32] carried out another crucible scale laboratory study to recover zinc
from industrial lead blast furnace slag. They investigated zinc fuming behaviour with
and without carbon addition by stirring the slag with air or argon. That study showed
that even without the carbon addition, zinc fuming took place by the diffusion of Fe+2
ions in the melt. They also reported that the type of stirring gas (air or argon) or the
flow rate did not influence the reaction mechanism. Lightfoot et al. [106] also
demonstrated some pilot plant scale trials in addition to crucible scale laboratory study
to investigate the zinc fuming behaviour from slag.
Gupta [107] proposed a method for the production of zinc from zinc oxide and
complex zinc concentrates. The two stage process proposed by Gupta [107] involves
oxidation of zinc sulphide to oxide and dissolution into slag and the fuming of zinc
from the slag by injecting carbonaceous materials into it to produce zinc vapours.
Gupta [107] carried out the experimental study on a laboratory scale TSL setup. The
author discussed the effect of the quantity of air, temperature and concentrate feed rate
on the production of zinc rich slag. The author also reported that smelting temperature
less than 1300oC have a detrimental effect on the viscosity of slag. Neira et al. [108]
carried out another study to recover zinc from the waste oxide generated by Ausmelt
Pyro-processing. In that research, a laboratory evaluation for electrowinning of zinc
electrolytes generated by Aumelt’s TSL pyrometallurgical process was carried out.
They suggested some special treatment to remove impurities before undertaking the
electrowinning process.
Miyake [109] described different aspects and optimized plant operating conditions of a
slag fumer at the Hachinohe Smelter using TSL furnace. Choi and Lee [33],
Matusewicz et al. [110], Sofra et al. [36], Hughes et al. [35], Hoang et al. [42]
described the Outotec’s Top Submerged Lance (TSL) technology for the processing of
secondary zinc feed materials, including zinc plant leach residues and EAF dust.
90
3.2.3 Other studies based on zinc extraction
Other than the studies based on conventional slag fuming and TSL slag fuming, there
had been numerous studies on thermodynamic analysis of zinc slag fuming and some
other modified way of slag fuming. In this section, some studies found in open
literature will be discussed briefly.
Jak and Hayes [111] carried out thermodynamic analysis for slag fuming using the
F*A*C*T computer package. They pointed out the effects of slag chemistry on the
liquidus temperatures, subliquidus phase equilibria and thermodynamic properties on
slag fuming by using that computer package. From their thermodynamic analysis, the
authors argued that initial stage of zinc fuming is principally equilibrium controlled
when the zinc concentrations in the slag is high. As zinc concentrations in the slag
decreases, the reaction becomes increasingly controlled by kinetic factors and far
away from equilibrium conditions. The authors compared their thermodynamic data
based on the plant measurements carried out by Grant [99]. Later, Jak and Hayes [112]
and Verscheure et al. [113] discussed the role of slag chemistry in the design of freeze
linings for slag fuming furnaces. The effect of sulphur on the rate of reduction of zinc
oxide from slags was studied by Dal and Rankin [114]. Fumed zinc from the slag bath
usually undergoes re-oxidation process above the bath in the post combustion zone.
Kinetics of zinc vapour oxidation was discussed by Lewis and Cameron [108].
Figure 3-1: Schematic of the submerged plasma process for the high temperature
fuming of zinc from zinc containing residues [Image taken from Verscheure et al.]
91
To treat zinc leach residues, electric arc furnace (EAF) dusts, and other zinc-
containing waste materials, Verscheure et al. [115-118] proposed a new high
temperature submerged plasma zinc fuming process. They developed the model by
using the FactSage thermodynamic databases and ChemApp thermodynamic software.
The proposed process is electricity based pyrometallurgical process that uses
submerged plasma torches to supply the necessary heat in the fuming bath. Inside the
plasma torches, the cold blast air is transformed into a high enthalpy plasma gas. The
plasma gas is then mixes with the natural gas (CH4) and injected into the slag bath.
Parallel to this, a mixture of leach residue, petroleum coke, and fluxes is fed
continuously into the slag bath. The investigators reported that the submerged plasma
arc process can operate at higher operational temperatures (1300o C – 1400
o C), as
compared to the conventional zinc fuming process (1200o C – 1250
o C).
Barcza et al. [119] and Latif [120] described the Enviroplas process of zinc recovery
from metallurgical wastes. As reported by Latif [120], the Enviroplas process was
developed at Mintek, South Africa, to treat certain metallurgical wastes, such as lead
blast furnace (LBF) slag, electric arc furnace (EAF) dust, and neutral leach residues
(NLR) from the zinc industry. The process involves smelting of those metallurgical
wastes in a DC arc furnace and subsequent recovery of the volatilized zinc in an ISP
lead splash condenser. In this process, the dry granulated slag are directly charged into
the fuming furnace. Metallurgical coke is employed as a reducing agent and is
delivered to the fuming furnace at a controlled rate. Coals, charcoal or other
carbonaceous materials, low in moisture and volatile content are also used as reducing
agents. Inside the fuming furnace, the zinc and lead oxides from the slags are reduced
to their metals at 1400o C – 1500
o C temperatures. The residual is tapped from the
furnace as required. The volatilized zinc and lead are collected through a refractory
lined duct and condensed in a lead splash condenser (operated at 500o C – 550
o C).
Figure 3-2 shows the schematic diagram of the Enviroplas process.
92
Figure 3-2: Schematic diagram of the Enviroplas pilot plant (Image taken from Latif
[104])
Tarasov and Besser [121] discussed the processing of lead and zinc raw materials to
extract from primary ores. Schwarz [122] carried out some study related to primary
zinc production by using computational fluid dynamic (CFD) modelling. That study
dealt with CFD modelling of thickeners only.
3.3 Research Objectives
The research objectives of the present study can be summarized from the above
discussion. As discussed in the primary research theme (section 1.1), combustion
behaviour and zinc fuming behaviour from a pilot plant scale TSL furnace was the
initial research theme. As an initial step, cold modelling analysis of the TSL system
was carried out by using CFD. Literature survey on the cold modelling investigations
of the gas injection system (section 3.1) revealed that the experimental studies dealt
with the effect of lance submergence level, gas injection rate and swirl intensity on the
change of velocity components and turbulence behaviour. None of the studies reported
about the bath mixing characteristics with the change in hydrodynamic parameters
(lance submergence level, flow rate and swirl intensity), splashing behaviour with the
93
change in hydrodynamic parameters and liquid properties (density and viscosity),
penetration depth of the injected gas jet for annulus air injection generally used in TSL
process. The numerical studies (Liovic et al.[76, 77], Schwarz and Koh [75]) carried
out for TSL investigation involved 2D axisymmetric grid and in some case the results
at the centreline were mirrored to get the overall behaviour. Due to the complex nature
of the flow structure involved in the gas injection system, detailed understanding of
the hydrodynamics of the gas injection system still needs thorough investigation.
Hence, the aim of the first phase of the present study is to investigate the physical
behaviour of the top submerged gas injection system and to predict the effect of swirl
intensity, lance submergence level and air injection rate on the overall bath mixing and
splash generation by using the 3D hybrid grid system. The vertical depth of
penetration of the air jet injected through the annulus of the lance into the liquid bath,
splashing behaviour inside the bath with the change of density and viscosity of the
liquid were also of the significant interests.
Although commercial slag fuming is well established, literature survey on the zinc
slag fuming process showed there have only been a few numerical modelling studies
on zinc fuming kinetics. Kellogg [90] assumed stepwise equilibrium in his computer
model of the slag fuming process during each micro-step (0.1 minute in a 90 minute
period). No CFD analysis has been found in the open literature regarding slag fuming
to date. The purpose of this study is detailed fluid dynamic analysis including
combustion behaviour, gas-liquid momentum interaction, generation of splashing due
to gas injection process in slag fuming furnace, analysis of fuming behaviour at
different locations of the furnace by using computational fluid dynamic (CFD)
modelling technique. Hence, the second stage of the present study emphasized on
revealing detail of the combustion behaviour, fuming kinetics in the slag bath,
splashing behaviour in a pilot plat scale combusting. Finally, in the third stage,
investigation on tuyere tip combustion dynamics, bath interaction conditions, jet
penetration length, generation of turbulence and coal utilisation behaviour inside the
bath were of significant interest.
94
Chapter 4
95
4 Modelling Techniques and Model Features
This chapter describes the governing equations for fluid motion and the numerical
methods used to solve these equations. The description covers the CFD modelling
approach, discretization methods, schemes, turbulence modelling, and difficulties
associated with the solution procedure and methods of overcoming them. A
comprehensive description of the boundary conditions used for the developed models
is presented. Details of meshing and the methodologies used in physical modelling
are also discussed.
4.1 CFD Modelling
CFD stands for Computational Fluid Dynamics. It is an iterative calculation procedure
to obtain the solution of Navier-Stokes equations. The Navier-Stokes equations are
derived from the principle of conservation of mass and momentum. Hence, the
cornerstone of computational fluid dynamics is the fundamental governing equations
of fluid dynamics – the continuity, momentum and energy equations. These equations
speak physics. They are the mathematical statements of three fundamental principles
upon which all of fluid dynamics is based:
The mass of fluid is conserved
Momentum is conserved, i.e. the rate of change of momentum equals
the sum of the forces on a fluid particle (Newton’s second law)
Energy is conserved, i.e. the rate of change of energy is equal to the
sum of the rate of heat addition to and the rate of work done on a
particle (first law of thermodynamics)
The Navier-Stokes equations can predict the fluid flow behaviour in its general form.
From the 1960s onwards, the aerospace industry has integrated CFD techniques into
the design, R & D and manufacture of aircraft and jet engines. Then the method is
being started to apply to simulate different processes in internal combustion engine,
96
combustion chambers of gas turbines and furnaces. Increasingly CFD is becoming a
vital component in the design of industrial products and processes. CFD has entered
into the wider industrial community since the 1990s [123]. With the advancement of
high performance computing facilities, Computational Fluid Dynamic (CFD)
modelling technique has evolved as a powerful tool for the researchers working in the
metallurgical field. CFD can predict flows ranging from simple single phase flows to
complex multiphase flows in high temperature combusting environment associated
with metallurgical process industries. Successful and efficient development of a CFD
model can predict the fluid flow behaviour, combustion behaviour, generation of
turbulence and splashing and other fluid dynamic parameters inside the furnace.
Fluid flow behaviour in a system and the related phenomenon like heat and mass
transfer can be represented by a set of non-linear partial differential (or integro-
differential) equations (PDE). Analytical solution of these equations is almost
impossible except in some special cases. To obtain an approximate solution
numerically, a discretization method is used, which approximates the differential
equations by a system of algebraic equations. These equations can then be solved on a
computer, providing a description of the flow field at discrete locations in space and
time. Much as the accuracy of experimental data depends on the quality of tools used,
the accuracy of numerical solutions is dependent on the quality of discretizations used
[124]. There are many different ways by which equations describing fluid flow and
heat transfer can be solved using computational methods. Most commercial and
research codes rely on the following:
Finite Volume
Finite Difference
Finite Element
Spectral Methods
Each of these methods requires the definition of discrete points in space at which
variables like velocity, pressure, temperature etc. will be computed. While the
governing equations are always the same, the particular geometry with initial and
boundary conditions determines a unique solution for each particular problem. This
97
research is based on the finite volume method, as it is the method used by most of the
popular CFD codes currently available. The other methodologies still commonly used
in industry and yield good result for certain types of application.
4.1.1 Finite Volume Method
The Finite Volume Method (FVM) starts with the integral form of the governing
equations, involving surface integrals (e.g. convective and diffusive fluxes) and
volume integrals (e.g. those describing sources and sinks). In case of a transient flow
(i.e. unsteady flow that changes over time), there is also a rate of change term. The
FVM represents the integration of the governing equations over (a finite number of)
contiguous control volumes (CVs) representing the solution domain. Since variable
values are computed only at discrete points, approximations must be used to express
the integrals in terms of unknowns at discrete locations. In this way one algebraic
equation per CV is obtained, linking variable value at the centroid of that CV with
those at neighbour CVs. For the solution domain as whole, a large system of algebraic
equations is obtained. Since these equations are in general non-linear and coupled, the
solution must be sought using iterative solution methods. Iterations means repeating a
sequence of operations over and over, until changes in computed variables becomes
negligible and we declare the process to as “converged”.
Most of the main commercial CFD codes, such FLUENT, STAR-CD, AVL FIRE are
based on FVM scheme. One of the reasons why FVM has succeeded over the other
methods is that it is inherently conservative: irrespective of errors in various
approximations, the discretized equations still fulfil the conservations laws exactly. In
other words, the errors introduced through various approximations affect only the
distribution of variables within solution domain without violating conservation
principles. The FVM is also easier to understand by engineers than some of the other,
more mathematically involved methods, since the term that need to be computed have
a clear physical meaning (e.g. mass or heat flux through a CV face, force at a CV
surface etc.).
98
4.2 Multiphase Flow Modelling
Multiphase flow in CFD can be referred to any fluid flow consisting of more than one
phase or component. In multiphase flow, a phase can be defined as an identifiable
class of material that has a particular inertial response to and interaction with the flow
and the potential field in which it is immersed. For example, different-sized solid
particle of the same material can be treated as different phases because each collection
of particles with the same size will have a similar dynamical response to the flow
field. Two phase flow is the simplest case of multiphase flow. Multiphase flow can be
classified according to the state of the different phases or components and therefore
refer to gas-solids flows, gas-liquid flows, liquid-solids flows or gas-particle flows or
bubbly flows and so on.
4.2.1 Approaches to Multiphase Modelling
Computational fluid mechanics (CFD) modelling in recent years have provided the
basis for further insight into the dynamics of multiphase flows. Currently there are two
approaches for the numerical calculation of multiphase flows: the Euler-Lagrange
approach and the Euler-Euler approach.
4.2.1.1 THE EULER-LAGRANGE APPROACH
The Lagrangian discrete phase model follows the Euler-Lagrange approach. This
approach is generally used for highly dispersed flows where the volume fraction of the
dispersed phase is small. The time-averaged Navier-Stokes equations is solved for the
fluid phase which is treated as a continuum, while the dispersed phase is solved by
tracking a large number of particles, bubbles, or droplets through the calculated flow
field. There is exchange of interfacial momentum, mass, and energy between the
dispersed and the continuous phase. A fundamental assumption made in this model is
that the dispersed phase occupies a low volume fraction, even though high mass
loading ( fluidparticle mm ≥ ) is acceptable. The particle or droplet trajectories are
computed individually at specified intervals during the fluid phase calculation. The
model is appropriate for the modelling of spray dryers, coal and liquid fuel
combustion, and some particle-laden flows, but inappropriate for the modelling of
99
liquid-liquid mixtures, fluidized beds, gas-liquid flow or any application where the
volume fraction of the secondary phases is not negligible.
4.2.1.2 THE EULER-EULER APPROACH
In Euler-Euler approach the fluid phases are treated mathematically as interpenetrating
continua. Fluids are treated in every computational cell with the concept of phasic
volume fraction. For a two phase flow situation, as the volume of a phase cannot be
occupied by the other phase, each of the phases is considered to occupy a fixed
volume fraction in a computational cell. These volume fractions are assumed to be
continuous functions of space and time and their sum is equal to one. Conservation
equations for each phase are derived to obtain a set of equations, which have similar
structure for all phases. These equations are closed by providing constitutive relations
that are obtained from empirical information, or, in the case of granular flows, by
application of kinetic theory.
The present study is based on Euler-Euler approach, as the research deals with gas
injection process of top submerged lance furnace and understanding the detailed fluid
dynamic and combustion behaviour inside the TSL furnace. Euler-Euler approach was
more suitable for the current research where gas and liquid volume fraction in a
computational cell was of fundamental interest rather than keeping track of the
lagrangian particle phase. The first stage of the research was based on gas (air)
injection in liquid (water) only and understanding the hydrodynamic parameters of the
process. In the second stage, the model was extended for submerged combustion of
gaseous fuel in molten ISF slag for zinc fuming TSL furnace and investigating the
combustion behaviour, reaction kinetics, fuming behaviour inside the furnace.
Gaseous fuel (CH4) was chosen to simulate the combustion behaviour for submerged
combustion at the lance tip. In the third stage, the model was extended for multiphase
submerged coal combustion in rectangular tuyere blow zinc fuming furnace, which
was also carried out in Euler-Euler approach. The aim was to investigate the
hydrodynamic parameters of zinc fuming behaviour from molten slag and understand
the kinetics of submerged coal combustion. The fundamental assumption of that
approach was that coal was assumed to be a separate scalar of the continuum liquid
100
phase instead of discrete particles. The detail of the coal combustion modelling is
discussed later (Section 4.3.3.4.1).
The present research was carried out by using the commercial CFD package AVL
FIRE (Version 8.52 and Version 2009.2). The FIRE Eulerian Multiphase Module
allows the use of the following models based on the Euler-Euler approach listed in the
order of increasing accuracy [125]:
Homogeneous (Equilibrium) Model
Multi-fluid Model
Volume-of-Fluid (VOF) Free-Surface Model
4.2.1.2.1 Homogeneous Model
The homogeneous model is the least accurate multiphase model based on the Euler-
Euler
approach. A volume fraction equation is calculated for each phase. However, only a
single
momentum equation is calculated for the phases in momentum equilibrium.
4.2.1.2.2 Multi-fluid Model
In the multi-fluid model, all conservation equations are solved for each phase. Since
the multi-fluid model requires by default the calculation of the complete set of the
conservation equations for each phase, it represents the basis for the Euler-Euler
schemes in the FIRE Eulerian Multiphase Module. The commercial software AVL
FIRE’s user-defined subroutines (UDF) allow customizing the calculation of the mass,
energy and momentum exchange.
4.2.1.2.3 VOF Model
The VOF model is a surface-tracking technique applied to a fixed Eulerian grid
proposed by Hirt and Nichols [126]. This model is designed for two or more
101
immiscible fluids where the accurate prediction of the interface between the fluids is
of interest. In the VOF model, a single momentum equation is shared by the fluids,
and the volume fraction of each of the fluids in each computational cell is tracked
throughout the domain. Applications of the VOF model include stratified flows, free-
surface flows, filling, sloshing, the motion of large bubbles in a liquid, the motion of
liquid after a dam break, the prediction of jet break-up (surface tension), and the
steady or transient tracking of any liquid-gas interface.
From the numerical perspective the Volume-of-fluid model is very similar to the
homogeneous model. A single momentum equation is calculated for all phases that
interact using the VOF model. However, the calculation of volume fraction equations
using VOF model is considerably more accurate allowing the sharp resolution of the
interfaces. One of the common defects of the VOF calculation can occur when the
interface is not resolved sharply despite the use of the high-order discretization
techniques for the volume fraction equation – in that case the VOF model degenerates
into the homogeneous model. This is quite common in many practical calculations. It
happens due to very high-resolution requirements of the VOF model that can be often
hard to fulfil.
In the following sections, details of the modelling procedures including model
geometry, solution procedures and governing equations solved for each geometry will
be discussed further.
4.3 Model Geometry and Computational Methodology
Numerical simulations were carried out in three different geometries by considering
three different gas-liquid combinations. First stage of the research considered a
laboratory scale top submerged lance gas injection system with air and water as the
fluid. Then the model was applied to a pilot plant scale top submerged lance zinc
fuming furnace with submerged CH4 combustion and reactions in slag and gas phases
included. In the final stage, a thin slice model of the real plant scale rectangular tuyere
blow zinc fuming furnace was developed. Details of the model geometry,
computational methodology, boundary and initial conditions, fluid properties are
discussed below.
102
4.3.1 Air water Model
A 3D CAD model of the one sixteenth-scale air-water model of a 150-ton steel
making ladle was developed by using CAD tool. The CAD model is similar to the
experimental model of Morsi et al.[1]. A schematic diagram of the model is shown in
Figure 4-1. The vessel has a diameter of D =230 mm and length Z=560 mm. The tank
was filled up to L=150 mm with water. A vertical lance with an annulus of inner
diameter di =12.2 mm and outer diameter d
o=17 mm was fitted at the centre of the
cylindrical vessel. Air was injected through the annulus of the lance into the water
bath. On the top of the cylindrical vessel the outlet was defined by Do= 60 mm. Mid
Plane cross sectional view of generated grid for CFD analysis is shown in Figure 4-2.
Figure 4-1: Schematic diagram of the air-water model
103
Figure 4-2: Mid Plane cross sectional view of generated grid for CFD analysis
4.3.1.1 MODEL FEATURES
The CFD modelling of top submerged gas injection involves multiphase simulation
where gas and liquid phases interact with each other and there is significant
momentum exchange between the phases. The model was developed using the finite
volume method in conventional Eulerian approach by using commercial CFD package
AVL FIRE 8.52. The model developed include the following features,
Unsteady state multiphase solution for momentum and continuity was
considered.
Standard k-ε turbulence model for the turbulence modelling was employed.
A cell centred finite volume approach was used to discretise the governing
equations and the resulting discretised equations were solved iteratively using
segregated approach.
Pressure and velocity were coupled using the SIMPLE algorithm [127].
Least square fit approach was used for the calculation of the derivatives
104
For momentum and turbulence, first order upwind differencing scheme was
used whereas central differencing scheme with second order accuracy was
used for the continuity equation
Swirl flow was injected through the annulus of the lance at 57.5o relative to
radial direction.
4.3.1.2 GOVERNING EQUATIONS
Basic Eulerian equations, describing multiphase non-combusting system are given by
the conservation equations for continuity and momentum equations. Other than the
basic conservation equations, interfacial exchange terms for momentum at the gas-
liquid interface were also modelled. For three-dimensional fluid flow, these
conservation equations can be expressed as:
4.3.1.2.1 Continuity
t
kk
∂
ρα∂+ ∇ kkρα⋅ v
k = ∑
≠=
ΓN
kl,1l
kl k= 1,……,N 4.1
Where, N is the number of phases, kα is volume fraction of phase k, k
ρ is density for
phase k, vk is phase k velocity, klΓ is the interfacial mass exchange between phases k
and l, for this air-water simulation, no interfacial mass exchange terms were
considered. Summation of the volume fractions of the phases present,
11k
k=α∑
Ν
=
4.2
4.3.1.2.2 Momentum conservation
t
vkkk
∂
ρα∂+∇ kkρα⋅ v
k v
k = - kα ∇ p + ∇ kα⋅ ( τ
k + t
kT ) + kα fkρ + ∑Ν
≠= kl,1l
Mkl 4.3
∑Ν
≠= kll ,1
Mkl represents the momentum interfacial interaction between phases k and l, f is
the body force vector which comprises of gravity (g), p is pressure. Detailed
105
description of the momentum interfacial interaction is discussed later. Pressure is
assumed identical for all phases:
p = kp k = 1, ……., N
The phase k viscous stress integral is divided into non-transposed and transposed
terms. It can be expressed as:
τk = kµ (∇ v
k + ∇ vT
k) 4.4
Where, k
µ is the molecular viscosity. For incompressible flow, Reynolds stress,t
kT ,
takes into account the effect of turbulence. According to the Boussinesq hypothesis, it
can be expressed as:
t
kT = - kρ kk vv ′′ = t
kµ (∇ vk + ∇ v
T
k) -
3
2kkkk δρ 4.5
Where, k
δ is the Kronecker delta function and t
kµ is the turbulent viscosity. For
continuous phase, turbulent viscosity has been calculated by adding shear induced
turbulent viscosity with Sato’s viscosity due to bubble induced turbulence [128].
t
cµ = SI,t
cµ + BI,t
cµ 4.6
Where shear induced turbulent viscosity for continuous phase can be expressed as,
SI,t
cµ = µρ Cc
c
2
ck
ε 4.7
Sato’s viscosity due to bubble induced turbulence can be expressed as [128],
BI,t
cµ = C sato cρ D
bv
r dα 4.8
Where, µC = 0.09 and Csato
= 0.6 are dimensionless constant, k is the turbulent kinetic
energy and ε is its dissipation rate which can be obtained by solving equations for the
standard k-ε turbulence model put forward by Launder and Spalding [129]. The
turbulent kinetic energy (k) equation can be expressed as:
∑≠=
+ερα−α+∇
σ
µ+µα⋅∇=ρα⋅∇+
∂
ρα∂ N
kl,1l
klkkkkkk
k
t
kkkkkkk
kkk KPkkvt
k 4.9
k = 1, ………, N
106
∑≠=
N
kll
klK,1
is the interfacial turbulence exchange between phases and Pk is the
production term due to shear.
The turbulence dissipation (ε) equation is,
k
2
kk2k
k
kk1k
N
kl,1l
klk
t
kkkkkkk
kkk
kC
kPCDv
t
ερα−
εα++ε∇
σ
µ+µα⋅∇=ερα⋅∇+
∂
ερα∂∑
≠=ε 4.10
Closure coefficients used in the current study are kσ =1.0, εσ =1.3, C1=1.44, C2=1.92,
µC =0.09.
∑≠=
N
kll
klD,1
represents interfacial dissipation exchange between phases. In the present
simulations turbulence level of the dispersed phase is assumed to be equal to the
continuous phase turbulence level. The turbulence interfacial interaction between the
two phases is thereby neglected.
4.3.1.2.3 Interfacial Momentum Exchange
Momentum interfacial exchange between gas and liquid was modelled by
implementing interfacial momentum source at the gas-liquid interface which includes
drag and turbulent dispersion forces [125]:
Mc = C
D8
1c
ρ iA ′′′ vr
vr + C
TD cρ
ck dα∇ = dM− 4.11
Where, c denotes continuous and d denotes the dispersed phase. The first term in
equation (4.11) represents mean contributions due to drag force and the second term
takes into account the turbulence effect. The turbulence effect is represented by a
global dispersion effect, which is proportional to the void fraction gradient (cited in
[130]).
107
The drag coefficient,D
C , is a function of the bubble Reynolds number, Reb. The
following correlation for drag coefficient,D
C , was used [125]:
( )687.0Re15.01
Re
24b
bDC += Re
b 1000≤ 4.12
=DC 0.438 Reb > 1000
Bubble Reynolds number, Reb, and can be defined as:
c
brb
DvRe
υ= 4.13
Where c
υ is the kinematic viscosity for continuous phase.
Relative velocity is defined as:
vr = v
d - v
c 4.14
The interfacial area density for bubbly flow can be expressed as [125]:
b
di
D
6A
α=′′′ 4.15
Where, Db = 0.01 mm is the bubble diameter and dα is dispersed phase volume
fraction. CTD
= 0.1 in equation (4.11) is the bubble dispersion coefficient.
4.3.1.3 BOUNDARY CONDITIONS
The following boundary conditions were applied in the computational domain for the
air-water model:
4.3.1.3.1 Inlet
All boundary conditions were chosen to match the flow condition of the experimental
study of Morsi et al. [1]. The values of the velocity components and other dependent
variables were prescribed at inlet boundaries. At inlet, mass flow boundary condition
was given for non-swirl flow case with the flow direction defined as normal to the
boundary. Velocity boundary condition was chosen to define swirl flow with a fixed
108
mass flow rate similar to the experimental data. Three velocity components ( v,uvr
and
wr
) were defined at the inlet with resultant velocity creating the swirl flow pattern at
57.5o relative to radial direction. Velocity vectors for swirl flow case at the inlet are
shown in Figure 4-3. Typical turbulence quantities at the inlet of the domain were
calculated from inlet velocities by considering turbulence intensity I = 0.05 where,
81
160−
≅′= (Re).U/uI inlet . The inlet values for dissipation rate was determined from
the Kolmogorov relation:
ε
µ≈εL
kC 23
b4
3
b 4.16
Where, εL is the length scale and bk can be calculated from the turbulence intensity
(I) at inlet by the following equation,
( )2
bb IU2
3k = 4.17
Figure 4-3: Velocity vectors (m/s) for swirl air injection at the lance tip (Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =57.5o)
4.3.1.3.2 Outlet
Outlet boundary conditions are used at the domain boundaries through which the fluid
leaves. The outlet boundaries should be placed sufficiently downstream from the
regions where the flow exhibits significant changes. In this air-water simulation, outlet
was selected at the top wall of the cylindrical vessel where a fully developed flow
109
exists. Static pressure boundary condition was applied at the outlet of the
computational domain with an outlet diameter of 60 mm (see Figure 4-1).
4.3.1.3.3 Wall
The walls were assumed smooth and impermeable. For real flows, the velocity of
fluid, which is in contact with the wall, is equal to the wall velocity. This is known as
a no-slip condition. This condition is usually enforced by specifying the wall velocity
components. The turbulence model implemented in the computational domain is of the
“high Reynolds number” type which means they are not applicable in the near-wall
region. The near-wall region is characterized by large variable gradients and dominant
molecular effects. In order to model the near-wall effects (e.g., viscous damping,
kinematic blocking of the velocity fluctuations normal to the wall) the standard wall
function was employed.
For mean momentum, the wall functions based on the assumed logarithmic velocity
and temperature distributions were used [131]. For turbulent mean velocity, the
following wall function was used:
( )** yEnk
1U l= *y >11.63 4.18
Where, µ
ρ= µ
P
21
p41
*yk
Cy
Where, Py denotes the normal distance from the near wall node “P” to the wall, k =
0.4187 is Von Karman constant, E = 9.0 is an integration constant that depends on the
roughness of the wall and index “P” denotes the values at the centre point of the wall-
nearest control cell. In the momentum equation, the near wall viscosity is defined as
µ=µ*
P
*
Pw
U
y 4.19
The production term in the ε−k equation is calculated from the following equation
for the near wall node “P”,
( )P
21
P
41
wPky
kCP
µτ=ρ 4.20
Where, wτ , wall shear stress, can be calculated from,
110
( ) ( ) ( )Ptwt*
P
21
P
41
P
w UUEyln
kkC−
ρ=τ µ
4.21
( ) ( ) ( ) ( )[ ] wwPwPwPtwt nnUUUUUUrrrrrr
•−−−=− 4.22
Subscript “t” in the above equation denotes the tangential direction, parallel to the wall
surface. Diffusion flux and the value of k at the wall were considered to be zero. No
transport equation is solved for the dissipation rate. Assuming the turbulence is in
local equilibrium, the value of the dissipation was calculated from the following term,
P
23
P
43
Pky
kCµ=ε 4.23
4.3.1.4 INITIAL CONDITIONS AND FLUID PROPERTIES
The flow was started from t = 0 second in all the simulations with small initial values
assigned to k and ε, which made the initial turbulent viscosity roughly equal to the
kinematic viscosity for water. The fluid properties for air and water were taken as the
properties at NTP (T = 293.15 K, P = 1 atm).
The calculation for different injection conditions were solved as unsteady state
problem with time steps of ∆t = 0.01 second. Total time period for each run was 180
seconds which was adequate to obtain time averaged steady state results and also it
ensured numerical stability. Mean values (time averaged) of the transport properties
were calculated for the total simulation period of 180 seconds for each single run. To
get a converged solution, the approach used was to reduce the normalised sum of
absolute residuals to a value of 1.0 x 10-4
for the transport properties. The whole
simulation was carried out by using Swinburne’s Supercomputer in one cluster of 8
Intel Quad Core CPU, each with 2.3 GHZ speed.
Results obtained from several runs for this model including the grid independency test
is discussed in Chapter 5.
111
4.3.2 Zinc Fuming TSL Model
A 3D model of the Outotec TSL zinc fuming pilot plant was developed using CAD
tool. A schematic diagram of the model is shown in Figure 4-4. The modelled furnace
has a diameter of D = 0.5 m and length Z = 1.68 m. The modelled furnace was filled
up to L=0.6 m with ISF slag of composition shown by point A in Figure 4-6. A
vertical lance with an annulus of inner diameter di = 30 mm and outer diameter d
o=
42 mm was fitted at the centre of the furnace. Air was injected through the annulus of
the lance and CH4 as fuel through the central hole into the slag bath. Necessary heat in
the bath for smelting and reduction of the slag is supplied by combusting CH4 at the
lance tip.
A schematic diagram of the modelled furnace is shown on Figure 4-4. Figure 4-5
shows the generated 3D coarse grid, fine grid is not shown here for visual clarity.
Figure 4-4: Schematic diagram of the modelled furnace for Outotec TSL pilot plant
112
Figure 4-5: Generated grid of the modelled pilot plant scale TSL furnace for CFD
analysis
4.3.2.1 SLAG COMPOSITION
In this investigation, the modelled furnace was filled with ISF slag of composition
shown by point A in Figure 4-6. Minor constituents of the ISF slag were not taken into
taken account to avoid complexity. For the given temperature, partial pressure, as well
as the lime contents, the slag constituents are shown in
Table 4-1.
113
Table 4-1: Slag composition for TSL Zinc Fuming Model
Slag Constituent Initial %
ZnO 18%
SiO2 45%
FeO 27%
CaO 10%
Figure 4-6: Simplified phase relationships for the reduction step in an Outotec TSL
furnace for the components FeOx, ZnO, CaO and SiO2 generated by FACT Sage
[101] for the given temperature, partial pressure as well as the lime content.
114
4.3.2.2 MODEL FEATURES
The multiphase flow simulation was based on Eulerian approach where gas and liquid
phases interact with each other and there was significant exchange of momentum and
energy between phases due to robust combusting environment inside the furnace. In
addition to momentum and energy exchange between the phases, there was also mass
exchange at the gas-liquid interface due to the chemical reactions in the slag bath.
Undertaking the simulation was convoluted in a sense that the computational domain
was filled with compressible gas and incompressible slag with an abrupt change of
density, viscosity and other fluid properties at the gas-liquid interface in every
computational cell. The model was developed by using commercial CFD package
AVL FIRE 2009.2 (AVL, Graz, Austria) coupled with a number of user defined
subroutines (UDF), as the available graphical user interface (GUI) of the commercial
CFD package used does not allow combustion in multiphase flow problems. The basic
model features and the subroutines developed in the model are mentioned below,
1. The 3-D governing equations of momentum, continuity, enthalpy and turbulent
flows were solved in unsteady state for an unstructured grid system.
2. Standard k-ε turbulence model [129] was used for the turbulence modelling.
3. A cell centred finite volume approach was used to discretise the governing
equations and the resulting discretised equations were solved iteratively using
segregated approach.
4. For continuity equation, the values of the variables at cell faces were calculated
by employing central differencing approximation scheme with second order
accuracy, which uses a linear interpolation to compute the cell face values. For
other equations such as momentum, turbulent, and energy equations, a first-
order accurate upwind scheme was used.
5. Semi Implicit Method for Pressure Linked Equations (SIMPLE) algorithm
[127] were used to couple the pressure and velocity. In this algorithm, the
velocity (u, v and w) and pressure (P) fields are solved separately and coupling
between these field variables are achieved via velocity and pressure
corrections.
115
6. The subroutines developed in this model are written in FORTRAN
programming language and Intel FORTRAN compiler (version 10.1.019) was
used to compile and couple it with the CFD package. The subroutines include,
a. Subroutine for submerged CH4 combustion and species transport
b. Subroutine for interfacial enthalpy exchange
c. Subroutine for interfacial mass exchange
d. Subroutine for chemical reactions in the slag bath
4.3.2.3 GOVERNING EQUATIONS
Basic Eulerian equations, describing multiphase combusting system are given by the
conservation equations for continuity, momentum, energy and species transport. Other
than the basic conservation equations, interfacial exchange terms for mass, momentum
and energy at the gas-liquid interface were also modelled. For three-dimensional fluid
flow, conservation equations for continuity and momentum have already been
described in the previous section (4.3.1.2). In the continuity equation (4.1), the
interfacial mass exchange between phases k and l ( klΓ ) were considered.
In the momentum equation, Sato’s constant (equation 4.7) was reduced for the zinc
fuming TSL model as Csato
= 0.1. For the high temperature combusting system, the
dispersed phase bubble diameter (equation 4.13) was also adjusted as, Db = 0.01 m.
The bubble dispersion coefficient used in equation (4.10) was, CTD
= 0.01.
In addition to the interfacial mass and energy exchange, details of the enthalpy and
species transport equations solved for the multiphase TSL submerged combusting
system are described below.
4.3.2.3.1 Enthalpy conservation
Total enthalpy conservation equation solved for the model can be expressed as,
∑∑≠=≠=
Γ++
∂
∂α+⋅τα⋅∇+′′′α++α⋅∇=ρα⋅∇+
∂
ρα∂
N
kl,1l
klk
N
kl,1l
kl
kkkkkk
t
kkkkkkkkkk
hH
t
pvq)qq(hv
t
h
4.24
116
where kq ′′′ is the enthalpy volumetric source, heat flux, kq , is defined as,
k
k,p
kk h
c
kq ∇= 4.25
where kk is the phase k thermal conductivity, kh is the phase k enthalpy. Turbulent
heat flux, t
kq , equals:
k
T
t
kt
k hq ∇σ
µ= 4.26
Where, t
kµ is the turbulent viscosity defined in equation (4.26) and the turbulent
Prandtl number is adjusted to Tσ =0.5 for better turbulent heat transfer [132].
klΓ and klH in equation (4.24) represents mass and energy interfacial exchange
between phases k and l.
4.3.2.3.2 Interfacial Energy Exchange
Heat generated due to the CH4 combustion at the lance tip was transferred to the
molten slag phase by considering interfacial energy exchange at the gas-liquid
interface. Heat transfer between the two phases was modelled by using the Ranz-
Marshall enthalpy exchange model [133] as follows:
( ) dcdi
b
cc HTTA*Nu
D
kH −=−′′′= 4.27
where, c
k is the thermal conductivity of the molten slag phase, b
D is the bubble
diameter and i
A ′′′ is the interfacial area density defined in equation (4.15).
Nu is the Nusselt number and can be expressed as [133]:
3
1
2
1
6002 PrRe..Nu b+= 4.28
where, bRe is the local bubble Reynolds number, and Pr is the Prandtl number.
117
4.3.2.3.3 Combustion Modelling
Six different species (CH4, O2, N2, CO2, CO, H2O) were considered for gas phase
reaction during the combustion process. By considering TSL pilot plant practice data,
(assuming CO/CO2 = 1), the following equation for CH4 combustion was considered
at the lance tip to supply the necessary heat for ZnO reduction in the slag bath.
( ) 222224 76.3782276.374 NOHCOCONOCH ×+++→++ 4.29
A species transport equation for every species was solved for gas phase reaction,
which can be expressed as:
( ) ( ) kk
t
tm,kkk Sy
ScD.yv.y
t+α
∇
µ+ρ∇=ρ∇α+ρ
∂
∂α
r gask....1k = 4.30
Where, k
y represents the mass fraction of an individual chemical species k, ρ is the
density of gas phase. gask is the total number of chemical species and kS is the mass
source. mk
D,
[m2/s] is the diffusion coefficient for each species k in the mixture and
70.Sct = is the turbulent Schmidt number. Species source terms, kS , in equation
(4.30) was determined by the well-established Eddy Break-up combustion model
[134].
Turbulence controlled combustion model, Eddy Break-up (EBU), is a popular and
efficient model in combustion calculations, which was firstly proposed by Spalding
[135] and modified later by Magnussen and Hjertager [134]. The mean reaction rate
can be written according to Magnussen and Hjertager [134],
+ρ
τ=ρ
⋅
S1
yC,
S
y,ymin
Cr
prprOxfu
R
fufu
4.31
The rate of consumption of fuel is specified as a function of local flow properties, thus
it is dependent upon the turbulent time scale ( Rτ ), which is defined as a ratio of
118
turbulent kinetic energy (K) to its dissipation rate (ε ). The first two terms of the
minimum value of operator simply check if fuel or oxygen is present in limiting
quantity, and the third term is used for a reaction possibility. Cfu and Cpr are empirical
coefficients, and the exact values for these coefficients are dependent on the fuel and
the detailed structure of the turbulent flow field. In the present simulation, Cfu and Cpr
were kept 3.0 and 0.5 respectively.
4.3.2.3.4 Chemical Reactions in the Slag Bath
Other than the combustion reaction in the gas phase defined by equation (4.29), a
number of chemical reactions in the slag bath were considered with the following
scalar transport equation (4.32) solved for every scalar. Scalars considered for slag
phase reaction are ZnO, FeO, SiO2, CaO, C and Zn which fumes off from the slag
bath.
kikikki
t
tm,kkikkkkikk S
ScD.v.
t+φ∇α
φ∇
µ+ρ∇=φρα∇+φρα
∂
∂ 4.32
where kα and kρ is the volume fraction and density of phase k respectively, kiφ is the
value of scalar i of phase k, ki
S is the source term for different scalars. The chemical
reactions in the slag bath are based on the kinetics equations developed by Richards
and Brimacombe [20]. In the mathematical formulations by Richards and Brimacombe
[20], rate of transport of ZnO and ferric iron from the bulk slag to the slag/bubble
interface was expressed by the empirical equation,
( )*
i
sl
iibi CCkAn −= 4.33
Where, bA is the surface area of the secondary bubble/char particle (m2), ik is the
mass transfer coefficient of species i. sl
iC and *
iC are the concentrations of the species
i at the bulk slag and slag/bubble interfaces respectively.
The present simulation takes into account the rate of change of concentrations of every
scalar with respect to time, their convection and diffusion in the slag bath through
119
Equation 4.32. In addition, effect of turbulence in the chemical reactions was also
taken into considerations by applying the effect of turbulent eddies and its dissipation
rate into the chemical reactions which controls the mass transfer of the slag scalars
from the bulk slag to the slag-gas interface. This was done through inclusion of
turbulent time scale ( )kε in the transport equation (Equation 4.32), where k is the
turbulent kinetic energy and ε is its dissipation rate. Diffusion coefficients used in
Equation 4.32 for ZnO and FeO at 1473 K was 2.3x10-10
m2s
-1 , taken from the
calculations carried out by Richards and Brimacombe [20]. However, this molecular
diffusion is quite negligible as compared to turbulent diffusion.
In the simulation, carbon is added in the slag bath as a reductant source. The reaction
between carbon and CO2 gas (Boudouard reaction) described in equation (2.6) has
been considered according to rate kinetics given by Skinner and Smoot [136] for
pulverized bituminous coal char (70 pct through 200 mesh). The same rate kinetics
was also used by Richards et al. [20] for their mathematical modelling studies of slag
fuming. The rate equation follows first order kinetics with respect to solid carbon and
in CO2 partial pressure.
( )20 COaB PRTEexpAr ⋅−= 4.34
Where,
( ) 116
0 s.kgkPa1013.3A −−=
K23600REa =
The model was developed based on gas-liquid Eulerian multiphase flow approach,
hence it considers only two phases – gas and liquid slag. Zn vapour in the gas phase
was considered as a gas phase scalar which transforms from the liquid phase scalar
(ZnO) through the mass interfacial exchange between gas and liquid phases. The coal
char was treated as one of the scalar of the slag phase. The coal char was not treated as
a uniformly distributed dissolved species. A full governing transport equation was
solved to obtain the variable distribution of coal char in the slag. The hydrodynamic
effect of coal particle size was not considered in this study. However, effect of coal
particle size on the chemical reaction was taken into account by using the appropriate
Boudouard reaction rate (equation 4.34) for the coal particle size mentioned above. It
120
was also assumed that only fixed carbon of the coal takes part in the reduction
reaction. Similar assumptions have been used by previous researchers where they
assumed that only fixed carbon in the coal takes part in the reduction reaction [2].
Hence, no evaporation and devolatilization models are considered in the calculation.
4.3.2.3.5 Interfacial Mass Exchange
It is important to emphasize that the interfacial mass exchange occurs at the gas-liquid
interface due to the chemical reaction in the slag phase. The simulation considers
appropriate mass balance in the gas and liquid phases. Liquid ZnO from the slag phase
is being transformed to metallic Zn vapour. Phase transformation between molten slag
and gas phase was considered by the following equation [125],
dcellc
n
d
n
cmxc VC ΓρααΓ −=⋅= 21 4.35
Where, c and d denotes the continuous and dispersed phase, mxC is the rate of phase
transformation, which is defined by the rate of zinc oxide reduction, determined from
the convection, diffusion, turbulence and chemical kinetics. cρ and α is the density
and volume fraction respectively and volume fraction exponent used in the model are
n1 = 1 and n2 = 0.
4.3.2.4 BOUNDARY CONDITIONS
The boundary conditions used in the pilot plant scale zinc fuming TSL furnace model
are summarized below:
4.3.2.4.1 Inlet
Unlike the air-water model, a combustion chamber is included at the lance tip (see
Figure 4-4 and Figure 4-5) where fuel mixes with air and initiates the combustion
process. All boundary conditions were chosen to match the flow condition of the pilot
plant trials of the Outotec TSL pilot plant. At the lance tip, which was the inlet of the
computational domain, the mass flow boundary condition was used. Details of the
121
mass flow rate at the inlet are provided in Table 4-. At the inlet, three different species
were injected: O2, and N2 through the annulus ring and CH4 through the central hole.
4.3.2.4.2 Outlet
The outlet boundary was selected at sufficiently downstream from the regions of the
slag bath where the flow exhibits significant changes. Hence, a gradient zero boundary
condition was applied at the outlet of the computational domain with an outlet
diameter of 160 mm (see Figure 4-4).
4.3.2.4.3 Wall
A no-slip condition was applied at all walls inside the modelled furnace which
includes furnace inside wall including side wall, top and bottom wall, outside wall of
the cylindrical lance. The wall temperature was set to 1500 K according to the data
obtained from pilot plant study. The CFD model was set up to allow heat flux across
the wall to maintain the wall temperature at 1500 K. Besides, the walls were treated
with Compound Wall Treatment (CWT) proposed by Popovac and Hanjalic [137] to
take care of the near wall effects like viscous damping, kinematic blocking of the
velocity fluctuations normal to the wall.
The Compound Wall Treatment (CWT) ensures a gradual change between viscous
sub-layer formulations and the wall functions. This wall treatment provides the
standard wall function for the large values of y+ as well as the integration of equations
up to the wall (ItW) for the very small values of y+. The CWT model is based on the
description of the mean velocity and temperature profiles proposed by Kader [138],
thus,
ΓΓ 11 −+−++ += e)Eyln(k
eyU 4.36
Where, +
+
+=
yP
)yP(.
r
r
3
4
51
010Γ
Where, k = 0.4187 is Von Karman constant, E = 9.0 is an integration constant that
depends on the roughness of the wall. The production of turbulence kinetic energy near
122
the wall is modified accordingly as low-Re model for small values of y+ and standard
wall function (SWT) for high values of y+.
4.3.2.5 INITIAL CONDITIONS AND FLUID PROPERTIES
The usual case of zinc fuming process involves heating up the furnace by ignition,
charging of slag, melting the charge and adding reducing agent before fuming starts.
One complete batch fuming cycle usually takes around 120 to 180 minutes, depending
on the process. The present simulation avoids some of the complexities by initializing
the simulation at high temperature. The simulation starts with a charged furnace of
460 kg molten slag at 1500 K (1227o
C) temperatures and initiation of combustion at
the lance tip to supply the heat necessary for the zinc reduction reaction. Air and fuel
flow rate at the lance tip were provided from real flow rate data from a pilot plant
scale trail. Each run of the computation simulates 0.5 minute of a 10 minute interval of
fuming cycle, which gives a total of 4.5 minutes simulation in a 90 minute industrial
fuming cycle. The timing of simulation is linked to the computational time limitations.
Higher computational power can lead to a longer time simulation. The computation
was carried out on an Intel Xeon Quad Core Z 400 machine with 8 GB RAM and each
processor has a speed of 2.67 GHz. The simulation was run on MPI (Message Passing
Interface) mode [131] which splits the computational domain into four sections, each
of which was computed by a separate processor. Every single run of the simulation of
30 seconds took approximately 20 days.
The flow simulation was started with small initial values assigned to turbulent kinetic
energy (k) and its dissipation rate (ε), which made the initial turbulent viscosity
roughly equal to the kinematic viscosity for molten slag. For the gas phase reaction
given in Equation 4.29, fluid and thermal properties of the different species involved
in the solution process (density, specific heat, dynamic viscosity, molecular weight,
thermal conductivity, diffusion coefficient) have been considered from the internal
thermodynamic database of AVL FIRE [139]. The fluid and thermal properties for
molten slag phase are listed in Table 4-2.
123
Table 4-2: Fluid and thermal properties of molten slag phase for TSL model
Density (kg/m3) [20] 3900
Specific heat (J/kg K) [20] 870
Dynamic viscosity (N s/m2) [20] 0.5
Thermal conductivity (W/m K) [20] 1.5
Turbulent Prandtl number [132] 0.5
Reference pressure (Pa) 100000
Reference temperature 1500 K (1227 oC)
Table 4-3: Injection conditions (CFD and Experimental)
Experimental
conditions (CZF5)
(Waladan et al. [2]
CFD
Calculations
Fuel type Fine Coal CH4
Reductant type Fine Coal Fine Coal
Combustion air (kg/s) 0.06 0.05
Fuel rate (kg/s) 0.035 0.0035
Reductant rate (kg/s) 0.035
Initial zinc content in the
slag (wt %)
10.7 18.0
Results obtained from several runs for this model including the grid independency test
is discussed in Chapter 6.
124
4.3.3 Conventional Tuyere blow model
In this stage of the present study, the developed model for the zinc fuming TSL
furnace was applied to a thin slice model of the conventional tuyere blow furnace.
Submerged combustion model for CH4 was modified for submerged coal combustion
in this investigation. Details of the model features are described below.
4.3.3.1 MODEL GEOMETRY AND COMPUTATIONAL MESH
A two tuyere thin slice model of the conventional tuyere blow slag fuming furnace of
company D mentioned by Richards et al. [21] as shown in Figure 4-7 was developed
using CAD. As mentioned by Richards et al. [21], the slag fuming furnace of company
D has a length of 4.57 m and width 2.44 m (4.57 x 2.44 m2), containing 30 tuyeres
with 56 mm ID for each tuyere. Owing to the computational time limitations, present
study dealt with a thin slice model with two opposing set of tuyeres. The principle aim
was to investigate tuyere tip combustion dynamics, coal utilization behaviour and zinc
fuming kinetics inside the bath. Keeping the width same as the original furnace
dimension (2.44 m), a slice of 0.3 m length was considered along the longitudinal
direction (along Y coordinate in Figure 4-7). The walls in the X – Z plane has been
considered as the symmetry wall. Two opposing set of tuyere were placed on both side
wall of the modelled furnace, the centre of which were placed at 0.1 m above the
bottom wall and 0.15 m along Y coordinate (Y = 0.15m, Z = 0.1 m). Both the tuyere
tips were extruded 0.1m from both the side walls (in Y – Z plane). The modelled
furnace was filled up to L=1.0 m with ISF slag. Figure 4-7 represents the schematic
outline of the developed model. Air as oxidant and coal as fuel was injected through
the tuyeres into the molten slag bath. Necessary heat in the bath for smelting and
reduction of the slag is supplied by combusting coal at the tuyere tip. A comparison of
the simulation and experimental conditions are summarized in Table 4-4 for more
clarity.
125
Figure 4-7: Schematic view of the modelled thin slice rectangular tuyere blow furnace
(Isometric Layout)
Table 4-4: Comparisons of the simulation and plant data
Parameter
Real Furnace of
Company D (Richards
et al. [21]) – Run 1
Present CFD
Simulation
Length (m) 4.57 0.3
Width (m) 2.44 2.44
No. of tuyeres 30 2
Tuyere ID, do (m) 0.056 0.056
Bath weight (tonnes) 45 2.855
Bath height (m) 1 1
Initial zinc content % 11.7 11.7
126
Coal
composition
FC % 60 60
Volatile
content %
20 20
Ash % 18 18
Coal rate through each
tuyere (kg/s)
0.032 0.032
Air flow rate through
each tuyere (kg/s)
0.158 0.158
(a) (b)
Figure 4-8: (a) Generated surface mesh, (b) Volume mesh for CFD analysis (course
grid)
127
4.3.3.2 SLAG COMPOSITION
In this investigation, the modelled furnace was filled with molten slag of composition
shown in Table 4-5. Minor constituents of the ISF slag were not taken into taken
account to avoid complexity. As presented by Richards et al. [21] (Figure 5, Cycle
D1), initial zinc, silica and lime contents are 11.7%, 26%, and 15.3% respectively.
Though, effect of ferric iron is likely to play some role in overall fuming efficiency,
current CFD investigation does not consider the effect of ferric iron, as an attempt to
avoid complexity in chemical reactions. Hence, ferric iron level was considered as
zero. The proportion of minor elements are also not mentioned by Richards et al. [21],
hence, the remaining of the slag composition was considered as FeO.
Table 4-5: Initial Slag Composition for Tuyere blow model
Slag Constituent Initial %
ZnO 11.7
SiO2 26
FeO 47
CaO 15.3
4.3.3.3 MODEL FEATURES
The multiphase flow simulation is based on the principle of interpenetrating continua,
i.e. Euler – Euler approach. In this approach, each phase is governed by the Navier-
Stokes equations. The existing phases share the same volume and penetrate each in
space and exchange mass, momentum and energy. Each phase is described by its
distinctive physical properties and has its own velocity, pressure, concentration and
temperature field. The model was developed by using commercial CFD package AVL
FIRE 2009.2 (AVL, Graz, Austria) coupled with a number of user defined subroutines
(UDF) for submerged coal combustion and gas phase species transport, chemical
reaction in the slag bath and interfacial mass and energy exchange (as described in
Section 4.3.2.2). The notable model features are kept same as the zinc fuming TSL
model (described in Section 4.3.2.2)
128
4.3.3.4 GOVERNING EQUATIONS
The governing transport equations describing Eulerian multiphase fluid flow
(continuity, momentum, energy and species transport) were solved in this model. In
addition, mass, momentum and energy interfacial exchange at the gas-liquid interface
were also considered, as described in the zinc fuming TSL model (Section 4.3.2.3).
Interfacial exchange terms (mass, momentum and energy) at the gas-liquid interface
were modelled by applying appropriate interfacial exchange models. Details of the
interfacial mass, momentum and energy exchange models are described in the zinc
fuming TSL CFD modelling work. Coefficients used in the interfacial exchange terms
are as follows:
For shear induced turbulence viscosity, µC = 0.09
Sato’s coefficient for bubble induced turbulence, Csato
= 0.1
Bubble diameter for interfacial momentum and energy exchange, Db = 0.01 m
The bubble dispersion coefficient, CTD
= 0.01
In this part of the research, the gaseous phase CH4 combustion model was modified
for coal combustion. The coal was injected through tuyeres along with air. Details of
the coal combustion modelling are discussed in the following sections.
4.3.3.4.1 Coal combustion
The numerical simulation of raw coal combustion is considered as a complex process
compared with the combustion of other fossil fuel sources because it includes several
complicated physical and chemical processes, which have not been completely
understood. The basic steps of coal combustion are thermal decomposition and the
consequent burnout of the volatile matter and the oxidation of char to leave the
incombustible ash as a final undesirable part. The physical properties (proximate and
ultimate analysis) of coal particle used in the present study are summarized in Table 4-
4. One of the assumptions of the present CFD investigation is that the injected coal
129
through the tuyeres has been considered as a continuum phase rather than discrete
particulate phase. The aim was to avoid complexity, as solid coal particle would
incorporate the third phase other than the gas and liquid, which would bring the
complexity of interfacial exchange terms. The injected coal was treated as one of the
scalars of the slag phase. The coal was not treated as a uniformly distributed dissolved
species. A full governing transport equation was solved to obtain the variable
distribution of coal char in the slag.
Scalar transport equation solved for the injected coal can be written as,
kkkk
t
tm,kkkkkkkk S
ScD.v.
t+φ∇α
φ∇
µ+ρ∇=φρα∇+φρα
∂
∂ 4.37
where kα and kρ is the volume fraction and density of phase k respectively, kφ is the
scalar value (injected coal) of phase k, kS is the source term for injected coal, which
was determined through the coal combustion model as described below.
In the third stage of this research, coal combustion was incorporated by three complex
reaction processes. The combustion of dry coal particles mainly includes two complex
reaction processes. The first reaction process is the devolatilization of the dry coal
particle, which includes the reaction of the released hydrocarbon fuel (volatile) which
combusts with the oxygen to produce essentially the water vapor (H2O) and carbon
dioxide (CO2) as final products. The second reaction process is the oxidation of the
residual char that is slower than the devolatilization process. In this study, injected coal
through the tuyeres was considered to participate both in the combustion and in the
reduction reactions within the bath. No water evaporation model was incorporated in
the model as the injected coal was assumed to have negligible moisture content. Total
coal reaction mechanism is shown in the following flow chart:
130
Figure 4-9: Coal combustion process flow chart
4.3.3.4.2 Devolatilization
In the present simulation, the single reaction model of Badzioch and Hawksley [140]
was applied to simulate the coal pyrolysis. The coal, considered in this model, was
assumed to have fixed fraction of volatile matters, char and ash, and its reaction
depends on the local temperature experience, as well as its temperature history. The
rate of production of the volatile is given by the first order reaction as follows:
( )VVKdt
dVfv −= 4.38
Where, V is the product of volatiles that have already released from unit mass of
pulverized coal at time t, fV is the ultimate product of volatiles and vK is the rate
constant given by the Arrhenius form as,
−
p
vv RT
EexpA , where R = 8314 J kmol
-1
K-1
is the universal gas constant, pT is the temperature of coal particle, vA and vE are
the pre-exponential factor and the activation energy constants, respectively, that are
131
determined experimentally for the particular coal. These factors are usually obtained
from the proximate analysis of the coal.
The good selection of the kinetic factors of Arrhenius expression can be considered as
a key issue towards the appropriate prediction of devolatilization rate, particularly in
the combustion zone [140, 141]. Therefore, the devolatilization kinetic factors of the
coal used were chosen to be 2.0x104 (s
-1) and 4.94x10
7 (J.kmol
-1) for the pre-
exponential factor and for the activation factor [142], respectively, in order to
overcome any difficulties during setting up the simulation cases and to get optimum
agreement with the available physical data.
4.3.3.4.3 Gas Phase Combustion
Species source term for CH4 was updated continuously, which is generated inside the
bath as a result of devolatilization process of coal. Detail of the coal devolatilization
process is described in the previous section (Section 4.3.3.4.2). Six different species
(CH4, O2, N2, CO2, CO, H2O) was considered for gas phase reaction during the
combustion process. The following equation for CH4 combustion was considered in
the slag bath,
( ) 222224 N76.37OH8CO2CO2N76.3O7CH4 ×+++→++ 4.39
A species transport equation for every species was solved for gas phase reaction,
which can be expressed as:
( ) ( ) kk
t
tm,kkk Sy
ScD.yv.y
t+α
∇
µ+ρ∇=ρ∇α+ρ
∂
∂α
r gask....1k = 4.40
Where, k
y represents the mass fraction of an individual chemical species k, ρ is the
density of gas phase. gask is the total number of chemical species and kS is the mass
source. mk
D,
[m2/s] is the diffusion coefficient for each species k in the mixture and
70.Sct = is the turbulent Schmidt number.
132
Where, mk
D,
[m2/s] is the diffusion coefficient for each species k in the mixture and
70.Sct = is the turbulent Schmidt number. Species source term, ky
S , in equation
(4.40) were determined by the well-established Eddy Break-up combustion model
[134], as described in Section 4.3.2.3.3. Empirical coefficients Cfu and Cpr for this
investigation were considered as 3.0 and 0.5 respectively.
As shown in the flow chart (Figure 4-9), the remaining char takes part in both
combustion and reduction reaction of the zinc oxide within the bath. A portion of the
remaining char also bypasses the bath un-combusted and un-reacted.
4.3.3.4.4 Char oxidation
The char oxidation rate is an important process in pulverized coal combustion. After
the devolatilization, the remaining char in the coal particle reacts slowly with the
surrounding gases. Therefore, the burnout time of the pulverized coal can be
determined in the furnace by this process [143].
In this study, the char combustion is modelled by a global reaction of order unity
(global power-law), which was proposed by Field et al. [144]. The diffusion rate of
oxygen is calculated by ( )sgd PPK − , where gP is the partial pressure of oxygen in the
bulk phase of the furnace (far from particles boundary layer), sP is the oxygen partial
pressure at the external surface of the particle and dK is expressed by the following
equation [145],
P
PTT
RK Agp
p
d
75.07
2
10*53.2
+=
−
4.41
Where: pR is the radius of the particle, pT is the temperature of the particle, gT is the
gas temperature in the far field, and AP and P are the atmospheric and local pressure
respectively.
133
The rate of char oxidation per unit area of the particle surface is described by sc PK .
The kinetic rate is expressed by the following Arrhenius expression,
−=
p
ccc T
EAK exp 4.42
Where, cA is the pre-exponential factor and cE is the activation energy. Due to the
lack of measured char reactivity data, the recommended values by Wall et al. [146] are
used for cA and cE in this simulation, which are 497 (kg.m-2
.s-1
.atm-1
) and 8540 K
respectively. Finally, the rate of the overall char reaction of a particle can also be
written as followsA
pg
cdP
PRP
KK
2
114
1π
+−−
and can be controlled by the smaller
rates of dK and cK .
4.3.3.4.5 Chemical Reactions in the Slag Bath
Some portion of the remaining char reduces the zinc oxide present within the bath.
Details of the chemical reaction modelling has been described Section 4.3.2.3.4. For
the scalars that involved chemical reactions in the slag bath, the following scalar
transport equation (4.43) was solved for every scalar. Scalars considered for slag
phase reaction are ZnO, FeO, SiO2, CaO, C and Zn which fumes off from the slag
bath.
kikikki
t
tm,kkikkkkikk S
ScD.v.
t+φ∇α
φ∇
µ+ρ∇=φρα∇+φρα
∂
∂ 4.43
where kα and kρ is the volume fraction and density of phase k respectively, kiφ is the
value of scalar i of phase k, ki
S is the source term for different scalars.
134
4.3.3.5 BOUNDARY CONDITIONS
All boundary conditions were chosen to match the flow condition of the plant data
(Company D) given by Richards et al. [21]. The boundary conditions used in the
model are summarized below:
4.3.3.5.1 Inlet
At the tuyere tip, which was the inlet of the computational domain, the velocity
boundary condition was used. Details of the flow rate at the inlet are provided in Table
4-4. At the inlet, both the gas and liquid phase were injected. The gas phase contains
two different species: O2, and N2 and the liquid phase contain coal only.
The plant data presented by Richards et al. [21], shows for company D1, total blast
was 4.0 m3/s and the blast temperature was 460 – 520
o C, which gives the average
blast temperature of 490o C. At each tuyere, volume flow rate of air at STP would be
0.133 m3/s and mass flow rate of 0.158 kg/s. With that specified volume flow rate and
blast temperature, inlet jet velocity was calculated as 139.6 m/s. Hence, at inlet, the
same velocity and blast temperature were kept.
4.3.3.5.2 Outlet
Outlet was defined at the top wall of the computational domain as shown in Figure
4-7. Static pressure boundary condition was applied at the outlet of the computational
domain.
4.3.3.5.3 Wall
All walls inside the modelled furnace, which includes furnace side wall, top and
bottom wall, were applied a no-slip condition. The temperature at the walls was
assumed as 1500 K (1227 oC). As mentioned before (Section 4.3.2.4.3), the CFD
model was set up to allow heat flux across the wall to maintain the wall temperature at
1500 K. Besides, the walls were treated with Compound Wall Treatment (CWT) to
take care of the near wall effects like viscous damping, kinematic blocking of the
velocity fluctuations normal to the wall. Detail of the Compound Wall Treatment
(CWT) has been described in the previous section (Section 4.3.2.4.3).
135
4.3.3.5.4 Symmetry
In the present simulation, it was assumed that symmetric state exists on the two
opposing sides of the furnace as shown by symmetry plane in Figure 4-7. This
treatment of the boundary condition corresponds to the physical assumption that, on
the two sides of boundary, the same physical processes exist. The variable values at the
same distance from the boundary at the two sides are the same. The function of such a
boundary is that of a mirror that can reflect all the fluctuations generated by the
simulation region. When the flow is bounded by a plane of symmetry, the velocity
component normal to this plane is set equal to zero, yielding zero convective flux. In
addition, the normal derivatives of all the remaining variables are set to zero which
implies zero diffusion fluxes.
4.3.3.6 INITIAL CONDITIONS AND FLUID PROPERTIES
The conventional zinc fuming process usually comprised of several steps. It begins
with heating up the furnace by ignition, followed by charging of slag, melting the
charge and adding reducing agent before fuming starts. One complete batch fuming
cycle usually takes around 120 to 180 minutes, depending on the process. The present
simulation avoids some of the complexities by initializing the simulation at high
temperature. The simulation starts with a charged furnace of 2855 kg molten slag at
1500 K (1227 oC) temperature and initiation of combustion at the tuyere tip to supply
the heat necessary for the zinc reduction reaction. Air and coal flow rate at the tuyere
tip were provided from plant data as mentioned in Table 4-. Each run of the
computation simulates 1.0 minute of a 10 minute interval of fuming cycle.
Computational time limitation restricts the longer simulation time. The computation
was carried out with a fixed time step of 5 x 10-4
second on an Intel Xeon Quad Core
Z 400 machine with 8 GB RAM and each processor has a speed of 2.67 GHz. The
simulation was run on MPI (Message Passing Interface) mode [131] which splits the
computational domain into four sections, each of which was computed by a separate
processor. Every single run of the simulation of 60 seconds took approximately 15
days.
136
For the gas phase reactions, fluid and thermal properties of the different species
involved in the solution process (density, specific heat, dynamic viscosity, molecular
weight, thermal conductivity, diffusion coefficient) has been considered from the
internal thermodynamic database of AVL FIRE [139]. The fluid and thermal
properties for molten slag phase are listed in Table 4-66.
Table 4-6: Fluid and thermal properties of molten slag phase for Tuyere blow model
Density (kg/m3) [20] 3900
Specific heat (J/kg K) [20] 870
Dynamic viscosity (N s/m2) [20] 0.5
Thermal conductivity (W/m K)
[20]
1.5
Turbulent Prandtl number [132] 0.5
Reference pressure (Pa) 100000
Reference temperature 1500 K (1227 oC)
Results obtained from several runs for this model including the grid independency test
is discussed in Chapter 7.
137
Chapter 5
138
5 Cold Flow CFD Model of the TSL Gas Injection Process
This chapter will focus on the detail hydrodynamic parameters of the cold flow CFD
model of top submerged lance (TSL) gas injection process. Effect of the different
injection parameters (gas flow rate, lance submergence level, swirl intensity) and fluid
properties (liquid density and viscosity) on the bath mixing will be discussed in the
subsequent sections. Details of the modelling techniques are discussed on Chapter 4
(Section 4.3.1).
The aim of the research presented in this chapter is to investigate the physical
behaviour of the top submerged gas injection system and to predict the effect of swirl,
lance submergence level and air injection rate into the liquid bath using the CFD
modelling technique. The present study is a numerical simulation of the cold model
experimental work of Morsi et al.[1]. In the present study, water was used as the
modelling fluid and air was used as the injected gas as it was the basis for the previous
experimental model of Morsi et al.[1].
Mixing within the bath was considered based on both micro-mixing (turbulence
mixing) and macro-mixing (volume exchange effectiveness) approach. A new
approach (volume exchange effectiveness) to express the degree of mixing in the
liquid bath in metallurgical process simulation has been proposed in the current study.
In addition, a modified semi-empirical equation is proposed to measure the vertical
depth of penetration of the air jet injected through the annulus of the lance into the
liquid bath based on the previous experimental study of Iguchi et al. [51].
5.1 Test of Grid Independence
Accuracy of every CFD analysis depends largely on the type of grid used. Hence, grid
generation needs special attention of the investigators. Generally, finer grid gives more
accurate results, but eventually it leads to higher computational expense. Therefore, an
optimum grid resolution is necessary for efficient CFD analysis. The purpose of grid
independency test is to determine the minimum grid resolution required to generate a
solution that is independent of the grid used.
139
In this air-water model, four grid resolutions were tested for grid independency test,
mainly increasing the number of cells in the water bath. Starting with a coarse grid,
number of cells was increased in the region of interest until the solution from each grid
was unchanged for successive grid refinements. All the cells in the calculation domain
were polyhedral with a large number of hexahedral cells. As the computational
domain consisted of hybrid unstructured meshes in curvilinear non-orthogonal
coordinate system with Cartesian base vectors and refined regions in some locations,
mentioning number of cells in each direction is complicated. The computational grid
(213344 cells) used in the present study is too dense for visual presentation. A cross
sectional view of the coarse computational grid in X-Z plane is shown in Figure 4-2,
which consists of total 89492 cells in 360o domain. Meshing procedure was done by
Fame Advanced Hybrid meshing technique [147]. Table 5-1 represents an overview of
the grid information studied in the grid independency test.
Table 5-1: Overview of computational grids
Name Grid Density No of
Computational cells
Grid 1 Coarse 89492
Gird 2 Medium 154072
Grid 3 Fine 213344
Grid 4 Very Fine 361024
Figure 5-1 shows the mean tangential velocity distribution (V) on X-Z plane. The
radial distance (r) is normalised by the radius of the cylinder (R=115mm) and the axial
distance z is normalised by the length of the cylinder (Z=560mm). Both Grid 3 and
Grid 4 gave a very close prediction compared with the experimental results. The
difference in predictions between Grid 3 and Grid 4 was small enough (around 5%) to
suggest that any further grid refinement would not yield a substantially different
profile in that plane. Hence, it was decided that the fine grid resolution (Grid 3) of
213344 cells was sufficient to obtain grid independent results.
140
Figure 5-1: Mean tangential velocity (m/s) distribution for different grid
configurations
(Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =0
o) for water model simulation
Overview of the simulation and experimental conditions are mentioned in Table 5-2.
The calculation for different simulation condition as mentioned in Table 5-2 were
solved as unsteady state problem with time steps of ∆t = 0.01 second. Total time
period for each run was 180 seconds which was adequate to obtain time averaged
steady state results and also it ensured numerical stability.
Table 5-2: Overview of the simulation and experimental conditions
Parameters Experimental
Condition
(Morsi et al. [1])
Present Simulation
Air injection rate
Q (Nm3/s)
Fraction of lance
submergence H/ L
Swirl Intensity
Ф (o)
1.50 x 10-3
2.67 x 10-3
1/3, 2/3
0o, 57.5
o
1.00 x 10-3
1.50 x 10-3
2.00 x 10-3
2.67 x 10-3
3.50 x 10-3
4.00 x 10-3
1/6, 1/3, 2/3
0o, 57.5
o
141
5.2 Results and Discussion
The results discussed here are presented in terms of the three major hydrodynamic
parameters: swirl intensity, gas injection rate and lance submergence level. The
simulation condition and corresponding figures that are described in this paper are
summarised in Table 5-3, referring to air-water system unless otherwise stated.
Table 5-3: Simulation conditions and corresponding figures
Air injection rate
Q (Nm3/s)
Fraction of lance
submergence H/ L
Swirl Intensity
Ф (o)
Figure number
2.67 x 10-3
2.67 x 10-3
2.67 x 10-3
2.67 x 10-3
1.50 x 10-3
1.50 x 10-3
1/3
2/3
1/3
2/3
1/3
1/3
0
0
57.5
57.5
0
57.5
4(d), 5(d), 6(d), 7(d), 16(a)
3, 4(b), 5(b), 6(b), 7(b),8,
12(a), 13(a)
4(c), 5(c), 6(c), 7(c), 16(b)
4(a), 5(a), 6(a), 7(a), 12(b),
13(b)
4(f)
4(e)
5.2.1 Effect of Swirl Intensity
The effect of swirl intensity on axial velocity is shown in Figure 5-2. The figure shows
that the instantaneous axial velocity (w) contours for swirl and non-swirl flow were in
the range of -0.3 to 0.4 m/s. Time instances are mentioned in the corresponding
figures at which contour plots are taken. The colour bar represents the velocity
magnitudes (m/s). The sign in the colour bar indicates the direction of the velocity
(either downward or upward). The axial velocity near the lance shows an upward trend
due to the buoyant force of the rising air bubbles. No significant change in axial
velocity was observed due to swirl. From Figure 4, swirl injection seems to have
larger penetration envelope in the case of 2/3 lance depth, whereas the reverse is true
142
for 1/3 lance depth. This observation can be attributed to the transient nature, sloshing,
and splashing phenomena in the water bath. By comparing Figure 5-2 (a) and (b), it is
clear that there is only a little increase in the axial velocity at the bottom of the tank
and near the top surface of the bath. This trend is also quite clear from Figure 5-2 (c)
and (d). There is some change in axial velocity due to the change in flow rate which
can be seen by comparing Figures 5-2 (c), 5-2 (e) and 5-2 (d), 5-2 (f). These velocity
contours showed reasonable agreement with the experimental results of Morsi et al.
[1] presented in Figure 5-3. Comparison of the Figures 5-2 (b) and 5-3 (b) shows that
there is around 5% discrepancy between the experimental and the simulation results
near the lance tip. Nevertheless, the simulation results predicted some higher value in
the remaining portion of the liquid bath as the movements of the fluid particles was
not uniform and there was generation of turbulence inside the bath. The transient
effect of the flow fields might be one of the factors that caused the discrepancies in the
contour plots between the simulation and experimental results. The contour plots
presented here from the simulation are instantaneous, but in Figure 5-3 the contour
plots from the experimental results of Morsi et al. [1] are time averaged. Time
averaged experimental results are shown here for qualitative validation purpose only.
This validation exercise represents the qualitative accuracy of the present study against
existing experimental data.
5-2 (a) 5-2 (b)
143
5-2 (c) 5-2 (d)
5-2 (e) 5-2 (f)
Figure 5-2: Axial velocity (w) distribution (m/s) for the water model simulation
(a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 57.5
o, t = 60 sec
144
(b) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =0
o, t = 60 sec
(c) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o, t = 60 sec
(d) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 0
o, t = 60 sec
(e) Q=1.5 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o, t = 60 sec
(f) Q=1.5 x 10-3
m3/s, H/L=1/3, Ф = 0
o, t = 60 sec
Figure 5-4 shows the instantaneous tangential velocity (v) contours on X-Z plane. The
effect of swirl is noticeable from the figures. Significant increase in tangential
velocities will only occur due to the presence of swirl component, which is revealed
by comparing the Figures, 5-4 (a), 5-4 (b) and 5-4 (c), 5-4 (d). The change in
tangential velocity component due to swirl is much higher than the change in axial
velocity component. The tangential velocity distribution in Figure 5-4(d) revealed the
so-called dead water region in the bath. It shows that the bottom half portion of the
liquid bath is almost unaffected by the gas injection process for the case of 1/3
submergence and non-swirl case. The simulation results for tangential velocity
contours are in good agreement with experimental results of Morsi et al. [1], as shown
in Figure 5-5.
5-3 (a) 5-3 (b)
145
5-3 (c) 5-3 (d)
Figure 5-3: Axial velocity (w) distribution (m/s) from experimental results of Morsi et
al. [1]
(a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 57.5
o
(b) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =0
o
(c) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф =57.5
o
(d) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф =0
o
5-4 (a) 5-4 (b)
146
5-4 (c) 5-4 (d)
Figure 5-4: Tangential velocity (v) distribution (m/s) for the water model simulation
(a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 57.5
o, t = 60 sec
(b) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 0
o, t = 60 sec
(c) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o, t = 60 sec
(d) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 0
o, t = 60 sec
5-5 (a) 5-5 (b)
147
5-5 (c) 5-5 (d)
Figure 5-5: Tangential velocity (v) distribution (m/s) from experimental results of
Morsi et al. [1]
(a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 57.5
o
(b) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =0
o
(c) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o
(d) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф =0
o
Figure 5-6 shows a comparison of mean tangential velocities between swirl and non-
swirl injection. The experimental data are also presented in Figure 5-6 for validation.
The radial distance (r) is normalised by the radius of the cylinder (R=115mm) and the
axial distance z is normalised by the length of the cylinder (Z=560mm). As expected,
the magnitudes of the tangential velocities are low under non-swirl conditions.
Tangential velocity near the lance shows a significant rise for swirl condition, but
drops off to around zero after r/R ≥ 0.20. The mean tangential velocities are calculated
at z/Z=0.92 which is just below the exit of the lance z/Z=0.91 for H/L = 2/3. The mean
tangential velocity distribution for non-swirl flow agrees well with the existing
experimental values of Morsi et al. [1]. The discrepancy between the two results lies
within a range of 0-10%. But for the case of swirl flow, present simulation shows the
peak value of mean tangential velocity 1.1 at a radial distance of r/R = 0.11, whereas,
from the experimental results of Morsi et al. [1], the peak was found to be 0.7 at a
radial distance of r/R = 0.128. The discrepancies between the experimental and the
simulation results may be attributed to the following,
148
(1) Differencing scheme used for momentum and turbulence is upwind, which
gives false diffusion in complex flow phenomena. However, this trend is quite
reduced by using fine grids in the liquid bath.
(2) The standard k-ε turbulence model [129] may give poor performance in a
number of important cases such as flows with large extra strains (e.g. curved
boundary layers, swirling flows) and rotating flows [148]. Still, the reason for
using this model is because it is well established, most widely validated
turbulence model and it gives excellent performance for many industrially
relevant flows.
(3) Inaccuracy of ± 6% associated with the experimental technique such as
optical component alignment, seeding, filtering, signal processing and
calibration [1].
Figure 5-6: Mean tangential velocity comparison between swirl and non-swirl flow
from the simulation results and comparison with water model experiment of Morsi et
al. [1] (z/Z= 0.92, H/L= 2/3, Q=2.67 x 10-3
m3/s)
5.2.2 Effect of Submergence Level
Different lance submergence level also plays a significant role on the fluid flow
characteristic in the top submerged lance (TSL) gas injection systems. The theory of
greater volume of splash generation put forward by Koh and Taylor [52] is revealed in
the present simulation data. In addition to 1/3 and 2/3 lance submergence level, the
149
simulation was extended for 1/6 submergence level. Figure 5-7 shows the time
averaged volume fraction of water, generated by splashing at 30 mm height above the
bath (z/Z = 0.68). No surface tracking method like Volume of Fluid was used in the
present simulation to quantitatively represent formation of each and every small
droplets generated from splashing. Tracking of each and every small droplets would
require massive computer resources and time. This was avoided in the present study
where the qualitative flow pattern in the liquid bath was of main interest. In the
present simulation, as the approach used was conventional Eulerian, time averaged
volume fraction is measured at certain heights above the liquid bath to get a qualitative
idea of splash. For a more quantitative analysis of the splash formation, surface
tracking methods like Volume of Fluid (VOF) have to be used. From the Figure it is
evident that increasing submergence level generates greater volume of splash. It is due
to an increase in penetration depth of air jet in deeper bath resulting in a greater
release of buoyancy energy, which produces more splashes. This result is also
consistent with the experimental study of Igwe et al. [53]. From the water model
experimental study of Igwe et al. [53], the authors qualitatively reported that the
degree of splashing increased with the increase in depth of submergence.
Figure 5-7: Average volume fraction of water at 68.0=Z
z for different submergence
level for the water model simulation (Q=2.67 x 10-3
m3/s, Ф = 57.5
o)
150
5.2.3 Effect of Air Flow Rate
The depth of penetration of air jet increases with increasing air flow rate, which can be
seen clearly by comparing Figure 5-2 (c), 5-2 (e) and 5-2 (d), 5-2 (f). Figure 5-2 (e)
and 5-2 (f) show the so-called dead water region near the bottom of the cylindrical
vessel used for water modelling. The increase in depth of penetration of air jet
provides better agitation into the bath hence better mixing. The penetration of the air
jet is a function of the term, rF ′ , which is a modification of the jet Froude number put
forward by Igwe et al. [53],
( )dg
vrF
g
g
ρρ
ρ
−=′
1
2
5.1
where 1ρ is the liquid phase density, gρ is the density of gas, g is the gravitational
constant, v is gas flow velocity and d is the orifice diameter. The higher the number,
rF ′ , the greater the jet penetrates into the liquid bath. Iguchi et al. [51] developed a
semi-empirical equation from an air-water experimental study to calculate the vertical
penetration distance of the air jet for top submerged lance gas injection which also
depends on air flow rate. According to Iguchi et al.[51], the semi-empirical equation
for vertical penetration distance of the injected air into the liquid bath can be
expressed as,
31
1.4 mnv FrdL = , 2< Frm < 6x103
5.2
where Lv is the vertical penetration distance of the injected air,
nd is the nozzle inner
diameter at the exit and mFr is the modified Froude number which can be expressed
as,
5
2
nL
gg
mgd
QFr
ρ
ρ= 5.3
where gρ is the density of gas, Lρ is the density of liquid, Qg is the gas flow rate and
g is the acceleration due to gravity. Equation 5-2, however, is not valid for the case of
the annulus air inlet as it is used in swirled lances. For air jet injection through the
annulus of the top submerged lance, a semi-empirical equation is proposed from the
151
present simulation data, based on the relationship proposed by Iguchi et al. [51], which
can be expressed as,
( ) 4745.0275.0 miova FrddL −= 5.4
where, vaL is the vertical penetration distance for air jet injected through annulus, od
and id is the outer and inner diameter of the lance respectively neglecting the
thickness of the lance wall, mFr is the modified Froude number which can be obtained
by Equation 5-3.
Figure 5-8 shows the relation between vertical penetration distance of the annulus air
jet (va
L ) and the modified Froude number (m
Fr ). The coefficients in equation 5-4 are
found from the fitted curve shown in Figure 5-8 with the correlation factor,
0.98R =2. Six different flow rates were used as mentioned in Table 5-2 for 1/3
submergence level and non-swirl flow. Vertical penetration distance of the air jet was
measured as the mean value for different time steps.
Figure 5-8: Relation between vertical penetration distance for annulus air injection
(va
L ) and modified Froude number (m
Fr ) as derived from the water modelling
simulation results (H/L=1/3, Ф = 0o)
152
5.2.4 Mixing in the Liquid Bath
Process kinetics in mixing phenomena in the real Top Submerged Lance (TSL)
smelting furnaces are quite complex. Mixing in the bath in real furnace scenario is
quite vigorous and there are several factors affecting the mixing process, a number of
which are mentioned below:
High temperature chemical reactions in the slag are dominant factors
affecting the mixing phenomena
Expansion of gases in the molten bath due to high temperature and air
injected through the lance accelerates the mixing
Sidewise and vertical movement of the lance in the molten bath affects
the total mixing process
Splashing phenomena at the free surface also increases the mixing
process in the bath
However, in the present simulation, only the isothermal cold model air-water system is
considered. Hence, many of the factors affecting the mixing phenomena are absent. In
the present simulation, the mixing phenomena studied are turbulence mixing through
the turbulent diffusion and macro mixing via convection.
5.2.5 Turbulence Mixing
Figure 5-9(a) and Figure 5-9(b) show the effect of swirl on turbulent kinetic energy (k)
distribution for the same flow rate and submergence level. The colour bar is showing
the magnitude range of the turbulence kinetic energy in the figure from 0 to 0.2 m2/s
2.
Figure 5-10(a) and Figure 5-10(b) show the turbulent kinetic energy distribution from
the experimental results of Morsi et al.[1]. The values obtained from the present study
are consistent with the values observed in the experimental study. Generation of
turbulence near the lance is increased in the case of 2/3 lance submergence level and
swirl flow. The maximum value of turbulent kinetic energy exists near the lance as
expected which was also revealed from existing experimental data. However, this
turbulence is significantly reduced with increasing distance from the lance tip and it
does not exist near the vessel wall. Though there is no noticeable change in turbulent
kinetic energy for swirl and non-swirl injection as shown in Figure 5-10(a) and Figure
153
5-10(b), our present simulation results show a noticeable change in the generation of
turbulent kinetic energy near the lance. The rising gas plume is extended radially from
the lance toward the wall as a result of swirl air injection.
5-9 (a) 5-9 (b)
Figure 5-9: Turbulent Kinetic energy (k) distribution (m2/s2)- (a) Q=2.67 x 10-3
m3/s,
H/L=2/3, Ф = 0o, t = 60 sec, (b) Q=2.67 x 10
-3 m
3/s, H/L=2/3, Ф = 57.5
o, t = 60 sec
154
5-10 (a) 5-10 (b)
Figure 5-10: Turbulent kinetic energy (k) distribution (m2/s
2) from experimental
results of Morsi et al. [1] ((a) Q=2.67 x 10-3
m3/s, H/L=2/3, Ф = 0
o, (b) Q=2.67 x 10
-3
m3/s, H/L=2/3, Ф =57.5
o)
As the relative density of air being very low as compared to water, the velocity of air
in this study does not give sufficient momentum to penetrate into the liquid water and
create turbulence in the whole bath. Formation of bubbles and hence high velocity
gradients, resulting from momentum transfer between the gas and the liquid phase,
provides higher turbulence near the lance. In the region above the lance exit, the rising
bubbles have lost most of their initial momentum. Only the buoyancy forces exerted
by the rising bubbles assist in the generation of turbulence near the lance wall.
155
Figure 5-11: Volume fraction for water after 180 seconds (Q=2.67 x 10-3
m3/s,
H/L=2/3, Ф =57.5o)
Figure 5-11 shows the volume fraction for water after 180 seconds for high injection
rate and 2/3 lance submergence. Figure shows significant asymmetry due to the
sloshing and splashing in the water bath, which represents the transient nature of the
simulation. This is the essence of the 3D transient multiphase flow simulation, which
can give more insights into metallurgical flows of interest. The volume fraction plot
for the non-swirl case (not presented in the paper) showed no significant difference
with the swirl case. Velocity vectors for the same condition are shown in Figure 5-12,
which shows the formation of a weak recirculating vortex just near the top level of the
liquid. Formation of this weak recirculation region is random and transient in nature,
as it was observed from the simulation results. The rising bubbles in the liquid induce
these short-lived vortices. Formation of the recirculating vortex inside the bath is quite
favourable for generation of uniform mixing inside the bath. However, no significant
vortex was observed inside the bath in the present study. The magnitude of the
156
velocity vectors near the wall of the vessel are negligible compared to the values near
the lance. Liquid near the bottom corner of the vessel is almost unaffected by the air
injection process, as it was observed from the present air-water simulation.
Figure 5-12: Velocity vectors for liquid phase (m/s) after 180 seconds
(Q=2.67 x 10-3
m3/s, H/L=2/3, Ф =57.5
o)
5.2.6 Mean Convective Mixing
To measure the convective mixing efficiency inside the physical models, traditional
tracer studies are generally used. In the present study, the mean convective mixing was
evaluated by using the “Volume exchange effectiveness” concept. This approach is
generally used in Heating, Ventilation and Air-conditioning (HVAC) process
simulation [149]. The term “Volume exchange effectiveness”, which is actually a
measure of the mean convective mixing, may be defined as the net exchanged volume
of fluid in each computational cell divided by the volume of that cell and can be
expressed as eefV where,
cell theof Volume
cell nalcomputatio a through rate flow Volume=eefV 5.5
157
It expresses the volume exchange rate through a cell which in-turn represents the
convective mixing. It has the unit of (time)-1
.
Figure 5-13: Volume exchange effectiveness along radial direction from present water
model simulation (Q=2.67 x 10-3
m3/s, H/L=1/3)
Figure 5-13 shows the time averaged volume exchange effectiveness ( eefV ) of the air-
water system for swirl and non-swirl injection at z/Z = 0.84 (which is 10mm below the
lance exit in the water bath) along the radial direction. As shown, swirl flow provides
greater convection mixing for up-to r/R = 0.1. The swirl flow sets up a centrifugal
force field, which has a favourable convection effect. Swirl flow dominates over the
non-swirl in the region ranging from r/R = 0 to 0.1. However, at distance r/R≥ 0.1,
there is no significant change in the convective mixing process due to swirling effect.
This mixing phenomenon is only valid for the air water system, not in the real furnace
scenario. Figure 5-14 shows the contour plots of the eefV for swirl and non-swirl flow.
The plots further confirm that the swirl flow can only increase mixing in the region
very close to the lance for the specific case of the air-water system.
158
5-14 (a) 5-14 (b)
Figure 5-14: Contours for volume exchange effectiveness from present water model
simulation
(a) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 0
o, t = 30 sec
(b) Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o, t = 30 sec
5.2.7 Effect of Density
To investigate the effect of density change on the formation of splashing, the liquid
density has been increased to 3 times of density of water. The new fluid has been
denoted as D3 ( 3
3 3000 mkgD =ρ ). The viscosity of that liquid was kept constant as
water. However, it should be noted that due to the lack of experimental data, these
results could not be validated and hence must be considered as exploratory. Therefore,
these results are presented here for discussion only.
159
Figure 5-15: Average volume fraction at 40mm height (z/Z=0.66) above the liquid
bath (Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o)
Figure 5-16: Average volume fraction at 60mm height (z/Z=0.625) above the liquid
bath
(Q=2.67 x 10-3
m3/s, H/L=1/3, Ф = 57.5
o)
160
The effect of density on the generation of splashing is shown in Figure 5-15 and
Figure 5-16. Figure 5-15 shows the average volume fraction at 40mm (z/Z = 0.66)
above the liquid bath and the generated splash pattern along the radial direction. As
expected, the degree of splash generation for the higher density liquid is reduced
significantly. At 60mm (z/Z = 0.625) height above the liquid bath (Figure 5-16), the
curves show similar trend. The bubbles from the injected air jet through the annulus of
the lance move radially outward and approach the free surface. This trend leads to the
creation of broad spouts in the free surface. The highest expulsion of the rising plume
was in the vicinity of the lance. When the bubbles collapse at the free surface,
formation of splashing occurs.
In the experimental study by Nilmani and Conochie [56], the authors investigated the
effect of different gas density by injecting Helium instead of air. They reported that
with a less dense gas, a greater volume flow is required to maintain the same injection
characteristics. However, no experimental study was found in the open literature on
the formation of splashing for higher liquid density for the case of top submerged
lance gas injection.
5.2.8 Effect of Viscosity
Effect of viscosity on splashing and slopping has been investigated experimentally by
Nilmani and Conochie [56], where the authors used three different liquids of different
viscosity (water, glycerol/water of viscosity 56 centipoise and glycerol/water of
viscosity 200 centipoise). They reported from their experimental study that splashing
and slopping were not as pronounced as in the air-water system. They also reported
that gas penetration of the viscous liquid on the lance axis was also small and
increasing liquid viscosity reduces gas dispersion.
The effect of viscosity change has also been investigated by Liovic et al. [76] on the
formation of splashing by numerical technique. In their numerical simulation, Liovic
et al. [76] used 95% glycerol solution to see the effect of viscosity on the generation of
splashing. They reported that high viscosity suppresses splashing and free surface
distortions. They also reported that high liquid viscosity of the glycerol solution also
damps out bulk bath motion and reduces back-penetration of liquid up the lance
significantly.
161
Chapter 6
162
6 Numerical Investigation of Zinc Fuming Bath in TSL
Furnace
This chapter will focus on the investigation of zinc fuming behaviour inside the top
submerged lance (TSL) furnace by using CFD modelling tool. The research findings
presented in this chapter are obtained from the developed CFD model that includes
submerged combustion in multiphase flow, reactions kinetics in the slag bath and heat,
mass and momentum interfacial interaction between the phases. Details of the model
development of the pilot plant scale TSL furnace including the geometry, generated
volume mesh, equations solved and interfacial exchange terms are discussed in
Chapter 4 (Section 4.3.2). This Chapter begins with the analysis of grid independency
test, followed by results and discussions on the detail zinc fuming kinetics inside the
TSL furnace.
6.1 Test of Grid Independency
Four grid resolutions were tested for grid independency test. Starting with a coarse
grid, number of cells was increased in the region of interest until the solution from
each grid was unchanged for successive grid refinements. Table 6-1 represents an
overview of the grid information studied in the grid independency test. All the cells in
the calculation domain were polyhedral with a large number of hexahedral cells.
Describing the number of cells in each direction is complex as the computational
domain consisted of hybrid unstructured meshes in curvilinear non-orthogonal
coordinate system with Cartesian base vectors and refined regions in some locations.
Regions for refinement in this model include combustion chamber at the lance tip,
area adjacent to the lance tip and surrounding the lance, near the wall, near the exit of
the furnace and in the slag bath. Figure 4-5 shows the generated course grid for the
CFD analysis. The meshing procedure was carried out using the Fame Advanced
Hybrid meshing technique [125].
163
Table 6-1: Overview of computational grids
Name Grid Density No of
Computational cells
Grid 1 Coarse 256138
Gird 2 Medium 337755
Grid 3 Fine 415017
Grid 4 Very Fine 581024
Figure 6-1: Mean tangential velocity distributions for different grid configurations (Qa
= 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
Figure 6-1 shows the mean tangential velocity distribution (V) on X-Z plane along the
radial direction at a vertical depth of z/Z = 0.774. The radial distance (r) is normalised
by the radius of the cylinder (R=0.25 m) and the axial distance z is normalised by the
length of the cylinder (Z=1.68 m). Both Grid 3 and Grid 4 gave a very close
prediction. The difference in predictions between Grid 3 and Grid 4 was small enough
(around 1%) to suggest that any further grid refinement would not yield a substantially
different profile in that plane. Hence, it was determined that the fine grid resolution
(Grid 3) of 415017 cells was sufficient to obtain grid independent results.
164
For validation exercise of the results presented in this chapter, a pilot plant scale
experimental study carried out by Waladan et al. [2] was considered. Comparisons of
the injection conditions between the present CFD simulation and the experimental
study by Waladan et al. [2] are mentioned in Table 6-2.
Table 6-2: Injection conditions (CFD and Experimental)
Experimental
conditions (CZF5)
(Waladan et al. [2]
CFD
Calculations
Fuel type Fine Coal CH4
Reductant type Fine Coal Fine Coal
Combust air (kg/s) 0.06 0.05
Fuel rate (kg/s) 0.035
0.0035
Reductant rate (kg/s) 0.035
Initial zinc content in the
slag (wt %) 10.7 18.0
6.2 Results and Discussion
Zinc slag fuming is a complex process to simulate and it is necessary to make a
number of assumptions to simplify the simulation. A large number of complex
reactions are involved, such as: reduction of zinc oxide and ferric iron, fuel
combustion, oxidation of ferrous iron oxide, gas-carbon reaction. In the present
simulation, oxidation of ferrous iron oxide is likely to be insignificant, as all the
supplied oxygen is combusted at the lance tip due to stoichiometric combustion. To
avoid complexity, the ferric iron level in the bath was considered as negligible. As the
numerical simulation is based on Euler - Euler approach, the model does not predict
the exact plume shape with a sharp gas-liquid interface for each small gas bubble. The
model could be improved, particularly, incorporating the effect of ferric iron and
providing a more realistic treatment of the behaviour of coal; in addition longer
simulation times could be used. However, the results here represent a significant step
in developing a more comprehensive model.
165
In the following sections, the results will be discussed in-terms of bath behaviour,
combustion behaviour, zinc fuming behaviour and effect of lance submergence level
on generation splashing and zinc fuming behaviour.
6.2.1 Bath Behaviour
Figure 6-2 shows the volume fraction distribution for molten slag phase of the
transient simulation at four different time steps (13th
, 13.5th
, 20.5th
and 32nd
seconds).
The cross sectional view along the vertical X-Z plane was considered in these figures.
The colour bar associated with each figure represents the volume fraction values (0 to
1). Red colour with volume fraction value of 1 indicates 100% slag phase and the blue
colour with volume fraction value of 1E-006 indicates 100% gas phase. The green
colour in between indicates the slag – gas emulsion zone. The generated plume shape
at the lance tip due to gas injection and combustion is clear from this figure. This
plume was not found to be constant in position, rather changing with time, as the
figure shows. There was significant amount of sloshing in the slag bath as well.
Sloshing is a free surface hydrodynamic phenomenon in the bath, where the liquid
phase generates a wavy motion on the free surface of the bath due to turbulence
created by the submerged combustion. On the other hand, splashing is the tearing of
liquid phase in the form of droplets from unstable liquid metal wave peaks in the
continuum bath surface. Slag droplets generated due to splashing may either fly inside
the bath (termed as slag in flight in this paper) or otherwise hits the furnace wall and
falls back to the bath. Figure 6-3 shows another cross section along the X – Y plane
from the top of the modelled furnace. These figures give a clear understanding of the
impact area of the injected gas at the lance tip inside the bath. The mean diameter of
the generated plume along the radial direction (in X – Y plane) was found to be 0.3 m
in the 0.5 m diameter furnace. The mean area of the impact zone at the lance tip was
calculated as 0.071 m2 from the transient simulation in a 0.196 m
2 area of slag bath.
Comparison of the Figure 6-3(a) and Figure 6-3(b) also indicates the transient nature
of the generated plume along the radial direction.
Figure 6-4 shows the iso-contour plot at the gas-liquid interface, indicating velocity of
the molten slag phase with a slag volume fraction of 0.5 (display attribute: slag
166
velocity, iso-surface data: slag volume fraction of 0.5, Qa = 0.05 kg/s, Qf = 0.0035
kg/s, LH ′′ = 1/3). The colour bar with the associated figures represents the slag phase
velocity inside the furnace varying from 0.07 to 0.7 m/s. Figure 6-4(a) and Figure
6-4(b) are captured at 25th
and 26.5th
second time of the transient simulation
respectively. Comparison of the two figures portrays the sloshing and splashing
phenomena inside the furnace in a time interval of 1.5 seconds. Generated movie from
the simulation results showed that the slag in flight (splashes) goes around 1 m high
above the bath surface. This sloshing and splashing phenomena creates a massive
agitation in the slag bath, which accelerates the gas-carbon and gas-slag reactions.
6-2(a) 6-2(b)
167
6-2(c) 6-2(d)
Figure 6-2: Volume fraction distribution for molten slag phase along the vertical cross
section in X-Z plane at four different time step of the transient simulation (Qa = 0.05
kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 13th
second, (b) 13.5th
second, (c) 20.5th
second and (d) 32nd
second.
168
6-3(a)
6-3(b)
Figure 6-3: Volume fraction distribution for molten slag phase along the cross section
in X-Y plane (top view of the modelled furnace) at different time steps of the transient
simulation (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 21.5th
second, (b) 22nd
second
169
6-4(a) 6-4(b)
Figure 6-4: Iso-contour plot of the molten slag phase at different time steps of the
transient simulation (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 25th
second,
(b) 26.5th
second
Figure 6-5(a) and Figure 6-5(b) shows the streamlines distribution of the molten slag
phase showing slag phase movement inside the furnace at two different time step of
16.5th
second and 27th
second respectively at a mid-plane cross section along the X – Z
plane. Velocity vectors inside the molten slag bath are shown on Figure 6-6. The
transient simulation showed some recirculation zones as marked in the figure.
Recirculation zones are favourable in improving the bath mixing. While the slag in
flight that falls back to the bath accelerates the bath surface reaction. Slag phase
streamlines from Figure 6-5 shows the slag in flight movement inside the furnace.
170
Some portion of the slag in flight strikes the furnace wall and falls back to the bath
along the wall, as shown in the Figure 6-4.
6-5(a) 6-5(b)
Figure 6-5: Streamlines distribution of the molten slag phase showing slag phase
movement inside the furnace at different time steps of the transient simulation (Qa =
0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3), (a) 16.5th
second, (b) 27th
second
171
Figure 6-6: Velocity vectors inside the furnace in the molten slag bath: (Qa = 0.05
kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
6.2.2 Combustion Behaviour
After the initiation of the combustion at the lance tip, the plume becomes larger due to
combustion and associated expansion of the gases in the slag bath. The CFD software
used - AVL FIRE 2009.2 (AVL, Graz, Austria) enables us to capture results at
millisecond time step. Figure 6-7(a) shows the species mass fraction (kg/kg)
distribution of CH4 after 27 seconds of the simulation, which shows CH4 being
combusted completely because of stoichiometric combustion. The colour bar with the
Figure represents the mass fraction (kg/kg) value. Because of the swirling effect in the
annular region of the lance, O2 in the air that comes through the annular region of the
lance tip, mixes well with the CH4 in the combustion chamber and ensures fuel-
efficient combustion at the lance tip. As indicated by the colour bar, CH4 is entering
into the combustion chamber through the central hole of the lance (red colour with
mass fraction = 1.0) and mixes with air, indicated by turning to a green colour with
mass fraction = 0.5 to 0.6. Before the supplied CH4 leaves the combustion chamber,
80% of it is combusted in the combustion chamber and the remaining 20% CH4
172
combustion takes place near the lance tip in the bath. Figure 6-7(d) shows the O2 mass
fraction distribution inside the furnace. The figure shows that the supplied O2 through
the annular region of the lance was completely combusted at the lance tip. The
products of combustion include CO, CO2, N2 and H2O vapour. Formation of CO2 and
CO in the bath is shown in Figure 6-7(b) and Figure 6-7(c) respectively. The colour
bar with the Figures represents the mass fraction (kg/kg) value of the respective
figures. As the figures show, mass fraction (kg/kg) of CO2 above the slag bath is
greater than mass fraction (kg/kg) of CO. This is attributed to the fact that, CO2 is
forming not only as a product of combustion of CH4, but also from the ZnO reduction
reaction (Equation 2-7). The CO above the slag bath contributes to increase the
fuming rate from the slag in flight (splash). Comparison of the Figure 6-2 and Figure
6-7(c) points clearly about phenomena that the CO from gas phase comes in contact
with the generated splash or slag in flight inside the furnace. The effect of slag in
flight (splash) on the fuming rate is discussed later (section 6.2.4.)
6-7(a) 6-7(b)
173
6-7(c) 6-7(d)
Figure 6-7: Species mass fraction (kg/kg) distribution for (a) CH4, (b) CO2 (c) CO and
(d) O2
(Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3)
Figure 6-8 is showing the temperature distribution of the molten slag phase only from
the transient simulation results. Temperature distribution inside the bath was not
uniform and it was governed by the enthalpy conservation and interfacial exchange
terms described in section G. In the boundary conditions the wall temperature was
considered as 1500 K and the initial temperature of the molten slag was considered as
1500 K for the simulation to start. The simulation results revealed that there is non-
uniform temperature distribution in the molten slag bath. This temperature profile
doesn’t imply steady state distribution. These non-uniform temperature distributions
in the slag bath are expected to be uniform with longer simulation time. In addition,
the transport properties of the molten slag phase influence this temperature
distribution in the slag. From the simulation results, the molten slag bath can be
174
divided into three zones: combustion zone near the lance tip, where the temperature
was highest due to the combustion process (1563-1573 K), bath zone below the lance
tip in the slag bath, where the temperature was almost uniform (1550-1563 K) and the
free surface zone away from the central lance of the furnace near the surface area of
the molten slag bath, where there is a thin layer of gas-liquid mixture (1540-1550 K).
In this gas-liquid emulsion zone, the temperature distribution is lowest, as the
endothermic reductions of ZnO by the CO (as mentioned in equation 2-7) takes heat
from that zone. The enthalpy of reduction of ZnO from the slag to the gas phase
(equation 2-7) has been considered in the simulation according to the data provided by
Richards et al. [21] ( 5.192H =∆ kJ/mol). However the model does not include the
post combustion zone and re-oxidation of Zn to ZnO vapour above the bath.
Figure 6-8: Temperature distribution for molten slag phase only (b) Species mass
fraction (kg/kg) distribution for O2 (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
175
6.2.3 Zinc Fuming Behaviour
The simulation results predicted that zinc fuming initiates from near the lance tip of
the molten slag bath in the combustion zone where CO forms as a product of
combustion of CH4 as mentioned by equation 2-7. CO thus formed initiates the first
fuming process in the bath. Figure 6-9(a) shows the initial stage (after 1 second) zinc
fuming process near the lance tip area. The fuming process accelerates from the bath
near the surface area where the slag phase is more in contact with the gas phase and
carbon reacts directly with CO2 to form CO. Figure 6-9(b) shows the fumed zinc
distribution inside the furnace from the molten slag bath after 30 seconds of
simulation time. The colour bar with each of the figures indicates amount of zinc
fumed in kg, on that specific mid plane section (along X – Z plane).
6-9(a) 6-9 (b)
Figure 6-9: Fumed zinc distribution inside the furnace (a) at initial stage (after 1
second) (b) after 30 seconds (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
Figure 6-10 shows the amount of ZnO present in the slag bath at different depth along
the radial direction from the lance. ZnO concentration in the slag bath below the lance
tip (l/L=0.1) is found to be 17.1 wt%, which indicates a highest fuming rate of 0.9
wt%/min, for a slag containing 18 wt% ZnO, where l indicates the depth from the
176
lance tip as shown in Figure 4-4 and L = (L’- 0.12) = 0.48 m for 1/5 lance
submergence level. Figure 4-4 is provided below for convenience.
Figure 4-4: Schematic diagram of the modelled furnace for Outotec TSL pilot plant
The highest mass fraction of CO was observed in the combustion zone below the lance
tip. As the present simulation is based on Eulerian approach, it does not consider any
sharp gas-liquid interface; rather the zone is a gas-liquid emulsion zone as shown in
Figure 6-2 (green colour between blue and red). The fuming rate near the bath surface
(l/L=0.1) was found as 0.34 to 0.9 wt%/ min, where availability of CO is more and
carbon reaction rate is higher. By comparing Figure 6-2(a) and Figure 6-10, it be
argued that highest fuming rate area is up-to the radial distance r/R = 0.3 in the
generated plume inside the bath. Significant fuming rates were also found at the depth
of l/L=0.2 ranging from 0.2 to 0.579 wt%/ min and 0.13 to 0.35 wt%/min at depth
l/L=0.3. The overall fuming rate from the whole bath is found to be 0.2 to 0.35
177
wt%/min for a slag containing 18 wt% ZnO using CH4 as fuel and pulverized coal as
reductant. Batch fuming rates for zinc fuming furnaces using pulverized coal as fuel
and reductant vary between 0.15 to 0.27 wt%/ min for slags containing about 10 wt%
Zn [150] and 0.09 wt%/ min for 9 wt% Zn slag (at 1275o C temperature) [102]. Thus
Figure 6-10 depicts the ZnO reduction behaviour at the gas – liquid interface from the
slag bath at different depth, in addition to the fuming rate.
Figure 6-10: ZnO distribution in the slag bath at different depths below the lance tip
along a line in radial direction from the lance (Qa = 0.05 kg/s, Qf = 0.0035 kg/s,
LH ′′ = 1/5)
The CO mass fraction (kg/kg) distribution along the radial direction at different depths
below the lance tip is shown on Figure 6-11. As indicated by the Figure, the mass
fraction is gradually decreasing at lower depth below the lance tip. The rate of
reduction of ZnO is also decreasing at lower depths, as shown on Figure 6-10. The
concentration gradient of ZnO near the gas-slag interface is indicative of the fact that
diffusion of ZnO from the bulk slag to the slag-gas interface and presence of CO plays
a very important role in overall kinetics. From the slag analysis of some zinc fuming
industrial trials, Richards et al. [21] also found that there was concentration gradients
of Zn and Fe adjacent to some carbon particles in the slag.
178
Figure 6-11: CO mass fraction (kg/kg) distribution along the radial direction at
different depths below the lance tip (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
Figure 6-12: Zinc fuming rate from some published experimental work
179
Figure 6-12 shows zinc fuming rate from some of the published experimental and
modelling work. Fuming rate for fuming cycle A1, B1, C1 and D1 are taken from the
mathematical modelling work of Richards and Brimacombe [20], where the authors
used some plant data for fuming rate to fit the model. The Figure shows that the
fuming rate curves decreases almost linearly with time and follows a zero order
pattern. A similar pattern of zinc fuming rate curve has also been predicted from the
present simulation results, as shown in Figure 6-13.
Figure 6-13 shows the CFD results for zinc fuming rate. The simulation results
showed zinc concentration in the slag decrease linearly with time. The linear time
dependence of the zinc elimination rate reveals the zero order reaction kinetics, which
is consistent with other published work (Figure 6-12). Experimental trials on pilot
plant top submerged lance zinc fuming process were carried out by Waladan et al. [2].
In those experimental trials, zinc fuming from Sulphide Corporation slag containing 8-
12% zinc were investigated using a range of fuel-reductant combinations [2].
Experimental results for fuming trial CZF5 are shown in Figure 6-13 for macro-step
validation purpose.
Figure 6-13: Fuming rate from present simulation results (Qa = 0.05 kg/s, Qf = 0.0035
kg/s, LH ′′ = 1/5) and comparison with experimental data (Curve regenerated from
CZF5 of Figure 11 from Waladan et al. [2])
180
6.2.4 Effect of Lance submergence level
As discussed earlier, the highest fuming rate was found to be below the lance tip in the
generated plume inside the slag bath (see results in Figure 6-10). The larger the plume
area, the higher is the fuming rate from the slag bath. In the slag bath, fuming rate was
found to be higher in area of the gas-liquid interface, where ZnO diffuses to the gas-
slag interface and being reduced by the CO present in the gaseous phase. It would be
interesting to see the fuming rate from the slag in flight (above the bath) or generated
splash. Therefore, three different submergence levels ( LH ′′ = 1/5, 1/4 and 1/3) have
been considered in the model to investigate the effect of lance submergence level on
fuming and splash generation. All the submergence levels were investigated with the
same air and fuel flow rate (Qa = 0.05 kg/s, Qf = 0.0035 kg/s). To elucidate the effect
of generated splash (i.e. slag in flight) on the fuming rate, a number of locations were
considered above the slag bath on the X-Y plane (Figure 4-4). The amount of splash
at those specific heights above the slag bath has been characterized by the amount of
volume fraction in flight. The volume fraction and the zinc fuming rates at those
specific heights above the slag bath are shown in Figure 6-14 for 1/3 lance
submergence level. Average volume fractions in flight are plotted on the left vertical
axis and zinc fuming rates (wt%/min) are plotted on the right vertical axis of the
curve. The horizontal axis of the curves shows the normalized radial distance from the
lance tip (r/R), where R = 0.25 m is the radius of the cylindrical furnace. Generated
splash and zinc fuming rates were measured at five different heights (h = 0.1, 0.2, 0.4,
0.6, 0.8) m above the bath. These heights were normalized as h/H = 0.09, 0.18, 0.36,
0.54, 0.73, where h is the height from the bath surface to the top wall of the furnace
and H = 1.08 m is the total distance between the bath surface and top wall of the
furnace (Figure 4-4).
181
Figure 6-14 (a): Zinc fuming rate and amount of splash at different heights above the
bath (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/3)
Figure 6-14 (b): Zinc fuming rate and amount of splash at different heights above the
bath (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/4)
182
Figure 6-14 (c): Zinc fuming rate and amount of splash at different heights above the
bath (Qa = 0.05 kg/s, Qf = 0.0035 kg/s, LH ′′ = 1/5)
Figure 6-14(a) shows largest volume of splash is at 100 mm above the bath (h/H =
0.09). The normalized radial distance on the horizontal axis of the curve indicates the
splash pattern. The largest volume fraction of the generated splash at that height was
found to be 0.37 at a radial distance of r/R = 0.45. The minimum zinc fuming rate
measured at that height is 0.03 wt%/min at a radial distance of r/R = 0.2. The zinc
fuming rate (continuous red line curve) gradually increased to 0.12 wt%/min as the
splash volume decreased gradually and gas comes more in contact with the slag phase.
At height h/H = 0.18, average volume fraction of the generated splash reduced to 0.18,
and zinc fuming rate increased to 0.08 – 0.24 wt%/min (dashed blue line curve).
Measured data at the other heights shows that generated splash volume gradually
reduces towards the upward direction of the furnace and zinc fuming rates at those
specific heights gradually increases. The zinc fuming rates presented on the Figure
6-14(a) at different heights are the predicted value of the fumed zinc or reduced ZnO
from the slag present at those specific heights only. Hence, the reader should not
confuse fuming rate from the slag at any plane with the overall fuming rate.
Comparison of the Figure 6-2 and Figure 6-7(c) indicates that reduced volume of
splash far above the bath has large interfacial area and comes more in contact with the
CO available above the bath. This may be more easily described with fact of large
interfacial area between the slag and the gas phase. Definitely, a large droplet or
183
splash of slag of 100 mm diameter has less interfacial area than 10 droplets of each 10
mm diameter. These predictions clearly indicate that more the slag phase is exposed to
the gas phase, the higher is the fuming rate. Therefore, a large quantity of fine slag
droplets will promote high fuming rates. Hence fine droplet generation and deeper gas
penetration in the slag bath will maximize the fuming rate. However, the availability
of CO and carbon limits the fuming rate in flight. Hence, the model predicts that the
mass transfer of ZnO from the bulk slag to the slag-gas interface and the CO present in
the gas phase plays a vital role in controlling the overall fuming rate for the specific
simulation time studied. The two other submergence levels ( LH ′′ = 1/5 and 1/4)
showed similar behaviour of splash generation and zinc fuming rate, as shown in the
Figure 6-14(b) and Figure 6-14(c).
For a quantitative comparison among the submergence levels studied, two other
graphs are presented on Figure 6-15 and Figure 6-16 at two different heights above the
bath (h/H = 0.09 and 0.73) respectively. As Figure 6-15 shows, the generated volume
of splash just 100 mm above the bath (h/H = 0.09) has increased for 1/3 lance
submergence level than 1/5 submergence level. The splash pattern has also changed
with the submergence level. Highest volume fraction for 1/5 lance submergence level
at h/H = 0.09 height was found to be 0.25 at a radial distance of r/R = 0.25. Whereas,
for 1/4 lance submergence level, splash volume were increased further away from
lance, at a radial distance of r/R = 0.80 to 1.0.
184
Figure 6-15: Comparison of zinc fuming rate and amount of splash at h/H = 0.09
above the bath for three submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s)
Figure 6-16: Comparison of zinc fuming rate and amount of splash at h/H = 0.73
above the bath for three submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s)
185
Not much noticeable change in overall fuming rate has been observed between the 1/5
and 1/4 submergence levels. Figure 6-16 shows a significant increase in the fuming
rate for 1/3 submergence level at a height of h/H = 0.73. Due to the deeper gas
penetration, noteworthy increment in the sloshing and splash generation with large
interfacial area was observed. The increment in the sloshing phenomena agitated the
gas-carbon reaction in the bath surface, which accelerates the production of CO in the
bath surface. This higher production of CO increased the fuming rate from the slag in
flight (splash). The highest fuming rate for 1/3 lance submergence level was measured
as 0.5 wt%/min at h/H = 0.73, which is 1.8 times higher than 1/5 lance submergence
level at the same height. The overall fuming rate for 1/3 lance submergence level were
found to be 0.25 to 0.4 wt%/min, which is around 1.3 times higher than 1/5 lance
submergence level. The bar chart shown on Figure 6-17 is the graphical representation
of the average fuming rates for the three lance submergence levels studied.
Figure 6-17: Overall zinc fuming rate comparison for three different lance
submergence levels (Qa = 0.05 kg/s, Qf = 0.0035 kg/s)
186
Chapter 7
187
7 CFD Modelling of Conventional Zinc Fuming Furnace
This chapter will focus on the application of the developed CFD model in other
furnace configuration. The developed CFD model for TSL furnace as described in the
previous chapter was extended for submerged coal combustion instead of CH4 and
applied to a thin slice model of the conventional tuyere blown furnace. Results were
analysed in-terms of combustion behaviour, jet penetration behaviour, bath agitation
and zinc fuming behaviour. Details of the model development including the geometry,
generated volume mesh, equations solved and interfacial exchange terms are discussed
in Chapter 4 (Section 4.3.3). This Chapter begins with the analysis of grid
independency test, followed by results and discussions on the detail of jet penetration
and combustion behaviour of the conventional tuyere blown furnace.
7.1 Test of Grid Independency
Accuracy of every CFD analysis depends largely on the type of grid used. Hence, grid
generation needs special attention of the investigators. Generally, finer grid gives more
accurate results, but eventually it leads to higher computational expense. Therefore, an
optimum grid resolution is necessary for efficient CFD analysis. In the present
simulation, only critical area for grid refinement was at the tuyere tip and the area
surrounding the tuyeres. The meshing procedure was carried out using the Fame
Advanced Hybrid meshing technique [125]. Three different grid resolutions were
tested for grid independency test. Initially with a coarse grid, then mesh refinement
was increased in the region of interest until the solution from each grid was unchanged
for successive grid refinements. The computational grid (176378 cells) used in the
present study is too dense for visual presentation. A figure of the coarse computational
grid (110843 cells) shown in Figure 4-8(b). To mention the number of control
volumes in each direction is complicated as the present computational domain
consisted of hybrid unstructured meshes in curvilinear non-orthogonal coordinate
system with Cartesian base vectors and refined regions in some locations. All the cells
in the calculation domain were polyhedral with a large number of hexahedral cells.
Other than the refinement, the number of cells is mentioned on Table 7-1. The Table
shows the number of the largest control volumes in X, Y and Z direction. In addition
188
to that, the computation domain has been refined in a number of locations. Regions for
refinement in this model include the tuyere tips, area adjacent to the tuyere tips and
surrounding the tuyere, near the wall, near the exit of the furnace and in the slag bath.
Table 7-1: Overview of computational grids
Name Grid type Grid Density No of
Computational cells
Grid 1 Medium (60 x 10 x 120) +
Refinements
110843
Gird 2 Fine (80 x 15 x 140) +
Refinements
176378
Grid 3 Very fine (90 x 15 x 150) +
Refinements
214911
Figure 7-1 shows the mean upward velocity distribution (W) on X-Z plane along the
axial direction at a vertical height of h/H = 0.02 and width (along Y direction) of d/D
= 0.5. The axial distance (l) is normalised by the length along X direction (L=2.44 m).
Both Grid 2 and Grid 3 gave a very close prediction. The difference in predictions
between Grid 2 and Grid 3 was small enough (less than 1%) to suggest that any further
grid refinement would not yield a substantially different profile in that plane. Hence, it
was determined that the fine grid resolution (Grid 2) of 176378 cells was sufficient to
obtain grid independent results.
189
Figure 7-1: Mean upward velocity, W (m/s) distribution of slag phase for three
different grids along axial direction (Uo = 67.8 m/s, d/D = 0.5, h/H = 0.02)
7.2 Results and Discussion
Computational fluid dynamic modelling of a multiphase combusting system
incorporating molten slag is a complex process to simulate. The model was simplified
by considering a number of assumptions. Reactions involved in the system are quite
complex, which includes reduction of zinc oxide and ferric iron, submerged
combustion, oxidation of ferrous iron oxide, gas-carbon reaction. In the present
simulation, oxidation of ferrous iron oxide was not taken into consideration. To avoid
complexity, the ferric iron level in the bath was assumed to be negligible. The rest of
the reactions, such as zinc oxide reduction, submerged coal combustion and gas-
carbon reactions were taken into account in the present simulation. Detail kinetics of
the chemical reaction is already described by Richards et al. [19-21]. The aim of the
present research is to investigate tuyere tip combustion dynamics, bath interaction
conditions and generation of turbulence at the tuyere tip inside the bath. Though, with
the available computational power and resources, these aims are hard to achieve
accurately. The actual tuyere tip interaction by tracking each and every single gas
bubble would require a very high powerful computer to solve millions of cells at the
tuyere tip. From the perspective of the CFD resources, the turbulence model itself is
190
not capable to predict the full turbulence behaviour accurately at the tuyere tip. In fact,
no turbulence models in the available literature have the capability to accurately
predict the full comprehensive turbulence behaviour within the bath. The standard k-ε
turbulence model [129] used in the present study may not provide expected
performance in a number of important cases such as flows with large extra strains (e.g.
curved boundary layers, swirling flows) and rotating flows [148]. Still, the model was
used as it is well established, most widely validated turbulence model and it gives
excellent performance for many industrially relevant flows. As the numerical
simulation is based on Euler - Euler approach, the model does not predict the exact
plume shape with a sharp gas-liquid interface for each small gas bubble. The model
could be improved, particularly, taking into account of the ferrous iron oxidation and
incorporating the effect of ferric iron in the bath, using highly dense computational
cells in the bath. However, the results represent a credible step in developing a more
comprehensive CFD model for analogous system.
7.2.1 Bath behaviour
From the CFD analysis of the developed model, a picture of the injection phenomena
emerged including jet and coal penetration into the slag bath, quiescent zones,
recirculation zones, tuyere gas stream behaviour, slag and gas phase velocity
distributions, turbulence behaviour.
Figure 7-2(a) and (b) shows volume fraction distributions of the molten slag phase at a
mid-plane cut along X-Z plane. The pictorial representation of the gas jet and coal
behaviour at the tuyere tip and shape of the tuyere gas column put forward by
Richards et al. [21] are also shown on Figure 7-3(a) and (b). The colour bar with the
associated figures indicates the volume fraction level (1 is indicated by the red colour,
which means 100% slag phase and 0 is indicated by the blue colour, which means 0%
slag phase). The green colour indicates the gas-liquid emulsion zone in the bath. The
simulation results showed that the tuyere gas columns at both the tuyere tip are
wavering in nature with respect to time and position. This is due to the tuyere tip
combustion and associated gas expansion from the combustion phenomena and also
the due to the sloshing phenomena created in the bath free surface. As soon as the gas
jet carrying coal phase injected into the bath, the jet moves horizontally into the
191
molten slag entraining with surrounding slag phase and subsequently losing its
horizontal velocity (U). The highest distance the jet travels horizontally into the
molten slag phase is known as jet penetration. Detail description about the depth of
penetration of the gas jet carrying coal will be discussed in the later sections. While
the jet penetrating into the molten slag, horizontal velocity component reduces and at
some point the upward velocity component (W) component becomes significant. The
gas phase mixes with the slag phase and creates the gas-slag emulsion phase, as shown
by the green colour around the jet core in Figure 7-2(a) and (b). Mean upward velocity
distributions of the slag phase along the axial line from the tuyere tip are shown in
Figure 7-1 for inlet condition, Uo = 67.8 m/s. The graph represents the mean upward
velocity (W m/s), which was calculated by time averaging of the velocity components
at every second from 20 seconds to 60 seconds of the simulation time. Highest mean
upward velocity component (W) was found to be 2.46 m/s at an axial distance of l/L =
0.09 from the tuyere tip. Beyond that point along the axial direction (l/L > 0.09), mean
upward velocity component (W) gradually decreases. At the distance of l/L = 0.283,
upward velocity component (W) was found to be 0.01 m/s. This velocity profile
remains almost linear with no major deviation until the axial distance of l/L = 0.75,
which is closest to tuyere 2 (Figure 4-7). Then, it rises again from l/L = 0.8 and
reaches the maximum of 2.48 m/s at a distance of l/L = 0.92. Upward mean velocity
profiles near the two tuyeres are almost similar. Figure 7-2(c) shows the iso-contour
plot at the gas-liquid interface, indicating velocity of the molten slag phase with a slag
volume fraction of 0.3 (display attribute: slag velocity, iso-surface data: slag volume
fraction of 0.3, tuyere inlet velocity Uo = 86 m/s). The generated bubble plume shape
in the tuyere gas column, free surface sloshing phenomena and generated metal fingers
are clear from the figure. From the mid plane cut along X-Z plane (Figure 7-2(b)), slag
volume fractions of 0.3 – 0.5 is visible at 1.5 m height above the bath surface (bath
surface is located at 1 m height from the bottom). Movie generation from the
simulation results showed that, the generated splash strikes both the side wall and the
symmetry walls and slides downward on the wall to the bath. Some splashes that do
not hit any wall, are ejected up-to 1.5 meters above the bath surface and come back to
the bath, which accelerate the mixing process and reactions in the bath surface.
192
7-2 (a) 7-2 (b)
7-2 (c)
Figure 7-2: (a) and (b) volume fraction distribution of slag phase at plane cut at X-Z
plane (b) Iso-contour plot of slag phase showing the slag velocity (Uo = 86 m/s)
193
Figure 7-3: (a) Diagram showing injection phenomena in zinc fuming, (b) schematic
representation of the sequence of reactions in the bath (both the figures are taken from
Richards et al.[21])
Figure 7-4(a) shows the coal mass injected into the molten slag bath through the
tuyeres and Figure 7-4(b) is the graphical representation of the average coal mass
along the axial direction (l/L>0.041) from the tuyere tip. By comparing the Figure
7-2(b) and Figure 7-4(a), scalar fraction of the injected coal shows a portion of the
injected coal entrains in the slag phase. Here, the coal scalar fraction is represented as
a continuum phase colour rather that discrete particulate phase, as the simulation is
based on Eulerian approach (phases are treated as continua). Other than two main
reaction zones pointed out by Richards et al. [21] (reduction in the slag by entrained
coal particles and generation of heat in the tuyere gas stream by combustion of un-
entrained coal), another reaction zone has been predicted by the simulation results,
which is in the free surface. Injected coal that carried away by the gas jet to the free
surface contributes to some extent in overall zinc elimination. Zinc elimination
behaviour from the slag near the free surface will be discussed in the later section.
As shown in Figure 7-4(a), some portion of the injected coal (light blue colour) is
being carried by the tuyere gas stream up-to the free surface of the molten slag bath.
Bath surface height (1m from the bottom wall) is also indicated in that figure. Some
portion of the injected coal is also entrained in the slag bath. To find out the coal mass
injected into the slag at the centre line joining the tuyere centre (at d/D = 0.5, h/H =
0.02 as shown on Figure 7-4(a)), average coal flow rates through the computational
194
cells along the axial direction have been calculated. This graph was plotted by time
averaging the coal mass along that axial line, as shown on Figure 7-4(b). Figure
shows, the injected coal penetrates up-to the distance of l/L = 0.18 along the centre
line from the tuyere tip. This portion of the injected coal in the slag phase take part in
the reduction reaction of ZnO.
Turbulent kinetic energy distributions at the two tuyere tips and inside the bath are
shown on Figure 7-5. Highest turbulence inside the bath was found in front of the
tuyere tips varying from 5 – 7 m2/s
2 over the entire range of simulation time. In the
tuyere gas stream, average turbulence kinetic energy was found to be 1 – 2 m2/s
2.
Generation of turbulence is favourable in accelerating the mixing process, however,
from both the tuyere tips, for l/L > 0.25, no significant generation of turbulence was
predicted.
7-4 (a)
195
7-4 (b)
Figure 7-4: (a) Injected coal mass at the tuyere tip (b) Average coal mass along the
axial direction from the tuyere tip(Uo = 86 m/s, d/D = 0.5, h/H = 0.02)
Figure 7-5: Turbulent kinetic energy distribution inside the modelled furnace (Uo = 86
m/s)
196
7.2.2 Jet Penetration
To investigate the tuyere tip jet behaviour, average jet penetration depth at the tuyere
tip and jet expansion angle was measured from the simulation results. Among all the
parameters describing the jet, penetration depth of the injected jet is the most
important one [151]. Some investigators studied horizontal gas injection behaviour in
fluidized bed, AOD, refining ladle etc. and suggested some correlations regarding jet
penetration depth of gas-solid fluidized bed [151-153]. Zhu et al. [154] compared the
mixing efficiency of different mode of horizontal gas injection in an air-water system.
Hoefele and Brimacombe [3] carried out laboratory experiments of horizontal jet
injected into water, zinc – chloride solution and mercury bath and commented that
penetration length increases with the modified Froude number ( FrN′ ) and the gas-
liquid density ratio
ρ
ρ
l
g . The authors suggested the following equation to predict
the penetration depth,
( ) ( ) 35.0
lg
46.0
Fr
o
P N7.10d
lρρ′= 7.1
Where,
( )[ ]ogl
2
og
Frdg
UN
ρ−ρ
ρ=′ 7.2
is the modified Froude number, od is the diameter of the tuyere, oU is the velocity of
the injected gas at the tuyere exit, g is acceleration due to gravity and the subscript g
and l represents gas and liquid, respectively.
197
Figure 7-6: Gas phase volume fraction at tuyere 1 tip showing jet penetration length
(lp) and jet expansion angle ( Θ ) for Uo = 86 m/s
In the present simulation, three different tuyere jet velocities (50, 67.8 and 86 m/s)
were simulated. To preserve the continuity of same mass flow rate as the plant data
D1 provided by Richards et al. [21] at different tuyere inlet velocity, blast temperature
at inlet were adjusted to maintain the same mass flow rate of 0.158 kg/s. For example,
at 100o C and 200
o C, the air density is 0.946 kg/m
3 and 0.746 kg/m
3 respectively.
Thus, tuyere inlet velocities of 67.8 m/s and 86 m/s maintains the same mass flow rate
of air (0.158 kg/s) at 100o C and 200
o C respectively. Penetration length (lp) and jet
expansion angle ( Θ ) were also measured for each of the three cases. As the present
simulation is based on Euler-Euler approach, where gas and liquid phases act as
interpenetrating continua, there is no sharp gas-liquid interface. Hence, to measure the
jet penetration depth, highest distance of the gas volume fraction of 0.8 from the
tuyere tip has been considered as the jet penetration depth, which has been widely
adopted by others researchers [151, 153]. Figure 7-6 shows the jet penetration length
(lp) and jet expansion angle ( Θ ) from the simulation results. Jet expansion angles
measured for the jet velocity of 67.8 m/s and 86 m/s are 85 to 110 degrees and 75 to
95 degrees respectively. Oryall and Brimacombe [155] measured that the jet
expansion angle for air-mercury system was 155o. As discussed by Zhu et al. [154] the
jet expansion angle would vary between 20o to 155
o based on the liquid density,
viscosity and surface tension. From the present simulation results it is clear that jet
expansion angle is also dependent on the jet injection velocity in addition to fluid
198
properties. Hence, for present the zinc slag fuming process considered, with the
injection conditions (Uo = 86 m/s, slag density of 3900 kg/m3), mean jet expansion
angle was measured as 85o.
Figure 7-7 (a): Comparison of the CFD results of tuyere jet penetration length (lp) with
correlation provided by Hoefele and Brimacombe [3] from experimental work
Figure 7-7(a) is the comparison of the present CFD results of jet penetration length (lp)
with the correlation provided by Hoefele and Brimacombe [3]. To validate the present
model, a no-coal test was carried out where only air jet was injected into the molten
slag bath at high temperature through the tuyeres and penetration depth were
calculated. The results were compared with the equation proposed by Hoefele and
Brimacombe [3], as shown in Figure 7-7(a). The simulation results showed close
correlation with the experimental studies of Hoefele and Brimacombe [3]. In Figure
7-7(a), jet penetration length (lp) is non-dimensionalised with the tuyere diameter (do).
Dimensionless penetration length parameter ( )op dl by using equation (7.1) for three
different tuyere inlet velocities (50, 67.8 and 86 m/s) were calculated both at NTP and
at present simulation conditions (different inlet temperatures and slag density 3900
kg/m3). The penetration length parameter ( )
op dl from present simulation results are
also presented in the Figure. Results from the present CFD simulation with coal
199
combustion shows large discrepancies with the correlation provided by Hoefele and
Brimacombe [3] (equation (7.1)). With the same velocity at NTP, gas has higher
density which increased the total mass flow rate resulting in increased momentum and
penetration length. The correlation in equation (7.1) was developed based on the
experimental work with air and three different liquid (water, zinc – chloride solution
and mercury). The injected jet in that experimental study did not carry any coal
particle, which would impart large amount of momentum on the liquid phase. The
injected coal particles with the jet would have been created larger penetration length.
The significant cause of this discrepancy is due to the combusting environment.
Injected coal imparts a greater momentum in the present simulation in addition to the
combustion at the tuyere tip, which causes the greater penetration distance.
Figure 7-7 (b) shows the decay of jet velocity with distance from orifice. The vertical
axis is showing the jet velocity ratio (Ux/U0), where, Ux is the velocity in the X –
direction and U0 is the inlet jet velocity at the tuyere tip. Dimensionless horizontal
distance from orifice (x/d0) is plotted on the horizontal axis, where x is the horizontal
distance from the orifice and d0 is the diameter of the tuyere (d0 = 0.056 m). Predicted
jet velocity decay for air – water results from Szekely et al. [13] are also shown on the
figure. For the present simulation condition with air – slag (slag density 3900 kg/m3),
the figure shows that the jet velocity decays to nearly zero at x/d0 = 4.129, whereas
for air – water, it decays to almost zero at x/d0 = 9.95.
200
Figure 7-7 (b): Decay of gas velocity with distance from orifice
7.2.3 Bath zones
Velocity vectors inside the molten slag bath are shown in Figure 7-8. A number of
recirculation zones have been observed inside the bath, as indicated in the figure. The
recirculation zones were transient in nature. Based on the slag phase velocity vector
observation, the slag bath can be divided into three zones, namely quiescent zone,
recirculation zone and tuyere gas stream zone. Among the three zones, highest
agitation was observed at the tuyere gas stream zone inside the bath followed by the
recirculation zones. Quiescent zones are found in a number of places such as area
below and above the tuyere near the wall (l/L < 0.041) and near the bottom wall of the
modelled furnace for distance 0.2< l/L > 0.8. Mixing, turbulence, combustion and
chemical reaction rate are found as the lowest in these zones inside the bath. However,
the length of the quiescent zone near the bottom wall mentioned here is quite variable
with respect to the gas injection velocity, coal particle size, coal injection rate and
injection conditions (i.e. pressurized or un-pressurized).
201
Figure 7-8: Velocity vectors of the slag phase inside the molten slag bath (Uo = 86
m/s)
7.2.4 Fuming Behaviour
Zinc slag fuming kinetics was modelled mathematically by Richards et al. [20] by
analysing some industrial zinc fuming plant data. Based on that mathematical model,
the rate controlling steps of zinc fuming process and coal combustion behaviour had
been well described by Richards et al.[19]. The CFD model of the zinc fuming TSL
furnace developed by the present investigators (Huda et al. [156]) revealed some
interesting insights for complex metallurgical flow inside the furnace. The developed
CFD model predicted the fuming behaviour and rate controlling steps in detail for
pilot plant scale zinc fuming top submerged lance (TSL) furnace. The CFD simulation
results by Huda et al. [156] predicted that mass transfer of ZnO from the bulk slag to
the slag-gas interface and rate of gas-carbon reaction (Boudouard reaction) plays the
dominant role in overall fuming efficiency for the simulation time studied. The model
showed, within the slag bath, fuming rate was found to be higher in area of the gas-
liquid interface, where ZnO diffuses to the gas-slag interface and is being reduced by
the CO present in the gaseous phase. The model further predicted that, above the bath,
a large quantity of fine slag droplets promotes a high fuming rate. Hence fine droplet
202
generation and deeper gas penetration in the slag bath will maximize the fuming rate
for a zinc fuming TSL furnace. However, the availability of CO and carbon limits the
fuming rate in flight.
Figure 7-9: ZnO distribution in the slag bath along the axial direction from the tuyere
tip at three different heights from the bottom wall (Uo = 86 m/s)
Figure 7-10: Schematic illustration of the tuyere tip jet and coal behaviour as observed
from the present simulation
203
In this section, zinc fuming behaviour from the present investigation will be discussed
in detail. Figure 7-9 shows the % of ZnO in the slag bath along the axial direction
from the tuyere tip (l/L > 0.041) at three different heights from the bottom wall (h/H =
0.02, 0.1 and 0.2, where h is the height in the bath from the bottom wall and H is the
total height of the modelled furnace, as mentioned in Figure 4-7(a)). At h/H = 0.02,
which is the line joining the tuyere centres, injected coal entrains in the molten slag
bath. However, within the simulation time studied, no significant reduction of ZnO
from the slag at that position was observed. The average fuming rate observed at this
plane is 0.07 wt%/min. From the plant data analysis, Richards et al. [20] found that
33% of the total injected coal entrains in the slag, 55% combusts in the tuyere gas
column and 12% bypasses the bath completely. Entrained coal particles seem to have
less influence on the overall fuming. Since the coal particles are non-wetting, it does
not react directly with the ZnO, but does so via the Boudouard reaction. Within the
simulation time studied, due to the shortage of oxygen around the entrained coal
particles, no CO/CO2 generation was observed. Hence, the model predicts that ZnO
reduction process is limited by absence of CO. This finding has already been
discussed by Huda et al. [156] for a TSL furnace, that the availability of CO at the gas-
liquid interface limits the fuming rate. The present simulation also predicts that the
major portion of the injected coal (50%) is combusted in the tuyere gas column. As a
result, the products of combustion (CO, CO2 and H2O) are predicted to be in the
tuyere gas column and gas-liquid emulsion zone around the column. The presence of
CO at the gas-liquid interface reduces the ZnO present in the slag phase. As Figure 7-9
shows, at a height of h/H = 0.1 (0.5 m from the bottom wall), ZnO reduction is
significant in the gas-liquid emulsion zone. The average zinc fuming rate found at this
plane is 0.11 wt%/min. Slag volume fraction plot (Figure 7-2 (a) and Figure 7-2(b))
shows that there are significant splash formations near the wall due to the shear force
exerted by the high velocity gas jet. The slag in flight falls back to the bath surface and
creates a cavity in the bath around mid surface. A portion of the injected coal also
carried away by the high velocity gas jet which reacts in the free bath surface and
forms CO/CO2. In addition, the entrained coal particles subsequently approach to the
free bath surface, which also undergoes the Boudouard reaction. The generated CO
and agitation due to the sloshing and splashes that comes back to the surface
accelerates the mass transfer of ZnO from the bulk slag to the slag gas interface and
204
formation of Zn by the reduction process. Surprisingly, the bypassed coal un-
combusted from the tuyere gas column seemed to play some role in overall zinc
fuming from the bath surface, as shown on Figure 7-9 at the height of h/H = 0.2. The
shape of the curve in Figure 7-9 at the height of h/H = 0.2 (1.0 m from the bottom
wall) indicates that the model predicts high ZnO reduction rate from the bath surface.
Figure 7-10 is the schematic illustration the tuyere tip gas jet and coal behaviour. At
h/H = 0.2 (1.0 m from the bottom wall), % of ZnO found for the length of 0.3<l/L >
0.7 are from 0.09 to 0.11, which gives an average ZnO of 10% within that length
mentioned after 60 seconds of simulation time studied. Hence, other than the two
reaction zones mentioned by Richards et al. [21] (reduction in the slag by entrained
coal particles and generation of heat in the tuyere gas stream by combustion of un-
entrained coal), one more reaction zone is marked within the bath that contributes to
significant fuming reaction which is in the bath surface as shown in Figure 7-10. The
overall zinc fuming rate predicted at this location is 0.115 wt%/min. Industrial study
by Richards et al. [21] showed that, for Company D (Figure 5 of Richards et al. [21]),
overall zinc fuming rate was 0.125 wt%/min. Other studies reported that, batch fuming
rates for zinc fuming furnaces using pulverized coal as fuel and reductant varies
between 0.15 to 0.27 wt%/ min for slags containing about 10 wt% Zn [150] and 0.09
wt%/ min for 9 wt% Zn slag (at 1275o C temperature) [102]. From the total simulation
time studied, the highest reduction rate of ZnO was observed in the gas-liquid
emulsion zone. Hence, from the simulation results, it is clear that for the time period
studied, zinc fuming is controlled the mass transfer of ZnO from the bulk slag to the
slag-gas interface and rate of gas-carbon reaction within the bath. The rate of mass
transfer is controlled by the level of agitation and generation of turbulence within the
bath.
205
Chapter 8
206
8 Conclusions and Recommendations
8.1 Conclusions
The observations of this research can provide useful information for improving the
process optimization and operating conditions of the slag fuming furnaces, in both
TSL and conventional approach. During the first stage of this research, gas injection
characteristic into water bath was investigated by using the Computational Fluid
Dynamic (CFD) modelling technique and swirled gas injection has found to play some
role in improving the mixing process in the near vicinity of the lance tip. The
simulation results for velocity fields and generation of turbulence inside the bath has
found to be in reasonable agreement with the previous water model experimental
study of Morsi et al. [1]. A semi-empirical equation was developed for the vertical
penetration distance of the annulus air jet into the water bath from the study. The idea
of mean convective mixing by volume exchange concept had revealed that mixing was
non-uniform and concentrated near the lance for the specific air-water system studied.
A study of the mixing process by turbulence mixing through turbulent diffusion also
revealed the similar phenomena. The formation of recirculation zone in the water bath
is quite favourable in creating uniform mixing; however, the recirculation zones
observed in this study were weak for air-water system. The air-water model was used
for higher density liquid by changing the water density as an exploratory step towards
the development of the next stage work.
The pilot plant scale CFD model of the zinc slag fuming process in top submerged
lance smelting furnace was developed in the next stage by using 3D Eulerian
multiphase flow approach. The simulation results give some interesting insights for
complex metallurgical flows and transient concentrations of slag components and
gaseous species inside the molten slag bath. The simulation results predicted the
generated turbulence, splashing and plume shape in the molten slag bath. The rate of
zinc fuming from the slag bath was validated through macro-step validation process
against the pilot plant trials of zinc fuming top submerged lance furnace by Waladan
et al. [2]. Simulation results showed that increased submergence level resulted in
increased fuming rate. Overall zinc fuming rate for 1/3 lance submergence level
( LH ′′ = 1/3) was found to be 1.3 times higher than 1/5 lance submergence level
207
( LH ′′ = 1/5). Increased residence time of the generated splash with large interfacial
area above the slag bath accelerates the fuming rate from the slag in flight. The
sloshing phenomena and splashes that come back to the bath surface increase the gas-
carbon reaction on the bath surface, thus increasing the formation rate of CO, which is
responsible for increased fuming rate from the slag in flight. The simulation results
revealed that the mass transfer of ZnO from the bulk slag to the gas-slag interface and
gas-carbon reaction both play vital roles in controlling the overall fuming rate.
The developed model in the second stage was further extended to submerged coal
combustion and employed to the conventional tuyere blown zinc slag fuming process.
The model gives a clear understanding of the interesting insights for complex
metallurgical flows, tuyere tip gas jet and coal combustion behaviour and overall zinc
fuming behaviour from the slag bath. The simulation results predicted that momentum
exerted by the coal combustion process is the dominant factor for larger jet penetration
length of a tuyere blown system. Three different zone within the bath are identified
which are tuyere gas stream zone, recirculation zone and quiescent zone. Other than
the two main reaction zones mentioned by Richards et al. [21], one more reaction zone
is observed in the bath surface that contributes to significant fuming reaction The
sloshing phenomena and splashes that come back to the bath surface increase the gas-
carbon reaction on the bath surface, thus increasing the formation rate of CO, which is
accelerates the ZnO reduction process from the bath surface. Tuyere jet penetration
length ( pl ) was compared with the equation provided by Hoefele and Brimacombe
from isothermal experimental work ( ( ) ( ) 35.0
lg
46.0
Fr
o
P N7.10d
lρρ′= ) and found 2.26
times higher, which can be attributed to coal combustion and gas expansion at high
temperature. The jet expansion angle measured for the slag system studied is 85o for
the specific inlet conditions during the simulation time studied. Highest coal
penetration distance was found to be l/L = 0.2, where l distance from the tuyere tip
along the centre line and L is the total length (2.44m) of the modelled furnace. The
model also predicted that 10% of the injected coal bypasses the tuyere gas stream un-
combusted and carried to the free surface by the tuyere gas stream, which contributes
to zinc oxide reduction near the free surface.
208
8.2 Recommendations for further work
A few recommendations are now made for future work to extend the completed
research. These recommendations aim to create better understanding of the reaction
kinetics as well as zinc fuming behavior in presence of some special slag constituents.
Coal combustion modelling in the tuyere blown furnace also needs modification. The
recommendations are listed below.
First of all, in the chemical reactions considered, the effect of ferrous iron
oxidation was not taken into consideration. Ferrous iron oxidation within the
bath supposed to play some role in the overall fuming rate, as suggested by
Richards et al. [20]. Hence, inclusion of the oxidation behaviour of ferrous iron
within the bath may lead to improvement of the developed model.
The ferric iron level in the bath was considered as negligible, which also plays
some role in the overall fuming kinetics, though the percentage is small in the
slag constituents. Inclusion of the effect of ferric in the bath may also improve
the results.
In the zinc fuming TSL furnace model, carbon was added in the bath as a
reducing source, assuming only fixed carbon takes part in the reduction
reaction. No devolatilization and char oxidation model was considered, which
would have affected the results to some extent. By providing a more realistic
treatment of the behaviour of coal as a reducing agent, the model can be
improved further.
Coal combustion model incorporated in the tuyere blown furnace model
considered the coal as a continuum phase, rather than discrete particulate
phase. In addition, the coal was considered to have no moisture content, hence
no evaporation model of was considered. Though the burn out time of the
injected coal is very small (in the order of ms), the hydrodynamic effects of the
bypassed solid coal particles would have affected the overall bath behaviour.
Thus, by implementing the coupling of Lagrangian and Eulerian approach to
incorporate the behaviour of solid coal particles can be an improvement of the
model in terms of computing time.
209
Last but not the least, longer simulation time should be considered for better
understanding of the overall process. In other way, zinc fuming rate at low %
of ZnO can be investigated further, which is the scenario of the nearly ending
time of a fuming cycle.
210
References:
1. Morsi, Y.S., W. Yang, B.R. Clayton, and N.B. Gray, Experimental
Investigation of swirl and non swirl gas injections into liquid baths using
submerged vertical lances. Can. Metall. Q., 2000. 39(1): p. 87-98.
2. Waladan, M., G.R. Firkin, J. Bultitude-Paull, B.W. Lightfoot, and J.M.Floyd,
Top Submerged Lancing for Recovery of Zinc from ISF Slag, in Lead-Zinc
Conf.' 901990: California, USA.
3. Hoefele, E.O. and J.K. Brimacombe, Flow Regimes in Submerged Gas
Injection Metall. Trans. B, 1979. 10(4): p. 631-648.
4. Floyd, J., Converting an Idea into a Worldwide Business Commercializing
Smelting Technology. Metall. Trans. B., 2005. 36(B): p. 557-575.
5. Guthrie, R.I.L., Fluid Flows in Metallurgy- Friend or Foe? . Metall. Trans. B.,
2004. 35(B): p. 417-437.
6. Irons, G.A. and R.I.L. Guthrie, Bubbling Behaviour in Molten Metals. Can.
Metall. Q., 1981. 19(4): p. 381-387.
7. Mazumdar, D. and R.I.L. Guthrie, Hydrodynamic Modelling of Some Gas
Injection Procedures in Ladle Metallurgy Operations. Metall. Trans. B., 1985.
16 B(1): p. 83 - 90.
8. Sahai, Y. and R.I.L. Guthrie, Hydrodynamics of Gas-Stirred Melts: Part I.
Gas/Liquid Coupling. Metallurgical and Materials Transaction B, 1982. 13(B):
p. 193-202.
9. Sahai, Y. and R.I.L. Guthrie, Hydrodynamics of Gas-Stirred Melts: Part II.
Axisymmetric Flows. Metall. Trans. B., 1982. 13(B): p. 203-211.
10. Floyd, J., Converting an Idea into a Worldwide Business Commercializing
Smelting Technology. Metall. Trans. B., 2004. 36(B): p. 557-575.
11. Scheepers, E., Y. Yang, M.A. Reuter, and A.T. Adema, A dynamic-CFD
hybrid model of a submerged arc furnace for phosphorus production. Minerals
Engineering 2006. 19(3): p. 309-317.
12. Stephens, D., M. Tabib, M.P. Davis, and M.P. Schwarz, CFD Simulation of
bath dynamics in the HIsmelt smelt reduction vessel for iron production, in 8th
International Conference on Computational Fluid Dynamics in the Oil & Gas,
Metallurgical and Process Industries2011: Trondheim, Norway.
211
13. Szekely, V.J. and N.J. Themelis, Rate Phenomena in Process Metallurgy1971,
New York: Wiley-Interscience.
14. Naser, J. Industrial Applications of CFD - A Review of Few Examples. in 7th
International Conference on CFD in the Minerals and Process Industries.
2009. Melbourne, Australia.
15. Mittoni, L.J. and M.P. Schwarz. Computational Fluid Dynamics as a Tool in
the Minerals Processing and Metal Production Industries. in 13th Australisian
Fluid Mechanics Conference. 1998. Melbourne, Australia.
16. Witt, P., C. Solnordal, L. Mittoni, S. Finn, and J. Pluta, Reduction in Fume
Emission from a Slag Fumer Charge Port Using CFD Modelling, in Third
International Conferrence on CFD in the Minerals and Process
Industries2003, CSIRO: CSIRO, Melbourne, Australia. p. 121-126.
17. Bockelie, M.J., et al. Computational Simulations of Industrial Furnaces. in
International Symposium on Computational Technologies For
Fluid/Thermal/Chemical Systems with Industrial Applications. 1998. San
Diego, California, USA.
18. Schwarz, M.P. and M.P. Davis. Use of CFD in the Development and Scale-up
of the HIsmelt Process. in Rod Guthrie Honorary Symposium on Process
metallurgy. 2011. Montreal, Canada: McGill Metals Processing Centre.
19. Richards, G.G. and J.K. Brimacombe, Kinetics of Zinc Slag-Fuming Process:
Part III. Model Predictions and Analysis of Process Kinetics. Metall. Trans. B,
1985. 16(B): p. 541-549.
20. Richards, G.G. and J.K. Brimacombe, Kinetics of Zinc Slag-Fuming Process:
Part II. Mathematical Model. Metall. Trans. B, 1985. 16(3): p. 529-540.
21. Richards, G.G., J.K. Brimacombe, and G.W. Toop, Kinetics of Zinc Slag-
Fuming Process: Part I. Industrial Measurements Metall. Trans. B, 1985.
16(B): p. 513-527.
22. Solnordal, C.B. and N.B. Gray, Heat transfer and pressure drop
considerations in the design of Sirosmelt lances. Metall. Trans. B., 1996.
27(2): p. 221-230.
23. www.zincworld.org. 2011; Available from: www.zincworld.org.
24. Habashi, F. Discovering the 8th Metal - A History of Zinc. Available from:
http://www.iza.com/publications.html.
212
25. www.en.wikipedia.org. 2011; Available from:
http://en.wikipedia.org/wiki/Zinc.
26. www.iza.com. 2011; Available from: http://www.iza.com/statistics.html.
27. Engh, T.A., Principles of Metal Refining1992: Oxford University Press Inc.,
New York. 473.
28. EPA. Zinc Smelting. October 1986; Available from:
http://www.epa.gov/ttn/chief/ap42/ch12/final/c12s07.pdf.
29. http://en.wikipedia.org/wiki/Zinc_smelting. 2011; Available from:
http://en.wikipedia.org/wiki.
30. Richards, G.G., Kinetics of the Zinc Slag Fuming Process, in Dept. of
Metallurgical Engineering1979, University of British Columbia: Vancouver,
Canada. p. 335.
31. Suzuki, R., S. Goto, and K. Azuma, On Fuming of Zinciferrous Slags. J. of
The Fac. of Eng., University of Tokyo (B), 1970. XXX(3): p. 247-287.
32. Reddy, R.G., V.L. Prabhu, and D. Mantha, Zinc Fuming from Lead Blast
Furnace Slag. High Temp. Mat. Processes, 2003. 21(6): p. 377-386.
33. Choi, C.Y. and Y.H. Lee, Treament of Zinc residues by Ausmelt Technology at
Onsan Zinc Refinery, in REWAS - Global Symposium on Recycling, Waste
Treatment and Clean Technology1999: San Sebastian, Spain.
34. Quarm, T.A.A., Submerged Smelting, in United States PatentJuly 1, 1975,
Bechtel International Corporation: USA. p. 9.
35. Hughes, S., R.W. Matusewicz, M.A. Reuter, and D. Sherrington. Ausmelt-
Extractive value from EAF Dust. in EMC 2007. Dusseldorf, Germany.
36. Sofra, D.J. and A. Heinz. Effective Treatment of Zinc Bearing Dusts &
Residues – Solution Ausmelt Technology. in EMC. 2003. Hannover, Germany.
37. Floyd, J., An update of Ausmelt Technology for Zinc and Lead processing., in
28th Annual Hydrometallurgical Meeting1998, The Metallurgical Society of
CIM: Calgary, Canada.
38. Floyd, J. The Third Decade of Top Submerged Lance Technology in Howard
Worner Int. Symp. Inj. Pyro. . 1996. Melbourne, Australia: TMS.
39. Hoang, J., M.A. Reuter, R. Matusewicz, and S. Hughes, Top Submerged Lance
Direct Zinc Smelting. Minerals Engineering, 2009. 22: p. 742-751.
213
40. Hughes, S., M.A. Reuter, R. Baxter, and A. Kaye, Ausmelt Technology for
Lead and Zinc Processing, in Lead and Zinc 20082008, SAIMM: Durban,
South Africa. p. 147-162.
41. Korea Zinc 2008 Annual Report. <http://www.koreazinc.co.kr/english>.
42. Hoang, J., M.A. Reuter, R. Matusewicz, and S. Hughes, Top Submerged Lance
(TSL) Direct Zinc Smelting, in Zinc Proc. Conf.2008: Brisbane, Australia.
43. Taniguchi, S., S. Kawaguchi, and A. Kikuchi, Fluid Flow and Gas-Liquid
Mass Transfer in Gas-injected Vessels. Applied Mathematical Modelling 2002.
26: p. 249-262.
44. Iguchi, M., Y. Eguchi, A. Kawasaki, S.-y. Kitamura, and K.-i. Naito, Swirl
Motions in a Cylindrical Bath Induced by Gas Injection under Reduced
Pressure on Surface. ISIJ Int., 1999. 39(8): p. 767-771.
45. Iguchi, M., S. Hosohara, T. Koga, R. Yamaguchi, and Z.-i. Morita, The Swirl
Motion of Vertical Bubbling Jet in a Cylindrical Vessel ISIJ Int., 1993. 33(10):
p. 1037-1044.
46. Iguchi, M., H. Kawabata, Y. Nakajima, and Z.-i. Morita, Continuous
Measurement of Bubble Characteristics in a Molten Iron Bath with Ar Gas
Injection. ISIJ Int., 1994. 34(12): p. 980-985.
47. Iguchi, M., A. Kawajiri, H. Tomida, and Z.-i. Morita, Effects of Viscosity of
Liquid on the Characteristics of Vertical Bubbling Jet in a Cylindrical Vessel.
ISIJ Int., 1993. 33(3): p. 361-368.
48. Iguchi, M., Z.-i. Morita, H. Tokunaga, and H. Tatemichi, Heat Transfer
between Bubbles and Liquid during Cold Gas Injection. ISIJ Int., 1992. 32(7):
p. 865-872.
49. Iguchi, M., K. Nozawa, and Z.-i. Morita, Bubble Characteristics in the
Momentum Region of Air-Water Verticle Bubbling Jet. ISIJ Int., 1991. 31(9):
p. 952-959.
50. Iguchi, M., J.-i. Tani, T. Uemura, H. Kawabata, H. Takeuchi, and Z.-i. Morita,
The Characteristics of Water and Bubbling Jets in a Cylindrical Vessel with
Bottom Blowing. ISIJ Int., 1989. 29(4): p. 309-317.
51. Iguchi, M., T. Uemura, H. Yamaguchi, T. Kuranaga, and Z.-i. Morita, Fluid
Flow Phenomena in a Cylindrical Bath Agitated by Top Lance Gas Injection.
ISIJ Int., 1994. 34(12): p. 973-979.
214
52. Koh, P.T.L. and R.N. Taylor. Liquid Splashes at the Surface of Baths Agitated
by Gas Injection Through Submerged Lances. in CHEMECA'90. 1990.
Auckland, New Zealand.
53. Igwe, B.U.N., S. Ramachandran, and J.C. Fulton, Jet Penetration and Liquid
Splash in Submerged Gas Injection. Metall. Trans. B., 1973. 4(August): p.
1887-1894.
54. Koria, S.C. and S. Singh, Experimental Investigation on the Design of Gas
Injecting Lances. ISIJ International, 1989. 29(8): p. 650-657.
55. Morsi, Y.S., W. Yang, and I. El-Katatny, Experimental Characterisation of
AusIron Top Submerged Injection System. Ironmaking and Steelmaking, 2008.
35(1): p. 69-74.
56. Nilmani, M. and D.S. Conochie. Gas Dispersion With Swirled Lances. in 4th
Int. Con. Inj. Met.: Scaninject IV. 1986. Lulea, Sweden: MEFOS.
57. Neven, S., B. Blanpain, and P. Wollants. Injection Dynamics in an Isasmelt
Reactor. in EPD Cong. 2002. 2002. TMS.
58. Dave, N. and N.B. Gray, Fluid Flow through Lances with Constant and
Variable Pitch Swirled Inserts. Metall. Trans. B., 1991. 22(B): p. 13-20.
59. Yokoya, S., S. Takagi, M. Iguchi, K. Marukawa, and S. Hara, Formation of
Fine Bubble through Swirling Motion of Liquid Metal in the Metallurgical
Container. ISIJ Int., 2000. 40(6): p. 572-577.
60. Iguchi, M., S. Hosohara, T. Kondoh, Y. Itoh, and Z.-i. Morita, Effects of the
Swirl Motion of Bubbling Jet on the Transport Phenomena in a Bottom Blown
Bath. ISIJ Int., 1994. 34(4): p. 330-337.
61. Ihira, H., M. Iguchi, and S. Hosohara, Swirl Motions of Two Bubbling Jets
Induced by Bottom Gas Injections. ISIJ Int., 2003. 43(8): p. 1283-1285.
62. Liow, J.-L., D.E. Morton, A.E. Guerra, and N.B. Gray. Dynamics of Splash
Formation in Gas Injected Systems. in The Howard Worner International
Symposium on Injection in Pyrometallurgy. 1996. TMS.
63. Liow, J.-L., W.H.R. Dickinson, M.J. Allan, and N.B. Gray, Study of Slopping
and Splashing in a Cylindrical Bath with Top-Submerged Injection. Metall.
Trans. B., 1995. 26(B): p. 887-889.
64. Davidson, J.F. and B.O.G. Schüler, Bubble Formation at an Orifice in an
Inviscid Liquid. Trans. Inst. Chem. Eng., 1960. 38: p. 335-342.
215
65. Snabre, P. and F. Magnifotcham, Formation and Rise of a Bubble Stream in a
Viscous Liquid. European Physical Journal B, 1998. 4: p. 369-377.
66. Komarov, S.V. and M. Sano, Bubble Behaviour and Heat Transfer in
Preheated Gas Injection into Liquid Bath. ISIJ International, 1998. 38(10): p.
1045-1052.
67. Kumagai, T. and M. Iguchi, Instability Phenomena at Bath Surface Induced by
Top Lance Gas Injection. ISIJ Int., 2001. 41(Supplement): p. 52-55.
68. Iguchi, M., K. Sasaki, K. Nakajima, and H. Kawabata, Coanda Effect on
Bubble Characterstics in a Bubbling Jet Rising Near the Side Wall of a
Cylindrical Vessel. ISIJ Int., 1998. 38(12): p. 1297-1303.
69. Diaz, M.C., S.V. Komarov, and M. Sano, Bubble Behaviour and Absorption
Rate in Gas Injection through Rotary Lances. ISIJ International, 1997. 37(1):
p. 1-8.
70. Rankin, W.J., F.R.A. Jorgensen, T.V. Nguyen, P.T.L. Koh, and R.N. Taylor.
Process Engineering of Sirosmelt Reactors: Lance and Bath Mixing
Characteristics. in Extractive Metallurgy. 1989.
71. Singh, R.P. and D.N. Ghosh, Cold Model Study of Mixing and Mass Transfer
in LBE Process of Steelmaking. ISIJ International, 1990. 30(11): p. 955-960.
72. Murthy, G.G.K. and S.P. Mehrotra, Mixing in Liquid Baths by Gas Injection.
Ironmaking and Steelmaking, 1992. 19(5): p. 377-389.
73. Diaz, M.C., T. Iida, S.V. Komarov, and M. Sano, Mixing Characterisitcs
under Gas Injection through Rotary Lance Submerged in Liquid Bath ISIJ
International, 1995. 35(5): p. 464-471.
74. Krishnakumar, K., N.B. Ballal, P.K. Sinha, M.K. Sardar, and K.N. Jha, Water
Model Experiments on Mixing Phenomena in a VOD Ladle. ISIJ International,
1999. 39(5): p. 419-425.
75. Schwarz, M.P. and P.T.L. Koh. Numerical Modelling of Bath Mixing by
Swirled Gas Injection. in 4th Int. Con. Inj. Met.: Scaninject IV. 1986. Lulea,
Sweden: MEFOS.
76. Liovic, P., M. Rudman, and J.-L. Liow, Numerical Modelling of Free Surface
Flows in Metallurgical Vessels. App. Math. Mod., 2002. 26: p. 113-140.
77. Liovic, P., J.-L. Liow, and M. Rudman, A Volume of Fluid (VOF) Method for
the Simulation of Metallurgical Flows. ISIJ Int., 2001. 41(3): p. 225-233.
216
78. Solnordal, C.B., F.R.A. Jorgensen, and R.N. Taylor, Modelling the heat flow to
an operating sirosmelt lance. Metall. Trans. B., 1998. 29(2): p. 485-492.
79. Turkoglu, H. and B. Farouk, Numerical Comutations of Fluid Flow and Heat
Transfer in Gas Injected Iron Baths. ISIJ International, 1990. 30(11): p. 961-
970.
80. Shin, M.K., S.D. Lee, S.H. Joo, and J.K. Yoon, A Numerical Study on the
Combustion Phenomena Occurring at the Post Combustion Stage in Bath-type
Smelting Reduction Furnace. ISIJ International, 1993. 33(3): p. 369-375.
81. Zhu, M.-Y., I. Sawada, N. Yamasaki, and T.-C. Hsiao, Numerical Simulation
of Three-Dimensional Fluid Flow and Mixing Process in Gas Stirred Ladles
ISIJ International, 1996. 36(5): p. 503-511.
82. Castillejos, A.H. and J.K. Brimacombe, Measuring of Physical Characteristics
of Bubbles in Gas-Liquid Plumes: Part II. Local Properties of Turbulent Air-
Water Plumes in Vertically Injected Jets. Metallurgical and Materials
Transaction B, 1987. 18(4): p. 659-671.
83. Bernard, R.S., R.S. Maier, and H.T. Falvey, A Simple Computational Model
for Bubble Plumes. Applied Mathematical Modelling, 2000. 24: p. 215-233.
84. Milelli, M., B.L. Smith, and D. Lakehal. Some New Approaches to Bubble
Plume Modelling. in 2001 ASME International Mechanical Engineering
Congress and Exposition 2001. New York: HTD.
85. Morsi, Y.S., W. Yang, D. Achim, and A. Acquadro. Numerical and
Experimental Investigation of Top Submerged Gas Injection System. in 5th
International Conference on Computational Methods in Contact Mechanics
2001. Seville, Spain.
86. Cross, M., T.N. Croft, G. Djambazov, and K. Pericleous. Computational
Modelling of Bubbles, Droplets and Particles in Metals Reduction and
Refining. in 3rd International Conference on CFD in the Minerals and Process
Industries 2003. Melbourne, Australia: CSIRO.
87. Alexiadis, A., Bubble Dispersion Patterns in Bubbly-Flow Released From a
Porous Plug into a Gas-Stirred Ladle. Applied Mathematical Modelling, 2007.
31: p. 1534-1547.
88. Ward, D.H., Lead and Zinc Technology for Survival, in Smelting and Refining
Operators Symposium1985: Queensland, Australia. p. 63-67.
217
89. Bell, R.C., G.H. Turner, and E. Peters, Fuming of Zinc from Lead Blast
Furnace Slags. TMS-AIME, 1955. 203: p. 472-477.
90. Kellogg, H.H., A Computer Model of the Slag-Fuming Process for Recovery of
Zinc Oxide. Trans. Metall. Soc. of AIME, 1967. 239: p. 1439-1449.
91. Kellogg, H.H., A New Look at Slag Fuming. E/MJ, 1957. 158(3): p. 90-92.
92. Quarm, T.A.A., Slag Fuming - Kinetic or Thermodynamic. E/MJ, 1968.
169(1): p. 92-93.
93. Quarm, T.A.A., A Submerged Smelting Process for Complex Zinc
Concentrates, in Mining MagazineFebruary, 1988. p. 120-122.
94. Richards, G.G. Optimization of the Zinc Slag Fuming Process. in EPD
Congress. 1994. TMS.
95. Richards, G.G. Continuous Fuming of Zinc - Containing Slags. in The Howard
Worner Int. Symp. on Inj. Pyro. 1996. Melbourne, Australia: TMS.
96. Cockcroft, S.L., G.G. Richards, and J.K. Brimacombe, Mathematical Model of
Lead Behaviour in the Zinc Slag Fuming Process. Can. Metall. Q., 1988.
27(1): p. 27-40.
97. Cockcroft, S.L., G.G. Richards, and J.K. Brimacombe, High Pressure coal
injection in Zinc Slag Fuming. Metall. Trans. B, 1989. 20(B): p. 227-235.
98. Chang, K., G.A. Irons, and Q. Jiao, A Mathematical Model of a DC Electric
Furnace for Recovery of Zinc from Lead Blast Furnace Slag. Can. Metall. Q.,
2005. 44(2): p. 281-288.
99. Grant, R.M. The Derivation of Thermodynamic Properties of Slags From Slag
Fuming Plant Data. in Aus./Jap. Ext. Metall. Symp. 1980. Sydney, Australia:
Australasian Institute of Mining and Metallurgy.
100. Scholey, K.E., G.G. Richards, and I.V. Samarasekera, Heat Transfer
Phenomena in Water Cooled Zinc-Fuming Furnace Jackets. Metall. Trans. B.,
1991. 22(B): p. 163-175.
101. Haralampiev, G. and N. Popov, Investigation of the Continuous Fuming
Process by Physical Model Can. Metall. Q., 1999. 38(1): p. 1-9.
102. Lehner, T. and R. Lindgren, Investigation of Fluid Flow Phenomena in
Boliden's Zinc Fuming Plant, in CIM Meeting1985: Vancouver, Canada. p. 99-
120.
103. Denholm, W.T., N.B. Gray, M. Nilmani, J. Malone, C.R. Fountain, R.N.
Taylor, and J.M.Floyd. Sirosmelt Investigation - An Example of Process
218
Evaluation Techniques. in Smelting and Refining Operators Symposium. 1985.
Queensland, Australia.
104. Waladan, M., Recovery of Zinc from Lead Blast Furnace Slag using a Mixture
of Natural Gas and Air in Top Submerged Lancing, in Lead and Zinc
19951995: Sendai, Japan. p. 520-531.
105. Waladan, M. and M. Nilmani, The Effect of Injection Parameters on Slag
Fuming. Can. Metall. Q., 1995. 34(4): p. 311-318.
106. Verscheure, K., M. Van Camp, B. Blanpain, P. Wollants, P. Hayes, and E. Jak,
Continuous fuming of zinc-bearing residues: Part I. Model development.
Metall. Trans. B, 2007. 38(1): p. 13-20.
107. Gupta, S.K., Kinetics of the Dissolution of Zinc Sulfide in an Oxidizing Slag.
Metallurgical and Materials Transaction B, 1990. 21(B): p. 867-872.
108. Langberg, D., X. Liu, and A. Kyllo, Behaviour of Coal Injected into Molten
Baths, 2000, Australian Coal Association Research Program (ACARP).
109. Miyake, M., Zinc-Lead Smelting at Hachinohe Smelter, 1995: Hachinohe p.
39-50.
110. Matusewicz, R. and J. Sofra, Lead Smelting, Copper Smelting And Copper
Converting Using Ausmelt Technology, in NMD/ATM 20012001:
Bhubaneswar, Orissa, India. p. 1-24.
111. Jak, E. and P. Hayes, Phase Equilibria and Thermondynamics of Zinc Fuming
Slags. Can. Metall. Q., 2002. 41(2): p. 163-174.
112. Jak, E. and P. Hayes, The Role of Slag Chemistry in the Design of Freeze
Linings for Zinc Slag Fuming, in Zinc Proc. Conf.2008: Brisbane, Australia.
113. Verscheure, K., M. Van Camp, B. Blanpain, P. Wollants, P. Hayes, and E. Jak.
Freeze Linings in Zinc Fuming Processes. in Sohn International Symposium on
Advanced Processing of Metals and Materials. 2006. TMS.
114. Dal, I. and W.J. Rankin. The Effect of Sulphur on the Rate of Reduction of Zinc
Oxide from Slag. in Int. Symp. - World Zinc'93. 1993. Hobart, Australia.
115. Verscheure, K., M. Van Camp, B. Blanpain, P. Wollants, and P. Hayes.
Modelling of a Submerged Plasma Process for Zn-Containing Residues
Treatment. in EMC 2005. 2005.
116. Verscheure, K., M. Van Camp, B. Blanpain, P. Wollants, P. Hayes, and E. Jak.
Investigation of Zinc Fuming Processes for the Treatment of Zinc - Containing
Residues. in John Floyd International Symposium on Sustainable
219
Developments in Metals Processing. 2005. Melbourne, Australia: NCS
Associates.
117. Schwarz, M.P., Simulation of gas injection into liquid melts. App. Math.
Model., 1995. 20: p. 41-51.
118. Vaarno, J., J. Pitkälä, T. Ahokainen, and A. Jokilaakso, Modelling gas
injection of a Peirce-Smith-converter. App. Math. Mod., 1998. 22(11): p. 907-
920.
119. Davis, M.P., R.J. Dry, and M.P. Schwarz. Flow Simulation in the HIsmelt
Process. in 3rd International Conference on CFD in the Minerals and Process
Industries. 2003. Melbourne, Australia: CSIRO Minerals.
120. Abdel-Latif, M., Fundamentals of zinc recovery from metallurgical wastes in
the Enviroplas process. Minerals Engineering, 2002. 15: p. 945-952.
121. Tarasov, A.V. and A.D. Besser. Processing of Lead-Zinc Raw Material. in
Sohn International Symposium on Adavanced Processing of Metals and
Materials. 2006. TMS.
122. Schwarz, M.P., Improving Zinc Processing using Computational Fluid
Dynamics Modelling, in Zinc Processing 082008: Brisbane, Australia.
123. Versteeg, H.K. and W. Malalasekera, An introduction to Computational Fluid
Dynamics. Second ed2007: Pearson Education Limited. 500.
124. Ferziger, J.H. and M. Peric, Computational Methods For Fluid Dynamics. 3rd
ed2002: Springerlink.
125. AVL FIRE CFD Multiphase Manual v 20082008, Graz, Austria: AVL.
126. W., H.C. and N.B. D., Volume of fluid (VOF) method for the dynamics of free
boundaries. Journal of Computational Physics, 1981. 39(1): p. 201-225.
127. Patankar, S.V. and D.B. Spalding, A Calculation Procedure for Heat, Mass
and Momentum Transfer in Three-Dimensional Parabolic Flows Int. J. Heat.
Mass. Trans., 1972. 15: p. 1787-1806.
128. Sato, Y. and K. Sekoguchi, Liquid Velocity Distribution in Two Phase Bubble
Flow. Int. J. Multiphase Flow, 1975. 2(1): p. 79-95.
129. Launder, B.E. and D.B. Spalding, The Numerical Computation of Turbulent
Flows. Com. Meth. App. Mech. Eng., 1974. 3(2): p. 269-289.
130. Chahed, J., V. Roig, and L. Masbernat, Eulerian-Eulerian two-fluid model for
turbulent gas-liquid bubbly flows. Int. J. Multiphase Flow, 2003. 29: p. 23-49.
131. AVL FIRE CFD Solver Manual v 2008. Vol. 2008. 2008, Graz, Austria: AVL.
220
132. Wilcox, D.C., Turbulence Modelling for CFD. 3rd ed1998, La Canada
Flintridge, CA: DCW Industries, Inc.
133. Ranz, W.E. and W.R. Marshall, Evaporation from Drops. Chemical
Engineering Progress, 1952. 48(3).
134. Magnussen, B.F. and B.H. Hjertager. On Mathematical Modelling of Turbulent
Combustion with Special Emphasis on Soot Formation and Combustion. in
Proceedings of the 16th Symposium (International) on Combustion 1976. The
Combustion Institute.
135. Spalding, D.B., Mixing and chemical reaction in steady confined turbulent
flames. Symposium (International) on Combustion, 1971. 13(1): p. 649-657.
136. Skinner, F.D. and L.D. Smoot, Heterogeneous Reactions of Char and Carbon,
in Pulverized Coal Combustion and Gasification, L.D. Smoot and D.T. Pratt,
Editors. 1979, Plenum Press: New York. p. 149-167.
137. Popovac, M. and K. Hanjalic, Compound Wall Treatment for RANS
Computation of Complex Turbulent Flows and Heat Transfer. Flow
Turbulence Combust, 2007. 78: p. 177-202.
138. Kader, B.A., Temperature and Concentration Profiles in Fully Turbulent
Boundary Layers. Int. J. Heat. Mass. Trans., 1981. 24: p. 1541-1544.
139. AVL FIRE CFD Species Transport Manual v 2008. Vol. 2008. 2008, Graz,
Austria: AVL.
140. Badzioch, S. and P.G.W. Hawksley, Kinetics of Thermal Decomposition of
Pulverized Coal Particles. Industrial & Engineering Chemistry Process Design
and Development, 1970. 9(4): p. 521-530.
141. F. C. Lockwood , T.M.a.M.A.Y., Simulation of pulverised coal test furnace
performance, Fuel, 77(12) (1998) 1329-1337.
142. Anthony, D.B.H., J. B.; Hottel, H. C.; Meissner, H. P. 15th Symposium
(International) on Combustion; The Combustion Institute: Pittsburgh, PA,
(1975) 1303.
143. Smith IW. In: Proceedings of the 19th Symposium (International) on
Combustion. The Combustion Institute, P., Pennsylvania. (1983) 1045–66.
144. M.A. Field, D.W.G., B.B. Morgan and P.G.W. Hawksley, Combustion of
pulverized coal. Leatherhead: The British coal utilization research
association, (1967) 186–210.
221
145. Baum, M.M. and P.J. Street, Predicting the Combustion Behaviour of Coal
Particles. Combustion Science and Technology, 1971. 3(5): p. 231 - 243.
146. Wall, T.F.P., W. J.; Bartz, S. Doc. F388/a/3; International Flame Research
Foundation: Livorno, Italy, 1976.
147. AVL FIRE CFD User guide Manual v 8.5, 2008.
148. Naser, J.A. and A.D. Gosman, Prediction of Compressible Subsonic Flow
Through an Axisymmetric Exhaust Valve Port Assembly J. Auto. Eng., 1995.
209(4): p. 289-295.
149. Daghigh, R., N.M. Adam, and B. Saharib, The Effect of Air Exchange Rate on
Human Thermal Comfort in an Air-Conditioned Office Under Different
Openning Arrangements. Eu. J. Sci. Resear., 2009. 25(2): p. 174-191.
150. Quarm, T.A.A., Physical Chemistry and Pyrometallurgy. Trans. I. M. M,
1980. 89: p. 139-144.
151. Hong, R., H. Li, H. Li, and Y. Wang, Studies on the inclined jet penetration
length in a gas-solid fluidized bed. Powder Technology, 1997. 92: p. 205-212.
152. Wang, F., Z. Yu, Q. Marashdeh, and L.-S. Fan, Horizontal Gas and Gas/Solid
Jet Penetration in a Gas-Solid Fluidized Bed. Chemical Engineering Science
2010. 65: p. 3394-3408.
153. Li, T., K. Pougatch, M. Salcudean, and D. Grecov, Numerical Simulation of
Horizontal Jet Penetration in a Three Dimensional Fluidized Bed. Powder
Technology, 2008. 184: p. 89-99.
154. Zhu, M.-y., I. Sawada, and M. Iguchi, Physical Characteristics of a
Horizontally Injected Gas Jet and Turbulent Flow in Metallurgical Vessels.
ISIJ Int., 1998. 38: p. 411-420.
155. Oryall, G.N. and J.K. Brimacombe, Physical Behaviour of a Gas Jet Injected
Horijontally into Liquid Metal Metallurgical and Materials Transaction B,
1976. 7 B(3): p. 391-403.
156. Huda, N., J. Naser, G. Brooks, M.A. Reuter, and R. Matusewicz, CFD
Modelling of Zinc Slag Fuming Process in Top Submerged Lance Smelting
Furnace. Metall. Trans. B, 2012. 43(1): p. 39-55.
222