UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Flash converting of chalcocite concentrate : a study of the flame Shook, Andrew A. 1992

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1992_fall_shook_andrew_a.pdf [ 4.97MB ]
JSON: 831-1.0078436.json
JSON-LD: 831-1.0078436-ld.json
RDF/XML (Pretty): 831-1.0078436-rdf.xml
RDF/JSON: 831-1.0078436-rdf.json
Turtle: 831-1.0078436-turtle.txt
N-Triples: 831-1.0078436-rdf-ntriples.txt
Original Record: 831-1.0078436-source.json
Full Text

Full Text

FLASH CONVERTING OF CHALCOCITE CONCENTRATE:A STUDY OF THE FLAME- ByAndrew A. ShookB.E., The University of Saskatchewan, 1983M.A.Sc., The University of British Columbia, 1987A THESIS SUBMITfED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Department of Metals and Materials Engineering)We accept this thesis as conformingto the required standardTHE COLUMBIAMay, 1992© Andrew A. Shook, 1992Signature(s) removed to protect privacyIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives, It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)____________________Department of_____________________The University of British ColumbiaVancouver, CanadaDate iDE-6 (2/88)Signature(s) removed to protect privacy11AbstractThe “flame” resulting from the flash converting of Inco chalcocite (or “MK”) copper concentratehas been studied, with the primary objective of reducing or eliminating the dust production associated with this process. Experimental trials with MK concentrate were performed in the UBCflash smelting pilot plant in which the response of the flash flame (as determined by solids andgas compositions) was examined as a function of process variables. In addition, solids and gassamples were recovered from the flash flame in the Inco Port Colbome pilot plant. These experiments indicated that the rate of dust production in the UBC pilot plant was significantly lowerthan that in the Inco Port Colbome pilot plant, and that the oxygen-to-concentrate ratio plays animportant role in the observed dust generation rate. A mathematical model of an individual MKconcentrate particle has shown that dust production is consistent with particle fragmentationinitiated by the boiling of copper within the particle. Particle size, oxygen concentration andtemperature were found to affect the ability of an MK concentrate particle to fragment by thismechanism.Incorporating this particle combustion study, a mathematical model of the chalcocite flash flamewas developed. This model utilized the boundary-layer form of the gas-phase conservation equations (modified to deal with recirculation and incorporating the effects of particles on the gas) todescribe the combustion phenomena occurring within the chalcocite flash converting flame. Themathematical model was verified first by non-reacting studies from the literature and then bycomparison with experimental data taken from the UBC flash furnace. The mathematical modelwas then extended to describe the behaviour of the Port Colborne pilot plant.The mathematical model has shown that the production of dust within the chalcocite flash flameoccurs due to the dissimilar spreading rates of the concentrate particles and the injected oxygen.111Two regions within the flame were identified as leading to dust generation : a ‘near-edge” regionof high local oxygen-to-concentrate ratio, and a ‘near-centre” region with high radial oxygentransport. The “near-edge” region has been suggested as being responsible for the dust producedby the UBC pilot plant, while the “near-centre” region is responsible for the dust produced in thePort Colborne pilot plant. The occurrence of these two regions could be eliminated by improvedburner design.ivTable of ContentsAbstract.iiList of Tables viiiList of Figures ixList of Symbols xvAcknowledgements xix1 Introduction 11.1 Flash Smelting 11.2 Inco Copper Cliff Operations: Current and Future 31.3 MK (Chalcocite) Flash Converting 42 Literature Review 102.1 Chalcocite Flash Converting 102.1.1 Theoretical and Laboratory Studies 102.1.2 Pilot Plant and Industrial Studies 112.1.3 Summary 122.2 Flash Smelting- Other Copper Suiphide Concentrates 122.2.1 Experimental Work 132.2.2 Mathematical Models 162.2.3 Summary 203 Objectives and Scope 223.1 Overall Project Ojectives 223.2 Experimental Objectives 233.3 Mathematical Modelling Objectives 244 Experimental 254.1 The UBC Pilot Plant 254.1.1 Plant Description Flash Reaction Shaft Gas Handling System Oxygen and Concentrate Feed Systems 304.1.1.4 Data Acquisition and Control Systems 314.1.2 Sampling and Analysis 354.1.2.1 Gas Sampling Equipment 354.1.2.2 Gas Analysis 364.1.2.3 Solids Sampling 374.1.2.4 Solids Analysis 384.1.3 Operating Procedure 384.1.4 Run Conditions 404.1.5 Run Data 424.1.6 Preliminary Data Analysis 434.2 Inco Port Colborne Pilot Plant 504.2.1 Plant Description 514.2.1.1 Flash Reactor 514.2.1.2 Concentrate Burner 514.2.2 Experimental Apparatus 53V4.2.3 Operating Procedure 544.2.4 Experimental Data 554.2.5 Preliminary Analysis 565 Chalcocite Reaction Kinetics 855.1 Literature Data 855.1.1 Thermodynamics 855.1.2 Chalcocite Oxidation Kinetics 865.1.3 Analysis 905.2 Kinetic Model Development 925.2.1 Simplifying Assumptions 935.2.2 Chemical Reactions Considered 945.2.3 Calculation of Reaction Rates 955.2.3.1 Reaction 5.1 : Low temperature oxidation to Cu20 955.2.3.2 Reaction 5.3 : High temperature oxidation to metallic copper 985.2.3.3 High Temperature Reaction of Copper with Oxygen 1005.2.4 Heat Transfer 1005.2.5 Heat Effects 1025.2.6 Copper Vaporization 1035.2.7 Solution Algorithm 1065.3 Kinetic Model Calculations 1075.3.1 Verification of Activation Energy and Pre-Exponential Constant 1075.3.2 Particle Temperature Predictions 1085.3.2.1 Predictions in Oxygen and Dusting Mechanism 1085.3.2.2 Predictions in 50 % 02 and Copper Vaporization Effects 1105.3.2.3 Predictions in Air and Particle Size Effects 1115.3.2.4 Discussion 1125.3.3 Particle Fragmentation Mechanism 1145.3.4 Prediction of Dusting Diagram for MK Concentrate 1166 Mathematical Model of the Chalcocite Flash Flame 1386.1 Model Objectives 1386.2 Model Description 1386.2.1 Simplifying Assumptions 1386.2.2 Gas and Particle Flow 1406.2.2.1 Gas Conservation Equations 1406.2.2.2 Turbulence Modelling 1436.2.2.3 Particle-Gas Interaction 1486.2.2.4 Particle Dispersion 1496. Particle Concentration Distribution 1506. Particle Composition Distribution 1536.2.3 Heat Transfer 1546.2.3.1 Gas-Particle Heat Transfer 1546.2.3.2 Wall-Particle Radiative Heat Transfer 1556.2.4 Chemical Reaction 1586.3 Model Development 1586.3.1 Discretization of Equations 1596.3.2 Solution Algorithm 1616.3.3 Flow Reversal Region 1626.4 Non-Reacting Model Verification 1646.4.1 Isothermal Free One-phase Jet - Axial and Radial Velocities 1656.4.2 Heated Free Single-phase Jet 1676.4.3 Two-phase Isothermal Turbulent Jet 1686.4.4 Isothermal Single-phase Confined Jet 1696.4.5 Isothermal One-phase Free Jet - Dissimilar Gas Injection 170vi6.4.6 Summary 1717 Mathematical Model Validation- Comparison with UBC Pilot Plant Data 1857.1 Determination of Furnace Wall and Recirculating Gas Temperatures 1867.2 Sensitivity Analysis 1887.2.1 Effect of Wall and Gas Temperatures 1897.2.2 Turbulence Modelling Parameters 1907.2.3 Mesh Size 1917.2.4 Summary 1927.3 Comparison With UBC Pilot Plant Data 1927.3.1 Base Case Trials 1957.3.1.1 Predicted Gas and Particle Temperatures 1957.3.1.2 Particle Reaction Degree 1977.3.1.3 Gas Composition: Comparison to Measured Values 1987.3.1.4 Solids Distribution : Comparison to Measurements 2007.3.1.5 Solids Samples Composition: Comparison to Measurements 2017.3.1.6 Dust Generation Rate 2057.3.2 Model Comparison with UBC Furnace : Non-Standard Conditions 2057.3.2.1 Smaller Particle Size 2057.3.2.2 Large Particle Size 2067.3.2.3 Excess Oxygen Trial 2077.4 Summary of Runs Simulating the UBC Flash Furnace 2098 Industrial Calculations : Inco Port Colborne Flash Reactor 2258.1 Experimental Data from the Port Colborne Flash Reactor 2268.2 Factors Complicating Comparison of the Model with Experiment 2268.3 Modifications to the Model to Simulate the Port Colborne Smelter 2308.4 Sensitivity Analysis 2318.4.1 Assumed Furnace Gas and Wall Temperatures 2328.4.2 Particle Blockage Effects 2338.5 Comparison Between Model and the Experimental Data 2338.5.1 Initial and Boundary Conditions 2338.5.2 Comparison Between Predicted and Measured Gas Composition 2348.5.3 Comparison Between Predicted and Measured Solids Composition 2348.5.4 Dust Generation 2358.6 Summary 2359 Discussion : Analysis of the MK Flash Flame 2429.1 Critical Factors Affecting the Combustion of MK Concentrate 2429.1.1 Burner Design - The Structure of the Two-Phase Jet 2429.1.1.1 Effect of Jet Structure on Solids Composition 2449.1.1.2 Effect of the Jet Structure on Dust Generation 2459.1.1.3 Dust Generation Mechanism - Port Colborne Flash Furnace 2479.1.1.4 Effect of Near-Centre Region in the UBC Flash Furnace 2489.1.2 Heat Transfer 2489.1.2.1 Location of the Primary Combustion Zone 2489. The UBC Flash Furnace 2499. The Port Colborne Flash Furnace 2509.1.2.2 Role of Recirculating Gas in Dust Generation - UBC Furnace 2509.1.3 Particle Size Effects 2519.1.4 Summary 2539.2 Dust Abatement: Improvements in Burner Design 25410 Conclusions and Recommendations 26311 References 265Appendix 1: Calculated efficiency of non-isokinetic solids sampler 273Appendix 2 - Derivation of Particle Blockage Effects 275A2.1 Volume Fraction Solids in UBC Flash Furnace 275A2.2 Relationship Between Volume Fraction and Area Fraction 275A2.3 Particle Blockage Effects - Normally Incident Radiation 276A2.4 Calculation of Mean Wall-Particle View Factor 277A2.5 Calculation of Back-Radiation Effects 278viiviiiList of TablesTable Page4.1 Comparison of Experimental Pilot Plants with Industrial Operations 264.2 Data Acquisition and Control Tasks Performed by the Main SystemComputer 344.3 Full-Scale Runs Performed in the UBC Pilot Plant 414.4 Chemical Analysis of Sized and As-Received Concentrate 414.5 Assays of Solids Samples from Run 6 424.6 Effect of Oxygen Concentration on Dust Formation 444.7 Chemical Analysis of Dust Produced at UBC and Port Colbome Pilot PlantFacilities 454.8 Comparison of Copper/Nickel and Copper/Iron Mass Ratios of Solids Input toand Recovered from the UBC Flash Furnace 464.9 Gas Analyses from three runs in the UBC Flash Furnace 484.10 Mass balances from all runs in the UBC Flash Furnace 494.11 Operating Conditions and Analyses of Gas Samples from the Port ColbomeFlash Reactor 604.12 Normalized Assays of Solids Samples from the Port Colborne Flash Reactor.... 615.1 Heat Capacities of Species as used by ‘kmodel’ 1025.2 Heats of Reaction 1036.1 Fitting Parameters Used in the Mathematical Model 1647.1 Experimental Run Conditions and Data Collected from the UBC Pilot Plant 1947.2 Predicted and Measured Hearth Composition from Run with Large ParticleSize 2078.1 Operating Conditions and Analyses of Gas Samples from the Port ColbomeFlash Reactor 2378.2 Normalized Assays of Solids Samples from the Port Colbome Flash Reactor.... 238ixList of FiguresFigure Page1.1 Schematic Diagram of Outokumpu Flash Reaction Shaft 61.2 Schematic Diagram of Inco Flash Furnace 71.3 Inco Copper Cliff Operations- Current Flowsheet. From Landolt et al. (10) 81.4 Inco Copper Cliff Operations- Bulk Concentrate Flowsheet. From Landolt eta!. (10) 94.1 Photograph of the UBC pilot plant facility 624.2 Schematic diagram of the UBC flash furnace shaft 634.3 Schematic diagram of the UBC pilot plant off-gas handling system 644.4 Schematic diagram of the UBC concentrate burner 654.5 Schematic diagram of gas sampling system as used in the UBC experimentaltrials 664.6 Schematic diagram of gas suction thermocouple 674.7 Schematic diagram of Solids Sampler Type 1 684.8 Schematic diagram of Solids Sampler Type 2 694.9 Predicted sampling efficiency of non-isokinetic solids samplers as a functionof particle size and velocity 704.10 Particle size analysis of regular MK concentrate 714.11 Particle size analysis of large size fraction of MK concentrate (Prepared byOrtech, Ltd.) 724.12 Particle size analysis of small size fraction of MK concentrate (Prepared byOrtech Ltd.) 734.13 Oxygen and concentrate feed rates from run of Dec. 13, 1989 (2 kg/mmunsized MK concentrate, stoichiometric oxygen ) 744.14 Reactor wall temperatures as a function of elapsed time from the experimentaltrial of Dec. 13, 1989 (2 kg/mm unsized MK concentrate, stoichiometricoxygen ), corrected for the effects of the alumina wool blanket 75x4.15 Gas chromatograph output from run of Dec. 13, 1989 (2 kg/mm unsized MKconcentrate, stoichiometric oxygen ) 764.16 Photomicrograph of dust produced at UBC pilot plant from run of Dec. 13,1989 ( 2 kg/mm unsized MK concentrate, stoichiometric oxygen) 774.17 Sketch of solids distribution from run of 27/4/89 784.18 Schematic Diagram of Inco Port Colborne Flash Reactor 794.19 Schematic diagram of Inco Port Colborne concentrate burner 804.20 Schematic Diagram of Gas Sampling Apparatus as employed in the PortColborne Flash Reactor 814.21 Approximate location of samples in Port Colborne flash furnace 824.22 Plot of all gas samples taken from trials in the Port Colborne reactor with the6.3 cm burner 834.23 Plot of all gas samples taken from trials in the Port Colborne reactor with the5.3 cm flash burner 845.1 Cu- S Binary Phase Diagram From (54) 1175.2 Cu- SO2- 02 Predominance Area Diagram. T = 1000 K. (55) 1185.3 Cu-SO2- 02 Predominance Area Diagram. T = 1500 K. (55) 1195.4 Cu-SO2- 02 Predominance Area Diagram. T = 2200 K. (55) 1205.5 MK mass loss data from Otero, Brimacombe and Richards (13), foras-received concentrate reacting in a stagnant gas furnace at varying temperature and oxygen concentration 1215.6 a. Still photograph of MK concentrate reacting in 100 % oxygen. b. Photograph of MK concentrate reacting in 50 % oxygen.From Otero et al. (13) 1225.7 High-speed photographs taken at 1 millisecond intervals showing an explodingMK concentrate particle. (From Otero, Brimacombe and Richards (13)) 1235.8 Data from Otero et al. (62):a. Particle temperature of combusting MK particleb. Energy emitted by combusting MK particle (as indicated by pyrometeroutput signal) 1245.9 Data from Otero et al. (62):a. Particle temperature of combusting MK Particleb. Energy emitted by combusting MK particle (as indicated by pyrometeroutput signal) 125xi5.10 Cu-O Phase Diagram From Schmid (69) 1265.11 Comparison between relative magnitudes of radiative (wall-particle) andconductive (gas-particle) heat transport mechanisms as a function of particlesize. Tgas = Twall = 1400 K 1275.12 Comparison between relative magnitudes of radiative (wall-particle) andconductive (gas-particle) heat transport mechanisms as a function of particlesize. Tgas = Twall = 1800 K 1285.13 Flowchart of computer program ‘kmodel’ 1295.14 Comparison between the mass loss measurements of Otero et al in air andpredictions of the kinetic model. (X = volume fraction copper) 1305.15 Comparison between the mass loss measurements of Otero et al. in oxygenand the predictions of the kinetic model. (X = volume fraction copper) 1315.16 Particle temperature predictions for a 20 micron particle combusting in oxygenat 1213 K, showing the effect of the surface covering parameter X, 1325.17 The effect of the parameter X, on particle temperature predictions forcombustion in a 50 % oxygen - 50 % sulphur dioxide atmosphere at 1213 K 1335.18 The effect of particle size Ofl particle temperature predictions for combustionin air at 1213 K (assuming that = 0) 1345.19 Predicted peak temperatures of particles reacting in oxygen, air and 50 %oxygen as a function of particle size 1355.20 Comparison between model predictions and mass loss data of Otero et al. (13)assuming that Xc = 0 1365.21 Predicted dusting diagram for MK concentrate combustion, as a function ofparticle size, temperature and reactor composition. (X = 0) 1376.1 Sketch of flow problem to be solved by the mathematical model 1726.2 Flowchart of the PSI-Cell algorithm (42) 1736.3 Schematic diagram of regions of radiative heat transfer in the UBC flashsmelting furnace 1746.4 Flowchart of the computer program ‘jini.f’ 1756.5 Comparison between the change of jet centreline velocity as a function of axialdistance predicted by the mathematical model with the measurements made forthis work, and the correlation of Tuve (99).I Constant temperature air jet intostagnant surroundings] 176xl’6.6 Comparison between the axial jet velocity as a function of axial and radialposition predicted by the model and the correlation of Tuve (99). [Constanttemperature air jet into stagnant surroundings] 1776.7 Comparison between jet radial velocity as a function of axial and radialposition and the equation derived from the Tuve (99) correlation and conservation of mass. [Constant temperature air jet into stagnant surroundings] 1786.8 Comparison between predicted jet centreline temperature and the experimentaldata of Wilson and Danckwerts (100). [Heated air jet into stagnant surroundings] 1796.9 Comparison between model predictions of centreline particle and gas velocityas a function of axial distance and the experimental data of Modarress et a!.(101) [Two phase air-particle jet I 1806.10 Comparison between centreline jet velocity as a function of axial distance aspredicted by the mathematical model and the predictions of FIDAP (FDI Inc.).[Single phase confined jet with recirculation] 1816.11 Comparison between axial jet velocity as a function of radial position aspredicted by the mathematical model and the predictions of FIDAP (FDI Inc.)for two axial positions (0.5 m and 1.0 m). [Single phase confined jet withrecirculationi 1826.12 Comparison between axial variation of centreline jet velocity predicted bymodel with the data of Keagy and Weller (102)[CO2jet into stagnant air ] 1836.13 Comparison between axial variation of centreline jet composition predicted bymodel with the data of Keagy and Weller (102) [ CO2jet into stagnant air] 1847.1 Radiation circuit diagram for the UBC flash reactor, illustrating the interactionbetween the energy emitted by the particles, the walls and the gas 2107.2 Calculated gas temperatures in the UBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen ] .... 2117.3 Calculated particle temperatures in the UBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen ] .... 2127.4 Regions in which the calculated particle temperatures in the UBC flash reaction shaft exceed the recirculating gas temperature (1200 K) for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen ] .... 2137.5 Calculated particle distribution in the UBC flash reaction shaft for the basecase operating conditions [2kg/mm unsized concentrate, stoichiometricoxygen ] 2147.6 Calculated oxygen concentration in the UBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen I .... 215xl”7.7 Calculated local oxygen-to-concentrate ratio in the UBC flash furnace for thebase case operating conditions [2kg/mm unsized concentrate, stoichiometricoxygen] 2167.8 Calculated solids composition in the UBC flash reaction shaft for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen ] .... 2177.9 Comparison between predicted and measured axial variation of gas composition in the UBC flash reaction shaft. Predictions are shown for stoichiomethcoxygen at radial positions of 0 (centre), and 4.3 cm, and for the centreline at 20% excess oxygen. Experimental data shown for original reaction shaft (stoichiometric oxygen) and new shaft (20 % excess oxygen) 2187.10 Comparison between predicted and measured solids distribution in the UBCflash furnace for the base case operating conditions [2kg/mm unsized concentrate, stoichiometric oxygen I 2197.11 Comparison between calculated and measured centreline composition of solidssamples for the base case operating conditions [2kg/mm unsized concentrate,stoichiometric oxygen] 2207.12 Comparison between calculated and measured radial variation of solidssamples recovered from reactor section C for the base case operating conditions [2kg/mm unsized concentrate, stoichiometric oxygen] a. Copper b.Sulphur c. Oxygen 2217.13 Comparison between calculated and measured radial variation of solidssamples recovered from reactor section D for the base case operating conditions [2kg/mm unsized concentrate, stoichiometric oxygen I a. Copper b.Sulphur c. Oxygen 2227.14 Measured values of flash furnace wall temperature from run with smallparticle size, corrected for the effects of the insulating blanket 2237.15 Measured values of flash furnace wall temperature from run with large particlesize corrected for the effects of the insulating blanket 2248.1 Predicted Gas Temperatures in Port Colborne Flash Reactor (Top View) [TgTw= 1600K,5.3cmburner] 2398.2 Comparison between predicted gas composition and experimental data fromPort Colborne Flash Reactor for 5.3 cm burner. [Tg = Tw = 1600 K I 2408.3 Comparison between predicted solids sample composition and experimentaldata from Port Colbome Flash Reactor for 5.3 cm burner. [Tg = Tw = 1600 K]2419.1 Schematic diagram of two-phase jet issuing from UBC and Port Colborneconcentrate burners 256xiv9.2 Normalized calculated velocity, concentration and particle density fieldswithin the UBC flash flame at 0.12 m from the burner tip [Base CaseConditions] 2579.3 Calculated radial variation of oxygen concentration within the UBC flashflame at two axial locations [ Base Case Conditions ] 2589.4 Calculated radial velocity and oxygen concentration within the Port Colborneflash flame as a function of radial position [at 0.50 m from the burner tip,prior to combustion] 2599.5 Ratio of heat transferred by convection to that transferred by radiation in theUBC flash furnace [ Base Case Conditions] 2609.6 Ratio of heat transferred by convection to that transferred by radiation in thePort Colborne Flash Reactor 2619.7 Calculated recirculating gas temperature necessary to produce dusting in theUBC flash furnace as a function of diluted gas composition 262A 1.1 Capture Efficiency of an Impingement Sampler 273A2. 1 Figure A2. 1: Diagram of Volume Fraction of Sphere, Radius r 276A2.2 Figure A2.2: Schematic Diagram Illustrating Particle Blockage Effect 276A2.3 Diagram illustrating the effect of radiation oriented at an angle 0 to the normalof the jet axis 279xvList of SymbolsSymbolA Pre-exponential constant (moles/(m2Pa s))Particle surface area (m2)B Interaction parameterC Concentration (mol/m3)Heat capacity (J/(kg K))Cd Drag coefficientC Constant in k — £ equationD DiffusivityDb Width of solids sampler for efficiency calculation (m)Effective (turbulent) diffusivity (m2/s)DOXb Bulk oxygen diffusivity (m2Is)D00 Diffusivity of oxygen through Cu20Particle diameter (m)Dt Turbulent particle diffusivity (rn2/s)D Width of sampler over which particles are collected (m)F Constant in turbulence model = 0.015F Total drag force on a particle (N)Wall-particle view factorABa Activation energy (J)h Convective heat transfer coefficient (W/(m2K))I Collision integrali Axial finite difference locationxvij Radial finite difference locationK Constant in Tuve (98) equation = 6.2k Production of turbulence energyk0 Effective thermal conductivity (W/(m K))kcUb Copper vapour mass transfer coefficient in bulk phase (mol/m2s)kOXb Oxygen mass transfer coefficient in bulk phase (mol/m2s)kr Reaction rate constant (mol/(m2s))K0,K12 Constants in radial velocity correlationim Mixing length (rn)Particle mass (kg)th Solids input rate (kgls)Molecular weights (g/mol) of Cu, Cu2S, Cu20M12 Mean molecular weight (glmol)N Particle number density (rn3)N Separation numberNu Nusselt Numbern Number of molesP Pressure (Pa)atm Atmospheric pressure (Pa)Vapour pressure of copper (Pa)P Partial pressure of copper at particle surfaceP02 Oxygen partial pressurePr Prandtl NumberQ0 Convective heat transfer rate (W)Qg Injected gas flowrate (m3/s)xviiQrad Radiative heat transfer rate (W)Qreac Rate of heat generation by reaction (W)R Gas constant ( 8.3 14 J mol’ K’)r Radial ordinate (m)r1 Velocity half width (rn)rCU2O Radial location of Cu20 (m)Rate of oxygen mass transport (moles! m2 s)r Particle radius (rn)Rrc Reaction rate (moles! m2 s)Rvap Vaporization rate (moles I rn2 s)Re Reynolds numberSc Schmidt numberSh Sherwood numberSm Particle momentum source/sink (N/rn3)SPg Particle source/sink of gas (kg / rn s)Se Particle heat source/sink (W/m3)S’, Particle sink of oxygen (moles! m3 s)T Temperature (K)u Axial velocity (mis)Umax Maximum stream velocity (m/s)U0 Axial velocity of gas at burner nozzle (mIs)u1 Terminal velocity of a particle (mis)v Radial velocity (mis)Velocity vector (mis)Particle volume (m3)xviiiFraction of particle surface covered by copperFraction of particle surface covered by suiphidez Axial ordinate (m)EmissivityIntermittency functionYCu Surface tension of copper (N/rn)Sampling efficiencyViscosity (kglrns)p Density (kg/rn3)Sefan-Boltzmann Constantt Turbulent shear stressParticle relaxation time (s)Subscripts and Superscriptsc at centrelinee effectiveg gasp particlet turbulentw wall* at particle surfacexixAcknowledgementsI would like to thank my advisors, Dr. G.G. Richards and Dr. J.K. Brimacombe for their guidance and encouragement. I would also like to thank the staff of Inco Research and in particularDr. Carlos Diaz for his interest and support.The efforts of Mr. Pat Wenman in the design, construction and operation of the UBC pilot plantare gratefully acknowledged.Mr. Justin Raskauskas of Inco Ltd. assisted in the gas and solids sampling in the Port Colbornepilot plant.Finally I would like to thank my wife, Leanne Shook, for preparing many of the figures, and forher tremendous patience and support.Financial support for this work was provided by Inco Ltd., the Natural Sciences and EngineeringResearch Council, and CANM ET.11 Introduction1.1 Flash SmeltingIn the forty years since its introduction, flash smelting has become the dominant industrialmethod for producing copper from suiphide ores (1). Originally developed by Inco Ltd. inCanada and Outokumpu Oy in Finland, flash smelting is a technique whereby finely-dividedsulphide concentrates (typically chalcopyrite, CuFeS2)are rapidly oxidized at high temperature with air, oxygen or oxygen-enriched air. The products of copper flash smelting areusually either a Cu2S-FeS matte (Inco and Outokumpu processes), or a mixture of high-gradematte and blister copper (Outokumpu process) and an iron-silicate slag.Schematic diagrams of an Outokumpu reaction shaft and an Inco flash smelting furnace (Figures 1.1 and 1.2) illustrate the basic design differences between the two processes. TheOutokumpu furnace injects concentrate and oxygen-enriched air downward into a cylindricalflash reaction shaft while in the Inco furnace, concentrate and technically-pure (95 %)oxygen are injected horizontally from both ends of a large open vessel.Flash smelting offers notable advantages over other smelting processes1. It is highly energy efficient the oxidation of the sulphur (and iron if present) generatesconsiderable heat, which facilitates the smelting reactions. Depending upon the oxygenconcentration in the input gas, and factors such as the desired matte grade and concentrate composition, the process can be operated autogenously, with no need for additional heat input from fossil fuels.22. The SO2 concentration in the off-gas is very high (particularly when using oxygen),making liquifaction or scrubbing of the gas stream much less expensive. This is ofparticular importance with regard to the current sulphur dioxide emission regulations.3. Most of the important reactions are carried out in a single process vessel, eliminatingthe need for roasters and reverberatory furnaces, and (in some cases) copper converters.Despite these undoubted advantages, the flash smelting of copper concentrates has someinherent difficulties. For example, it has been reported (2,3) that the flash smelting of copperconcentrates in an Outokumpu furnace generates considerable quantities of dust, whichrequires additional handling equipment and can cause operational problems. However, Incodoes not seem to experience severe dust problems when smelting their conventional copperconcentrate (4).Furthermore, although the flash smelting process itself is simple and efficient, treating theproducts of the flash furnace (matte and slag) can cause additional difficulties. For example,with the Outokumpu process, sufficient oxygen is usually added to produce a matte that ishighly concentrated in copper (60-70 % vs. 40-50 % for Inco). This high matte grade has theadvantage of reducing the number of converters required and increasing overall efficiency,but unfortunately also causes much copper to report to the slag. This in turn necessitates anadditional slag-cleaning step in order to prevent excessive copper loss (5,6).With the Inco matte-smelting process, the flash furnace slags are low in copper and can bediscarded without additional treatment. However, converting the matte in conventionalPeirce-Smith copper converters produces an off-gas which is difficult and expensive to treat(large volumes of gas of fluctuating sulphur dioxide content), thus partially negating one of3the major benefits of flash smelting. To circumvent this difficulty, Outokumpu and Kennecott have introduced the “flash converting” process (6), whereby mattes produced by a flashfurnace are cooled, granulated and then further oxidized to blister copper using tonnageoxygen. Worldwide however, the Peirce-Smith converter still remains by far the mostcommon means of converting copper mattes (9).1.2 Inco Copper Cliff Operations: Current and FutureA schematic diagram of the current process flowsheet of the Inco operations at Copper Cliff,Ontario is shown in Figure 1.3. It consists of two partially separated processes one forcopper, the other for nickel. The two metals are separated initially by flotation, producing acopper concentrate and a nickel concentrate. However, the unavoidable contamination of thecopper concentrate with nickel and the nickel concentrate with copper complicates the flowsheet (10). As described above, the copper concentrate is flash smelted to a matte, while thenickel concentrate is smelted conventionally using multiple-hearth roasters, reverberatoryfurnaces and Peirce-Smith converters. The product of the nickel smelting is called Bessemermatte, and is approximately 50 % nickel and 25 % copper. Slow cooling and separation ofthe resulting phases produces a number of products, among which are nickel sulphide andcopper (I) suiphide (Cu2S). The nickel suiphide is treated further, while the wet (12 % moisture) copper concentrate (mainly chalcocite, and termed ‘MK” concentrate) is flash convertedto form “semi-blister” copper, i.e. nickel contaminated sulphur saturated copper. Thissemi-blister is then transferred to Peirce-Smith copper matte converters. (Note the origin ofthe term “MK” to describe the chalcocite concentrate is obscure). The process used at IncoCopper Cliff has been proven to be extremely reliable and efficient. Few problems with this4process have been reported; copper and nickel losses to slags are slight, and the smelter hascontinued to operate virtually unchanged (apart from expansions and minor improvements)for decades.However, the desire to minimize sulphur dioxide emissions has caused Inco drasticallyre-design its flowsheet drastically (Figure 1.4). In the new process, the roasting and reverberatory smelting operations (which are responsible for most of the SO2 emissions) will beentirely eliminated. Instead, the bu’k copper-nickel concentrate will not be separated intocopper and nickel concentrates, but will be flash smelted and converted to a Bessemer matte.Inco have done considerable work on the flash smelting of Cu-Ni concentrates on a pilot andproduction scale (10,11), and have found no significant operational problems. As before, theBessemer matte will be slow-cooled and separated to form nickel-copper metallics, nickelsulphide, and a chalcocite (approximately 75 % Cu, 3-4 % Ni, 21 % S, 0.3 % Fe) “MK”concentrate. The nickel sulphide will be treated conventionally, but the chalcocite will not besent to Pence-Smith converters. Instead, it will be flash converted with tonnage oxygen tosemi-blister copper, containing approximately I % sulphur. The semi-blister will be finishedto blister copper, i.e. copper containing less than 0.7 % nickel and 100 ppm sulphur, in atop-blown vessel developed by Inco.1.3 MK (Chalcocite) Flash ConvertingThe flash converting of this dry MK concentrate has been studied on a pilot scale for severalyears at Inco’s Port Colborne research facility. However, studies at this pilot plant haveshown that the MK concentrate, when flash smelled, does not behave like a conventionalchalcopyrite concentrate. In particular, oxidation of the MK concentrate produces approximately twice as much dust (copper, copper oxides and suiphates), which is both difficult and5expensive to handle and recycle. It is clear that the production-scale MK flash converter willrequire a great deal of dust handling equipment, if it performs similarly to the Port Colbornefacility. If possible, a means of reducing the dust generated by flash converting the MKconcentrate would clearly be of considerable importance. At the same time, there is also astrong need for basic information on the combustion of this chalcocite concentrate, whichcould lead to further improvements in the design and operation of this converter.6ConcentrateburnersTopping holeFigure 1.1 :Schematic Diagram of Outokumpu Flash Reaction Shaft7CHALCOPYRITECONCENTRA and SANDFigure 1.2 :Schematic Diagram of Inco Flash Furnace8Ni concentrate Sulfur dioxide cu concentrateTo LiquidplantBlistercoppertoCopperRefineryMarketandrefineriesFigure 1.3: Inco Copper Cliff Operations- Current Flowsheet. From Landolt et al. (10)9Sulk concentrate Sulfur dioxideMarketBlistercoppertoCopperRefineryMarketandrefineriesFigure 1.4: Inco Copper Cliff Operations - Bulk Concentrate Flowsheet. From Landolt etal. (10)102 Literature Review2.1 Chalcocite Flash ConvertingDue to the fact that, up to the present time, very few commercial pyrometallurgical processesproduce copper directly from chalcocite concentrates, the flash convening of chalcocite hasnot been extensively studied, and very little experimental data is available on chalcocite flashreactions. Several studies of the kinetics of the oxidation of chalcocite pellets and lumps doexist, and these have been reviewed in detail in Chapter Theoretical and Laboratory StudiesAmong the first to study the combustion of finely divided chalcocite on a small scalewere investigators at Inco Ltd. (12). Utilizing a laboratory scale flash furnace, they established a direct link between the dust generation rate and the oxygen-to-concentrate ratio.Particle size was also observed to be an important parameter affecting dusting, with fineparticles observed to dust more readily than coarse particles.Otero, Brimacombe and Richards (13) compared the combustion behaviour of the IncoMK concentrate with conventional chalcopyrite concentrate. High-speed and still photographs of chalcopyrite and MK concentrate combustion in a stagnant gas reactor showeda considerable difference between the behaviour of the two concentrates. Thechalcopyrite particles were seen to explode into a relatively small number of particleswhile the MK concentrate particles seemed to explode into an extremely fine dust. It washypothesized that the low iron content of the MK concentrate was responsible for itsgreater tendency to fragment into dust. With chalcopyrite, the oxidized iron may act as a‘glue”, binding the combusting particle together, and reducing the number of fragments11formed if a particle does explode. Lacking this iron oxide “glue, the MK concentrateparticles simply fragment into a fine dust of copper and copper oxide. This phenomenonmay, therefore, provide a possible explanation for the high rates of dust generationobserved when flash converting chalcocite concentrates. The experiments of Otero et al.will be discussed in more detail in Chapter 5.Warczok et al. (14) carried out an experimental and numerical study of the combustion ofboth chalcocite and the concentrate treated by the “Glogow” smelter. Predictions of asimple model of a single combusting chalcocite particle compared well with experimentaldata. Particle size was seen to have a strong effect on the degree of particle reaction, withvery large particles found to remain almost completely unreacted. This information wasapplied to suggest modifications to the concentrate feed to the Glogow smelter whichresulted in performance improvements.2.1.2 Pilot Plant and Industrial StudiesThe only pilot plant work exclusively devoted to chalcocite flash converting has beenperformed by Inco. As discussed in the previous chapter, studies of the flash reaction ofchiacocite concentrate have been carried out by Inco over a number of years in their PortColborne pilot plant. In this facility, the chalcocite concentrate was found to generate dustmuch more readily than the conventional chalcopyrite concentrate.Two full scale flash smelters currently in operation also treat low-iron concentrates. TheOlympic Dam smelter in South Australia treats a copper-uranium-gold sulphide ore,which is smelted directly to blister copper (15,16). A conventional Outokumpu flashsmelter is used, but no information on burner design or dust generation rates has been12presented. More information is available on the Glogow II smelter in Poland (14,17). AnOutokumpu flash smelter is operated to treat a low-iron (2- 6 %) chalcocite concentratecontaining a significant amount of coal (5- 7 %) and gangue materials (17 - 23 % Si02,and approximately 15 % of CaO , MgO and A1203). The concentrate does not need to befluxed and is simply dried and injected into the flash furnace with oxygen-enrichedpreheated air. The concentrate is smelted directly to blister copper, and the high-copperslag (7-16 % Cu) is reduced in an electric furnace with coke. The dust generation rate ofthis flash furnace appears to be very high : the recycled dust is approximately 12 % of thetotal concentrate feed rate. The authors cite some improvement in blister quality broughtabout by improvements in concentrate burner design and the concentrate feed system.2.1.3 SummaryIt is clear from this review that the field of chalcocite flash converting is not at all wellresearched. A number of valuable studies have been carried Out, but relevant industhal,pilot plant, and laboratory data are scarce. Apart from the study of single particle combustion by Warczok et al. (14) no modelling studies of chalcocite flash smelting have beenpublished. A great deal of work remains to be done in a number of areas, including theCu2S-0 reaction kinetics, the mechanism of dust formation, and the behaviour of chalcocite in a flash furnace flame.2.2 Flash Smelting- Other Copper Suiphide ConcentratesDue to the lack of information on chalcocite flash smelting, the flash reaction of othersulphide materials has been examined. This information has been analyzed to provide insightand guidance for the study of chalcocite flash smelting.132.2.1 Experimental WorkIn contrast with the research on chalcocite concentrates, there has been considerableexperimental work done on the flash smelting of other suiphide concentrates, particularlychalcopyrite and bornite, but also nickel and lead concentrates.Jorgensen (18-21) was the first to study the ignition of individual particles of varioussulphide concentrates. He employed a laminar flow furnace to observe the combustion ofconcentrate particles under controlled conditions. A two-colour pyrometer measured thetemperature of the reacting particles, and mass loss measurements were made to determine particle ignition temperatures. A few of Jorgensen’s findings with chalcopyriteconcentrates are summarized below:1. Volatilization of copper and copper oxide limits the maximum temperature attainedby reacting chalcopyrite particles to about 2000 °C.2. Particles frequently fragment or form hollow spheres (cenospheres) when combusting.3. The time taken to heat a particle to ignition is strongly dependent upon particle size.In the geometry under study, Jorgensen found that heat transport by conductionbetween the particle and the surrounding gas dominates over radiative heat transferbetween the particle and the reactor walls.4. There was no significant reaction rate increase with oxygen concentrations above60%.14Dunn and Smith (22) carried out flow visualization experiments in a physical model of anOutokumpu reaction shaft and examined a large number of different, and original, burnerdesigns to determine the most efficient means of distributing solids through the reactor. Itwas found that there was considerable gas recirculation ocurring within the model, bothin the reaction shaft and in the gas uptake. This study also revealed that the most efficientmeans of dispersing the concentrate particles is by means of radially oriented gas jets atthe burner tip, which is essentially in agreement with the research performed byOutokumpu (23,24).Kimura at al. (25) of Sumitomo employed a one ton/hour pilot plant (similar to anOutokumpu reactor) to study the flash smelting of copper concentrate. Analysis ofquenched samples withdrawn from the flash flame indicated that the size of particlesactually increased with distance down the reaction shaft. This contrasts directly with theobservations made by Jorgensen (21) and others who observed fragmentation of concentrate particles undergoing combustion. After some criticism of their sampling technique(a possible tendency to preferentially select larger particles) the Sumitomo group carriedOut an exhaustive series of trials and showed that their samples were, in fact, representative (26). Once this growth of particles had been verified, a mechanism for dust formationin the Outokumpu reactor was proposed:1. The agglomeration of small molten particles causes larger particles to form. Collisions between solid particles do not cause particle growth.2. These larger particles are easily captured by the bath at the bottom of the reactionshaft.153. The dust collected therefore consists of solid, small, less oxidized particles, whichare carried out of the reactor by the off-gas stream.A number of small-scale flash smelting experiments of copper concentrates have beenperformed by Sohn and co-workers (27). In these studies, the main effort of the researchwas to validate a mathematical model being developed by others, and as a result, the testconditions under which these experiments were performed did not need to be accuratemodels of industrial practice.Munroe (28) and Munroe and Themelis (29) studied the flash reaction of chalcopyrite in asmall electrically heated pilot plant (0.127 m inside diameter x 2 m tall cylindrical shaft),simulating the central core of an Outokumpu flash furnace. The degree of particle reaction, oxygen concentration, and gas concentration were measured as a function of axiallocation and process variables such as particle diameter and oxygen enrichment. Withoutparticles, the injected gas stream was found to heat very rapidly by convection from thehot walls. Increasing the oxygen-to-concentrate ratio (or reducing the particle size)caused increased combustion intensity at smaller distances from the exit of the concentrate burner. Particle fragmentation increased with increasing oxygen concentration, andthis was explained as follows : “. . .gas formation within the molten core plays a veryimportant role in fragmentation and oxygen enrichment would result in faster rates ofreaction, which would increase gas evolution and therefore fragmentation’. Largerparticles (over 70 microns) were observed to be more likely to fragment than smallerparticles. The particle size increase observed by Kimura et al. (25,26) was not seen in thisfacility. This effect was attributed to the influence of low reactor wall temperatures whichprevented the concentrate particles from remaining molten, and hence prevented particle16growth by agglomeration. Munro and Themelis (29) cite further experiments at ColumbiaUniversity in which reactor wall temperatures above 1573 K have resulted in particleagglomeration.In two related papers, Kemori et al. (30) and Inami et al. (31) of Sumitomo have investigated improved copper flash smelting burners both on an industrial and a pilot plant scale,with the objective of reducing dust generation rates and increasing oxygen efficiency.This work has suggested that concentrate particles should not be well dispersed, in orderto maximize particle-particle collisions, and enhance particle growth. In addition, causingall of the particles to combust together in the upper region of the shaft was found toreduce overall dust emissions.2.2.2 Mathematical ModelsMany models of the flash smelting of copper concentrates and the flash converting ofcopper mattes have been published. Virtually all published work has been on theOutokumpu flash furnace - there are no published models of the phenomena occurringwithin the Inco flash flame. However, these studies should serve as a guide for the development of a mathematical description of chalcocite combustion in the new Inco flashfurnace.Ruottu (32) was the first to attempt a model of the flash smelting flame. He utilized thecomplete time-averaged turbulent Navier-Stokes equations, written in cylindrical coordinates, with swirl, to describe the flow of the gas and particles in an Outokumpu reactor.Descriptions of heat transfer, mass transfer and particle kinetics were quite complex- atotal of seventeen simultaneous partial differential equations were written to describe the17system. However, a number of greatly simplifying assumptions were incorporated intothe model, which in some ways defeated the purpose of such a complex mathematicaldescription. For example, the particles and gas were assumed to be at the same temperature at all times and the gas was assumed to be incompressible - density differences dueto the very large temperature gradients in the reactor were neglected. These assumptionsintroduce a considerable degree of approximation into the calculations, while the considerable number of empirical fitting parameters used to “tune” predicted values to experimental measurements greatly restrict the validity of the model.The most complex model of copper flash smelting in an Outokumpu furnace has beendeveloped by Hahn and Sohn (33-40). They considered the flow of concentrate and gas inthe reaction shaft to be two-dimensional (axisymmetric, no swirl). The complete time-averaged steady Navier-Stokes equations described the flow of the gas, but the conthbution to the gas-phase momentum caused by gas expansion was negelected. This allowedthe well-known TEACH code of Gosman and Pun (41) to be used to solve the flowequations. The particle trajectories and contribution to the gas-phase momentum werecomputed using the PSI-CELL method proposed by Crowe et. al (42), whereby the gasand particle momentum equations are solved separately in an iterative manner untilconvergence has been attained. The Boussinesq approximation together with the k — Emodel for turbulent viscosity were used to model the gas phase turbulence. The modellingof the heat transfer processes occurring within the flash reaction shaft was also quiteextensive, with particular emphasis being placed on a complex model of radiative heattransfer. Hahn and Sohn justify this effort by stating that “radiation is the dominant modeof heat transfer in the flash-smelting furnace’. Only heat transfer among the particles, thegas, and the walls was considered: particle-particle heat transfer was neglected. Particlereaction kinetics were treated based on the studies of Chaubal and Sohn (43).18The mathematical model developed by Hahn and Sohn represents a very significant technical acheivment in that they have attempted to describe the flash smelting of chalcopyrite concentrates as rigorously as possible, with few simplifying assumptions. The gas andparticle flow solutions coupled with the particle-gas reaction model have providedsignificant insights into these phenomena. However, it is possible that a few well-foundedsimplifying assumptions could have significantly reduced the difficulty of the problemwith little additional approximation. For example, the considerable mathematical effortmade by Hahn and Sohn to model radiative heat transfer in an Outokumpu furnace maynot be entirely necessary, since calculations made by Jorgensen (19) suggest that conductive/convective heat transfer between the gas and particles is likely to dominate. This ispartially borne out by the experimental and theoretical studies of Walsh et al. (44), andKolb et al. (45) who studied the combustion of a coal-water slurry in a test chamber. Theapparatus which was employed was geometrically quite similar to an Outokumpu reactionshaft - in which combustion air and fuel entered the chamber through a burner mounted inthe top. These investigators found that the entering fuel and air mixture were heated tocombustion almost entirely by entrainment of recirculated combustion products. Based onthis and previous work, they determined that radiation accounted for a maximum of 20 %of the heat required to ignite the coal-water slurry droplets. Predictions of a mathematicalmodel (44), in which radiation was neglected, successfully fitted their experimental data.The mathematical model of Hahn and Sohn also does not consider such phenomena asparticle fragmentation, dust generation and particle agglomeration. This is unfortunate asthese have been shown experimentally to be of considerable industrial importance inchalcopyrite smelting.19Perhaps due to the lack of available industrial data, the findings of the modelling effort ofHahn and Sohn may have limited applicability. For example, two of the major conclusions of this model were that radially directed gas jets will disperse the concentrate morethoroughly and that most of the combustion in the Outokumpu furnace occurs near thetop of the reaction shaft (and hence many furnaces could be made much shorter). The factthat radially directed jets will disperse the concentrate has been well establishedpreviously (22), and forms a major part of several Outokumpu patents (23,24). Moreover,if the Sumitomo group is correct, the particle-particle agglomeration occurring in theOutokumpu reaction shaft after combustion occurs is an important mechanism forreducing dust formation, and hence reaction shafts perhaps should not be made anyshorter.A considerably different approach has been taken by Kim and Themelis (46) in theirmathematical model of copper flash smelting. Initially employing a one-dimensionalmodel, Kim and Themelis proposed a mechanism for particle fragmentation and madepredictions for gas and particle compositions and temperatures as a function of distancedown the reaction shaft. Despite the simplicity of this approach, the predictions made bythe model were quite similar to those made by the much more complex formulation ofHahn and Sohn (33-40).Themelis, Wu and Jiao (47) were the first to attempt to predict the particle size increaseocurring in an Outokumpu furnace as reported by the Sumitomo group (25,26). Utilizingisothermal jet theory together with an equation originally developed for predicting waterdroplet formation in clouds, Themelis et al. modelled particle agglomeration and conse20quent particle size increase. Prior to the experimental data of Inami et al. (30) and Kemoriet al. (31), they suggested that a burner design which promoted the agglomeration ofparticles may be effective in reducing dust formation.A two-dimensional model of the Kennecott-Outokumpu flash converting process hasbeen developed by Jiao, Wu and Themelis (48). As with the model of Hahn and Sohn(33-40) the k— e model was used for turbulence, but a considerably simpler model forradiative heat transfer was employed. Unlike the model of Kim and Themelis (46),particle fragmentation was not considered to occur. Jiao et al. state that the lower ironcontent of the mattes causes a lower reaction rate, and hence a reduced tendency forparticles to explode. While plausible, this is not consistent with the experimental work ofOtero et al. (13) in which low-iron concentrates were seen to explode readily.2.2.3 SummaryThe extensive studies of chalcopyrite flash reactions have shown that reduced particlesize or increased oxygen-to-concentrate ratio can lead to enhanced rates of reaction. Theformation of dust when flash smelting chalcopyrite concentrates has been found to bedominated by two competing factors particle fragmentation and particle agglomeration.Studies have shown that enhanced rates of particle agglomeration may be capable ofreducing dust emissions from Outokunipu flash furnaces.The complex mathematical model of chalcopyrite flash smelting which has been developed by Hahn and Sohn (33-40) has attempted to describe the combustion of chalcopyriteconcentrates from first principles. As a result, this modelling effort represents aconsiderable achievement, which has clarified the analysis of the problem by deriving allof the equations governing this complex process. The more directed approach ofThemelis and co-workers has also attempted to predict the heat, mass and momentumtransfer and reaction kinetics ocurring in the Outokumpu reaction shaft, with similarsuccess.21223 Objectives and Scope3.1 Overall Project OjectivesThe survey of the available literature has shown that little fundamental information exists onthe high-intensity oxidation reactions occurring within the ‘flame of a chalcocite flashsmelting furnace. Mathematical models have not been formulated to describe the behaviourof chalcocite in an Inco flash furnace. Moreover, the relatively high rates of dust formationwhich have been observed in industrial pilot plant trials have remained largely unexplainedand uncorrected. As described by Munroe (28) for the case of chalcopyrite flash smelting,there are two separate mechanisms responsible for the emission of dust from a chalcociteflash reactor:1. The ‘chemical’ production of dust within the chalcocite flash flame by some (as yetunknown) reaction or series of reactions.2. The ‘mechanical’ entrainment of this dust by gas leaving the smelter.The mechanical entrainment of dust, although complex, is predictable for a given flashfurnace based on well-defined fluid mechanics principles. In contrast, the ‘chemical’ production of dust is not yet understood, and is explainable only by a complete analysis of the chalcocite flash flame and a thorough understanding of the chalcocite reaction kinetics. Further, ifit were possible to eliminate the chemical production of dust, there would certainly be littleneed to worry about mechanical entrainment. Clearly, the fundamental mechanism of significance is the means by which this dust is formed in the chalcocite flash flame.Because both the reduction of dust generation and further optimization of the chalcocite flashreaction require fundamental knowledge which is as yet not available, the primary objectives23of this research project were as follows:1. To determine critical phenomena affecting the behaviour of the chalcocite flash smeltingflame.2. To determine the mechanism of dust formation when flash smelting chalcocite concentrates.3. To utilize this information in order to recommend improvements to the existing chalcocite flash smelting reactor and burner which would minimize dust formation.To accomplish these objectives, a number of experimental investigations have beenperformed both in the UBC pilot smelter and in the Inco Port Colborne pilot smelter. Thisexperimental data has then provided fundamental information necessary to the developmentof a mathematical model of the chalcocite flash smelting flame. In turn, this mathematicalmodel has assisted the analysis of the experimental work and aided in recommendations forimprovements in industrial practice.3.2 Experimental ObjectivesThe objectives of the experimental work were as follows:1. To investigate the effect of process variables (oxygen-to-concentrate ratio, particle size,burner design) on the fundamental behaviour of the chalcocite flash smelting flame. Inthis manner, experimental data would be provided both as input data and as test data forthe development of the chalcocite flash flame model.2. To determine factors which are important to the generation of dust when flash smeltingMK concentrate.243.3 Mathematical Modelling ObjectivesThe objectives of the mathematical model were as follows1. To predict, within a reasonable degree of error, the experimental data generated by theUBC flash furnace when supplied with the appropriate operating data.2. To analyze the behaviour of the UBC flash furnace in order to determine the criticalphenomena affecting the behaviour of the flash flame. In particular, the means by whichdust is formed in the chalcocite flash flame is of interest.3. To provide a means whereby the knowledge gained in the UBC pilot plant can be utilizedto suggest improvements in the industrial process.In order to accomplish these objectives, the mathematical model must be as flexible aspossible, and capable of dealing with a wide variety of possible operating conditions. In addition, the mathematical model will need to serve as an important source of insight into thefundamental behaviour of the chalcocite flash flame. Taken togther, these requirements implythat simple ‘tuning” of the mathematical model to one set of experimental conditions is notof significance to this study. That is, the model will be much more valuable if it is capable ofpredicting (within a reasonable margin of error) the behaviour of the chalcocite concentrateover a wide variety of conditions, rather than accurately describing only one set of experimental data. As a result, the total number of “fitting” parameters utilized by the model mustbe kept to an absolute minimum, and not be altered from one set of conditions to another.254 ExperimentalTo accomplish the experimental objectives which were described in the previous chapter, avariety of chalcocite flash converting experiments were conducted at both the UBC pilot plantand the Inco Port Colbome pilot plant. These experiments examined the effects of process variables (concentrate size, oxygen-to-concentrate ratio, concentrate exit velocity, etc.) on the behaviour of the chalcocite flash flame. The UBC and Inco Port Colborne pilot plant facilities havebeen discussed below, together with a description of operational procedures and examples oftypical data. The analysis of this data has been performed with the aid of the mathematical modeland has been presented in Chapters 7,8 and 9.4.1 The UBC Pilot PlantThe UBC flash smelting pilot plant was constructed between 1985 and 1986 and was originally designed to investigate the flash smelting of lead concentrates. At the commencementof this work, the pilot plant was converted to operate with copper concentrate bymodifications to the concentrate feeder and gas handling system. A total of seven preliminarytrials were carried out with both chalcopyrite and chalcocite concentrates to verify that thereactor performed correctly, and to allow design and construction of appropriate data collection equipment.4.1.1 Plant DescriptionThe UBC flash smelting pilot plant was constructed over a one year period in the Department of Metals and Materials Engineering, at the University of British Columbia andconsists of a cylindrical vertical reaction shaft connected to an off-gas handling train. A26photograph of this facility is shown in Figure 4.1 which illustrates the large size of thisplant, although it is of small scale when compared to the actual industrial MK flashconverting process (Table 4.1).Table 4.1: Comparison of Experimental with Industrial Scale Flash Smelting PlantProcess MK Feed Rate Oxygen Feed RateInco “Mini” Pilot Plant- Mississauga 1 kg/hr 0.2 kg/hrUBC Pilot Plant 120 kg/hr 24 kg/hrInco Port Colborne 2000 kg/hr 300-400 kg/hrPilot PlantFull Scale Process 30000 kg/hr 6000 kg/hr(1994) (estimated) (estimated)Nevertheless, this study has demonstrated that the UBC pilot smelter is large enough toprovide useful insights on the actual industrial process, while allowing far greater control,accuracy and flexibility than can be obtained industrially. Flash Reaction ShaftThe reaction shaft is 1.8 m tall and has an inside diameter of approximately 0.5 m(Figure 4.2). The shaft is composed of four sections, each of which has two ports forsolids and gas sampling and three Pt/Pt- 10% Rh thermocouples mounted in the refractory wall. From the top down, the four reactor sections have are designated A,B,C &D respectively. During the experimental trials, all unused ports into the reactor weresealed to minimize air infiltration, but later gas analyses have shown that this effortwas not entirely successful.27The reactor was constructed of 12.7mm mild steel plate and lined with 114 mm ofchrome-magnesite brick, backed up by 89 mm of chrome-magnesite castable refractory and 12.7mm of alumina wool against the steel shell. As Figure 4.2 shows, thereaction shaft differed slightly from a perfect cylinder at the top and the bottom of thereactor. The “well” at the top of the shaft is used for mounting the natural gas preheatburner, while the refractory “lip” at the bottom of the shaft was originally designed todirect molten combustion products into a refractory crucible. No such crucible wasused in the chalcocite flash converting trials, and solid and liquid combustion products were allowed to collect in the hearth mounted below the reaction shaft. Thehearth was lined with similar refractory to the reaction shaft (chrome magnesite brickand castable) but had 2.5 cm of additional alumina wool refractory lining to minimizeheat losses.The reactor is cooled by air drawn upward through an annulus formed by the outershell of the reactor shaft and a steel shroud mounted over the flanges of each shaftsection. Air flowrates of up to 140 m3/rnin were used to maintain the outer temperature of the reactor below 473 K. The outer steel shell temperature was monitored bytwelve chromel-alumel thermocouples (three per reactor section) which were mountedon the outside of the reaction shaft.Originally constructed to investigate the flash smelting of lead concentrates, the UBCpilot plant reactor was designed to operate autogenously, with no need for additionalheat input to maintain combustion temperature. However, the low heat generation rateof MK concentrate (approximately 10 kW net heat at a feed rate of 2 kg/mm)compared to lead concentrates caused problems with maintaining temperatures ininitial chalcocite converting trials. To prevent excessive heat losses, an additional 25.428mm of sacrificial alumina wool refractory was cemented to the inside of the reactionshaft prior to each run. This reduced the reactor inside diameter to approximatelyO.45m, and made interpretation of the wall thermocouples behind this blanket difficult, but did allow chalcocite combustion to proceed.After most of the experiments in the UBC reactor had been performed for thepurposes of this work, a completely new reaction shaft was constructed. This reactionshaft had identical dimensions to the original shaft, but great care was taken inconstruction to ensure that all sections were tightly sealed to minimize air infiltration.The new shaft was lined with 17.8 cm of “Claycast 60” refractory, backed up by 5.1cm of “Kfac-19” alumina wool block. This new refractory provided much better heatretention in the reaction shaft, and also allowed the reactor wall thermocouples tomeasure the true inside wall temperature, as they were no longer covered by the sacrificial alumina wool lining. In association with this new reaction shaft, a number ofadditional improvements were made to the off gas handling and concentrate feedingsystems. Two experimental trials were successfully performed in this new reactionshaft prior to the termination of this study. Gas Handling SystemThe pilot plant gas handling system is shown schematically in Figure 4.3, and consistsof a cooling air inlet line, a refractory-lined cyclone, an impingement-type solidsseparator, a heat exchanger, a baghouse and a packed bed scrubber. The gas handlingsystem is a complex and vital component of the UBC flash smelting facility, allowingthe flash smelting of concentrates to be studied on a large scale in a university envi29ronment. However, the gas handling system is of secondary importance to this studyand will only be briefly described here. A more detailed description in availableelsewhere (49).Gas and suspended solids leaving the bottom of the reaction shaft passed though thesettling chamber, which allowed larger particulates to leave the gas under the influence of gravity. The remaining fines and the off-gas were then quenched with approximately 1.3 m/min of cold air before entering the cyclone. The refractory-linedcyclone was found to capture particulates greater than 5-10 microns quantitatively.Finer particulates left the cyclone with the gas stream, and were finally captured bythe impingement separator and the baghouse. The solids which were recovered fromthe flash smelter as ‘dust” were found in the cyclone, the impingement separator andthe baghouse.During all experimental trials, the entire system was held under vacuum at all times toprevent sulphur dioxide emissions. Gas analysis performed in preliminary runs withthe original reaction shaft indicated that if the reactor vacuum exceeded 500 Pa belowatmospheric pressure, large quantities of air would be drawn into the reactor throughthe numerous openings in the steel shell. To prevent this, a reactor vacuum controlsystem was constructed that limited the reactor vacuum by regulating the flow ofcooling air to the gas handling system. With the reactor vacuum control system, theair infiltration rate into the flash reactor was measured to be less than 10 std I/mm.304.1.1.3 Oxygen and Concentrate Feed SystemsA loss-in-weight screw feeder, controlled by a dedicated computer system (bothmanufactured by Western Scale, Inc.), was installed to deliver concentrate to thereactor at feed rates up to 2 kg/mm. The screw feeder was mounted on calibrated loadcells which were continuously read by the controlling computer, thus ensuring aconstant mass flow rate of concentrate. A computer-controlled bucket elevator systemwas used to replenish the feeder as needed. The output of the screw feeder wasconnected to a vibratory feeder, which transported the concentrate 0.45 m horizontallyto the concentrate burner.A diagram of one of the two concentrate burners used in the U.B.C. pilot reactor isshown schematically in Figure 4.4. The other burner was of identical construction,but had an inside diameter of 2.23 cm which provided twice the cross-sectional areaof the smaller burner. Concentrate and oxygen are mixed together in the upper regionof the burner, and then injected into the flash reactor. The concentrate burner wasconstructed from type 316 stainless steel with an inside diameter of 1.58 cm and anoverall length of 0.45 m. Approximately 5 1/mm of cooling water was used to maintain the temperature within the burner below 30 °C. As shown in Figure 4.2, theconcentrate burner was centred at the top of the reaction shaft, and was mountedconcentrically within the natural gas preheat burner. The tip of the burner was positioned so that it was level with the top of the reaction shaft. In most trials, the smallerconcentrate burner was used, in order to obtain a superficial gas exit velocity of 25rn/s (similar to the exit velocity in the Inco Port Colborne pilot plant) when 2 kg/mmof chalcocite concentrate was reacted with stoichiometric oxygen.31The mixing of the oxygen and the concentrate originally caused considerable experimental difficulties, since the positive oxygen pressure which developed when theoxygen and concentrate were mixed frequently caused the screw feeder to shut down.To prevent this, a positive back pressure of air was maintained over the concentratefeeder using a high pressure air pump. Exactly 28 1/mm of air (less than 10 % of thetotal oxygen flowrate) was used both to maintain this positive pressure and to supplycooling air to the ultraviolet flame detector (part of the natural gas preheat system).A liquid oxygen storage facility (provided by Canadian Liquid Air, Ltd.) suppliedoxygen to the pilot reactor. Oxygen control was performed by a cascade P1 flowcontroller connected to a pneumatically actuated control valve. The setpoint of thiscontroller was automatically fixed by the main system computer. Oxygen flowmeasurement was accomplished by a Micro-Motion coriollis-type mass flowmeterwhich gave a direct reading of oxygen mass flow, with no temperature or pressurecompensation required. The calibration of this meter was checked periodically and themeter was zeroed prior to each run. Data Acquisition and Control SystemsVirtually all data acquisition in the pilot plant was carried out automatically by anIBM PC/AT compatible computer (1 MB RAM, 12 MHz) connected to a KeithleySeries 500 intelligent front end, which provided AID, D/A and digital input capabilities. The Series-500 as used had provision for 16 single-ended high level analoginputs, 48 thermocouple inputs, 32 digital 1/0 lines and 4 0-20 mA analog outputs.The thermocouple boards used an isothermal block with a calibrated thermistor toprovide on-board cold junction compensation. The A/D conversion was performed by32a 14-bit high-speed successive approximation analog to digital converter which had amaximum conversion rate of 50 kHz. In practice, the actual conversion rate waslimited to about 1 kHz by the settling time of the board and channel select multiplexers. Communication between the computer and the Keithley Series 500 wasaccomplished using a memory-mapped interface (128 bytes in high RAM) whichallowed both high speed and flexibility. The main system computer also communicated with the concentrate feed computer via an RS-232 line, which allowed instantaneous control and display of the concentrate feed rate.The form of the output from a number of devices had to be altered to allow reading bythe computer/Series-500 system. For example, the 4-20 mA outputs of the oxygenmass flowmeter and the pressure transducers were passed through 250 ohm precisionresistances to obtain voltage outputs. In addition, a small voltage divider circuit wasconstructed which caused the operation of the gas chromatograph to generate avoltage pulse which in turn automatically triggered high-speed data acquisition of thisdevice.The Series-500 and the RS-232 line were driven by the main system computer using aprogram developed with the ASYST programming language (ASTI inc.). The ASYSTlanguage is a FORTH-like interactive incrementally compiled language whichprovides the speed and power of assembly language with some of the ease ofprogramming of a high-level language. ASYST also supports interrupt-driven multitasking under MS-DOS, which allowed the various data acquisition tasks to be run inthe background at synchronized rates, while the computer simultaneously responded33to commands from the keyboard. The computer program which controlled the systemand read, displayed and stored all of the experimental data was developed over fouryears and comprised several thousand lines of executable code.The acquisition and storage tasks which were performed by the computer system areshown in Table 4.2 along with the sample rate used for each device. It was necessaryto sample the gas chromatograph at high rates due to the high speed at which the gaspeaks emerged from this device. Software was developed which automatically pickedand integrated the gas peaks in real time, while still performing all of the other dataacquisition and control tasks.In addition to collecting, displaying and storing all of the experimental data, thecomputer system also monitored and displayed the state of various critical systemcomponents such as the two system blowers and the cooling water flowrate to theconcentrate burner, if so ordered by the operator, or if any of these components failed,the oxygen and concentrate feeds were shut down by the computer.The accuracy and integrity of the data acquired by the computer was verified either bycomparison with data collected by other means (Eg. a chart recorder and integratorwas used to check the reading of the gas chromatograph) or by comparison with directvoltage measurements taken using a hand-held multimeter (this method was used forthe thermocouples). After the end of an experimental trial, a second computerprogram was used to convert the raw data files into a format which was directly readable by Lotus 1-2-3 and similar spreadsheet-type software packages. This programalso converted all of the thermocouple millivolt readings to degrees Celsius using aseries of third order curve-fit equations.34Table 4.2: Data Acquisition and Control Tasks Performed by the ComputerTask Sample Task LocationFrequencyRead Oxygen 2Hz BackgroundMass FlowmeterCheck System 1 Hz BackgroundComponentsRead GC 2 Hz BackgroundTriggerRead GC 20 Hz Background -When Enabled by TriggerRead Reactor once per minute BackgroundThermocouplesRead JR 1 Hz Background:Analyzer Initiated by OperatorRead Concentrate 0.2 Hz ForegroundDelivery (Automatically)Read 0.1 Hz BackgroundPressureTransducersMonitor 0.1 Hz BackgroundScrubber EffluentsRead Suction 2 Hz Background:Thermocouples Initiated by OperatorAccept Commands Continuous ForegroundFrom OperatorStore all Data Once per minute Foreground(Automatically)354.1.2 Sampling and Analysis4.1.2.1 Gas Sampling EquipmentThe system which was used to withdraw gas samples from the flash reactor is shownschematically in Figure 4.5. It consisted of a water-cooled stainless-steel tube (5 mminside diameter) with a reverse-mounted nozzle connected to a vacuum pump andfilter arrangement. The gas sampling probe was inserted through the windows in theside of the reactor and was positioned radially by measurement. Samples werenormally taken both at the centreline of the reactor and near the reactor wall. Gas waswithdrawn from the reactor at a rate of approximately 4 std. L/min until a sufficientvolume of sample had been collected. The total volume of the entire gas samplingsystem was approximately 0.5 1, and hence it was necessary to sample for at least oneminute prior to analysis in order to ensure that the gas sampling system was thoroughly flushed with the new sample. The gas sampling probe usually became cloggedwith solids after only two to three minutes of sampling, but occasionally cloggedmuch sooner. If this occurred, the gas sample was rejected.Owing to the lack of information on the exact gas velocities within the flash furnace,it was not possible to sample the gas truly iso-kinetically. An attempt at iso-kineticsampling was made by setting the sampling rate to a value which would approximatethe gas velocity in the region of interest. However it is likely that this gas samplingvelocity was frequently too high when samples were taken at the centre of the reactor.As a result of this, it is likely that many of the gas samples were contaminated by thesurrounding recirculating gas being withdrawn from the reactor, along with the36desired gas sample from the oxygen/concentrate jet. This error has been reflected inthe small change of gas composition with axial location which was observed in manyof the gas samples.A number of attempts were made to determine the true gas temperature within theflash furnace as a function of axial and radial position. To accomplish this, four water-cooled suction thermocouples were constructed (Figure 4.6), in which gas was suckedat velocities of 15 mIs over a shielded thermocouple, and then rapidly quenched by awater-cooled stainless steel coil. In practice, the suction thermocouples were notsuccessful, as they rapidly clogged with solids which accompanied the gas. The thermocouples only operated reliably when the outer shield was removed and no suctionwas performed. In this configuration, the thermocouples were inserted just past thesacrificial alumina wool refractory to provide an estimate of the approximate walltemperature. Gas AnalysisGas samples withdrawn from the reactor were analyzed with an on-line, infra-red SO2analyzer (manufactured by Horiba and AnalyGas ) and a Varian gas chromatograph.The gas chromatograph and the SO2 analyzer were calibrated prior to each run usinggas standards. The infra-red analyzer allowed percentage quantities of SO2 to be readinstantly while the gas chromatograph required two minutes to analyze for SO2, 02,and N2. The gas chromatograph operated at 80 °C and used a column toseparate the SO2 from an 02/N composite peak and a molecular sieve column to37divide the°2 from the N2. The outputs of both the gas chromatograph and the SO2analyzer were automatically recorded by the main system computer which also integrated the gas chromatograph peaks. Solids SamplingTwo different sets of apparatus were used to withdraw solids samples from the flashflame. The first of these is shown in Figure 4.7 and consisted of a partitioned stainlesssteel trough 2.54 cm x 2.54 cm x 24.0 cm mounted to a support pipe and fitted with astainless steel cover. The solid sampler was filled with water and inserted into thecentre of the reaction shaft at appropriate times during a pilot plant trial. The lid wasthen withdrawn and solid material was allowed to collect in the sampler for a periodof 30-45 seconds. After this time, the lid was replaced and the sampler was withdrawnfrom the reaction shaft. The resulting quenched solids samples were dried, partitionedradially and sent to Inco for chemical analysis.Unlike the water-filled trough, the second solid sampler covered the entire diameter ofthe reaction shaft. This device is illustrated in Figure 4.8 and consisted of two water-cooled stainless steel tubes welded together. Solids sampling with this device wascarried out as described previously, except that this sampler, having no lid, wassimply inserted into the flame and then withdrawn. The solids collected from thissampler were also partitioned radially for analysis, and provided an indication of thesolids disthbution within the flash flame.Various attempts were made at iso-kinetic solids sampling, all of which proved futiledue to excessive solid loadings. This was unfortunate, as theory suggests that a static38impingement-type collector of the design described above will tend to sample largerparticles preferentially, with fines following the gas flowing around the sampler. Anapproximate calculation based on filtration theory has been performed (Appendix 1)to estimate the magnitude of this effect. As Figure 4.9 indicates, this effect is probablynot severe; only particles smaller than 4 microns have a sampling efficiency below50%, and this only occurs in low velocity regions of the reaction shaft. Nevertheless,this does indicate that very fine particles will not be effectively collected by thissampler and the subsequent analysis of the solids samples must take this effect intoaccount. Solids AnalysisThe analysis of solids samples withdrawn from the UBC pilot reactor was performedat the Inco J. Roy Gordon Research Laboratories in Mississauga, Ontario. Solidssamples were analyzed for copper, sulphur, iron, nickel, oxygen and sulphate. Inaddition, many of the samples were also examined using the scanning electron microscope and X-ray diffraction.4.1.3 Operating ProcedureThe procedure followed in the UBC pilot plant experiments was as follows1. The reaction shaft was preheated with natural gas over two to three days until theoperating temperature of 800-900 C (determined by the wall thermocouples) hadbeen attained.392. All analytical equipment (gas chromatograph, infra-red analyzer, pH meter, sulphurdioxide detector) was checked and calibrated.3. The natural gas burner was shut off, and the oxygen and concentrate feeds were initiated. Air was bled into the gas handling train to maintain the reactor vacuum at 255Pa below atmospheric pressure.4. Data collection began immediately. Gas and solid samples were taken repeatedly ateach axial location; solid samples were partitioned radially, labelled and set aside.5. In addition to collecting data during each run, it was necessary to monitor the effluentpH and SO2 concentration of the scrubber, the system vacuum, the oxygen andconcentrate feed rates and the gas temperature entering the baghouse in order toensure that the system was working correctly, and neither clogging with solids norventing SO2 to the outside. In addition to this, the cyclone had to be emptied periodically of collected solids and the main concentrate feed hopper had to be continuouslyreplenished. These duties required the full-time attention of four operators.6. Once all data had been collected (typically after 90 to 120 minutes), the system wasshut down and allowed to cool down over two to three days. The material which hadaccumulated in the various parts of the gas handling train was sampled, weighed andrecorded. All solids samples were then sent to Inco for assay.Due to the complexity of this system, problems frequently arose during an experimentaltrial, some of which occasionally caused premature shutdown before all data had beencollected. The concentrate and oxygen feed systems were the two biggest culprits in this40regard. In addition, the total amount of sodium hydroxide available for scrubbing thesulphur dioxide was limited and eventually caused most runs to terminate after amaximum of 150 minutes.The experimental trial itself actually comprised a very small fraction of the overall UBCpilot plant campaign. Constructing, repairing and calibrating equipment, cleaning up andpreparing for a new run could take between 2 to 24 weeks depending on circumstances,with an average of 8 weeks required.4.1.4 Run ConditionsTable 4.3 summarizes the nine full scale runs which were carried out in the UBC pilotplant. As Table 4.3 shows, most trials were performed with standard MK concentrate asprovided by Inco; a particle size analysis of the MK concentrate as received from Inco isgiven in Figure 4.10. A typical composition of the unsized MK concentrate is 73.5 % Cu,4.55 % Ni, 0.26 % Fe, 20.1 % S, 1.04 % 0. Because the composition of the MK concentrate received from Inco tended to vary slightly with time, an analysis of the inputconcentrate was performed prior to each run.For two trials, the MK concentrate was separated into large ( > 30 im ) and small (<30jim ) fractions by Ortech Ltd of Mississauga, Ontario. The particle size analysis of thelarge and small size fractions is given in Figures 4.11 and 4.12 respectively. To ensurethat sizing the concentrate did not alter the chemical composition, an analysis of sized andunsized concentrate was performed and is summarized in Table 4.4. It is clear from thistable that sizing the concentrate had little or no significant effect on chemical composition.41Table 4.3 : Full-Scale Runs Performed in UBC Pilot PlantRun Date Run Conditions Burner Diameter02/ConcentratePerformed (cm) Mass Ratio1 27/4/89 Regular MK 1.58 0.2as received2 12/5/89 RegularMK 1.58 0.2as received3 22/6/89 SizedMK 1.58 0.2fine fraction4 27/7/89 Regular MK 2.23 0.4100 %_excess_°25 1/8/89 Regular MK 2.23 0.350 % excess 026 13/12/89 Regular MK 1.58 0.2improved burner7 3/5/90 SizedMK 1.58 0.2large fraction8 3/7/91 RegularMK 1.58 0.24New Shaft9 3/4/92 RegularMK 1.58 0.2Table 4.4: Chemical Analysis of Sized and As-Received ConcentrateConcentrate Cu (%) Fe (%) S (%)Typical Unsized 69-75 .1-2 18-22Large Size Fraction 73.4 .03 21.3Small Size Fraction 71.1 .32 18.6424.1.5 Run DataThe data obtained in a typical pilot plant run consisted of the following:1. Oxygen and concentrate feed rates.2. Reactor wall temperatures.3. Output from gas chromatograph and infra-red SO2 analyzer.4. Suction thermocouple output.5. Assays of solids samples collected from the reactor.Examples of data from run 6 with standard MK concentrate are given in Figures 4.13 to4.15 and in Table 4.5. The data from all other runs are discussed in Chapter 7.Table 4.5 : Assays of Solids Samples from Run 6.Sample Location Cu (%) Ni (%) Fe (%) S (%) 0 (%) SO4 (%)MK - As Received 70.0 3.58 2.63 20 4.01 0.29Hearth 84.9 3.15 2.08 3.48 3.6 0.89Hearth 75.8 5.33 3.68 1.03 11.7 0.63Cyclone 75.9 1.47 1.76 2.52 16.3 7.19Solid Sample - Level C 74.7 3.79 2.78 8.2- 0.25Solid Sample - Level D 71.2 3.98 2.55 10.2- 0.32(0-5cm from centre)Solid Sample - Level D 77.9 3.92 2.73 4.29- 0.27(5-7cm from centre)Solid Sample - Level D 82.6 3.79 2.57 1.33 7.6 0.15(7-9cm from centre)Solid Sample- Level D 83.3 3.93 2.86 0.38 7.2 0.12(9-12cm from centre)Solid Sample- Level D 73.4 4.04 2.74 7.86 12.0 0.29(0-9cm from centre)Solid Sample- Level D 84.1 4.07 2.59 0.9 6.41 0.1(9- 12cm from centre)Solid Sample- Level D 82.3 3.94 2.75 0.56 5.5 0.09(12cm from_centre-)43As described earlier, the additional sacrificial alumina wool refractory placed inside theoriginal reaction shaft prevented the wall-mounted thermocouples from measuring thetrue wall temperature. This difficulty was overcome by the following technique:1. The wall heat loss in each reactor section was calculated from the temperature difference between the steel shell and the inside wall temperatures as measured.2. This heat loss (10 kW total or approximately 2.5 kW per section) was used togetherwith the thermal conductivity data of the blanket (supplied by the manufacturer) tocorrect these inside wall temperatures. Normally, the temperatures beyond thealumina wool refractory were calculated to be 200-300 K higher than the valuesmeasured by the wall thermocouples.Wall temperatures which have been “corrected” by this technique have been plotted inFigure 4.14. Interestingly, reactor wall temperature measurements from the runs with thenew reactor shaft (without the alumina wool refractory) agree very closely (within 5-10K) to these corrected values, indicating that the error associated with correcting thetemperatures in this manner is small.4.1.6 Preliminary Data AnalysisA complete analysis of this experimental data has been performed with the aid of a mathematical model and will be discussed in Chapters 6 and 7. Nevertheless, a number ofgeneral observations may be made about the performance of the UBC pilot plant at thistime44Dust Generation:The dust produced in each of the experimental runs was collected in the cyclone,impingement separator and the baghouse. For all runs in which stoichiometricoxygen was used, the dust generation rate of the UBC pilot smelter was usuallyvery low (Table 4.6) : 5-6 % of the total mass input, and approximately 2.5-3.5 %of the total copper input. This compares with dust rates of approximately 10 % oftotal mass fed in the Port Colborne pilot plant (51), and significantly higher rates inthe MK melter at Copper Cliff. As Table 4.6 shows, the dust generation rate in theUBC pilot plant was found to be very much higher for the runs with excess oxygenso much dust was formed that the system was forced to shut down prematurely inthese runs.Table 4.6: Effect of Oxygen Concentration on Dust FormationFDate 02/Concenirate Run Length Concentrate Dust DustRatio (mm) Input Collected Collected(kg) (kg) (%)27/4/89 0.20 84 168 7.0 4.112/5/89 0.20 67 132 5.0 4.013/12/89 0.20 85 173 5.0 3.03/4/92 0.20 126 252 15.6 6.027/7/89 0.40 14 28 7.0 25.0(100% excess)1/8/89 0.30 40 80 6.5 8.1(50 % Excess)The chemical composition of the dust produced in these runs is quite consistent,and is similar to that produced in the Inco Port Colborne Pilot smelter (Table 4.7).X-ray diffraction analyses performed by Inco and at UBC have shown this dust tobe largely copper metal, copper oxides and sulphates.45Table 4.7 Chemical Analysis of Dust Produced at UBC and Port ColbomeSample Source Cu (%) Ni (%) Fe (%) S (%) 0 (%) SO4 (%)UBC Run 6, Cyclone 75.9 1.47 1.76 2.52 16.3 7.19Port Colborne 73.4 4.7 1.1 2.8 N/A 8.8(Campaign #11)As Figure 4.16 indicates, this dust is extremely fine, and is mainly composed ofagglomerates of small spheres. These agglomerates had a mean particle size of 3-7microns, while the constituents of the agglomerates were much smaller (many << 1micron). Studies by Otero et al. (62) have shown that this dust is very similar inappearance to that produced at both the Port Colborne pilot reactor and the MKmelter in Copper Cliff.Interestingly, as Table 4.8 shows, the copper-to-nickel and copper-to-iron massratios of the dust in the UBC flash furnace was very much greater than the inputvalues for all runs. This indicates that copper was being preferentially concentratedin the dust by some (as yet unexplained) mechanism. This also implies that all ofthe dust recovered from the UBC flash furnace is due to ‘chemical’ effects, andhence there exists at least the possibility of eliminating this dust by improvedburner or smelter design.2. Combustion LocationThe reactor wall temperatures indicate that most of the MK concentrate combusts inreactor section D, approximately 1.40 to 1.80 m from the burner tip. For example,Figure 4.14 shows a plot of wall temperatures from run 6. (Note that these temperatures have been corrected for the effects of the alumina wool blanket, using the46Table 4.8 Comparison of Copper/Nickel and Copper/Iron Mass Ratios of SolidsInput to and Collected from the UBC Flash FurnaceRun Cu/Ni Ratio Cu/Fe RatioInput Cyclone Input Cyclone1 15 69.8 80 1262 26.3 73.6 222 2475 26 69.1 200 3656 18.2 51.6 24.8 43.1technique described above). In this plot, it is evident that section ‘1D” is consistentlythe hottest, most likely reflecting the highest local heat generation rate, and therefore indicating that the majority of combustion occurs in this region.Observations made through the windows in the UBC flash reactor substantiate thisobservation. Under conditions of stoichiometric oxygen and unsized MK concentrate, a very intense reaction was observed in section D, and a very bright whitelight could be observed through the surrounding fume. This reaction was notobserved in any of the other reactor sections.3. Gas AnalysisGas analysis data from three experimental trials is summarized in Table 4.9, whilethe remainder of the data is discussed in Chapter 7. As this table shows, dependingupon the sample and the run, very high concentrations of nitrogen were occasionally observed in the flash reactor (up to 40 %), while the normal nitrogen concen47tration measured was approximately 10-20 %. As described above, approximately28 std 1/mm of air were introduced into the flash reactor in each run to facilitateconcentrate feeding and to cool critical system components such as the ultravioletflame detectors. Air infiltration measurements suggest that an additional 15-251/mm of air may have entered the reactor through various loose fittings and leakingseals. However, even considering that some 42 1/mm of N2 may be entering thereactor, it is not possible to obtain nitrogen concentrations of 40 % if the reactionCu2S + 02 = Cu + °2 occurs completely, since some 300 1/mm of SO2 would beproduced (leading to N2 concentrations of 12 to 15 %). As the correspondingoxygen concentrations in the gas samples were very low, it is unlikely that this dataresults simply from contamination of the gas analysis system with ambient air.Therefore, it is very likely another chemical reaction is also occurring in the UBCflash smelter which removes oxygen but does not produce SO2, resulting inmoderate to high nitrogen concentrations. As Table 4.9 shows, the severity of thiseffect clearly depends upon the run conditions.At the temperatures most commonly employed, the most likely chemical reactionswhich could produce this effect would be2Cu + 0.5 02 = Cu20 (4.1)Cu + 0.5°2 = CuO (4.2)Cu2S + SO2 + 3°2 = 2CuSO4 (4.3)that is, the over-oxidation of some chalcocite concentrate to produce copper oxidesand the formation of sulphates. The occurrence of any of these reactions wouldreduce the oxygen concentration but would not produce additional SO2, resulting ina net increase in the measured N2 concentration. An additional effect produced by48these reactions is that the occurrence of reactions 1 and 2 will cause a reduced rateof overall mass loss, while reaction 3 will actually tend to result in an overallmeasured net mass increase observed in the hearth.Table 4.9 Gas Analyses From Three Runs in the UBC Flash FurnaceRun Sample Location°2 % 02 % N2 %8 A 27 70 3.1(Base Case) B 55 40 5.0C 65 29 6.0D 86 5.1 9.09 A 50 45 5(20 % Excess) B 60 34 6C 73 18 9D 80 12 87 A 20 41 39(Large Particles) D 42 16 424. Material BalanceTo aid in the analysis of the UBC flash reactor, an overall material balance wasattempted after each trial. Unfortunately, it was not always possible to collect all ofthe solids produced during every run, due to the accumulation of material on thesacrificial alumina wool blanket lining the reactor walls. A more serious problemwas that the solids collected in the hearth below the reaction shaft were found tovary greatly in composition. For example, as Table 4.5 indicates, solids in a particular region of the hearth in a given trial could have greatly different compositionscompared to those found in other regions.49Since the majority of material collected in the flash smelter was found in the hearth,the difficulty with obtaining reliable hearth compositions made performing anoverall mass balance very difficult. Therefore, in each run, a number of assays ofthe hearth material were taken and the average was used to obtain an overall massbalance. Although this method is imperfect, the impossibility of obtaining a representative sample of the hearth made this technique necessary. The material balanceswhich have been obtained in this manner are shown in Table 4.10. Note thatcomplete conversion of the MK concentrate is associated with a 20 % net massloss, caused by the removal of sulphur as SO2.Table 4.10: Mass balances from all runs in UBC Flash FurnaceMass Collected by Unit (kg) Mass Totals (kg) Cu, Ni, Fe Input Cu, Ni, Fe(kg) Collected(kg)Run Hearth Walls Settling Cyclone Baghousel Collected Input Cu Ni Fe Cu Ni FeChamber1 81 30.0 13.0 7.0 0.0 131 150 107 4.5 0.6 107 3.9 0.76 113 33.5 5.5 4.0 0.0 156 173 120 6.2 4.5 119 6.3 4.67 138 0.0 0.0 0.0 0.0 138 116 87 3.5 0.5 100 4.1 0.78 114 10.3 25.7 11.3 8.4 170 175 124 5.3 0.7 127 5.1 0.89 176 40.0 6.6 11.8 3.7 239 252The radial variation of solids composition with location in the hearth was noticed inthe first MK trial of April 27th, 1989. A sketch was made of the hearth compositionand is shown in Figure 4.17. As this sketch indicates, there were three distinctregions or ‘rings” of solids found in the hearth, the middle ring being largelymetallic copper, while the inner and outer rings were composed of a darker material. These “rings” were not always observed in the hearth in every MK trial, but thevariability of solids composition with location in the hearth was always present.505. Solids SamplesIn all of the experimental trials, it was very difficult to obtain solids samples fromthe top two reactor sections (A,B), possibly indicating that the reacting MK concentrate was not molten in this region, and that the high velocities present in theoxygen-concentrate jet in the upper regions of the reaction shaft simply blew theconcentrate off the impingement-type solids collector. This is in agreement withqualitative observations made in the pilot smelter - the solids sampler was occasionally found to be coated with a thin film of dust in this region, which was visually similar to the input concentrate.The data shown in Table 4.5 clearly indicates a strong radial variation in samplecomposition, with solids toward the edge of the reactor more thoroughly reactedthan those toward the centre. This effect has been observed previously in the UBCflash smelter when smelting lead concentrates (49). The trend in the axial composition data is somewhat less clear, with solids from reactor section D being onlyslightly more reacted than those at section C.4.2 Inco Port Colborne Pilot PlantThe second series of experimental trials which was performed for this work was carried outat the Inco pilot plant at Port Colborne, Ontario. The Port Colborne pilot plant is very muchlarger than the UBC facility, and simulates the actual industrial process much more closely.A total of fourteen experimental trials were carried out, which investigated the effects ofburner diameter (gas exit velocity), dust recycle and supplementary heat addition on the flashflame composition.514.2.1 Plant Description4.2.1.1 Flash ReactorThe Port Colborne flash reactor is shown schematically in Figure 4.18 and consists ofa horizontal cylindrical shaft with a vertical gas offtake. MK concentrate and oxygenare injected horizontally at the north end of the vessel, while additional heat isprovided by natural gas burners mounted at both the north and the south ends. Thesetwo natural gas burners were each operated at rates of approximately 0.6 scm/mm ofnatural gas, which maintained the furnace temperature at approximately 1350 °C. Dueto the experimental nature of this facility, and the fact that the vessel has undergonemany different modifications, the exact internal dimensions of the flash reactor aredifficult to obtain. The best estimates available are shown in Figure 4.18.For all of the flame sampling trials without dust recycle, concentrate and oxygen feedrates of 2000 kg/hr and 3.8 std m3/min respectively were used, resulting in an oxygento concentrate mass ratio of about 0.16. This low oxygen to concentrate ratio was usedby Inco because sufficient additional oxygen was provided by other (previouslyunmeasured) sources to oxidize the concentrate. Concentrate BurnerTwo concentrate burners (or “guns”) were used at Port Colborne for these researchtrials with inside diameters of 5.3 cm and 6.3 cm (termed “2 inch” and “2 1/2 inch”burners respectively). Depending on the oxygen and concentrate feed rates, the bulkgas exit velocities from these burners varied between 25 and 40 rn/s. One of the two52concentrate burners is shown schematically in Figure 4.19. The MK concentrateenters at the rear through the top of the burner, while oxygen is injected horizontallyunder a confining baffle. Each burner is constructed of mild steel and is water-cooled.The confinement produced by this baffle causes the oxygen to accelerate to superficialvelocities of 100 rn/s or higher and produces strong suction in the concentrate feedtube, thus facilitating concentrate delivery. However, cold modelling studies sponsored by Inco (50) have shown that this baffle produces a number of other effects,some of which may not be desirable:1. The acceleration and expansion of the oxygen past the confining baffle produces ahighly turbulent jet-like flow, which directs the concentrate against the walls ofthe burner. This results in considerable erosion of the burner wall, and requiresfrequent burner changes when using hard or abrasive concentrates. This is less of aproblem when running with unfluxed MK concentrate.2. This ‘jet” caused by the baffle may also result in a recirculation occurring withinthe burner. Cold measurements indicated that gas outside of the burner could bedrawn inside by this recirculating flow. The superficial gas exit velocity used inthe Port Colborne trials (obtained by dividing the gas flowrate by the burner crosssectional area) bears little resemblance to velocities actually produced by theburner jet. Pitot tube measurements have shown that for a superficial exit velocityof 25 mIs, gas velocities of 70 rn/s to -10 rn/s (into the burner) were measured nearthe burner tip.53Owing to these fundamental differences, the performance of the Port Colborne burneris difficult to compare exactly with that of the UBC burner, despite the similarity ofthe bulk gas exit velocities.4.2.2 Experimental ApparatusA schematic diagram of the apparatus which was used to sample the flash flame is shownin Figure 4.20 and consisted of a water-cooled stainless steel sampling probe connected toa gas handling train. The sample probe had an inside diameter of 0.63 cm, and slid withina flanged locking collar which both supported and sealed the probe. The maximum lengthof probe which could project beneath the collar was 107 cm.The probe was connected to the gas handling train via 3 m of 0.63 cm diameter rubberhose, which was in turn connected to a solids trap, the gas sample vial, a vacuum pumpand a bubbler, which served to verify flow. Hose clamps and stopcock grease were usedto eliminate leaks, and thin-walled latex rubber tubing was used to connect the variouscomponents of the handling train. This thin tubing collapsed under a vacuum of approximately 1000 Pa or more, which in the case of clogging of the sample probe both providedan indication of clogging and simultaneously prevented excessive vacuum from suckingoutside air into the system. The vacuum pump had a capacity of 18 Vmin, ensuring thatthe entire gas sampling system (which had a volume of approximately 0.5 1) was rapidlyand thoroughly flushed when sampling.Analysis of the gas samples for N2, C02,°2’ and SO2 was carried Out using the Variangas chromatograph available at Port Colborne. Standards were injected before, during,and after injection of gas samples, thus ensuring reliable analysis. A slight complication54was encountered with some high N2 samples, which produced data beyond the range ofthe available standards. A technique to cope with such samples was developed (dilutionwith He and re-normalization) which allowed accurate analysis, but the N2 analysis of asmall number of the earlier samples (runs 1-3) may have a slightly higher margin of errorassociated with them (± 5 % vs ± I %).4.2.3 Operating ProcedureTable 4.11 summarizes the reactor operating conditions under which samples were takenfrom the flash flame together with the gas analysis data. A complete summary of theoperations of the Port Colborne smelter during the time that these samples were taken canbe found in the weekly summaries for the weeks ending October 21 and October 28,1990. The run conditions were varied in order to determine the effect of the natural gasburners and flash gun diameter on the composition of the flash flame. Unfortunately, dueto the constraints of time and scheduling it was not possible to investigate completely theeffects of dust recycle.During these trials, every attempt was made to locate the gas sampling probe on thecentreline of the flash gun at each sample port used. However, the (approximately) 8degree inclination of the gun, and the difficulty in determining the exact location of thegun relative to the refractory and the sample ports made exact location impossible. Therefore samples were taken at a number of radial positions at each sample port - the approximate locations of these samples are shown in Figure 4.21. However, the fact that a largeamount of solids was recovered from the outside of the probe indicates that the gassampling probe was probably quite near the centre of the jet.55The procedure used to take gas and solid samples was as follows1. The probe collar was set to give the desired immersion depth - typical depths usedwere 0.91 m, 1.0 m, 1.07 m.2. A leak check of the probe and handling train was performed.3. The probe was inserted into the reactor, and the vacuum pump was started.4. The sampling continued for 30-45 seconds, until the apparatus had been completelyflushed with sample.5. The sample pump was shut off, the probe was removed from the reactor, and a newgas sample vial was put in place.6. Solids which had frozen to the outside of the probe were removed and put in labelledbags identifying their position. The probe was then cleaned and prepared for anothersample.In an attempt to verify the fact that the sample vial was thoroughly flushed with newsample and not contaminated with residual quantities of air, helium was back flushedthrough the entire system prior to runs 15 and 16. No difference in gas concentration wasobserved, indicating that the experimental technique employed was valid.4.2.4 Experimental DataA complete summary of the gas and solids analysis data is given in Tables 4.11 and 4.12.In addition to this data, a number of other observations have been made:561. Four separate temperature measurements at the centre of the MK/oxygen jet at port #1indicate that with the south burner off, the temperature in this region was possiblyquite low (980 - 1100 CC). These temperature measurements did not use any suction,and must therefore be regarded as only an estimate of the gas and solids temperaturein the oxygen/concentrate jet.2. With the south burner on, the temperature in this region was much higher. This wasapparent from the water in the probe boiling rapidly and the blinding white lightemitted by the flame under these conditions.4.2.5 Preliminary AnalysisAs with the data produced in the UBC pilot plant, a complete analysis of the PortColborne experimental data will be performed with the aid of the mathematical model.Nevertheless, a number of general observations can be made about the Port Colborne dataat this time.A plot of all of the gas analysis data with the 6.3 cm and 5.3 cm flash burner has beenpresented in Figures 4.22 and 4.23 respectively. A number of observations can be madeabout these plots1. The reproducibility between sample readings made at the same port at different timesis very good, with many points plotting on top of one another.2. There appears to be little variation in gas composition as a function of radial position.3. There appears to be little qualitative difference between the 5.3 cm and the 6.3 cmburner data.574. The nitrogen content of the gas was extremely high at the first sample port, slightlylower at the second, but dropped quickly by the fourth sample port. The sulphurdioxide concentration followed the reverse trend.5. Even with the south burner off, there was still some carbon dioxide found in the gas atthe first sample port.6. The oxygen concentration was quite low (< 10% ) everywhere, indicating that mostof the oxygen had either reacted or been diluted by the time it has reached the firstsample port.The high nitrogen content in the gas at sample port #1 is at first rather worrying, possiblyindicating severe contamination of samples with air. However, the consistency of the gasanalyses is slightly reassuring, since it is unlikely that air leaks would be so consistent. Inaddition, the rise in sulphur dioxide concentration from the second sample port to thefourth sample port suggests that some additional reason rather than sample contaminationis more likely.The initially high nitrogen levels in the Port Colborne flash reactor can be best explainedby air infiltration into the reactor along with uneven combustion of MK concentrate. Thatis, if some of the MK concentrate which combusts keeps on reacting to oxide while someMK concentrate remains unreacted, then this would result in both low SO2 and low°2concentrations in the gas at the first sample port (while entrained air would supply thenitrogen). The over-reacted copper oxide (either CuO or Cu20) and the unreacted MKwould fall into the bath and react according toCu2S + 2CuO = 6Cu + SO258Cu2S + 2CuO = 4Cu + SO2It is this SO2 emerging from the bath which was probably responsible for the sharp rise inSO2 concentration in the region of the fourth sampling port. Some confirmation of thishas been obtained from Port Colborne personnel, who observe large volumes of SO2coming off the bath during tapping operations.There are a number of locations where air could enter the Port Colborne flash reactor, butthe largest single source is likely to be the central gas offtake. It is interesting to note thatthere was a considerable amount of CO2 found at port #1 (5-6 %) even when the southburner was turned off. When the north burner was turned off as well, this value droppedto 0.3 % or less, indicating that the flash flame is capable of sucking CO2 produced thenorth burner all the way from the opposite end of the reactor, even past the gas offtake.Based on these gas samples, and knowing the air, natural gas and oxygen feed rates intothe furnace, it is possible to make calculations of the air infiltration rates into the PortColborne flash reactor. Depending upon which samples are chosen to make this calculation, the air infiltration rate was between 4 and 5 scm/mm. This indicates that fully 25 %of the oxygen required to smelt the concentrate was provided by infiltrated air.Examining the (S02+)/C ratio from port #1 with both natural gas burners on alsoprovides some information. Since approximately 1.2 scm/mm of natural gas are put intothe Port Colbome flash reactor through both natural gas burners and about 3.8 scm/mmof oxygen, one would expect this ratio to be (at an absolute minimum) 3.8/1.2 or about3.2. For the 2 1/2” burner, this ratio is usually lower than 3, while for the case of the 2”burner, it is usually 3 or greater. This indicates that the 2 1/2” flash burner over-oxidizes a59considerable fraction of the MK concentrate when both natural gas burners are on. It isnot possible (despite any air leakage into the sample if it occurred) for this ratio to be lessthan 3.2 without losing some oxygen and not generating SO2 - the only way of accomplishing this is by over oxidizing concentrate. This ratio is somewhat higher for the 2”burner, indicating either a reduced entrainment of CO2 from the north burner, increasedair infiltration, or reduced over-oxidation of concentrate.This is corroborated by evidence from the solids samples. As Table 4.12 shows, onlyapproximately 25 % of the sulphur in the concentrate has been removed by the time theMK reaches the first sampling location, while a high molar concentration of oxygen isfound in the solid samples at this point. This indicates that much of the concentrate is stillunreacted by the second sample port, while some concentrate has been over-oxidized tocopper oxides.60Table 4.11: Operating Conditions and Analyses of Gas Samples from PortColborne Flash ReactorRun Sample Probe Flash South North 02 % N2 % SO2 % CO2 % SumPort Depth Gun ID. Burner Burner(cm) (cm)1 1 108 6.3 on on 1.4 54.5 24.8 19.1 99.81 1 108 6.3 on on 1.2 60.1 23.5 15.2 1002 1 108 6.3 on on 2.9 31.2 22.8 43.1 1002 1 108 6.3 on on 3.3 48.6 28.8 80.73 1 108 6.3 off on 7.3 59.7 22.5 11.7 101.23 1 108 6.3 off on 6.2 53.4 26.1 10.9 96.64 4 76 6.3 on on 2.9 32.5 32.1 32.1 99.65 1 91 6.3 off on 6.6 55.1 24.7 13.6 1005 1 91 6.3 off on 7.3 55.9 21.3 12.8 97.36 2 108 6.3 on on 5.8 53.9 31.2 9.2 100.18 1 108 6.3 off on 16.6 62.9 18.2 1.8 99.59 1 108 6.3 off off 15 60.9 21.2 0.2 97.39 2 91 6.3 off off 4.9 57.7 37.5 0.3 100.49 4 91 6.3 off off 3.2 43.2 49.3 0.4 96.112 1 107 5.3 on on 15.3 58 24.2 10.6 108.112 1 91 5.3 on on 7.7 55 30 12.6 105.314 1 91 5.3 off on 8.4 61.9 26.3 3.4 10014 1 99 5.3 off on 9.6 72 19.3 2.7 103.614 1 107 5.3 off on 16.4 56.7 24.2 2.6 99.914 1 107 5.3 off on 10.5 56 23.9 2.6 9314 1 91 5.3 off on 5.1 58.3 32.7 3.8 99.914 2 107 5.3 off on 13.5 59 21.3 2 95.814 2 91 5.3 off on 4.6 58.6 33.5 3.3 10014 4 107 5.3 off on 1.7 47 47.6 5 101.314 4 91 5.3 off on 4.7 43 48.4 5.6 101.715 1 107 5.3 on on 20.3 57.1 16.3 6.3 10016 1 107 5.3 off on 13.4 58.8 24.9 3 100.116 1 91 5.3 off on 6.7 63.5 24.9 3.7 98.816 1 99 5.3 off On 6.9 62.2 28.2 4.8 102.116 1 107 5.3 off on 7.3 58.1 32.9 5.2 103.516 2 107 5.3 off on 14.8 50.9 28.4 3 97.116 2 107 5.3 off on 3.2 59.6 34.7 5 102.516 4 107 5.3 off on 2.1 41.5 51.1 5.6 100.316 4 91 5.3 off on 2.3 42 50.1 5.6 10061Table 4.12: Normalized Assays of Solids Samples from Port Colborne Flash ReactorRun Sample Probe Flash South North %Cu %SPort Depth Gun I.D. Burner Burner(cm) (cm)2 1 108 6.3 on on 74.8 12.5 12.72 1 108 6.3 on on 83.6 7.1 9.33 1 108 6.3 off on 72.9 5.4 21.76 2 108 6.3 on on 89.4 5.5 5.07 2 108 6.3 off on 78.9 17.5 3.67 4 108 6.3 off on 86.9 0.6 12.59 2 91 6.3 off off 83.7 11.5 4.811 2 107 5.3 off on 79.7 16.0 4.314 2 107 5.3 off on 78.0 17.2 4.814 4 107 5.3 off on 87.5 1.7 10.815 1 107 5.3 on on 83.7 16.3 0.016 2 107 5.3 off on 80.0 20.0 0.062Figure 4.1 :Photograph of the UBC pilot plant facilityABCDFeed dehvery tube63Oxygen supplygas burnerair exitFigure 4.2 :Schematic diagram of the UBC flash furnace shaft64Air Bleed I Preheat ExhauatJEZReactor CycloneBaghouaeHeat ExchangerScrubberFigure 4.3 :Schematic diagram of the UBC pilot plant off-gas handling system65ConcentrateOxygenCooling Water InI Cooling Water ReturnI__________________ _____________I I III I IIII IIII IIII IIII IIII IIII IIII IIII IIII IIII I III III I I I OuterBurnerTube:3.81 cmO.D.316SSII IIII I III II IIII IIII IINOT TO SCALE Inner Burner Tube:1.58cm ID.Figure 4.4: Schematic diagram of the UBC concentrate burner.66FillerFlotameterCooling Water InSchematic diagram of gas sampling system as used in the UBC experimentaltrials.GasCooling Water ReturnIR. Analyzer G.C.Figure 4.5:Ii C C) :3- CD 2 C) CD 1 -J CD 1 -J‘C) ‘S 21*a a a a aC *C 2.—a aai IC C) CC’68Lid24 cmFigure 4.7 :Schematic diagram of Solids Sampler Type 1.69Cooling Water In1.3 cm tubing, 316 SSI Ii2.5 cm tubing, 316 SSCooling Water Return1.15m025 mNOT TO SCALEFigure 4.8 :Schernatic diagram of Solids Sampler Type 270I.)C-)>-C-)C-)C)Figure 4.9: Predicted sampling efficiency of non-isokinetic solids samplers as a functionof particle size and velocity.Particle Diameter (microns)71‘1)0I0Cl)Cl)0>zUEquivalent Spherical Diameter (jim)I ‘IlIlilIllIllIllI IlIllilIlIllitIl 11111 _IIIlIIIII;III1IIlIt[;;r:tIr1 initi,ii iiTT ;Il;lIlIIIpII4IIIlIIIIIII;I;IIIlII1IIIIIl:III;,’ I,?,I 1,1,chtI1flL4 ‘II)II1’iiII Ii 111111 I IllliIlIilIilIIIlIllhllliI} iIi1llI1lI iIIi i 4 1 ii III I 1t llihIH 11 It[M1 IAh1llh11I1llhIllhIllllh111fllAl J It I t h V60rn {jj llhiliiIiIl IIWIULI lllllulll{Mllhl{llllhilDulhlflhi{l!I hililil It 1 I’’ I’f ft tllhlI1IIiI 111111111 i{filMIillillllMllh1llllh1I 1111111] Ii I !jII H U701’t$1th1WI1ft llhi{1llh11IMffl4iIi’iffi1ffliUd II B’’ t’‘II. ! lb i.i I — tUfflIffltfflhIII4tllIflttDWtTh4l flfltNtt[tH I I IJT!IIflmM : I 1 1M,’j1 II I I’4 ujiniuifJil!i I itIII!!hi ThlhHi H1I P20 4‘ bi:tIIhij1jLI -100 80 6000 40 30 20 10 8 6 5 4 3 2 I 0.8 0.6 0.5 0.4 0) 02Figure 4.10: Particle size analysis of regular MK concentrate72ci)ci)C,)ci)>UEquivalent Spherical Diameter (urn)Figure 4.11 Particle size analysis of large size fraction of MK concentrate (Prepared byOrtech, Ltd.)-F’1.ii..j1iiIItttlIIIi liiiWIi’UIPL.‘ IulIuititIiitiuItI1IIIitIt1titiIttttItrt1 i IIIIIlIITIIFIIflHIIlfITlI II.iII,I:II!;&II:I——nirI:.,.I..I..,,I:I..1 ii I - - IiI I I’ I I ii t tlllhIll1llllllrnu1t4milluhtllnfl w it1 ii iiI_I.. i II--IbiJlIIiiTii Ill ...,I’IJJlIllIf’ILLIJl L[E•JIII[1I1HIfl[I1IHFJ iIiiilIliiIH:H riJtIlI14W41t1ItiHtHt4 Ill4ttIllhiH :1 It IifflIWflhlIlluulTIIIlIlIllll 11111111111 H I:Jall::HI::!IiBl.• —;:Il!IIII!1HiH::.1 I’I:It:III;tI1II 1IiiiIiiii1iiI1tOO OO 60 50 40 30 20 50 • 5 4 3 2 I 0.8 0.6 0.5 4 0 3 0 273C11.)0I.0Cd)Cd)0>L)Equivalent Spherical Diameter (.tm)Figure 4.12: Particle size analysis of small size fraction of MK concentrate (Prepared byOrtech Ltd.)I iIIIIIIJiI11I1IIfi 1‘‘‘‘‘ !l!I!1TTIlIT1lTJ1’’t IIlIIIIlIIitIIItIIIIlItli° fgllfI If iiiiiiiiiiititiniT-1.i.i:iiii.iiIIIITIflIITIIIIITTTTTtI1 1111““ I’’!”’ I3’I?tt t lJlJ1flflltJ!fli!iFf+ uhlllJllllJJllillllInJrrntl JIllJllIJBIl!IlIillh1iillIJIllllllUIiUJllJllJII1Bi ilitti‘IIH]1I1flhIII_.I—I—IIN4IItIIIIIl IIIHhIllhIIIIllUIItIitftI1 LII LIULIIIIHIIIIII I I I II..i.,.s.i.s 1k.il II l!IIui1IIIlillhi!IilIIIilif1tttI litI.jIIIIIIIIIIIIIIIIIFIILIIIIIt I lIltIllIllIflI 11111111.l.IltH—l1—4 Iil1JlllIHullrt11 1111jIIIIiIIIiIiIII!tI II III II IBiItiIiI1iIiIiILIIitiItIII IIiIiIiiIiiiiiiiIjfflI I I I IIItiiIiiI— I l!IlIt It .iLJIIlItttIiILiILIIi_f4ttiIIJlIIJJII1lfI’I LIlT{IIIIIIIIIIIIIIIIIIIIII III.! I—’ ‘IIIIIIIIIIIIIITtlliiiIIi!I iiiiitniiIiiiit.f ft11—rr.1IIIIJI1IJJII illllhilliItI1J 1J;JJjiiJI I I IIIIlIIlIIIlIIIIIlIIlIIUHIIIIIIIIII1II[II1IIII III IIIIIIIIIIIIIII IT iiiiii ;tiiiIIiIIITI I’lillIliiIiIiil I I IIIII IIII1IIII 111 I301, II’’’! I’’’ I ‘IIIIIIIIIIIIIIII II.I1I I!IiIIIIHiIII! ii IIH-+IIIiii!!iIl IIi:I::II :1TiIflItI’ 11111 III l’IIIIII I‘IIoII’!II’ ii I 1111,IiIIiIIIIIIIiI]I ‘IL I IiiIfillhIllhIf!i 1. I I: IiIIILIIIIHInuI1In4-wIIHIIiII: II tt HIll II. .1II)iIIIIIiIi I I1I.1iIIi IIIi!iIIIIIIIIIIIIIIIIIIIIII[II I I IIIIIIIIIWIIlIItIlt4IIittIIIIIIIIiIIiIIIIiIJJl1itli.IIIIIIIINIIIILiIIIIlI II LII FIIHIlIIIlIIliI IT’I liILIIIi1-T-‘ IJLiiIIIIIIIiJIiiIlII I I I: I IUIIIIIIIIIIII[IIIIIIIIIIILI_f ii iii: 111111Li‘ lIif hull LII III-t-1 111-IIIIIIII IIIIIIIflIIIIIIIII1II1I1I rtrrlttIttlIiIIItlItI IIIIiullLItIIIttiIttlIi’iL4biIjllIiIjIIliIIf tiff I + . I!ltIIIIIIILIILIIlIIIIIIL INIIILILIIIIIIIH‘I; IILIiLIIIIIII-i-ihIIIIIIIIIIIIIIIIuIIIIIIn J 111111 IHIILH III FL III I F II ll III ILIILLIIIILII I I I IIIIHLIIIII IIIhllIlllIllililIflflItttlI liii IIIllilt ii liiiI IIII’’I I’ ‘I“1”U!UIIIIIfIiIlIiIHIIILH IfiIUlill‘ II:LIIIIIIIII;IIIIIII,‘ IIIIIIIIIIIIIIIIIi lIlt I IInii4 + hIIIIIlflIIILIIIIIIIfI1I1 IlllIlIf 11111 fill I 14_I_I ‘ IIffIlIIIfIiIIIfiIIi fill IflllIiIifIIi jjI_ IiIIiIiIiIIIIIIIII LIII!IiIiiiIiiiIIItII I. TIIIIIIIIlIIIIIIIIi’ IIlIllIllIlil I’ 11111’ IlI’ III]ILIIIIIIIII 111111100•0oO50 0I80706050404.)00C-)3020100IOxygen and concentrate feed rates from run of Dec. 13, 1989 (2 kg/mmunsized MK concentrate, stoichiometric oxygen).740 5 10 15 20 25 30 35 40Elapsed Time (miii)Figure 4.13:a)Ia)C-a)HCUa)Ca)Figure 4.14 : Reactor wall temperatures as a function of elapsed time from the experimental trial of Dec. 13, 1989 (2 kg/mm unsized MK concentrate, stoichiometric oxygen ), corrected for the effects of the alumina wool blanket.75Elapsed Time (mm)7694009200900088008600840082008000GC Output Reactor Section *IDI1 40Figure 4.15: Gas chromatograph outpLlt from run of Dec. 13, 1989 (2 kg/mm unsizedM K concentrate, stoich iometric oxygen ).1.4-)z0C-)0 20 40 60 80 100 120Elapsed Time (see)77Figure 4.16: Photomicrograph of dust produced at UBC pilot plant from run of Dec. 13,1989 (2 kg/mm unsized MK concentrate, stoichiometric oxygen).78Metallic Material>90% CuRed-black material(10% 0)Figure 4.17: Sketch of solids distribution from run of 27/4/89. This figure is not to scale,but indicates the general distribution of solids in the hearth at the conclusionof the flash run.Dark powder with the appearanceof unreacted concentrate.> 10 % SNorth SouthFlash Gun Tip LocationI NOT TO SCALE79OO5m [017mS1Figure 4.18: Schematic Diagram of Inco Port Colbome Flash ReactorNOT TO SCALEOxygen Inlet80MK Concentrate In’et91.4cm— 70cmFigure 4.19: Schematic diagram of Inco Port Colbome concentrate burnerWater in/outCollarSchematic Diagram of Gas Sampling Apparatus as employed in the PortColborne Flash Reactor.81Solids RemovalSample Probe1.14 m Long1.0. = 0.635cmSample Vial300 ccVacuum PumpBubbler-lie BackFiushGas InFigure 4.20:82Port #4 Port 112 Port #1_4__Ctre Line of OxygeMKt= Approximate Gas Sampler LocationFigure 4.21 Approximate location of samples in Port Colborne flash furnace8390_________________Oxygen80• NitrogenX Sulphur Dioxide70* Carbon Dioxideti)60ei05 15 25 3Axial Position (m)Figure 4.22: Plot of all gas samples taken from trials in the Port Colborne reactor withthe 6.3 cm burner.848&x7&60 c Oxygen M• NitrogenC-)50 X Sulphur Dioxide X* Carbon DioxideC: 40CI30 a aIC0.1000 05 1’.5 25 3Axial Position (m)Figure 4.23: Plot of all gas samples taken from trials in the Port Colborne reactor withthe 5.3 cm flash burner.855 Chalcocite Reaction KineticsIn order to develop the mathematical model of MK flash converting, it was essential to investigate the thermodynamics and reaction kinetics of chalcocite combustion, with particular regardto potential dust generation mechanisms. Using the literature data available, the kinetic model ofchalcocite combustion was then developed and employed as a subroutine in the overall model ofchalcocite flash converting.5.1 Literature Data5.1.1 ThermodynamicsA review of the thermodynamic literature available on the Cu-S system has beenperformed by Elliott (52). As Figure 5.1 shows, Cu2S melts at approximately 1403 K andis stable at temperatures up to the boiling point of copper (2836 K). Plots of predominance area diagrams for the system Cu-S-O have been produced using the F*A*C*T (55)database and are shown in Figures 5.2 to 5.4. Figure 5.2 shows a plot of the stable phasespresent at 900 K and indicates that the sulphate (CuSO4)and basic sulphates are preferredat low temperatures and high oxygen potentials. At moderate temperatures (1700 K),Cu20 becomes more thermodynamically favoured, while at higher temperatures andlower oxygen partial pressures, copper metal is preferred (Figure 5.4).This information therefore suggests a possible reaction sequence for a chalcocite particlecombusting in the oxygen-concentrate jet of an MK flash smelter:1. The particles are injected into the furnace at a low temperature (298 K). As they heatup to the furnace temperature, sulphates and oxides will tend to form on the surface ofthe particles.862. This oxidation is highly exothermic and results in both more rapid heating of theparticle and a subsequent lowering of the partial pressure of oxygen at the particlesurface due to removal by chemical reaction.3. The increase in particle temperature and the lowering of the oxygen partial pressure atthe particle surface will then cause copper metal to be formed preferentially overCu20. Any Cu20present may react with the surrounding Cu2S to form copper metaland sulphur dioxide.4. When all of the sulphur ‘fuel” has been combusted, the partial pressure of oxygen atthe particle surface will increase and the particle may either cool or react withsurrounding oxygen or sulphur dioxide to form oxides or sulphates.This reaction sequence is consistent with the phenomena observed industrially: a collection of metallic “semi-blister” copper is found in the bath, while the dust removed fromthe smelter consists of copper metal and copper sulphates and oxides. It is highlyprobable however that reaction kinetics and associated heat and mass transportphenomena are equally important factors determining the actual reactions occurringwithin an operating MK flash furnace.5.1.2 Chalcocite Oxidation KineticsA number of investigations (56-58) have been performed on the low temperature roastingof solid chalcocite to copper oxides and sulphates. However, due to the highly exothermicnature of this reaction, earlier studies have been flawed by poor temperature control,frequently causing any measured reaction rates and activation energies to be inaccurate. Astudy by Rao and Abraham (59) of an oxidizing chalcocite pellet has shown that at87furnace temperatures of 1223 K the oxidation reactions were oxygen mass transportcontrolled. A similar observation was made by Asaki et a1. (60), although gas diffusionthrough the Cu20 layer on the particle was also found to be a significant resistance toreaction.To date, the most accurate determination of the activation energy of chalcocite oxidationis that which has been performed by Kim and Themelis (61). These workers utilized acylindrical pellet of sintered chalcocite suspended in a carefully controlled gas stream tomeasure the reaction rate as a function of temperature and particle porosity. An Arrheniusplot of the initial oxidation rates of chalcocite was used to determine this activationenergy, thus eliminating the effects of particle temperature increase on reaction rate. Kimand Themelis found that below 1096 K, the reaction is chemically controlled, with anactivation energy of 125 kcal/rnol. Above this temperature, oxygen mass transport wasfound to limit the reaction rate. Although several different particle porosities were tested,little variation of initial reaction rate with particle porosity was reported, due to the evensurface finish of all Cu2S pellets caused by sintering.The only published study on the ignition of MK concentrate particles was performed byOtero, Brimacombe and Richards (13). These investigators used a stagnant gas furnace tostudy the combustion of both chalcopyrite and MK concentrate particles as a function ofparticle size, oxygen concentration and furnace temperature. In this study, pre-weighed2.0-g charges of sized concentrates were introduced into a preheated tube furnace whichhad been filled with gas at the desired composition. The reacted concentrate passing outof the bottom of the tube furnace was then collected, weighed and analyzed. Very smallparticles such as fume could not be collected by this apparatus. A plot of the results ofOtero et al. for MK concentrate is shown in Figure 5.5; the ‘ignition” temperature of the88MK concentrate can be seen as the sharp inflection points in this graph where the weightloss accelerates suddenly. This temperature was found to be between 1023 K and 1123 Kdepending upon gas composition, but was not strongly affected by particle size. Otero etal. have also shown that the observed mass loss after ignition was frequently very muchhigher than could be explained simply by the loss of sulphur from the sample. Forexample, the lines labelled as “CuO’, ‘Cu20”, and “Cu” in Figure 5.5 represent weightlosses corresponding to a theoretical conversion of the concentrate to CuO, Cu20, and Curespectively. Data points below the “Cu” line indicate that some mechanism other thanthe formation of copper from copper sulphide must be responsible for the observed massloss. It was suggested by Otero et al. that this additional mass loss represents ‘dust”formation, as the sampling apparatus at the bottom of the stagnant gas furnace could notquantitatively capture very fine particulates. As Figure 5.5 shows, this tendency of theMK concentrate to generate dust (as determined by the difference between the observedmass loss and the mass loss associated with conversion to copper) declined withdecreasing oxygen concentration.Otero et al. also utilized high-speed and still photography to observe the combustion ofMK particles after ignition (Figure 5.6). As Figure 5.6 indicates, the MK particles werefound to explode violently into a large number of small fragments, both in oxygen and in50-50S02-oxygen mixtures. When compared with Figure 5.5, it can be seen that theseexplosions are associated with a low weight recovery and hence a high dust generationrate.This contrasts with the observed behaviour of copper concentrates (chalcopyrite) whichtended to fragment (if at all) into a smaller number of larger particles. The reduced fragmentation observed with the chalcopyrite concentrate was explained by the presence of89iron, which may act as a glue’ to bind the chalcopyrite together. The MK particle,lacking this iron would hence tend to fragment more completely. It was suggested thatthis enhanced fragmentation may be a primary cause of the higher dust formation ratesobserved when flash converting MK concentrates.Recently, Warczok et al. (14) have modelled the oxidation kinetics of chalcocite particleshaving a diameter of 150 to 300 microns. They considered that the following reactionoccurred:Cu2S + 1.5°2 Cu20 + SO2 (5.1)The shrinking core model was utilized, and the reaction rate was assumed to be determined by oxygen mass transport, diffusion through the oxide layer, and chemical reaction. Model predictions agreed well with their experimental measurements of oxidation asa function of furnace temperature and oxygen concentration. It was discovered thatparticles 300 microns in diameter could take very much longer to ignite compared toparticles smaller than 150 microns diameter. Peak particle temperatures were predicted tobe near 2000 K, which were compared with measurements of 1500 to 2300 K byJorgensen (21) for chalcopyrite.The only temperature measurements available of combusting chalcocite particles arethose carried out by Otero, Tuffrey, Brimacombe and Richards (62). These investigatorsused a laminar flow furnace and two-colour pyrometer developed by Tuffrey (63) tomeasure the temperature of both chalcopyrite and MK concentrate particles combustingunder controlled conditions. Interestingly, the peak particle temperatures of the reactingMK concentrate were consistently much higher than those reported in the literature forchalcopyrite. For example, Figures 5.7a,b and 5.8a,b show both peak MK particle90temperatures together with the total energy output detected by the pyrometer. In thesefigures, the peak particle temperature is approximately 2800-3000 K, and coincides witha sharp peak in the total energy emitted by the burning particle. This peak in thetemperature and energy appears to last approximately I to 2 ms. It has been reported byTuffrey (63) in his study of combusting lead and iron concentrates that sharp peaks of thistype frequently accompany particle explosion, with ejection of hot fragments and acorresponding increase in the apparent diameter of the particle. It is also interesting tonote that high speed photographs of exploding MK concentrate particles taken by Otero etal. (13) (Figure 5.9) have shown that the observed explosion of the MK concentrateparticles typically takes I to 2 ms to complete, which agrees with this pyrometer data.5.1.3 AnalysisThe experiments of Otero et al. (13) have shown that the fragmentation of the MKparticles is likely to be associated with dust generation, but the mechanism by which agiven MK particle fragments has not yet been elucidated. It is clear that internal pressuremust be responsible for this fragmentation, as the particles were consistently seen toexplode outwards. However, it is difficult to envisage a reaction mechanism whereby gasis suddenly generated at the centre of a reacting chalcocite particle. Decomposition orboiling of the Cu2S is not a likely source of gas, since the available thermodynamic data(64) suggests that Cu2S is a stable liquid at temperatures up to and exceeding 3000 K. Inaddition, the reaction:Cu2S(l) +°2(g) = 2Cu<I) + SO2(g) (5.2)91generates sulphur dioxide, but only at or near the particle surface. The reaction of copperoxide with unreacted Cu2S is a potential candidate for sulphur dioxide evolution within aparticle:Cu2S + 2CuO = 6Cu + SO2 (5.3)However, it is unlikely that sufficient quantities of Cu20 are present in the particle, and itis also probable that this reaction would take place at or near the particle surface. Inaddition, Byerley et al. (65) have shown that this reaction has an appreciable reaction rateeven at quite low temperatures (1473 K), indicating that if this mechanism were responsible, particle fragmentation would probably occur at much lower temperatures thanchalcocite particles are known to attain (62). Nevertheless, this reaction remains as one ofthe few potential sources of gas within a combusting particle.The particle temperature measurements of Otero et al. (62) in high oxygen atmospheressuggest one other possible source of gas within a combusting MK particle: coppervaporization. These measured peak particle temperatures consistently reached 2800-3000K, which corresponds closely to the boiling point of copper at 2836 K. Moreover, thestudy carried out by Tuffrey (63) has shown that combusting pyrite particles are capableof reaching the boiling point of iron, and PbS particles were found to attain the boilingpoint of galena. Particle fragmentation caused by the vaporization of the internal materialhas also been observed for the combustion of aluminum particles (66) and fuel droplets(67,68).The data from the UBC flash furnace does provide some experimental evidence thatsuggests that copper vaporization may play a role in the dusting mechanism. A photomicrograph of the dust recovered from a typical run with MK concentrate was shown earlier92in Figure 4.16, and indicated that the “dust” is mainly composed of agglomerates of manyfine (sub micron) spheres of copper and copper oxide. The physical size and compositionof this dust is therefore consistent with the formation of particulates from copper vapour.In addition, greatly increased (compared to the input concentrate) copper-to-nickel andcopper-to-iron mass ratios were seen in the dust recovered from the cyclone (Table 4.8).This is also consistent with a copper vaporization hypothesis, as nickel and iron would beleft behind by the vaporizing copper, resulting in high relative copper concentrations inthe dust in the cyclone. This mechanism has been described by Munroe (28) for the caseof an Outokumpu furnace, whereby volatile components tend to be concentrated in theOutokumpu flue dust. However, Munroe also states that copper is not normally one ofthese volatiles.Therefore, it is likely that copper vapour may both lead to enhanced particle fragmentation (“dust”) and as a condensed phase may comprise much of the dust generated whenflash converting MK concentrate. To investigate this hypothesis, a mathematical modelhas been developed which calculates the temperature and composition of a combustingMK particle. The primary objective of this mathematical model has been to calculate themaximum temperature which is attainable by a combusting MK particle, and so determine if the boiling of copper inside the particle is a possible means of particle fragmentation. The development and predictions of this model are described in detail in thefollowing sections.5.2 Kinetic Model DevelopmentUtilizing data from the literature as a guide, a model of the combustion of an individual MKconcentrate particle has been developed. The model predicts the temperature and composi93tion of a chalcocite particle as a function of time, furnace temperature, reaction gas composition and particle size. The model considers the case of a single particle combusting in afurnace of constant gas composition. The furnace is assumed to be vertical, with theconcentrate particle introduced at the top (an idealized representation of the apparatus ofOtero et al.(13) ). The motion of the particle is computed as it moves though the furnaceunder the influence of gravity and is restrained by the viscous drag of the surrounding gas.The particle temperature is calculated continuously from a heat balancedT (5.4)C,,=rad + cony + reacThe heat transfer and heat generation due to chemical reaction are calculated from theequations described below.5.2.1 Simplifying AssumptionsThe following assumptions have been made to simplify the development of the kineticmodel:1. The particle is assumed to be a perfect sphere.2. Temperature gradients through the particle are neglected. The Biot modulus of aspherical particle of Cu2S is given by : -. For small particles at low velocities, thec2sNusselt number is close to 2, and hence the Biot modulus in this case simplifies toInserting typical values for these thermal conductivities gives a value for the Biot944. The MK particle is assumed to be composed of pure chalcocite. The presence ofresidual nickel and iron in the concentrate is neglected.5.2.2 Chemical Reactions ConsideredFive separate chemical reactions are assumed to describe the phenomena occurring in areacting chalcocite particle, depending upon particle temperature, and consistent with thepredominance area diagrams of Figures 5.2 to 5.4.1. At temperatures below the melting point of Cu2S only reaction (5.1) is assumed tooccur- the oxidation of solid chalcocite to solid Cu20. The rate of this reaction isassumed to be controlled both by chemical reaction and by oxygen mass transport asdescribed in section If the particle reaches the melting point of Cu2S, the reaction of liquid Cu2S withCu20 (5.3) is assumed to occur instantaneously, producing metallic copper andsulphur dioxide. Byerley et al. (65) have shown that this reaction has a finite rate atthis temperature (1403 K), and hence this assumption is not completely accurate.More probably, the Cu20reacts with the Cu2S over a temperature range at a rateincreasing with particle temperature. However the effect of this approximation on theparticle temperature predictions is likely to be insignificant, since these calculationshave shown that only a very small fraction (<5 %) of Cu2S reacts to Cu20 before themelting point is reached in ignited particles.953. After all of the copper oxide is removed by reaction (5.3), the copper sulphide isassumed to react directly to metallic copper according to reaction (5.2). The rate ofthis reaction is assumed to be controlled only by oxygen mass transfer. If the particlereaches the boiling point of copper, the rate of boiling is calculated continuously.4. Once all sulphur had been removed, the copper in the particle was allowed to reactwith surrounding oxygen (if present). Based on the anaysis of the Cu-O system bySchmid (69) and the Cu-O phase diagram shown in Figure 5.10, this reaction ofcopper with oxygen was assumed to occur as followsa) At oxygen concentrations below the miscibility gap, oxygen was simplyassumed to dissolve in the liquid copper:02(g) = 2[Oi (5.5)b) Beyond the miscibility gap shown in Figure 5.10, Cu20 was assumed to formby:2 Cu + 0.5 02 = Cu20 (5.6)5.2.3 Calculation of Reaction Rates5.2.3.1 Reaction 5.1 : Low temperature oxidation to Cu20The overall low temperature oxidation rate of Cu2S to Cu20can be calculated by acombination of the resistances provided by chemical reaction rate and oxygen masstransport rate96dncc 1 (5.7)— dt=(._(—)Chemical Reaction Rate:The work by Kim and Themelis (61) has indicated that the low temperature chemicalreaction of chalcocite with oxygen is likely to be first order in oxygen. Therefore thechemical reaction rate, Rrcc, can be calculated by:Rrea(. = kTPO2APXCi4S (5.8)In this equation, the surface area of the particle available for oxidation is reduced bythe fractional conversion of sulphide to oxide following the model of chalcopyrite smelting developed by Hahn and Sohn (37). This parameter represents anattempt to model the reduction in area caused by the coating of the particle surfacewith oxides. Thus the product A represents the surface area of the particlewhich is available for chemical reaction. Following the suggestions of Hahn and Sohn(37), this parameter was initially assumed to be equal to the volume fraction ofsuiphide in the particle:1c2sMcs (5.9)Pc2s= flc2oWc nc2sMcs+PCu2S97The chemical reaction rate constant k, was calculated from an Arrhenius expression:- (5.10)kr=Ae RTAn activation energy of 125 kcal/rnol was used for this reaction, as suggested by Kimand Themelis. An estimate of the value of the pre-exponential constant, A, was alsodetermined from the data of Kim and Themelis. At 1068 K, the initial reaction ratewas observed by Kim and Themelis to be chemically controlled, and was reported tohave a value of approximately 5 x g S021s. Substituting this reaction rate intoEquation (5.10) together with the activation energy and the values of surface area,temperature, and partial pressure of oxygen provided by Kim and Themelis gives apre-exponential constant of 2.52 x i0’ moles m2 Pa’ srn’.Oxygen Mass Transport Rate:The oxygen mass transfer rate to the particle, RDOX, was assumed to be determined bytwo combined rate effects : diffusion of oxygen from the bulk phase through theboundary layer to the particle surface, and diffusion of oxygen through the oxide film(if present) coating the particle:i ( i (5.11)C -C iiR \ 2 2’k Dox,h 2° ox,Cu20 ,The oxygen bulk phase mass transport coefficient kOX,b was calculated from theSherwood number Sh = which was predicted from the following correlation98Sh =2+0.6Re112Sc3 (5.12)The diffusivity was estimated on the basis of equimolar counter-diffusion of oxygenand sulphur dioxide and was predicted by the Chapman-Enskog relation (70):BTflM12 (5.13)DOXb= PJr2The diffusivity of oxygen through the oxide layer was estimated based on experimental values determined by Asaki et al (60) (1.5 x i0 m2/s at 1060 K ). The radiusof the oxide layer at any time was calculated assuming that all of the Cu20accumulated at the surface:( v0”3 (5.14)r0=1—VP J r1,, Reaction 5.3: High temperature oxidation to metallic copperOnce the particle had become molten, the reaction rate was assumed to be limitedonly by the rate of oxygen mass transfer:dnc s (5.15)U2 c— dt P c2sL-O °2As written, this equation assumes that the surface area available for chemical reactionhas been reduced by the parameterX- the fraction of the particle surface area thatis covered by suiphide. However, in this case the influence of the parameter Xc islikely to be small, since in a molten particle both the copper and the sulphur in the99particle are able to travel quite freely. This parameter therefore becomes an approximate description of the relative tendency of the copper sulphide to migrate toward theparticle surface. Equating to the volume fraction of copper sulphide in theparticle (as was done with a solid particle) is equivalent to assuming that both thecopper and the copper sulphide in the particle have equal tendencies to remain at theparticle surface. However, for the case of copper-sulphur melts having a low sulphurconcentration at 1573 K, it has been shown by Elliott (52) that the concentration ofsulphur at the surface of the melt is approximately 600 times that in the bulk phase.Elliott (52) also reports thatContrary to what might be expected because of the relatively high concentration of asurface-active solute spccies on the surface of a metal, the time that is required to formthe equilibrium layer of surface-active atoms [sulphur in this case] on a freshly formedsurface of a liquid metal is relativel’ small. If the diffusion coefficient of the species inthe bulk phase is of the order of 10 cm2/s, the formation time is in the range of 10-100IIS.Studies of oxidation of (unstirred) liquid copper suiphide melts by low velocity airjets blown onto the melt surface (73) tended to confirm this by showing that theprimary resistance to oxidation is due to oxygen mass transfer in the gas phase, and isnot influenced by depletion of copper suiphide from the surface of the melt. In addition, surface tension differences in a melt have been shown by Brimacombe andco-workers (74,75) to greatly increase liquid mass transfer rates due to stirring causedby the Marangoni effect.This information indicates that ftr the case of an MK concentrate particle, suiphideoxidation probably occurs over the entire particle surface area, even at low suiphideconcentrations. Clearly the suiphide has a greater tendency to remain at the surface ofthe particle and possesses the ability to move rapidly to the surface. Therefore, in thecalculations presented below, the parameter in equation 5.13 was simplyassumed to be 1.1005.2.3.3 High Temperature Reaction of Copper with OxygenThe rate of reaction (5.5) was assumed to be limited only by oxygen mass transport,with the equilibrium partial pressure of oxygen at the surface of the particle calculatedfrom data presented by Schmid (69). Once Cu20 was formed as a separate phase, therate of reaction (5.6) was determined as follows1. Above the melting point of Cu20 (1500 K) the rate of reaction (5.6) was assumedto be limited by oxygen mass transport.2. Below the melting point of Cu20 but above the melting point of copper (1356 K),the reaction rate was assumed to be determined by the combined effects of oxygenmass transport and diffusion through the solid Cu20 in a manner similar to Equation (5.7). The Cu20 was assumed to accumulate on the surface of the particle.5.2.4 Heat TransferRadiative heat transport to the particle was calculated from the following equation:Qrad GE,,FA(T — T) (5.16)while conductive/convective heat transport was calculated by:Q(.c,n=hAp(TgTp) (5.17)The convective heat transfer coefficient was calculated from the Nusselt number,hi)Nu = using the following correlation applicable to spheres:Nu = 2.0+0.65RePr3 (5.18)101It has been shown by Jorgensen (19) that for particles smaller than 1mm, the convective/conductive heat transport mechanism dominates, while radiation only becomes significant for larger particles. For example, Figure 5.11 shows a plot of the relative magnitudesof convective and radiative heat transfer as a function of particle size and particletemperature, assuming equal gas and wall temperatures of 1400 K. The conductive heattransfer coefficient was calculated assuming a Nusselt number of 2, while the radiativeheat transfer assumed the view factor and emissivity to be both 1. This figure clearlyindicates that conduction will dominate for small (<< 1mm) particles while radiationincreases in importance with increasing particle size. The reasons for this are relativelysimple : assuming a Nusselt number of 2, the convective heat transport equation willdecrease with increasing particle size, i.e. heaD’, and hence the overall rate of heatconduction, hAct(D1)(D2)will vary linearly with diameter. In contrast, the radiativeheat transport term, given by Equation (5.16) will increase according to the square of theparticle diameter. At higher wall temperatures, the radiative effects become significant atsmaller particle diameters, but even assuming gas and wall temperatures of 1800 K,Figure 5.12 still shows conduction to dominate for particles below 150 microns (theaverage diameter of the MK concentrate is 20-25 microns).In view of this fact the radiative heat transfer term was therefore treated quite approximately, with the particle emissivity and view factor assumed to be 1.0 for all of thecalculations. It is not likely that this assumption had adverse effects upon the kineticmodel predictions, since the maximum furnace temperature tested was only 1273 K.Furthermore, this assumption will tend to over-predict the rate of heat loss from theparticle at high particle temperatures. As a result, if the particle is predicted to attaintemperatures above the ambient furnace temperature with this approximation, then thetrue values for particle emissivity and wall view factor must result in identical or higher102particle temperatures being predicted. As one of the primary objectives of this model wasto determine whether or not a combusting particle is capable of reaching the boiling pointof copper (2836 K), this assumption provides a severe test of the ability of the particles toreach this temperature.5.2.5 Heat EffectsThe effect of temperature and composition on all gas and particle properties (thermalconductivity, viscosity, heat capacity) was considered in the calculations, while the gaswas assumed to behave ideally. The effect of temperature on the heat released by thechemical reactions was also considered, altering these values by:CT T (5.19)Al-I = Al-I298— J CPprodii.cdT + J CPreacta,dT298 298The heat capacities of each of the various species involved in the chemical reaction arepresented in Table 5.1, while the Al-I values of the various chemical reactions arepresented in Table 5.2. These values were obtained from references (55,64,69,77).Table 5. 1 : Heat CapacitiesChemical Species Heat Capacity (J/mol K)CU(S) 22.6 + 0.0063 TCu(I) 31.3Cu2S T < 376 K : Cp = 81.6376K<T<623K : Cp=97.3T>623 K :Cp= 85.0Cu20 62.3 + 0.0239 T°2 29.96 + 4.18 x 103T - 1.67 x l0 T2SO2 46.19 + 7.87 x i0 T - 7.70 x iO T2103Table 5.2: Heats of ReactionReaction AH kJ/moleCu2S + 1.5 02 = Cu20 + SO2 -398.3Cu2S + 2Cu0= 6Cu + SO2 +5 3.5Cu2S + 02 = 2Cu ± SO2 -229.9°2 = 2{01 -74.12Cu±0.50=Cu0-157.8The changing composition of the particle due to chemical reaction was followedthroughout the calculations to re-calculate particle volume, surface area and densitycontinuously.5.2.6 Copper VaporizationThe vaporization of copper and copper compounds below the boiling point has beenstated by Jorgensen (18,20,21) as limiting the temperature attained by reacting chalcopyrite particles in air to a maximum of about 2273 K. This volatilization during the combustion of chalcopyrite or copper matte has been considered to limit particle temperatures inthe models of Hahn and Sohn (38), and Warczok et al. (14). Typically, the rate ofvaporization has been described by a mass transfer equation, assuming that the partialpressure of copper at the surface of the particle is equal to the vapour pressure of copperat the particle temperaturedn (5.20)-=A b(CCUh — C)104where the total surface area of the particle, A, is frequently used in this calculation.However, Hahn and Sohn (37) reduced the total particle surface area for copper vaporization in Equation (5.20) by a surface blocking factor fç which was assumed to be equal tothe volume fraction of suiphides present in the particle.Equating the rate of copper leaving the surface of the particle to the mass transport ratemay omit a significant additional resistance: the copper volatilization rate itself. As hasbeen shown elsewhere, (72) the steady-state volatilization rate of a substance can becalculated from the Langrnuir-Knudsen equationRvap = (2RMc TP)”2AP(P(U—P;) (5.21)Therefore, depending upon the difference between the partial pressure of copper at theparticle surface and the vapour pressure of copper at a given temperature, the rate ofvolatilization from the particle surface may be predicted. In effect, by assuming that theonly limitation to copper volatilization is due to mass transport alone, previous investigators are assuming that the rate of equation (5.21) is fast compared to mass transport.However, it is clear that if the partial pressure of copper near the surface of the particle istruly exactly equal to the vapour pressure of copper at the temperature in question, thenthe net rate of Equation (5.21) would be zero, and no copper could vaporize at all(equilibrium conditions would prevail). Clearly, the actual copper vaporization rate fromthe particle surface is determined by the combined effects of mass transport and thevolatilization rate determined by the Langmuir-Knudsen equation.The area term in the Langmuir-Knudsen equation again causes some concern, as it isunclear what fraction of the particle surface is occupied by copper. As describedpreviously, employing a surface blocking factor and assuming this to be equal to the105volume fraction of suiphide may over-estimate the true area available for copper vaporization, due to the greater tendency of the copper suiphide to migrate to the particlesurface. It is therefore quite likely that a very large fraction of the particle surface area iscoated with suiphide at all times, and hence the total area available for vaporization isprobably much lower than the entire particle surface area.In the kinetic model of chalcocite combustion, which has been developed for this work,the copper vaporization effects are described as follows1. The volatilization rate of copper from the particle is calculated from the Langmuirequation (5.21). The area term in this equation was assumed to be determined by:A = where X is the fraction of the particle surface area occupied by copper.This value is not truly independent of Xç (previously assumed to be I in Equation(5.15) ), since +X = 1. However, independent values of X were tested in thecalculations in order to determine the effect of this parameter. The sensitivity of thecalculations to this factor has been investigated and is summarized below. The vapourpressure of copper as a function of temperature is calculated from thermodynamicdata tabulated in (64).2. The diffusion of the copper away from the particle is calculated according to theChapman-Enskog theory (Equation (5.16)). In this case, the copper vapour is assumedto be distributed over the entire surface of the particle. The Sherwood number iscalculated according to Equation (5.12), and the copper is assumed to diffuse freelyaway from the particle. The equimolar counter-diffusion occurring between theoxygen and sulphur dioxide is assumed to have no effect on this diffusion rate. Theambient copper vapour concentration in the surrounding gas is assumed to be zero.1063. The overall vaporization rate is then predicted by equating the copper volatilizationrate to the rate of diffusion of copper away from the particle (along with the assumption of quasi steady state). The only unknown in this calculation is the partial pressureof copper at the particle surface. When this value is determined, the overallvaporization rate can be calculated and hence the rate of heat loss due to vaporizationpredicted.This prediction of copper vaporization rate has been performed assuming that steady stateis maintained as the particle heats up, allowing a balance between volatilization rate anddiffusion rate to be performed. However, with the rapid temperature increases caused byparticle ignition it is possible that periods of unsteady-state behaviour occur and that theoverall copper vaporization rate may be somewhat lower than that predicted on the basisof an analysis of this type.5.2.7 Solution AlgorithmThe solution of Equation (5.4) was performed numerically using simple Euler integration(a fourth order Runge-Kutta integration was tested, but provided no additional accuracywith the small time steps employed in the integration). The computer program used tocarry out this integration is called ‘kmodel’, and a flowchart of this program is shown inFigure 5.13. Calculation proceeds from the entry point of particles into the furnace untilthe furnace exit, or until all of the sulphur or copper in the particle has been combusted. Ifthe particle attained the boiling point of copper then the rate of copper vaporization wascalculated.Very small time steps are employed in the model so that the particle temperature change107between successive calculation times did not exceed one degree Kelvin. This ensureedthat iteration within a time step did not need to be performed due to any change of material properties with temperature. The temperature change between successive time stepswas continuously monitored, and the time step was adjusted within the program tomaintain this low rate of temperature increase.5.3 Kinetic Model Calculations5.3.1 Verification of Activation Energy and Pre-Exponential ConstantTo verify both the computer program and the assumed values of activation energy andpre-exponential constant, the kinetic model predictions were compared with the mass lossdata of Otero et al. (13) in an attempt to predict the ignition temperatures determined bythese investigators. To make the predictions, the kinetic model was run with particle sizesof 10 to 100 microns, and the resulting series of computed mass loss values were averaged to give an estimate of the mass loss to be expected from the as-received concentrate.In these calculations, the parameter X, - the fraction of the particle surface from whichcopper may vaporize- was assumed to be equal to the volume fraction of copper. Thisparameter was found to have no effect on the predictions of the ignition temperature,either in oxygen or in air.The predictions of the model for the combustion of chalcocite in air and in oxygen areshown in Figures 5.14 and 5.15 respectively. These figures illustrate that the model iscapable of predicting the ignition temperature of the MK concentrate, thus simultaneously108verifying the computer code and also indicating that the values of the pre-exponentialconstant and activation energy obtained from the data of Kim and Themelis (61) arereasonably applicable to this system.The model appears to be less successful in predicting the total mass loss after ignition,especially for the case of the combustion in oxygen (Figure 5.15). In both cases, theexperimental points record a mass loss greatly in excess of that which can be predicted bythe combined loss due to desuiphurization and copper vaporization. However, the dashedline of Figure 5.15 indicates points in which the predicted particle temperatures havereached the boiling point of copper, indicating that this mechanism may be responsiblefor the discrepancy. It is significant that the point of departure between the model predictions and the experimental data closely corresponds to the onset of the region at which thecopper within the particles is predicted to boil. The boiling of copper inside the particle asa potential mechanism of particle fragmentation will be discussed in greater detail below.Note: Due to sampling errors, the measurements of Otero et al. (13) showed a consistent mass loss of approximately 10 % prior to ignition. To facilitate comparisonwith the mathematical model (and following the practice of Otero et al.), theeffects of this sampling error have been removed from the experimental datapoints plotted in Figures 5.14 and 5.15. This slight adjustment had no effect onthe location of the ignition temperatures, either in oxygen or in air.5.3.2 Particle Temperature Predictions5.3.2.1 Predictions in Oxygen and Dusting MechanismThe predicted particle temperatures for a 20 micron chalcocite particle combusting in109pure oxygen is shown in Figure 5.16 and therefore copper vaporization below theboiling point does not limit the peak particle temperature predicted by the model.Even assuming that the copper was capable of leaving from the entire surface of theparticle at all times (setting X, = 1) had no effect- the combusting particle continuedto attain the boiling point of copper (2836 K).Based on the literature survey and the particle temperature predictions of the model,the following observations can be made1. Otero et al. (13) have observed that as-received MK concentrate particles explodeviolently in both 100 % and 50 % oxygen atmospheres.2. Utilizing a two-color pyrometer Otero et al. (62) have measured the temperaturesof combusting MK particles in oxygen. These measured peak particle temperatures compare closely to the boiling point of copper ( 2836K).3. Figure 5.16 shows that MK concentrate particles combusting in oxygen will reachthe boiling point of copper.4. Figure 5.15 shows that the predicted onset of copper boiling in oxygen corresponds closely to the increased mass deficit measured by Otero et al. (13).Therefore, it follows that the explosion of the MK particles is due to the boiling ofcopper inside the particle.1105.3.2.2 Predictions in 50 % 02 and Copper Vaporization EffectsAssuming that copper boiling does lead to explosions, then MK particles reacting in a50 % oxygen atmosphere would also be expected to attain the boiling point of copper,since the data of Otero et al. indicates that these conditions lead to particle explosions(Figure 5.6) and an increased mass deficit associated with dusting (Figure 5.5).A plot of the predicted temperature of a 20 micron particle reacting in a 50 % oxygenatmosphere is shown in Figure 5.17 and indicates that the parameter X (fraction ofsurface available for copper vaporization) has a strong effect on the calculations.Setting X equal to unity or equal to the volume fraction of copper limited thepredicted peak particle temperature to below the boiling point of copper. Either ofthese assumptions produces similar results, since the vaporization of copper onlybecomes significant at high particle temperatures, which are in turn associated withhigh volume fractions of copper. Therefore merely setting the surface coverage factorequal to the volume fraction of copper (as has been done by Hahn and Sohn (38))is essentially equivalent to assuming that the surface is completely coated by copperat all times. Clearly, either of these two assumptions greatly over-estimates the actualconditions1. As outlined previously, there is a sound experimental and theoretical basis tosuspect that the copper sulphide would tend to migrate rapidly toward the particlesurface in Equation (5.15) has been assumed to be 1). This presence of thesuiphide at the particle surface would prevent much of the copper within theparticle from vaporizing until most of the suiphide had been removed. As a resultof this, the parameter X, is likely to remain close to zero until all of the suiphidehas reacted.1112. The link between the boiling of copper and the particle explosions has been established by the calculations for particle combustion in oxygen. Figure 5.17 indicatesthat the only way in which these particles can reach the boiling point of copper ina 50 % oxygen atmosphere (and produce the observed explosions) is if X is setto some very small value (0-0.05).Therefore, it is clear that copper vaporization below the boiling point cannot controlthe particle temperature for the case of particle combustion in 50 % oxygen, andhence the surface coverage parameter is likely to be close to zero. Predictions in Air and Particle Size EffectsAssuming that X, is zero leads to the predictions of particle temperature shown inFigure 5.18 for particle combustion in air. This figure shows a strong particle sizeeffect for combustion in air: particles 20 microns and smaller will reach the boilingpoint of copper, while those greater than 20 microns will not. The peak temperaturesattained by a reacting particle have been plotted as a function of particle size for air,oxygen, and 50% oxygen atmospheres in Figure 5.19.Figure 5.19 may also provide an explanation for the relative degree of mass lossobserved by Otero et al.(13). As Figure 5.5 shows, the dust” measured by the experiments of Otero et al. (indicated by mass loss in excess of conversion to copper)declines with decreasing oxygen concentration : for combustion in air, a greater massof solids is recovered than in 50 % or 100 % oxygen. Figure 5.19 indicates that thismay be due to particle size combined with the effect of oxygen concentration, since alarger fraction of the as-received MK particles will explode in oxygen than in air. For112example, the particle size distribution of the as-received MK concentrate has beenplotted in Figure 4.10, and indicates that approximately 50% of the particles aregreater than 20 microns in diameter. Based on Figure 5.19, all of the particles largerthan 20 microns would be expected to explode (ie. reach the boiling point of copper)when combusted in oxygen, while none should explode in air. This would thereforeaccount for the reduced dusting tendency shown in Figure 5.5 for combustion in airrelative to oxygen.Setting XQ = 0 and re-calculating the mass loss data of Otero et al. (originally plottedfor X, volume fraction suiphide in Figure 5.14) leads to the model predictions ofFigure 5.20. In this plot, the dashed line indicates a region in which particles 20microns and less are predicted to attain the boiling point of copper. As with Figure5.16, this plot shows that the predicted onset of the boiling of copper coincides closelywith the increased mass loss (‘dust”) observed by Otero et al. DiscussionThe measurements of Otero et al. (13) have shown that, at sufficiently high furnacetemperatures, reacting MK concentrate particles appear to lose much more mass thancan be explained by desulphurization. These investigators have shown that thisapparent mass loss is due to the production of very fine particulates and fume whichcould not be captured by their sampling apparatus. This production of fine particulateswas shown to be associated with particle fragmentation.The calculations of the kinetic model have shown that the particle fragmentation isconsistent with copper boiling within a combusting MK particle. The model has also113indicated that the fragmentation of the particles must be responsible for most of themass “loss” recorded by Otero et Al., and is not due to any losses caused by vaporization of copper below the boiling point. This is illustrated by Figures 5.14 and 5.15 inwhich the maximum possible copper vaporization rate from the surface of the particle(X = 1) resulted in predicted mass recovery far in excess of that actually recorded byOtero et Al.Due to their very small size, fragments resulting from the particle explosions couldpossess an enhanced copper vaporization rate. As the Kelvin equation (78) indicates(p “i (5.22)mi— 1=P°} RTp,rpsmall diameter particles (r << I O m ) will have a vapour pressure in excess of thevalue determined by temperature alone (P°). Calculations made with Equation (5.22)suggest that this enhanced vaporization is only significant for particles much smallerthan 1 micron in diameter. This could therefore help to explain the high copper/nickeland copper/iron ratios observed in the dust recovered from the UBC flash furnace:exploding particles result in many fine fragments from which the copper preferentially vaporizes. As the off-gas is cooled, the copper vapour condenses and thecyclone dust is thus composed of small spheres of copper and copper oxide.The particle temperature predictions shown in Figures 5.15 to 5.17 are considerablyhigher than those presented by Warczok et Al. (14) for chalcocite combustion, whichmay be a cause for concern. However, it is possible to show that the higher particletemperatures which result from this work are due to introduction of the resistance tocopper vaporization imposed by the Langmuir equation, and are not due to errors in114the calculations. For example, setting X, to 1, neglecting the resistance to coppervaporization as described by the Langmuir equation, and assuming that the coppervaporization rate can be predicted by Equation (5.18) alone (oxygen mass transport)results in predicted peak particle temperatures of approximately 2000-2300 K(depending on particle size and furnace temperature). This agrees well with thepredictions for chalcocite combustion presented in the literature. Therefore withoutthe rate limiting imposed by the Langmuir equation, the rate of copper volatilizationwould greatly limit the peak temperature attained by a combusting particle, asreported by Jorgensen (18,20,21) and Warczok et al. (14).The high temperature oxidation of copper was not found to be a significant factoraffecting the ability of particles to attain the boiling point of copper. Under noconditions were reactions (5.5) or (5.6) found to increase the predicted peak particletemperature significantly over that achieved by the oxidation of sulphur alone. Rather,particles cooled slowly as the copper reacted to the oxide.5.3.3 Particle Fragmentation MechanismAssuming that the formation of copper vapour within the particle is responsible forparticle fragmentation, the actual mechanism by which this fragmentation occurs is notimmediately obvious, for if copper vapour is generated in the particle it will be formed ata partial pressure of 1 atmosphere, and hence it would seem that no driving force forparticle fragmentation is present. However, if copper vapour continues to be formed inthe particle by continued addition of heat, then one of two events must occur:1151. The surface tension of the particle is strong enough to confine the gas, and the particlecontinues to increase in temperature while the saturated vapour within the particleincreases in pressure and temperature. This condition is analogous to boiling a vesselfull of water with the lid tightly fixed. Ultimately, with a sufficiently high rate of heatinput, the pressure of the confined gas will increase until the particle is no longercapable of containing it. The particle walls will rupture catastrophically, ejecting fragments in all directions.2. The particle walls are weak and expand outward with the gas at a rate sufficient tomaintain a low pressure (approximately 1 atmosphere) within the particle as thecopper continues to boil.Calculations made with the program ‘kmodel’ have shown that either condition leads tosimilar results. For example, the maximum pressure which can be generated within aMK particle has been estimated based on the total amount of copper which could beboiled from a combusting particle. This pressure (assuming that the copper vapour iscompressed within the whole of the particle diameter) is in excess of 70 atmospheres for a20 micron particle. It is most unlikely that a molten droplet could successfully resist pressures of this magnitude for long.Calculations of rates of copper vapour production from a boiling particle have indicatedthat the rate of volume increase is of the order of magnitude of several particle volumesper microsecond. Any particle expansion at this rate would be virtually indistinguishablefrom an internal explosion.1165.3.4 Prediction of Dusting Diagram for MK ConcentrateAssuming that the boiling of copper within the particle is responsible for particle fragmentation and therefore dust formation, it is possible to use the kinetic model to predict,as a function of particle size and oxygen concentration, the furnace temperature at whichan ignited particle will reach the boiling point of copper. This data has been plotted inFigure 5.21. In the region labelled ‘dust’, the particles reach the boiling point of copper,while in the region labelled ‘no dust’ they do not.This figure may help to explain the different behaviour observed between the UBC pilotplant and the Inco Port Colborne pilot plant. For example, the majority of the MKparticles in the UBC furnace combust relatively far from the burner tip, in a region of lowoxygen concentration and relatively low furnace temperature, and hence tend not toproduce dust. On the other hand, particles which react in the Inco Pilot Furnace tend to doso quite close to the burner tip in a region of relatively higher reactor temperature andoxygen concentration, and hence may have a greater tendency to produce dust. This analysis is considerably simplified, since the actual path that an individual MK concentrateparticle takes while moving through each of these flash furnaces is very difficult todescribe by an isothermal plot of this type, for a fixed oxygen concentration. In reality theconcentrate particles are subjected to continuously varying temperature and gas concentration conditions which change both radially and axially. Nevertheless, it is clear fromthis figure that if the MK concentrate is subjected to both high temperature and highoxygen concentration, it will have a greatly increased tendency to produce dust.117Atomic Percent Sulfuro 10 20 30 40 50 60 70 80 100I1400L1+L1200r” uos°c 2019• 1130±2°C1067°C ‘L2+L3813°C6 800-Z241l)1G) 6002a56_________________-507±2°C4-35°C400200 03±2°C-1032.±05°C75±3°C-415°C 11522°C— (Cu) Ch- An (8)0-0 10 20 30 40 50 60 70 80 00 100Cu Weight Percent Sulfur SFigure 5.1 :Cu- S Binary Phase Diagram From (54).118-10.0-10.00.0log(p0log(pSO2) 0.0Figure 52: Cu- SO2- °2 Predominance Area Diagram. T = 100() K. (55)1190.0log(p02)-10.00.0-10.0 log(pSO2)Figure 5.3: Cu- SO2- °2 Predominance Area Diagram. T = 1500 K. (55)1200.0-10.0-10.0 0.0log(p02)log(pSO2)Figure 5.4: Cu- SO2- °2 Predominance Area Diagram. T 2200 K. (55)Figure 5.5: MK mass loss data from Otero, Brirnaconibe and Richards (13), foras-received concentrate reacting in a stagnant gas furnace at varying temperature and oxygen concentration.1212.0Exp lossesCu 0Cu20Cu-oa)000.a,3E(I)I02 SO2 N2- r• 21— 790 50— 50V (0— 90a 90— 1050 50—A 90 (0—S (0 90—I.0—0.5—0—3000500I I I700Reaction gas temp (°C)900122Figure 5.6: a. Still photograph of MK concentrate reacting in 100 % oxygen.b. Photograph of MK concentrate reacting in 50 % oxygen.From Otero et a!. (13)a. b.MK (2ms)123Figure 5.7: High-speed photographs taken at I millisecond intervals showing anexploding MK concentrate particle. (From Otero, Brimacombe and Richards(13))MK COrns)MK (Ims)MK (3rns)>0E>‘C-124(a)Time (milliseconds)(b)Figure 5.8 : Data from Otero et al. (62)a. Particle temperature of combusting MK particleb. Energy emitted by combusting MK particle (as indicated by pyrometeroutput signal).C>‘C-450040033503a3003C-25002000125UTime (milliseconds)(a)Time (milliseconds)(b)Figure 5.9: Data from Otero et al. (62)a. Particle temperature of combusting MK Particleb. Energy emitted by combusting MK particle (as indicated by pyrometeroutput signal).12614O I S I-5 -4 -35 -3 -2 -1 Q Iog(p02/ tm)5’I I 5’I I—.__ IiI I———_ I II——-———,— IIII III j I II ‘ I II0 it I1343 ‘ ‘ otoI ‘ Ill‘ ii vR€431III l Ref 32I ‘I \I . .Ref.33Øp021otmJ/ oRef34I. I I, L,Ret 35- 7 I I‘1 “‘_cc’ I I I.I -, po01otmRet27I / “ I 02-1otm J0 II/‘ II‘ I I/ I.o1250 iI-spinottIine‘ II Io127c 4 1228o I 1222II_____________I’I I\ o.og70 03J81LI2.D0 — I-I0(I0 I I12 Cut)The I at p24.4atmI 1 Cu20-0L Cut)11339O3 4 0.3825 C.02(g,) -10650.0170 Cu20 • CuO1L.(Cu) (Cu) Cu70I I I_______________________—Cu 01 02 0.3 Cu20 04 CuDMe frtn oxygen. N0Figure 5.10 :Cu-O Phase Diagram From Schmid (69)127Figure 5.11: Comparison between relative magnitudes of radiative (wall-particle) andconductive (gas-particle) heat transport mechanisms as a function of particlesize. Tgas = Twall = 1400 KC’,Particle Diameter (microns)128Figure 5.12: Comparison between relative magnitudes of radiative (wall-particle) andconductive (gas-particle) heat transport mechanisms as a function of particlesize. Tgas = Twall = 1800 KParticle Diameter (microns)129Figure 5i3: Flowchart of computer program ‘kmodel’.Read in initial data9O0 idooFurnace Temperature (K)Figure 5.14: Comparison between the mass loss measurements of Otero et al in air andpredictions of the kinetic model. (X = volume fraction copper)1302.5-21.510.5I>CC’)Cl)DCCModel Predictionsj Data of Otero et al.7O0 8O0 11’OO 12’OO 13009O0 idooFurnace Temperature (K)Figure 5.15: Comparison between the mass loss measurements of Otero et al. in oxygenand the predictions of the kinetic model. (X = volume fraction copper)1312.521.51•0.5>CC,,Copper within particlesgj boils beyond this point————DIIID- Model PredictionsData of Otero et al.7ö0- 8öö 1100 12b0 1300I132Time (s)Figure 5.16: Particle temperature predictions for a 20 micron particle combusting inoxygen at 1213 K, showing the effect of the surface covering parameterFigure 5.17 : The effect of the parameter XCu on particle temperature predictions forcombustion in a 50 % oxygen - 50 % sulphur dioxide atmosphere at 1213 K.13311)IcL)ITime (s)134Figure 5.18: The effect of particle size Ofl particle temperature predictions for combustion in air at 1213 K (assuming that = 0).Particle Diameter (microns)135Figure 5.19: Predicted peak temperatures of particles reacting in oxygen, air and 50 %oxygen as a function of particle size.Particle Diameter (microns)960 idooFurnace Temperature (K)Figure 5.20: Comparison between model predictions and mass loss data of Otero et al.(13) assuming that X = 0.2.521.510.5Copper in particlesboils beyond this point-e1O)>0(IDModelPredictionsj Data of Otero et al.1361300760 860 11100 12b0I137Mole Fraction OxygenFigure 5.21 : Predicted dustin2 diagram for MK concentrate combustion, as a function ofparticle size, temperature and reactor composition. (X = 0)1386 Mathematical Model of the Chalcocite Flash Flame6.1 Model ObjectivesAs stated previously, the objectives of the mathematical model, which has been developedfor this study, were as follows:1. To analyze the experimental data produced in the UBC pilot plant in order to determinecritical phenomena affecting the flash smelting of chalcocite concenuate, with particularemphasis on the investigation of the dust generation mechanism.2. To utilize this information to suggest improvements in burner and smelter design whichcould lead to a reduction in dust generation.6.2 Model DescriptionFigure 6.1 shows a description of the overall problem which is solved by the mathematicalmodel which attempts to predict the heat, mass, and momentum transfer coupled with chemical reaction occurring in a two-phase confined, non-isothermal turbulent jet, of mixed gascomposition. The mathematical model comprises some 4000 lines of FORTRAN-77 codearranged in 60 subroutines and functions. The program was developed in stages, the first ofwhich was the solution of the fluid flow equations describing a turbulent compressible jet. Asa result of this, the overall flash smelting model is called ‘jim’ from (j)et (i)mplicit (m)ethod.6.2.1 Simplifying AssumptionsThe primary assumptions which were made to simplify the development of this model areas follows1391. The problem is considered to possess axial symmetry and hence is two-dimensional incylindrical coordinates.2. The flash furnace is assumed to be at steady state.3. The “boundary layer” forms of the various governing equations were adopted. Thisform of the equations assumes that there is a “primary” flow direction, with flowperpendicular to this direction being less important. Only the axial momentum equation was applied, while the radial component of velocity was calculated from theconservation of mass equation.4. Input concentrate particles were assumed to be mono-sized spheres. Particles wereallowed to change size during the calculations as their composition altered due tochemical reaction, but they were assumed to remain spherical.5. Particle-particle collisions were not considered.6. Buoyancy forces were neglected.7. The effect of pressure gradients Ofl gas temperature was ignored.8. Chemical reaction was assumed to proceed as outlined in Chapter 5.9. Air infiltration into the reactor was ignored. The only gas species considered were 02and SO2.1406.2.2 Gas and Particle Flow6.2.2.1 Gas Conservation EquationsThe gas flow equations in the model are as followsConservation of Momentum3u 3u a 1 (6.1)Conservation of Mass(6.2)—(rpu)+--(rpv) =SConservation of EnergyT 1 a ( a (au 2 (6.3)+S’Conservation of SpeciesC, 1 a ( (6.4)pu + pv =--1rDe — J + SGlobal Mass(6.5)Qg=J mpurdr141Equations (6.1) and (6.2) are the turbulent, time-averaged boundary layer form of theNavier-Stokes equations for a compressible fluid in cylindrical coordinates. Thesymbol tin Equation (6.1) represents the sum of viscous and turbulent shear stresses.Similarly, Equation (6.3) is the boundary layer form of the energy equation for aturbulent fluid. Equation (6.4) describes the transport of the concentration of injectedspecies C1 throughout the medium. This transport is assumed to occur largely by theaction of turbulent eddies, and a turbulent mass diffusion coefficient (Do) has beenadopted to model this effect. Finally, Equation (6.5) describes the mass flowconstraint for internal flows the flow of mass across any cross-section of the vesselis constant and equal to the injected mass flowrate minus the net loss of mass tooxide.The boundary layer form of the conservation equations has been applied successfullyby many workers (79,80) to predict the fluid flow and heat transfer within freelyexpanding turbulent jets. Despite the fact that, in the UBC pilot flash reactor, the gasflow consists of a confined turbulent jet, these equations should still apply, due to thesmall diameter of the burner relative to the vessel diameter. For example, Morrison,Tatterson and Long (81) have demonstrated experimentally that the centrelinevelocity and turbulence parameters possessed by a confined jet with a largeduct/nozzle area ratio are virtually identical to those of a free jet for a distance of upto 3 duct radii downstream. These experiments were performed with a duct-to-nozzlearea ratio of 534 to 1 whereas the UBC pilot flash smelter has a duct-to-nozzle crosssectional area ratio of over 1000 to 1, indicating that the UBC pilot reactor is likely toexhibit even greater flow similarity to a free jet. The application of the boundary layerequations also has been extended successfully to accommodate regions of recircula142tion with internal flow. For example Kwon and Pletcher (82,83) have shown that therecirculation resulting from a symmetric sudden expansion can be predicted very wellusing the boundary layer equations alone.The effects of neglecting buoyancy on the predictions for the two-phase flow regimeare not likely to be serious due to the small magnitude of the buoyant forces relativeto the inertial forces. As described by Kreith (84), buoyancy has an importantinfluence on the velocity distribution when the ratio of Gr/Re2 is greater than 1. Dueto the continuous input of momentum from the injected solids, it is probable that thedensely loaded two-phase oxygen-concentrate jet retains relatively high Reynoldsnumbers even at the bottom of the reaction shaft and the ratio GrfRe2 is likely to bemuch less than 1. In the region of recirculating flow outside of the two-phase jet,however, velocities are significantly lower and there is no input of momentum fromsolids to the gas. As a result, Reynolds numbers are much lower in this region andhence buoyant forces may be of significance, despite the fact that axial temperaturegradients are not as high as for the central two-phase jet. However, the experimentalstudies presented in Chapter 4 have shown that the reacting solids which are injectedinto the flash furnace disperse radially only to a limited extent, and do not enter thesezones of recirculation. Therefore, because the objective of the model was to investigate the chalcocite flash flame, these slight errors in predicting the velocity fieldoutside of the flame were accepted to reduce the computational effort.The various source/sink terms (S,,S/’,S,’) in Equations (6.1), (6.3) and (6.5) describethe effects of particles on the gas phase. The source term in the momentum equation isdue to particle-gas drag, while the source term in the heat equation representsparticle-gas heat transfer. The source term in the species conservation equation143describes mass transport to/from the particles under the influence of chemical reaction. Each of these terms are themselves dependent upon the local number concentration of particles in the region of interest. Turbulence ModellingThe jet Reynolds numbers of the two-phase oxygen-concentrate jets in the UBC andPort Colborne flash smelters are 26,000 and 106,000 respectively. These values areboth much larger than the transitional value of 2,000 and hence, in both cases, the jetsare highly turbulent. In order to model turbulent flows of this type with continuumequations such as Equations (6.1) to (6.5), it is necessary to utilize some type of turbulence model, which describes the effect of turbulent fluctuations on the time-averagedvelocity and pressure fields. A common initial approximation is to assume thatturbulent shear stresses are proportional to time-averaged velocity gradients via aturbulent viscosity (the Boussinesq approximation)/ , 3u (6.6)t—la-——puv =Qi+i)—It is then necessary to find equations which describe the turbulent viscosity in terms ofquantities which can be measured or calculated.Employing an analogy with the kinetic theory of gases, Prandtl (85) was the first topropose a turbulence model which predicted turbulent shear stresses in terms of a“mixing length”:1442 Ju u (6.7)t= P/m I Ior,2 iu (6.8)Pe = Imwherein the mixing length im cannot be larger than the vessel enclosing the flow. Forturbulent jet predictions, a mixing length that is proportional to the jet width (a linearfunction of distance) is frequently adopted. One undesirable feature of Prandtl’soriginal mixing length model is that it predicts an effective viscosity of zero at thecentreline of an axially symmetric turbulent jet, since is zero in this location.Experiments have shown this to be untrue, with a finite effective viscosity at the jetcentreline in most cases (86).A number of other later investigators have proposed similar “eddy viscosity” modelswhich do not have this difficulty. Instead, the effective viscosity is predicted directlyfrom mean flow parameters= Cb P(Umax — Umin) (6.9)where C is a ‘constant” for the flow considered, determined by comparison withexperiment and b is the width of the mixing layer. These simple eddy viscositymodels have been applied with considerable success to describe the flow of freeturbulent jets (79,82,83). However, these models have been criticized (88) as145requiring excessive “ad hoc” adjustment when moving from one type of flow toanother. For example, different proportionality constants are frequently required forround jets than and wakes.In an attempt to address this criticism and to provide for greater generality ofapplication, various other (much more complex) models of turbulence have beendeveloped including the so-called “one equation” and “two equation” models, andReynolds stress transport models. By far the most popular turbulence model of theseis the k — E model popularized by Spalding and co-workers (88,89). In this model, theturbulent viscosity is calculated from a balance between the “production” (k) and“dissipation” (e) of turbulenceCpk2 (6.10)lit =where k and are themselves described by two additional transport equations.Over the last two decades, the k— e model has replaced the simpler models ofturbulence in popularity due to its generality and apparent applicability to all types offlow. In addition, the k — E model has the added benefit that very little prior information about the flow regime appears to be needed in order for calculations to be made.However, despite its widespread use, the k — E model still suffers from a number ofinherent limitations :1. The model requires a large number (6-8) of “universal constants” which have beenshown by many (90) to be neither universal nor constant- these parameters areusually simply “tuned” to the desired flow regime. Unfortunately, therefore, one146tends to be forced into a similar type of ad hoc adjustment as with simpler turbulence models but with a greatly increased number of ‘constants”. The effect ofparticles or temperature on these “constants” has not been well studied.2. The model requires initial conditions for each of the two turbulence equations.Measurements of the turbulence energy and dissipation are rarely available forindustrial flow regimes, and hence these parameters are often simply guessed.It has also been shown (90) that the predictions of the k — turbulence model are notnecessarily more accurate (and are sometimes less accurate) than a simple mixinglength model for computing ducted turbulent jets.Finally, despite its complexity, the k — E model remains only an effective viscositymodel and is incapable of describing flows for which the Boussinesq approximationdoes not apply. In this regard it still suffers from the same inherent limitations as thelower order turbulence models. In addition, recent developments in the area of chaostheory have indicated that the logic behind the development of two equation turbulence models is inherently flawed. It has been shown by Mandelbrot (91) that theturbulent behaviour of fluids is essentially chaotic, and therefore it is unlikely thatturbulence can ever be predicted by any equations, however complex, which utilizeonly the time-averaged flow properties.In an attempt to reduce both the total number of “fitting” parameters and modelcomplexity, a simple eddy viscosity model of turbulence was employed in the chalcocite flash smelting model. This model was originally suggested by Hwang (87) basedon that proposed by Prandtl (85).147In the initial region (potential core) the mixing length hypothesis was adopted, withthe mixing length given by:Im = 0.0762gm (6.11)Beyond the potential core, (once the central jet velocity has declined below theinjection velocity) the effective viscosity was calculated by a simple algebraic model:Pgi(trnaxmin) (6.12)Where the parameter ‘ is an intermittency function defined byr (6.13)y=1 whenO—0.8r1(r (6.14)y= (0.5)2 and z = — when >0.8The parameter r1 is the radial distance at which the velocity is one half that at thecentreline, and is termed the “velocity half width”.As will be seen below, the predictions made by the model compare quite well withmeasurements on non-reacting flows. The effects of particles, chemical reaction andhigh temperature on the eddy viscosity are much more difficult to assess. Melville andBray (92) and Hahn and Sohn (33-40) have suggested that the presence of particles ina two-phase jet dampens the turbulent fluctuations and hence tends to lower the effective viscosity of the gas. In an attempt to quantify the magnitude of these effects onthe model predictions, a sensitivity analysis was has been performed (and issummarized in Chapter 7) in which the effective viscosity was varied widely.148The effective thermal conductivity and the effective mass diffusivity in Equations(63) and (6.4) were related to the effective viscosity via a turbulent Prandtl numberand a turbulent Schmidt number respectivelyPeC’pPr =eSc =DePgIn studies of free non-isothermal jets (93) it has been shown that typical values ofthese parameters are in the region of 0.7 to Particle-Gas InteractionThe effect of particles on the gas phase momentum has been modelled using theParticle Source in Cell (PSI-Cell) method of Crowe et al. (42). A flowchart of thistechnique is shown in Figure 6.2, where the gas phase equations are solved firstwithout the particles, and then the particle momentum sources and sinks are added ina simple iterative manner.Viscous drag was assumed to be the sole means of transporting momentum betweenthe particles and the gas. The total force on a single particle in the region of interestwas then calculated byI —. —. (6.14)Fp=5Cd(Vp_Vg)IVp_VgIAp+mpgwhere the particle drag coefficient, Cd, was computed from (94)14924 7 (6.15)Cd=—-—(l +O.l5Re°68 )Gas properties were evaluated at a mean film temperature determined by an arithmeticmean of the particle and gas temperatures.The total momentum source/sink to the gas was then calculated by multiplying thisforce per particle by the number of particles per cubic metre in this regionS=FN (6.16)where A and N are the particle cross-sectional area and number concentration ofparticles respectively in the region of interest. The sources and sinks of heat and massfrom the particles to the gas were calculated similarly. This iteration was assumed tohave converged when the change in each of the predicted values of the gas velocity,temperature and composition was very small (<0. 1 %). Particle DispersionThe prediction of the radial dispersion of particles through a two-phase jet poses twoseparate problems determining the particle location and predicting the particlecomposition. The mechanical distribution of the particles under the influence of thegas was predicted by applying a conservation equation in a manner analogous to theconservation of species equation used to compute the oxygen concentration. However,the composition of the particles as they disperse is much more difficult to describe,because chemical reaction is occurring thereby altering the radial composition ofparticles simultaneously with this radial transport. These two problems have beenconsidered separately.1506. Particle Concentration DistributionThe radial transport of particles through the gas jet has been observed experimentally in the solids samples taken from UBC pilot plant. That is, when the concentrate is injected into the UBC flash reactor, it originally occupies only the area ofthe burner nozzle, but then spreads radially and occupies a larger fraction of theshaft cross-section by the time it reaches the bottom of the shaft. The experimentaldata have shown that this radial transport is relatively small (approximately0.1-0.2 m by the bottom of the reaction shaft), but it must be taken into account ifaccurate calculations are to be made.The mechanism of the solids dispersion is complex ; due to their relatively smallsize the particles will tend to follow turbulent radial fluctuations of the gas to acertain extent, and will therefore spread throughout the two-phase jet. In addition,at various locations within the jet, the time-averaged gas radial velocity is finiteand directed outward, which also spreads the particles. The model of chalcopyriteflash smelting developed by Hahn and Sohn (33-40) was used as a guide in thedevelopment of the chalcocite flame model, since they studied the effect ofparticles on confined gas jets in great detail.The approach which was followed by Hahn and Sohn was to disperse the particlesradially through the gas according to a particle conservation equation analogous tothe conservation of species equation. In boundary layer form this becomesN, JN, I (1 3N (6.17)rD’—3z r rJr ° 3r151In this formulation, the particles are dispersed radially according to the time-averaged turbulent velocities of the gas and depending on a particle diffusivity(Dot). In their work, Hahn and Sohn calculated this diffusivity from a turbulentparticle Schmidt number (assumed to be constant at 0.35) and an effective particleviscosity. This “viscosity” was, in turn, related to the effective viscosity of the gas(altered to account for particle effects) through the particle relaxation time (tn) andthe turbulence time scale (t1)— Pe (6.18)Scp(l +(which were calculated by:mp (6.19)3iqid______(6.20)I pV’V’The particle relaxation time is the length of time necessary for a particle at rest toattain 36.7 % (lie) of the gas velocity.Because experimental observations in the UBC flash smelter have revealed thatthe particles tend to disperse radially only to a very small extent, a simpler andmore approximate technique has been followed in the development of the mathematical model for this work. The initial particle number density in the UBC flashsmelter was calculated from the oxygen and concentrate input rates and the (true)density of the concentrate152th (6.21)pp 4Qg7trpppand was assumed to be constant over the diameter of the burner. The turbulentparticle diffusivity was assumed to be a constant fraction of the gas effectiveviscosity, using a turbulent Schmidt number for the particles. Measurements of theturbulence parameters of a free turbulent jet presented by Harsha (95) wereadapted to obtain the turbulent time scale, while the particle relaxation time wascalculated by Equation (6. 19). Together these values provided estimates of theturbulent diffusivity.For a 20 micron MK particle (approximately the mean particle size) in the UBCflash furnace, the relaxation time is approximately 7 milliseconds. For a free jet,the turbulence time scale is a function of axial distance, but has an approximatevalue of I - 2 milliseconds, while the turbulent Schmidt number for the particles,as suggested by Hahn and Sohn, was 0.35. Substituting these values into Equation(6.18) gives a ratio of the particle diffusivity to the gas effective viscosity ofapproximately 2.5, indicating that the particles will not spread radially as quicklyas the gas. Interestingly, when this value is substituted into the correlationsuggested by Goldschmidt and Eskinazi (96) for two-phase jets, the predictedradial spread of the particles by the bottom of the reaction shaft is approximately60 % of that for the gas. Assuming the gas to behave as a free jet (expanding witha jet angle of 20 degrees), the maximum predicted particle radial spread istherefore approximately 15 cm, which agrees reasonably well with the experirnental data.153The solution of Equation (6.17) employing these assumptions led to predictions ofGaussian particle distributions throughout a relatively small region in the centre ofthe UBC flash reactor at all axial locations which also agreed well with experimental findings. In these calculations, the edge” of the particle jet was definedwhen the particle number concentration fell below 0.1 % of the centreline value. Particle Composition DistributionThe composition of the concentrate particles changes with axial and radial location due to the effects of chemical reaction as Cu2S in the particles is converted toCu20 and Cu. However, in a flash reactor, the composition of particles at a givenradial location must also be affected by radial input from adjacent regionspossessing particles of a differing composition. For the purposes of this mathematical model the effect of dispersion on concentrate composition was neglectedexcept at the ‘edge’ of the two-phase region. In this region, if the edge of theparticle jet advanced radially by one computational cell from the previous axiallocation, the particles at this‘tnew” location were assumed to be of identicalcomposition to those at the original edge of the jet. Particle reaction, gas drageffects, and heat and mass transfer were then allowed to proceed normally at thisnew location.Neglecting the effect of radial transport on particle composition in the bulk of thejet will tend to under predict the extent of reaction since under-reacted concentratein the centre of the jet is not allowed to move outwards. This effect would be154stronger toward the bottom of the reaction shaft where there are more reactedparticles. However, if the radial dispersion is small then most particles will followthe spatial and composition trajectories computed by the model.6.2.3 Heat TransferThe temperature of the particles was calculated from the heat balance, in an identicalmanner to that described in Chapter 5. For the purposes of the mathematical model of thechalcocite flash flame, gas-wall and wall-wall heat transport were ignored. The oxygenand concentrate were assumed to encounter a recirculating gas of constant compositionand temperature as they descended the reaction shaft.This formulation required that the model be supplied with appropriate input values offurnace wall and recirculating gas temperatures, as the mathematical model was notcapable of predicting them. Since the overall objective of this model was to investigatethe flash smelting flame, and because good experimental data was available, this assumption introduced little error while greatly simplifying the problem. Gas-Particle Heat TransferThe heat transfer between the gas and the particles was assumed to occur only byconduction/convection. The convective heat transfer coefficient between the gas andthe particles was calculated from the Ranz-Marshall correlation (84):Nu =2+ O.6Re°5Pr°33 (6.22)155As the slip velocity between the gas and the particles tends to zero, this equationpredicts the Nusselt number to approach a value of 2 - the case of heat transport byconduction alone. Assuming a Nusselt number of 2 and a 20 micron particle, a typicalgas-particle conductive heat transfer coefficient would be of the order of iO W/m2K,whereas the radiative gas-particle heat transfer coefficient (assuming emissivities ofunity and gas and particle temperatures of 1200 K and 298 K respectively) isapproximately two orders of magnitude lower. WaIl-Particle Radiative Heat TransferAs Figure 6.3 indicates, the rigorous description of the wall-particle radiative heattransfer in the UBC flash furnace is a complicated problem radiative exchangebetween the walls and the particles is complicated by the presence of a particle-ladenrecirculating gas, and absorption/emission caused by the particles themselves.In their model of chalcopyrite flash smelting, Hahn and Sohn (33-40) presented anelaborate means of predicting the absorption, emission and scattering of radiatedenergy caused by particles. However, in this case, the fundamental values required forradiative heat transfer calculations are unknown, and hence highly detailed scatteringand absorption calculations of this type are not justified. For example, to performcalculations in the manner suggested by Hahn and Sohn, values for the emissivities ofthe gas, the reactor walls, the particles and the fume are required as a function oftemperature. This data is not available, and therefore only estimates of these valuescould be applied. The error introduced into the calculations by this estimation wouldthen invalidate such a complex approach.156Both to reduce the number of “fitting’ parameters required by the model, and tosimplify the analysis of wall-particle radiative heat transfer, a more approximatemethod was employed in the development of the model. This technique required twofundamental assumptions1. The walls of the flash reactor were assumed to be at a constant temperature.2. The emissivities of the particles and the walls were both assumed to be 1.Although neither of these assumptions is completely correct, taken together they serveto place all that is unknown about the wall-particle radiative heat transfer (emissivities, scattering, view factors etc.) into a single parameter: a mean reactor wall temperature. Thus the wall-particle heat transfer was modelled by assuming that the particlesexchange heat with a large enclosing source at a constant temperature. In this way, thefurnace wall temperature is represented by a global mean temperature - driving all ofthe radiative heat transfer between the walls and the particles.As a result of these assumptions, the experimentally measured mean wall temperatureserves only as an estimate of the value required by the mathematical model. Forexample, if the emissivity of the walls and the particles is significantly lower than 1,or particle scattering in the recirculating gas reduces the radiative heat transfer rate,then the effective wall temperature experienced by a particle on heat up will be lowerthan the experimentally measured mean value. Conversely, if the bottom or the top ofthe reaction shaft are significantly hotter than the walls, then this effective temperature will be higher than the experimentally measured mean.Thereftre, the wall-particle heat transfer rate was calculated by157Qrad — T) (6.23)The wall-particle view factor, presents further difficulties. If particles within theflash flame do not significantly block wall-particle radiation, then is constant at 1(the case of a small particle in a large enclosure). Under stoichiometric conditions, themaximum volume fraction of solids within the UBC flash furnace is 0.12% and thecalculations presented in Appendix 2 suggest that, due to this low volume fraction,much of the radiation from the walls reaches the particles. These calculations indicatethat more than 90 % of the energy emitted normal to the axis of the flame reaches theparticles at the centre of the UBC furnace. However, at angles off the normal (FigureA2.3), it is much more probable that a significant fraction of the incident radiation isblocked by the presence of particles. The integration of the particle blockage effectpresented in Appendix 2 indicates that the mean value of is approximately 0.8.This mean value was applied to Equation (6.23) for all axial and radial particlelocations.For many regions of the flash flame, the effects of this approximate treatment of theradiative heat transport rate are likely to be quite small, since the radiative term itselfis frequently small compared to gas-particle convection. This effect has beendiscussed both in the previous section and in Chapter 5. However, in regions wherethe particles and the gas have similar (low) temperatures, the high gas-particle heattransfer coefficient is irrelevant, and hence radiative transport from the walls may bethe most significant means of heat input to the particles. Such regions are most likelyto occur near the centre of the oxygen-concentrate jet close to the nozzle of theburner, where the entrainment rate of the hot surrounding gas is not significant andthe high solids concentration provides a significant ‘sink’ of heat from the gas.1586.2.4 Chemical ReactionThe kinetic model which was used to predict the progress of chemical reaction has beendescribed in Chapter 5. The progress of reaction in each finite difference regioncontaining particles was continuously monitored and the effects of changing compositionon particle size, density, and heat capacity were taken into account.The sink terms in equations (6.2) and (6.4) were calculated depending upon the reactionstoichiometry. For example, a particle undergoing reaction of Cu2S to Cu20 by:Cu2S + = Cu20 + SO2 (6.24)produces a net sink of oxygen in Equation (6.4) at a rate equal to 1.5 times the reactionrate of Cu2S. However, the sink of mass in Equation (6.2) is only 0.5 times the reactionrate, since 1.5 moles of oxygen produce 1.0 moles of SO2. All other reactions weretreated similarly.Heat generation caused by chemical reaction was treated in an analogous manner, andwas applied to Equation (6.3) through the particle heat source/sink term.6.3 Model DevelopmentA significant advantage of the boundary layer form of the conservation equations is thatwhen approximated with finite differences, the resulting equations are relatively easily solvedwith a “marching’ type procedure. That is, the solution can advance axially from the knowninitial conditions (burner outlet) until the end of the reaction shaft, calculating radial variation159as the solution proceeds. This method is very economical in computer storage and time, isvery stable numerically, and also allows dynamic adjustment of the axial step, from fineincrements in regions of rapid change to coarser increments where axial gradients are small.6.3.1 Discretization of EquationsEquations (6.1) through (6.5) and the solids transport and reaction equations constitute aset of simultaneous non-linear partial differential equations. A direct analytical solutionof these equations is not possible and hence requires a numerical solution. In the literature, the most common technique for the numerical solution of equations of this type isthe finite difference technique whereby the governing equations are applied at discretelocations within the computational domain. The derivatives are then replaced with differences, and the discrete equations are solved simultaneously for each of the desired variables. The fully implicit finite difference forms of the governing equations are:Conservation of Momentum— u_1,) (u,+1 — u_1) (6.25)+P1V2Ar =Ap 1 (______________+2rArri,J+1R,J+I Ar Ar JJ+SmConservation of Mass(pu)1—(pu) (pv)11—pv1,+=(6.26)Az 2Ar r160Conservation of Energy(7 — 1) (T1+ —T11) (6.27)pJU1JCAz+ pjJVjJC2Ark k (i.J1+SPAz 2rAr i,j+J i,j+1 Arr1,_ ij1 JJConservation of Species(C11 C11) (C11,— C11 ) (6.28)p..u.. +p.v:: =If Z,j Az If 2Ar1 ( (C1.. -C . (C1. .-c.2rjAriJDi+1 J_j_ij_1 ArGlobal Mass(6.29)th= j1it(pu)1yAr = constantThe discretization of the UBC flash furnace is shown schematically in Figure 6.1 and isaxially-symmetric in two dimensions. Various mesh densities were tested in the radialdirection, the most frequently employed being 143 nodes over the 0.25 m reactor radius.This resulted in 6 nodes being allocated to the initial radius of the burner (diameter =0.0158 m) and an inter-node radial spacing of 1.7 mm. The 1.8 m reactor length wasdivided into 1,800 intervals, each being 1.0 mm long. Therefore up to 257,000 nodes161were utilized to cover the 0.25 m x 1.8 m UBC flash reactor. Without the “marchingt’ability of the boundary layer formulation, mesh densities of this size are prohibitivelylarge.6.3.2 Solution AlgorithmA flowchart of the technique employed in the solution of these equations is shown inFigure 6.4. The fully implicit form of the equations was used because it leads to verystable solutions, and one is able to employ the rapid tridiagonal solution technique.However, the solution of Equations (6.25) to (6.29) is complicated by nonlinearitiesarising from convective terms. For example, in Equation (6.25), the term -‘- presentsdifficulties for linear solution techniques because it is second order in the unknown u. Acommonly employed approximation (82) is to ‘lag” the coefficients, using the velocitiesfrom the last axial position as approximations to those at the new position. This term thenbecomes u_11 and implicit solution for the velocity at the new axial position u canproceed normally using the tridiagonal solution algorithm. The coefficients can then beupdated iteratively using the newly calculated value for velocity at this axial location.Convergence of the coefficients in this case usually took less than 10 iterations. Once thecoefficients had converged and the values of u and v at the new axial location hadbeen calculated, the other conservation equations could then be solved.The particle source and sink terms of heat, mass and momentum then were computed viathe PSI-Cell algorithm whereby the motion and behaviour of particles in each appropriatefinite difference cell were used to calculate the effects on the gas. The particles wereallowed to react with the surrounding gas and change composition and temperature asthey moved. Since these changes were usually very rapid, even with respect to the short162time necessary for a particle to traverse a finite difference cell, a number of sub-steps (2to 10) were employed to move the particles. The total aerodynamic drag force experienced by the particles in their flight was summed together with the particle-gas heattransfer and chemical reaction effects and multiplied by the particle number density toproduce the source and sink terms in the gas conservation equations. This computationrequired that the exact chemical composition, position, velocity, density, size and temperature of the particles be monitored at all times.Once the source and sink terms had been calculated, Equations (6.25) through (6.29) weresolved again, and the entire process was repeated until the predicted values did notchange. Finally, the velocity and density of the gas were integrated across the cross-section of the reactor and compared with the overall mass input rate. If there was a sufficient excess or loss of mass, the pressure gradient was adjusted appropriately and theentire solution was repeated. Either a variable-secant (82) or an interval halving rootfinding method was used to determine appropriate adjustments to the pressure gradient.Final convergence at the next axial location was deemed to have been achieved when theerror in the overall mass flow rate was less than ± 0.1 %.6.3.3 Flow Reversal RegionRegions of flow reversal or recirculation usually pose significant stability problems forsolutions to the boundary layer flow equationS (82). This is due to the fact that a “marching” type of solution does not lend itself to regimes in which information flows backward against the primary flow direction. Multiple pass techniques in which the marchingdirection is reversed have often been used (97) successfully to deal with these regions.However, Rehner and Flugge-Lotz (98) introduced a technique termed the FLARE163approximation which allows calculation of these regions while preserving the straightforward one-pass marching method of the boundary layer equations. As described inAnderson et al. (82), this method is based on the observation that regions of flow reversalare usually characterized by relatively low axial velocities, such that the term u can bereplaced with C I u I , where C is assumed to be zero or a small positive constant (-0.1). These regions of flow reversal can then be treated normally, without any stabilityproblems.As indicated above, the implementation of the FLARE approximation does introduce anadditional approximation, that of low recirculating velocities. However, since the primaryobjective of the chalcocite flash smelting model was to describe the phenomena occurringwithin the flash flame, and since the chalcocite flash flame itself did not occur in a regionof flow recirculation, it is unlikely that the effect on the overall predictions was severe.Provided that the recirculating velocities are low, the transport of momentum (and therefore of heat and mass) from a recirculating flow region into a turbulent jet differs verylittle from that occurring between a turbulent jet and stagnant surroundings. This isillustrated by the studies of Morrison, Tatterson and Long (81) who found that a confinedjet (with recirculation) behaves very similarly to a free jet if the nozzle diameter is smallrelative to the duct diameter (i.e. low recirculating velocities). In this case, the high axialand radial velocity gradients at the edge of the mixing layer mask any secondary effectscaused by (small) negative axial velocities outside of this region.The fact that recirculation occurs is certainly of great overall significance to the flashflame, as heat and mass enter the jet through this process. If there were no recirculationoccurring within the UBC flash furnace, then the oxygen-concentrate jet could not beheated or diluted by gas entrainment. However, it is likely that the mass entrainment ratewith this recirculation does not differ significantly from that of a freely expanding jet.164Qualitative observations made in the UBC flash furnace indicate that recirculating velocities are certainly very low (< 1 mIs) compared to the centreline velocity (25 mIs). Acheck on the error introduced by the FLARE approximation is presented below, through acomparison made by ‘jim’ for a confined recirculating flow to that made by the FIDAPfluid flow package (FDI, Inc.), which solves the complete Navier-Stokes equations.6.4 Non-Reacting Model VerificationPrior to comparison with the data of the UBC flash smelting reactor the model was runagainst a series of simple test cases (without chemical reaction) to verify the reliability of thecomputer code and to provide a check on the various fitting parameters required by themodel. These parameters have been summarized in Table 6.1, together with a description ofthe means by which they were verified.Table 6.1 : Fitting Parameters Used in the Mathematical ModelParameter Use VerificationEffective Tiirbu lence modelling. Free jetsViscosity All gas predictions 1,2-phase, heatedTurbulent Temperature predictions Heated 1-phase free jetsPrandtl No. (Gas Mixing)Turbulent Gas composition C02-Air jet mixingDiffusivity (Gas Mixing)(Gas)Turbulent Particle Dispersion 2-phase jetDiffusivity UBC Pilot Plant(Particles)1656.4.1 Isothermal Free One-phase Jet - Axial and Radial VelocitiesThe first test case considered to verify the model was a freely-expanding, isothermal airjet. This case has been extensively studied by a large number of investigators and therforeprovides a good initial test of the mathematical model. A correlation of the existingexperimental data for high Reynolds number subsonic jets has been proposed by Tuve(99):For x/D0 < 8U0 (Potential core region)For x/D 8:KDQU0 (6.30)U=CAs Figure 6.5 shows, the measured axial velocity of a single-phase air jet emerging fromthe UBC pilot burner is well predicted by this correlation, which was thereftre selected asthe basis of comparison with the mathematical model. The model predictions were madeby neglecting the pressure gradient and dropping the overall mass flow constraint as isappropriate for a freely-expanding jet. The turbulence model described by Equations(6.11) to (6.13) was used to make these predictions. The comparison between the axialvelocity predicted by the model and that from Equation (6.30) is shown in Figure 6.5 andthe agreement is very good.166A con-elation for the jet axial velocity as a function of radial position has also beenproposed by Tuve (99):(U0”\ 1r’2 (6.31)logIU) ZThe comparison between the model predictions and this correlation is shown for anumber of axial locations in Figure 6.6 and again shows good agreement.Correlations for the radial jet velocity as a function of axial and radial location are notgenerally available in the literature. In an attempt to produce a correlation of this typefrom the Tuve equations, Equations (6.30) and (6.31) were combined and substituted intothe boundary layer continuity equation for incompressible flow. At any one axial location, u and v are functions of r alone, and hence Equation (6.2) becomes:du v dv (6.32)+ +By differentiating the Tuve correlation for the axial velocity u, and substituting, Equation(6.29) becomes a first order linear ordinary differential equation for the radial velocity asa function of radial position. The necessary boundary condition is that at r=0, v=0 due tosymmetry. After much algebra, the solution of this equation yields:2____‘ (K (6.33)v=K1exp(—K0r167401n(1O)where K0= 26.21) UK1=6.21) UK2= 80 In(1 0)The comparison between Equation (6.33) and the model predictions is shown in Figure6.7, and again the agreement is seen to be quite reasonable. The model and the correlationdiffer to the greatest extent beyond the region of maximum radial velocity (correspondingto the axial velocity half-width). This is due to slight errors in the predictions of the turbulence model for this region - the decline of the effective viscosity to the edge of the jet isnot perfectly predicted by the intermittency function in Equation (6.12). Nevertheless, theagreement between the model and Equation (6.33) differs by at most 10 %, which is wellwithin the inherent error of this or any other turbulence model (82).6.4.2 Heated Free Single-phase JetThe case of a heated air jet discharging into stagnant, unheated ambient air has also beenextensively studied. The data of Wilson and Danckwerts (100) were chosen for comparison, in which ajet of air at 498 K was injected from a 0.0127 m diameter nozzle into stillair at 298 K. The predictions of the mathematical model has been compared with this datain Figure 6.8. The turbulence model which has been used to make these predictions is thesame as that employed previously. The temperature field was calculated by Equation(6.3) with a turbulent Prandtl number of 0.7 and, as Figure 6.8 shows, the agreement isagain very good. The maximum error in the predicted centreline temperature is consistently less than one degree Celsius.1686.4.3 Two-phase Isothermal Turbulent JetThe data used for comparison in this case were those of Modarress, Wuerer and Elghobashi (101) who used laser-doppler measurements to determine the velocity of both gasand particles in a two-phase free jet. This study was chosen as a test of the mathematicalmodel because it considered mono-sized particles in a densely loaded gas stream (solidsto-gas mass ratio of 0.8). As a result, this case simultaneously acts as a test of the gas-phase turbulence model, the particle-gas drag calculations and the particle dispersionmodel. As described previously, the turbulent diffusivity of the particles was calculatedfrom Equation (6.18), based on the particle response time (1.48 s) and the turbulent timescale for a free jet.The comparison between the model predictions and the measurements of Modarress et al.is shown in Figure 6.9. It appears that these experimental gas and particle velocities arevery well predicted by the mathematical model.Isothermal particle dispersion predictions made by the model were not compared withisothermal particle dispersion data from the literature for the following reasons1. No data was found for conditions approximating the UBC or Port Colborne flashfurnaces : densely loaded (free or confined) single-entry two-phase turbulent jets witha mean particle size of 20 microns.2. There is considerable disagreement in the literature (101,103-107) on the effects ofparticle size, density, and loading on the rates of solids dispersion, and much of theavailable experimental data appears to be contradictory. Therefore any “verification”of the model with data from the literature may have very little applicability to the169UBC flash furnace. The studies of particle dispersion by Smoot and co-workers(108,109) are the most consistent, but these investigators used a double-entry system,very different from either the UBC or the Port Colborne flash burners.3. Most importantly, good particle dispersion data has been taken from the UBC flashfurnace and is available for comparison.Therefore, the particle dispersion predictions of the model were verified by comparisonwith data from the UBC flash furnace (Chapter 7), obtained under actual operating conditions (chemical reaction, non-isothermal).6.4.4 Isothermal Single-phase Confined JetAs described previously, the boundary layer form of the flow equations which has beenapplied in the development of this mathematical model is not commonly used for internalflow due to numerical instabilities which are frequently encountered in the recirculatingflow region. It was, therefore, necessary to verify whether the predictions made by theseequations are comparable to those made by the complete Navier-Stokes equations.The model predictions were compared with those made by FIDAP (FDI mc)- a commercially available finite-element computational fluid mechanics package which solves thecomplete Navier-Stokes equations. The problem considered was that of isothermalincompressible injection of oxygen into the UBC flash furnace. A simple constant effective viscosity model was used in all cases.Figure 6.10 shows the comparison between the model calculations of the centreline axialjet velocity with those made by FIDAP, and clearly the agreement is very good. The170calculated jet axial velocity as a function of radial position has also been compared fortwo axial positions (z = 0.5 m and z = 1.0 m) in Figure 6.11. Again, the agreementbetween the model and the calculations made by FIDAP is reasonable. Most importantly,the calculated (negative) recirculating velocities agree very closely, which indicates thatthe boundary-layer equations (and the FLARE approximation) are capable of predictingrecirculating flows which agree well with results from the complete Navier-Stokes equations. The savings in computer time associated with solving only the boundary-layerequations are very considerable. The calculations in Figures 6.10 and 6.11 made by theprogram ‘jim’ took approximately 30 Cpu minutes to complete, while those made byFIDAP took 1000 CPU minutes and required 50 times more memory.The slight disagreement between the two sets of predictions in Figure 6.11 is most likelydue to minor errors in FIDAP’s overall mass balance. When the velocity profiles inFigure 6.11 are integrated and compared with the mass input rate, the calculations madeby ‘jim’ agree to within 0.01 %, while those made by FIDAP are approximately 5 % low.This is not unexpected, since elliptic solvers, such as FIDAP employs, tend to terminatecalculations when the mean divergence of each element is small, which can result significant errors in the overall flowrate. These errors can be improved significantly (using afiner mesh and employing more iterations), but only at the expense of greater CPU time.6.4.5 Isothermal One-phase Free Jet - Dissimilar Gas InjectionThe effect of mixing and dilution of the incoming oxygen-concentrate concentrate streamwith the surrounding hot sulphur dioxide in the UBC flash furnace must also be accurately calculated by the model. In effect, this is the case of a jet mixing with surroundingswhich are less dense.171To evaluate the performance of the model in this regard, the experimental data of Keagyand Weller (102) for a free jet of carbon dioxide injected into air, were selected. Theresulting velocity and concentration fields are shown in Figures 6.12 and 6.13, togetherwith the model predictions. In this case, a turbulent Schmidt number of 0.7 was used tocompute the effect of entrainment on the centreline composition, and the agreement withthe experimental data is reasonable.6.4.6 SummaryThe mathematical model has been ftund to predict accurately both isothermal and nonisothermal flow of free and confined jets, with and without the presence of particles. Toverify the model considering the cumulative effect of all of these conditions, togetherwith chemical reaction and particle dispersion, it was necessary to compare the modelpredictions with the experimental data obtained from the UBC pilot plant. This has beendone in the following chapter.RAz1Axis of symmetryApproximate edge ofoxygen-concentratejet172zFigure 6.1 :Sketch of flow problem to be solved by the mathematical model173*YesSignificant Change?Solve flow equationswithout particle sourcesUse new data to calculateparticle sources/sinksCompute revised valueswith particle contributionsFigure 6.2 : Flowchart of the PSI-Cell algorithm (42)-t CDCfC/)z C) CD 0 -t CDIll175Initialize, read in data and optionsAdd particle drag tenns anddo chemical reaction, heatand mass sources and sinksRe-calculate new velocitiesconvergence NoIntegrate total flow across[actor width.Solve boundary layerI equations for velocities I[at new position4,Solve for new temperaturescompositions, densities[ir dpldxj1No>YesWrite out data and moveto next axial positionFigure 6.4 : Flowchart of the computer program ‘jirn.f’176C0C>‘U‘UU‘UNIFigure 6.5: Comparison between the change of jet centreline velocity as a function ofaxial distance predicted by the mathematical model with the measurementsmade for this work, and the correlation of Tuve (99).[Constant temperatureair jet into stagnant surroundings IAxial Position (m)C,,>-.C-)0>1771411Figure 6.6: Comparison between the axial jet velocity as a function of axial and radialposition predicted by the model and the correlation of Tuve (99). Constanttemperature air jet into stagnant surroundings].S>-0C178Radial Position (m)Figure 6.7 : Comparison between jet radial velocity as a function of axial and radial position and the equation derived from the Tuve (99) correlation and conservationof mass. [Constant temperature air jet into stagnant surroundings].179Figure 6.8: Comparison between predicted jet centreline temperature and the experimental data of Wilson and Danckwerts (TOO). [Heated air jet into stagnantsurroundingsAxial Distance (Nozzle Diameters)1807 Model Predictions - GasModelPredictions - ParticlesGas Velocity Measurements - Modarress et al.Particle Velocity Measurements of Modarress et al.0:z:2ZDimensionless Axial Distance (z/D)Figure 6.9: Comparison between model predictions of centreline particle and gas velocityas a function of axial distance and the experimental data of Modarress et al.(lOfl.[Two phase air-particle jet I181C,)>0C1)>ti.)zC.)ccUFigure 6.10: Comparison between centreline jet velocity as a function of axial distance aspredicted by the mathematical model and the predictions of FIDAP (FDIInc.). [Single phase confined jet with recirculation]0.8Axial Distance (m)Figure 6.11 Comparison between axial jet velocity as a function of radial position aspredicted by the mathematical model and the predictions of FIDAP (FDIinc.) for two axial positions (0.5 m and 1.0 m). [Single phase confined jetwith recirculation 1182cJC-)0>><Radial Position (rn)0>004-aU0Nc0zFigure 6.12: Comparison between axial variation of centreline jet velocity predicted bymodel with the data of Keagy and Weller (102)[ CO2jet into stagnant air 1183Axial Distance (Nozzle Diameters)ci)ci)Uci)><0C.0U1ci.)C.)‘1)1184Axial Distance (Nozzle Diameters)Figure 6.13 : Comparison between axial variation of centreline jet composition predictedby model with the data of Keagy and Weller (102) [CO2jet into stagnantair]1857 Mathematical Model Validation - Comparison with UBC Pilot Plant DataIn the previous chapter the predictions of the mathematical model were verified by comparisonwith (non-reacting) data from the literature. The central objective of this chapter is to demonstrate, by comparison with the UBC pilot plant data, that the mathematical model is capable ofquantifying the behaviour of the chalcocite flash flame. Each of the various run conditionsperformed in the experimental trials (stoichiornetric and excess oxygen, sized and unsizedconcentrate) were tested by the model and compared to the available data. This data includedparticle ignition location, gas and solids sample compositions, and dust generation rates. Thevarious constants within the model (turbulence parameters, activation energy, pre-exponentialconstant, etc.) were not fitted to the experimental data : all predictions were made with the samevalues of these parameters, which were identical to those determined previously. The sensitivityof the model predictions to all of the input parameters has been investigated and is presentedbelow.The mathematical model was run on a Silicon Graphics 4D/220 GT Project Supercomputer (33MIPS). Depending upon the problem, the total CPU time required varied between I and 5 days(86,000 to 424,000 CPU seconds). Every effort was made to reduce this computer time byprogram optimization, but this was not very effective. The sheer complexity of the problemmeant that many calculations had to be performed, which consumed large amounts of computertime. Without the advent of high speed computers at reasonable cost, such as the SGI 4D/220,the mathematical model would have been prohibitively expensive in computer time. By way ofcomparison, the Silicon Graphics 4D/220 GT computer is between 60 and 200 times faster than a20 MHz 386/387-DX PC, with much faster disk and memory access times. One of the larger runswould have taken over one year to calculate on a fast PC.186The large volume of output generated by the computer program (millions of numbers for eachrun) made it essential that automated data analysis tools be used to examine the model predictions. To accomplish this analysis, a Silicon Graphics 4D/80 GTB workstation, running the‘DataView’ software package was employed. The graphics workstation and the DataViewsoftware allowed all of the data produced by the mathematical model to be displayed simultaneously on the computer screen using colour-coded contouring. This display also aided modeldevelopment by rapid identification of ‘bugs’ or errors in the computer program. Hardcopy ofthe colour plots was produced by a Seiko-5 504 thermal transfer colour printer.7.1 Determination of Furnace Wall and Recirculating Gas TemperaturesIn Chapter 6, the various parameters necessary for the model to describe the flow of anisothermal, confined, two-phase jet were determined by comparison with literature data.Similarly, the primary factors in the kinetic model (activation energy and pre-exponentialconstant) were obtained from the literature and were verified by comparison with additionalliterature data in Chapter 5. To allow the model to predict the behaviour of the chalcociteflash flame in the UBC pilot plant, two additional parameters were required: the effectivefurnace wall temperature and the recirculating gas temperature.As described in Chapter 6, the development of the model was simplified by assuming that thefurnace wall temperature and the recirculating gas temperature were constant. The calculations presented in Appendix 2 have suggested that, regardless of their axial or radial position,the particles can “see’ approximately 80 % of the enclosing walls of the UBC flash furnace.All experimentally measured wall temperatures for the base-case conditions (2 kg/mm187unsized concentrate with stoichiometric oxygen) were averaged to obtain a mean value of1160 K. The sensitivity of the model predictions to changes in this value have been summarized in section 7.2.Obtaining an accurate value for the recirculating gas temperature was somewhat more difficult, because experimental measurements were not available (as described in Chapter 4,solids in the recirculating gas clogged the suction thermocouples). Therefore, the temperatureof the recirculating gas was derived from the assumed wall temperature through a heatbalance. The experimental data has established that approximately 10 kW of heat is lostthrough the refractory walls, and hence the gas-wall heat transfer rate must be approximately10 kW (assuming negligible direct particle-wall heat transfer). Based on the experimental gasanalysis data in which the recirculating gas was shown to be 90 % SO2, and the resultingvalue of the mean beam length, the emissivity of the gas in the UBC flash furnace was calculated to be approximately 0.2. By applying a mass balance, the velocity of the recirculatinggas was estimated to be approximately 1-2 mIs, and the resulting gas-wall convective heattransfer coefficient was calculated to be 8-9 W/rn2 K . With these values, the mean gastemperature was calculated to be approximately 1200 K. It is interesting to note that themajority of the heat transferred to the walls from the gas is due to gas radiation - gas/wallconvection being only a very minor component. Therefore in each of the cases below, boththe furnace gas and wall temperatures were assumed to be constant, and had values of 1200K and 1160 K respectively.The radiation circuit diagram showing the relationship among the gas, the walls, and theparticles is shown in Figure 7.1. With assumed gas and wall temperatures of 1200 K and1160 K respectively, and a gas emissivity of 0.2, it can be shown that neglecting gas-particle188radiation (as assumed in the development of the model) does not introduce serious error. Thatis, with the contribution of the gas incorporated (view factors of 1 assumed), the radiativeheat transfer to the particles per unit area is determined byGEgFgp(T — T) + (l— ç)Fw (T — T)(7.1)Neglecting the contribution of the gas, the radiative heat transfer to the particles per unit areais given byq (7.2)Substituting the calculated values of gas emissivity, wall temperature and recirculating gastemperature, and assuming a particle temperature of 298 K, gives a maximum error of only2.6 %, (which is far less than the uncertainty in the values of the temperatures or the gasemissivity).7.2 Sensitivity AnalysisTo determine the response of the model to the various input data, and to examine the assumptions made in the development of the model, a sensitivity analysis was performed. This studyinvestigated the effects on the model predictions which were produced by large changes incritical parameters, and hence provided an estimate of the reliability of the model predictions.In each of the cases described below, all quantities have been maintained constant except thesingle value being investigated. The flash reaction conditions assumed by the model were thestandard’ operating conditions of the UBC pilot plant: 2 kg/mm of unsized concentrate with189stoichiometric oxygen. Unless stated otherwise, the furnace gas and wall temperatures wereassumed to be 1200 K and 1160 K respectively as indicated earlier, and the particle diametertested was 20 microns.7.2.1 Effect of Wall and Gas TemperaturesAs described previously, the furnace wall and gas temperature are not truly independent,but are linked by the overall heat balance. Therefore any change in one of these variableswill cause a corresponding change in the other. However, to determine the individualeffect of each of these values, the gas and wall temperatures were changed independentlywith no regard for the overall heat balance.Unexpectedly, the model was found to be relatively insensitive to variations in theassumed temperature of the recirculating gas, as a change of± 50 K had very little effecteither on the predicted combustion location or the degree of particle reaction. This wasfound to be due to the effect that gas temperature had on the recirculating gas density,which compensated for the change in gas temperature. That is, an increase in the assumedrecirculating gas temperature reduced the density of the gas opposing the motion of theoxygen-concentrate jet. In turn, this allowed the jet to maintain a higher axial velocity,and consequently reduced the particle and gas residence times. (A similar effect has beendemonstrated in Chapter 6, in which a C02-air jet was observed to maintain higher axialvelocities than an air-air jet). The reduction in residence time caused by this effectreduced the impact of the increased temperature, resulting in little overall change.Changes of± 50 K in the assumed value of the furnace wall temperature were found tohave very little impact on the combustion of the particles at the edge of the jet. This190parameter did have a slight effect on the behaviour of the solids at the centreline ; anincrease in the assumed wall temperature of 50 K moved the centreline combustion zoneupward by 0.1 m. This effect was due to the relatively low temperature of the gas in thisregion (shown below), and hence the high gas-solid convective heat transfer coefficient isof little relevance until combustion initiated at the edge of the jet generates high local gastemperatures.The wall-particle view factor was varied from 0.5 to 1.0 and was observed to have asimilar effect to the furnace wall temperature. Increasing the particle-wall view factor to1.0 raised the primary combustion zone upward by approximately 0.2 m (15 % ), whiledecreasing this factor to 0.5 moved the combustion zone downward by 0.22 m.7.2.2 Turbulence Modelling ParametersAs described in the previous chapter, a variety of parameters were employed to predictthe turbulent flow of a two-phase jet. These parameters included the turbulent Prandtlnumber, the turbulent Schmidt numbers for the gas and the particles and the turbulentviscosity. The effects of changes (of± 50 %) to the turbulent Prandtl number and thegas-phase turbulent Schmidt number were investigated and found to be insignificant. Thatthe two-phase jet is very insensitive to changes in these critical values is due almostentirely to the presence of particles, which both affect the flow of the gas due to viscousdrag and act as very significant heat and mass sources and sinks. As a result, the effects ofthe heat, and mass transfer parameters- SO critical to the flow of a single phase jet -become much less significant for a densely loaded two-phase jet of this type.191The turbulent Schmidt number of the particles, describing the relative tendency of theparticles to spread radially, was found to have a very significant effect on the modelpredictions. Decreasing Sc (increasing the spreading rate) caused the particles to spreadgreatly, and resulted in the particles reacting much more thoroughly. In this case, thepredicted particle distribution no longer matched the experimental results presentedbelow. Presumably, this effect would be observed experimentally if air jets or somemechanical means were employed to enhance the radial distribution of the particles.The sensitivity of the predictions to the effective viscosity was more difficult to investigate, because the turbulent viscosity simultaneously affects all values determined by theturbulent Prandtl and Schmidt numbers. Therefore, to perform this investigation, thethermal and mass diffusivities were held constant while the turbulent viscosity waschanged. Under these conditions, changing the effective viscosity by ± 50 % was found tohave little effect on the calculations.7.2.3 Mesh SizeThe effects of varying axial and radial mesh size have been examined and were found tobe insignificant, provided that the mesh was not reduced below 1000 nodes axially and110 nodes radially. At lower mesh densities, the assumptions of insignificant particle andgas temperature increase over one axial node no longer held, and erroneous predictionsresulted. At mesh densities much larger than 2000 axial nodes by 300 radial nodes, thecomputer time required to perform the calculations increased dramatically, with no corresponding increase in accuracy. In the interest of economy and accuracy, 1800 axial nodesand 143 radial nodes were used. This provided 6 nodes to cover the burner radius, andresulted in an axial inter-node spacing of I mm.1927.2.4 SummaryThe predictions of the model are clearly quite insensitive to many of the assumed inputparameters, due mainly to the presence of particles. The assumed furnace wall temperature was found to have a significant effect on the axial location at which the majority ofthe concentrate ignited, while the gas temperature proved to be less important to theconcentrate at the centre of the reactor.Aside from altering the predicted location of the combustion zone by a few centimetres,none of the changes made to any of the input parameters had any effect on the following:1. The predicted dust generation rate.2. The general reaction degree of the solids - particles near the edge of the jet werealways predicted to combust more thoroughly than those at the centre.7.3 Comparison With UBC Pilot Plant DataA series of calculations were performed with the model in an attempt to simulate each of therun conditions tested in the UBC pilot plant (Table 7.1). The first flash converting simulationwas the base case operation - 2 kg/mm of MK concentrate (as received from Inco) with stoichiometric oxygen. In all of the model predictions with stoichiometric oxygen, the gas recirculating within the UBC flash reactor (outside of the jet region) was assumed to be 100 %SO2. The presence of infiltrated nitrogen in this recirculating gas was ignored (the gasanalyses from the UBC flash furnace have shown that the recirculating gas is normally 85-90% SO2).193In each simulation presented below, the particles and oxygen were assumed to enter thereactor with constant velocity, temperature, and composition distributions across the burnerdiameter. A total of 6 finite-difference nodes were used to cover the burner radius (0.0 158m), resulting in a total of 143 nodes covering the entire 0.25 m cross-section of the flashreactor. Axially, the 1.8m reaction shaft was divided into 1800 nodes, each 1mm thick. Up to10 sub-steps were used to move the particles within these 1mm sections.Table 7.1 shows the total amount of pilot plant data which is available for comparison. Forthe “base case’- 2 kg/mm of unsized concentrate with stoichiometric oxygen- there is alarge amount of data available for verification of the model. As described in Chapter 4, theUBC pilot plant operated smoothly under these conditions (the conditions for which it wasdesigned) and produced reliable data with few operational problems. However, less data isavailable for conditions other than the base case, due in large part to operational difficultieswith the UBC flash furnace caused by the abnormal nature of these run conditions. Nevertheless, these “problems’ with the UBC flash furnace represent unequivocal responses to theseoperating conditions, and as such represent valuable data in themselves. For example, with100 % excess oxygen, the dust generation rate in the UBC flash furnace increased by 5 timesto 25 % of the input rate, which forced early run termination. Similarly, an input concentratewith a larger mean particle size caused very little reaction at all the UBC flash furnace.Therefore, although much of the comparison between the mathematical model and the nonstandard experiments must necessarily be more qualitative, the model must still be capable ofpredicting the response of the UBC flash furnace to these conditions. The primaryquantitative validation of the model however will be by comparison to “base case” runs.194Table7.1:Experimental RunConditionsandDataCollectedRunRunConditionsWallSolidsRadialSolidsGasDustingRatesTemperaturesSamplesPartitioningDistributionSamples(solids)IBaseCaseXB,C,DC,DA-DX2XA-DX3XC,DDA,C,DX4100%ExcessXX550%ExcessXx6LargeParticleXxSize7SmallerParticlesXx8NewShaftXA-DXA-DX(20%_excess)NewShaftXA-DDXA-DX9BaseCase1957.3.1 Base Case TrialsThe first runs of the mathematical model attempted to simulate the “base case” conditions, that is, 2 kg/mm of MK concennate as received from Inco, with stoichiometricoxygen (oxygen-to-concentrate mass ratio of 0.2). This resulted in an oxygen flowrate of0.4 kg/mi corresponding to a superficial gas exit velocity of 26 m/s (burner diameter0.0158 m). The input concentrate was assumed to be composed of spherical particles 20microns in diameter. This particle size (20 microns) was the mean diameter of theconcentrate received from Inco, as indicated by the particle size analyses presented inChapter 4. In accordance with the assumptions made while developing the mathematicalmodel, the input concentrate was assumed to be perfectly mono-sized, with no finer orcoarser particles present. The furnace wall and recirculating gas temperatures determinedin section 7.1 (1160 K and 120() K respectively ) were applied to make the calculationsshown below. Predicted Gas and Particle TemperaturesContour plots of the predicted gas and particle temperatures within the UBC flashfurnace for this first run are shown in Figures 7.2 and 7.3. As these figures indicate,the majority of the particles are predicted to combust in one localized region, whichoccurs quite far down the reaction shaft and corresponds to reactor section D. Thisprediction agrees very well with the experimental observations, in which themaximum wall temperatures for these conditions were consistently found to lie insection D. Visual observations made through the windows into the reaction shaftsubstantiate this with a well defined region of combustion, characterized by brilliantradiant emission, being observed at the viewport in section D.196Figure 7.2 shows that the maximum predicted gas temperature in this region was 1750K, but rapidly dropped to approximately 1300-1400 K by the exit of the reactor. Aswould be expected, the peak particle temperatures (Figure 7.3) are considerablyhigher than the gas temperatures (2200 K) but are still well below the boiling point ofcopper (2836 K).A plot of the regions in which particle temperatures are predicted to exceed the recirculating gas temperature is shown in Figure 7.4 (the black background in this figurecorresponds to regions in which there are no particles or the particle temperature isbelow 1200 K). This figure shows that the particles are predicted to combust in twowidely separated regions. The majority of the particles react in the region corresponding to the maximum gas temperatures of Figure 7.2, but there are a smallnumber of particles which combust closer to the burner nozzle and in so doing, attainthe highest predicted particle temperatures. The particle number concentration in thisinitial region is shown by Figure 7.5 and is clearly very low. This explains why thereis no corresponding gas temperature increase in this region, since the few particlespresent cannot a provide a large source of heat to the surrounding gas.The high particle temperatures in this region can be explained by examining the ratesat which the oxygen and the concentrate in the jet spread radially, and the resultinglocal oxygen-to-concentrate ratio. Under the “base case” conditions the initial oxygento-concentrate ratio of the jet was 0.2 kg 02/kg MK concentrate. However, as Figure7.6 shows, the oxygen which is injected into the flash furnace spreads rapidly as itentrains the surrounding sulphur dioxide gas. This gas entrainment is accompanied byrapid heating of the oxygen near the edge of the jet as it mixes with the surroundingsulphur dioxide at 1200 K. Since the concentrate particles spread radially to a much197smaller extent than does the oxygen, only a few particles are transferred into thisregion, and high local oxygen-to-concentrate ratios result. This is illustrated by Figure7.7, in which the edge of the jet maintains a high local oxygen-to-concentrate ratio.The high gas temperatures at this location result in rapid ignition of the concentrate,while the high local oxygen concentration ensures that the reacting particles attain avery high temperature.Figure 7.7 also indicates that by the time the bulk of the concentrate ignites, the localoxygen-to-concentrate ratio at the centre of the jet has dropped below 0.2, due toentrainment of the surrounding sulphur dioxide. This means that the bulk of theconcentrate cannot generate very high particle temperatures, since the combustion ofthe concentrate will rapidly deplete the surrounding oxygen and hence limit the reaction rate. This effect is clearly seen toward the bottom of Figure 7.6, where thecombusting concentrate has removed all of the oxygen at the centre of the reactionshaft. Particle Reaction DegreeThe predicted concentrate composition as a function of axial and radial position hasbeen plotted in Figure 7.8, where white indicates completely unreacted concentrate,red corresponds to copper metal and green represents Cu20. As was indicated by theparticle temperatures, Figure 7.8 shows that some of the concentrate reacts veryquickly to copper and copper oxide. The majority of the concentrate, however,remains clumped together in the centre of the jet, and does not react until muchfurther down the reaction shaft. It is interesting that Figure 7.8 predicts the formationof ‘rings1’of varied composition at the bottom of the reaction shaft, with largely198unreacted concentrate surrounded by concentric rings of copper metal and copperoxide. This prediction corresponds very closely to the “rings” which were observed inthe UBC flash furnace and which were depicted in Figure 4.18.As with the particle temperatures, the formation of these “rings” can be explained byexamining the temperatures and oxygen potential encountered by the concentrate. Atthe outermost edge of the jet the gas temperature is high, but the oxygen concentrationis low, and hence the concentrate reacts slowly to form copper oxide. Slightly closerto the centreline of the furnace, the oxygen concentration is higher and the localoxygen-to-concentrate ratio is high, causing rapid combustion to form metalliccopper, and then further reaction of this copper to form Cu20. At the centre of thereactor, the local oxygen-to-concentrate ratio at the point of ignition is below the stoichiometric value of 0.2, and hence complete combustion to metallic copper is notpossible. Once the concentrate ignites, it foms metallic copper as described inChapter 5. However, the rapid local depletion of oxygen accompanying the oxidationof the copper sulphide forces much of the concentrate at the centre of the shaft toremain unreacted. Gas Composition : Comparison to Measured ValuesThe composition of the gas in the UBC flash reactor as predicted by the mathematicalmodel has been compared with the available experimental data in Figure 7.9. Twosets of model predictions are presented in this figure1. Stoichiometric oxygen injection (an oxygen/concentrate mass ratio of 0.2).2. 20 % excess oxygen, (an oxygen/concentrate mass ratio of 0.24).199The predictions of the centreline gas composition resulting from these two conditionscorrespond to the solid lines in Figure 7.9, while the dashed line represents modelpredictions for stoichiometric oxygen at 4.3 cm off-centre.Three sets of experimental data are presented in this figure which were taken underthe following conditions1. Runs in the original reaction shaft with stoichiometric oxygen.2. Runs in the new reaction shaft with:a) stoichiometric oxygenb) 20 % excess oxygenThe model predictions agree very well with the data from the new reaction shaft, butthe agreement is not as satisfactory for data from the original shaft. This is most likelydue to experimental difficulties1. The positioning of the gas sampler was quite inaccurate for all of the runs with theoriginal reaction shaft. As a result, it is likely that many of the gas samples weretaken some small distance (a few centimetres) off the centreline of the reactor.This error was due mainly to irregularities of the original shaft and the presence ofthe additional alumina wool refractory, which made exact positioning of thesampler difficult. The large change of gas composition with radial position calculated by the model and indicated by the dashed line in Figure 7.9 shows that anyslight error in positioning would result in large errors in the measured gascomposition, particularly in reactor section A.2002. Much more accurate positioning of the gas sampler was achieved with the newreaction shaft. When this shaft was constructed and installed, great care was takento ensure that the gas sampler could be accurately positioned at the centre of thereaction shaft.The calculations presented in Figure 7.9 indicate that a 4.3 cm error in the radial positioning of the sampler could result in a 50-80 % error in the measured centreline gascomposition at section A. The closer agreement between the model predictions andthe data from the original reaction shaft at sections C and D is observed because thewidening of the jet reduces radial composition gradients, and hence minor errors inpositioning do not have as strong an effect on the measured gas composition at thelower sections of the reaction shaft.Figure 7.9 therefore indicates that the values of gas effective viscosity and turbulentSchmidt number used by the model are reasonable, and that the model is capable ofpredicting the composition of the gas within the chalcocite flash flame. Solids Distribution : Comparison to MeasurementsPlots of the predicted radial distribution of solids in the UBC flash furnace arecompared with experimental results in Figure 7.10. There is no experimental datafrom the first reactor section as it was never possible to obtain any amount of solids inthis region, possibly due to locally high gas velocities and low solids temperatures.Figure 7.10 shovs that the predicted particle distribution agrees reasonably well withthe small amount of experimental data available. However, in reactor section B (0.675m from the burner tip), the model fails to predict the bimodal solids distribution201encountered experimentally. It is likely however that this distribution is an artifact ofa poor sampling technique and is not an accurate representation of the true solidsdistribution. As shown in Figures 7.3 and 7.4, particles near the edge of the jet willhave a much higher temperature than those at the jet centre. As a result, particles atthe edge of the jet are much more likely to be partially molten and hence will tend toadhere more readily to the solids sampler. Further down the reaction shaft when manymore particles are much hotter, this bimodal distribution is not observed.Despite this minor disagreement, Figure 7.10 indicates that the model is apparentlycapable of predicting the general distribution of solids in the reaction shaft. The solidsare tightly grouped near the centre of the reactor and slowly move radially outwardwith distance down the shaft. However, even at the bottom of the flash furnace, themodel predicts that the solids will only occupy about 20-30 % of the total shaft cross-sectional area. Clearly if one were concerned with the utilization of the shaft area, asolids distribution of this type would be most unsatisfactory. Radially directed gas jetsor a solids distribution cone as employed in some Outokumpu flash furnaces (23,24)could presumably overcome this problem. Solids Samples Composition Comparison to MeasurementsThe single most critical test of the mathematical model is the comparison with thecompositions of the solids samples, because the fundamental objective of the model isto predict concentrate oxidation. To facilitate comparison with the mathematical202model, the assay values were normalized with respect to Cu, S, and 0, thus removingthe effects of the minor amounts of Fe (< 1 %) and Ni (<6 %) present in the concentrate.An additional factor complicating the comparison of the model predictions with theexperimental data is that both the solids composition and the total mass flux varystrongly with radial position. The solids samples recovered from the UBC flashfurnace were collected over varying radial positions using non-isokinetic sampling,and hence the sample assays reflect a mass-averaged composition. For example, mostof the solids designated as being collected at the centre of the UBC flash furnace werein fact recovered from the centremost 2.5-3.8 cm of the solids sampler. In order forthe predictions of the mathematical model to be compared with this data, the modelpredictions of the solids composition between the centreline and 3.8 cm must becombined with the corresponding predicted local mass flux for each location, summedand then averaged to yield the predicted mean solids composition at this location.The predictions of the centreline solids compositions derived using this techniquehave been plotted together with the assays in Figure 7.11. As this figure indicates, themodel predicts a sharp decline in sulphur (and a corresponding increase in copper andoxygen) between the third and fourth sample locations, resulting from combustion ofthe bulk of the concentrate in this region. Considering that the model predictions havebeen made for a mono-sized concentrate of 20 microns diameter, and that no effort to‘tune’ the model to the data has been made at all, the agreement between the modeland the experiments is quite good. The most serious disagreement between the modelpredictions and the measurements occurs at the second reactor section, where thepredicted percentage of sulphur is too high by approximately 20-25 %. This disagree-203ment is most probably an artifact of poor sampling rather than an indication that themodel is greatly in error. It was always very difficult to obtain a solids sample at thesecond reactor sampling port (and impossible to obtain one at the first sample port),regardless of the length of time the sampler was allowed to remain in the reactionshaft. As explained above, this was due to the low temperature and high velocity ofthe solids in these regions. Molten solids would tend to adhere to the sampler, whilecompletely solid particles would most probably simply bounce off. Before ignition,and the subsequent heating of all solids, the material captured by this sampler is therefore not representative of the solids as a whole, with hotter, more completely reactedparticles, being preferentially sampled relative to colder and less well reacted solids.The radial variation of solids composition as calculated by the model has beencompared with the experimental data in Figures 7.12a-c and 7.13a-c. As shown inTable 7.1, experimental data on the radial variation of solids sample composition wasonly available for the last two reactor sections. As before, the model predictions havebeen derived by mass-averaging over finite radial regions to facilitate comparisonwith the experimental data. The experimental data from all runs with stoichiometricoxygen have been averaged together to yield the solid lines in Figures 7.12a-c and7-1 3a-c, with the range of experimental values indicated by error bars. The experimental data represents assays of solids collected over a range of radial positions, andtherefore this method was adopted to present the data rather than plotting discretepoints.In general, Figures 7.1 2a-c and 7.1 3a-c indicate that there is good agreement betweenthe model predictions and the small amount of experimental data. It is interesting to204note that in most cases, the model clearly predicts the strong variation in solidssample composition with radial position which is shown by the experimental data. Infact, the radial variation of sample composition is often much greater than the axialvariation. The mathematical model suggests that there are two major reasons for this:1. Particles near the edge of the oxygen-concentrate jet have a lower axial velocitythan those near the centre, and hence have a longer residence time in the reactor.2. As discussed previously, there is a strong radial variation in both oxygen concentration and temperature within the oxygen/concentrate jet. The particles at theedge of the jet encounter mostly sulphur dioxide at high temperature, and theoxygen concentration in this region is quite low. As a result, particles at the edgeof the jet react slowly - ‘roasting”- to form copper and copper oxides. Particles atthe centre of the jet experience low temperatures and residence times in association with a rapidly diminishing oxygen concentration, as oxygen is removed byreaction with the high local concentration of particles. Again, this results inrelatively low levels of oxidation. In intermediate regions between these twoextremes, both temperatures and local oxygen-to-concentrate ratios are muchhigher, which results in almost complete combustion to copper metal, followed bysubsequent reaction to the oxide if sufficient oxygen is present.Figures 7.1 2a to 7.1 2c indicate that the solids at the edge of the jet have reacted toform Cu20 by the time they have reached the third reactor section. However, Figures7.1 3a to 7.1 3c also show that much of the concentrate at the centre of the reactorremains largely unreacted, even at the last reactor section. The bath reaction between205the sulphur-rich and oxygen-rich concentrate fractions is clearly important if significant quantities of metallic copper are to be produced, since much of the oxygenentering the reactor reaches the hearth as Cu20. Dust Generation RateThe experimental trials with stoichiornetric oxygen have shown a consistent dustgeneration rate of approximately 5 % of the input mass. In Chapter 5 it was suggestedthat dust generation was associated with particle fragmentation caused by boilingcopper within individual concentrate particles. The results presented in Figure 7.4show that none of the concentrate particles are predicted to attain such high temperatures, and therefore the total dust generation rate for stoichiometric oxygen ispredicted to be zero. The causes of this disagreement with the experimental data areanalyzed in Chapter Model Comparison with UBC Furnace : Non-Standard Conditions7.3.2.1 Smaller Particle SizeThe effects of a smaller particle size was investigated using the mathematical model.The particle size distribution of this concentrate has been presented in Figure 4.10,and has a mean particle size of approximately 10 microns. As shown by Table 7.1,wall temperature measurements and approximate dust generation rates are availablefor comparison.The conditions which were used to perform the calculations were identical to those206employed previously, except that the input particle diameter was assumed to be 10microns. In this case, the bulk of the concentrate was predicted to combust 0.8-1.0 mfrom the burner tip, corresponding to reactor sections B-C. A plot of the reactor walltemperatures from the pilot plant run with the smaller particle size is shown in Figure7.14. This figure indicates that the small size fraction concentrate does indeed burn ata reduced axial distance compared to the regular MK concentrate, and in reactorsections B-C. Large Partc1e SizeThe effects of increased mean particle size was also investigated by the mathematicalmodel and compared with experimental data. The size analysis of the larger particlefraction as used in the pilot plant trial was shown in Figure 4.11, and as this figureindicates, the mean particle diameter of this concentrate was approximately 40microns. The model calculations for this larger particle size were therefore madeassuming a mono-sized concentrate of 40 microns diameter.The mathematical model has predicted that concentrate particles of this size will reactonly to a very limited extent in the UBC flash furnace. The model indicates thatparticles in the central region of the reaction shaft (the majority) will not ignite, butthat some concentrate at the edge of the jet will react to form copper metal and copperoxides. The hearth composition predicted by the mathematical model has beencompared with assays of the hearth composition in Table 7.2. This data shows goodgeneral agreement between the model and experiment, and also illustrates the consid207erable quantity of unreacted sulphur entering the hearth in this run. The reactor walltemperatures from this run have been plotted in Figure 7.15, and show a generalcooling of the flash reactor associated with low combustion rates.It is interesting to note that flO dust at all was produced during this run. This was theonly trial in the UBC flash reactor which produced no dust whatsoever. As with thebase case run, the mathematical model predicted that no particles would reach theboiling point of copper, and hence the predicted dust rate for this case was zero.Table 7.2: Predicted vs Measured Hearth Composition from Run with Large Particles%Cu %S %OMeasured 80.7 16.8 2.5Predicted 81.5 17.7 Excess Oxygen TrialAs described in Chapter 4, two runs were performed in the UBC pilot plant withexcess oxygen to investigate the effects of oxygen on dust formation rates. Theseconditions tested the effects of 50 % and 100 % excess oxygen, and utilized a largerdiameter burner to provide similar exit velocities to previous experimental trials.Unfortunately, neither solids nor gas samples were taken in the experimental trialsdue to their short duration, and hence the only data available for comparison betweenthe experimental data and the mathematical model are the dust generation rates. Therate of dust generation in the runs with excess oxygen was much greater (at least afactor of five) than that observed in the runs with stoichiometric oxygen.208The run with 100 % excess oxygen was selected for comparison with the mathematical model, as this experimental trial both showed the most dramatic effects on dustgeneration and maintained the same gas exit velocity as in the trials withstoichiometric oxygen (26 mIs). The input conditions to the model were as follows:D = 20 t - monosizedBurner Diameter = 0.0224 rnOxygen-to-concentrate Ratio = 0.4The composition of the recirculating gas in the UBC flash smelter poses some problems for the mathematical modelling of these experiments, as it was not measured.Because of the large quantities of oxygen present in this run, it is likely that most ofthe concentrate input to the reactor ultimately forms Cu20, depleting some of theexcess oxygen. As a result, the recirculating gas is probably more deficient in oxygenthan the 50 % S02-5 0 mixture suggested by combustion to metallic copper. Inthe absence of experimental data, the recirculating gas composition was assumed toresult from complete reaction to Cu20, resulting in a gas composition of 66.6 % SO2and 33.3 % 02 (effectively the lowest possible oxygen concentration).Under these conditions, the model predicted that the dust generation rate would be 10% of the mass input rate. This behaviour is in contrast to the previous runs with stoichiometric oxygen, and accurately reflects the increase in dust generation observedexperimentally when the UBC flash furnace was run with excess oxygen. However,this predicted dust generation rate was only approximately one-half of the experimentally observed rate (25 % of the mass input rate). This discrepancy will be discussedin detail in Chapter 8.2097.4 Summary of Runs Simulating the UBC Flash FurnaceWithin an acceptable degree of uncertainty, the mathematical model has been found to havebroad agreement with most of the data obtained from the UBC flash furnace. This dataincludes the location of the ignition region of the concentrate, the gas composition, and, mostimportantly, the composition of the solids as a function of both axial and radial position. Thisagreement indicates that the considerable number of simplifying assumptions which aided thedevelopment of the model appear to be generally valid, and that the model is a reliablepredictor of the solids composition entering the hearth. The model also successfully predictsthe effect of increased or decreased mean particle size on the behaviour of the UBC flashfurnace.The model is somewhat more limited in its ability to predict the observed dust formationrates in the UBC flash reactor. There is general qualitative agreement for the case of excessoxygen, in which the model successfully predicts that these conditions will result in a sharpincrease in dust generation. However, no dust whatsoever was predicted for the runs withstoichiometric oxygen, which does not agree with the experiments, in which 3-5 % of theinput mass was normally recovered as dust. For the trials with stoichiometric oxygen, themodel predicts that the maximum particle temperatures (and hence the region most likely togenerate dust) should occur close to the burner nozzle and near the edge of the oxygen/concentrate jet, where the local oxygen-to-concentrate ratios are high. The role this region mayplay, and the influence of other factors such as recirculating gas temperature and particle sizeon dust formation is examined in Chapter 9.Figure 7.1 Radiation circuit diagram for the UBC flash reactor, illustrating the interaction between the energy emitted by the particles, the walls and the gas.2101A F,p (1 - Eg)Eb,wall Eb,particle1A Fw,g Eg1A Fg,p EgEbgas2111.8 mSection ASection CTemperature Sc ale1750 K875KFigure 7.2: Calculated gas temperatures in the UBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen]1313 KSecon D438K212Section BSecion CTemper attire Scale2200 K1725 K1250 K775 KFigure 7.3 Calculated particle temperatures in the UBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiometric oxygen]Section ASectianD298K213Section ASection BSecnonCTemperature Scale2200 K1725 K1250 K775 K298KFigure 7.4: Regions in which the calculated particle temperatures in the UBC flashreaction shaft exceed the recirculating gas temperature (1200 K) for the basecase operating conditions [2kg/mm unsized concentrate, stoichiometricoxygen jSection 1)214Section ASection 3Section CParticle Number Density142/ mm371/ mm3Figure 7.5: Calculated particle distribution in the UBC flash reaction shaft for the basecase operating conditions [ 2kg/mm unsized concentrate, stoichiometricoxygen I284/ mm3213 / mm3Section 1)0215Section ASection BSection C0% 50225 % 50250% SO275 %SO100 %S02100 % 075% 0250% 0225% 020% 02Figure 7.6: Calculated oxygen concentration in the JJBC flash furnace for the base caseoperating conditions [2kg/mm unsized concentrate, stoichiomethc oxygen]Gas CompositionSectionD216Section ASection BSection COxygen-to - ConcentrateMass Ratio> 0.400.300.20 = Stoichiometric0.10Figure 7.7 Calculated local oxygen-to-concentrate ratio in the UBC flash furnace for thebase case operating conditions [2kg/mm unsized concentrate, stoichiomethcSectionD<0.01oxygen I217Edge of JetSection BSection CSolids Composition2-Figure 7.8: Calculated solids composition in the UBC flash reaction shaft for the basecase operating conditions { 2kg/mm unsized concentrate, stoichiometricoxygen]Section ACoSection D4—.,0.90.5-0.4-0.3-0.2-0.1 -a-Figure 7.9:4E2181.8Comparison between predicted and measured axial variation of gas composition in the UBC flash reaction shaft. Predictions are shown for stoichiometricoxygen at radial positions of 0 (centre), and 4.3 cm, and for the centreline at20 % excess oxygen. Experimental data shown for original reaction shaft(stoichiometric oxygen) and new shaft (20 % excess oxygen).% Excess Oxygen0.8>o>NIModel PredictionsCentreline CompositionComposition : r 4.3 cmExperimental DataStoichiometric Oxygen7New Shaft• 20 % Excess Oxygen J* Stoichiometric Oxygen Original ShaftStoichiomethc Oxygen*0 02 0.4 0.6 0.8Axial Position (m)-[2 1.4 1.62190z.0COCCO00InNI0zFigure 7.10: Comparison between predicted and measured solids distribution in the UBCflash furnace for the base case operating conditions { 2kg/mm unsizedconcentrate, stoichiometric oxygenRadial Position (m)22090Cu8070.0Experimental Data60÷ Mass Percent Copper50s Mass Percent Sulphur0Mass Percent Oxygen40__________________0_ __ _ ____ _ __ _ _ __ _ ____30 — Model Predictions— So___________00 0.2 0.4 O’.6 0:8 1’.2 1’.4 1’.6 1.8Axial Distance (m)Figure 7.11 Comparison between calculated and measured centreline composition ofsolids samples for the base case operating conditions [ 2kg/mm unsizedconcentrate, stoichiometric oxygen 1908988&87C-)86C.8584____________________________838281obi ob2 0b3 ob4 abs o.b6 o.b-i 0.Radial Position (m)12C10CFigure 7.12: Comparison between calculated and measured radial variation of solidssamples recovered from reactor section C for the base case operating conditions I 2kg/mm unsized concentrate, stoichiometric oxygen I a. Copper b.Sulphur c. Oxygen221— Experimental Data- Model PredictionsC1614(a)8(b)I42— Experimental Data- Model Predictionsobi 0.b2 0.ö3 0.b4 abs 0M6 0.b7 0.Radial Position (m)CCoa-’aC.Radial Position (m)94 —Figure 7.13:90aa.(3 888684Comparison between calculated and measured radial variation of solidssamples recovered from reactor section D for the base case operating conditions j 2kg/mm unsized concentrate, stoichiometric oxygen j a. Copper b.Sulphur c Oxygen92222rLz Model Predictioos(a)12(b)8280o.b o.b4 o.b6 o.bs 0.Radial Pos,uon (m)14 -I10a.642— Expeameatal Data. Model Prodictionso.b2 0.b4 0.t)6 01)8 0.1Radial Position (m)0.120Radial Position (m)(DC—)FFigure 7.14: Measured values of flash furnace wall temperature from run with smallparticle size, corrected for the effects of the insulating blanket.22330Elapsed Time (mm)224Figure 7.15: Measured values of flash furnace wall temperature from run with largeparticle size corrected for the effects of the insulating blanket.z4)UI______Section B_______Section CSectionDElapsed Time (mm)2258 Industrial Calculations : Inco Port Colborne Flash ReactorIn this chapter, the predictions of the mathematical model will be compared with the experimental data obtained from the Port Colborne Flash Reactor (previously discussed in Chapter 4).This comparison between the model and the experiments had two overall objectives:1. To determine whether the mathematical model (developed for and validated by the UBC pilotplant) could successfully predict the combustion of MK concentrate in the Port Colbornesmelter.2. To compare the similarities and differences of the Port Colborne pilot reactor with respect tothe UBC pilot plant.To accomplish these objectives, the mathematical model was modified slightly to accomodatethe different geometry and operating conditions of the Port Colborne smelter. Once modified, themathematical model was run under all conditions tested experimentally, and the output of themodel was compared with the available data (solids and gas samples).The large scale of the Port Colborne flash reactor relative to the UBC pilot plant provides aunique opportunity to test the mathematical model under conditions more closely approximatingindustrial practice. Because of the large scale of this process, and because the Port Colbornesmelter is primarily a production unit, the data obtained from this vessel is, of necessity, somewhat less precise than that originating from the UBC pilot furnace. However, if the model is tobe accepted as providing a valid description of chalcocite combustion, then it must be capable ofpredicting, within the experimental error, the general composition of the solids and gas samplestaken from the Port Colborne smelter.2268.1 Experimental Data from the Port Colborne Flash ReactorThe technique by which gas and solids samples were recovered from the Inco Port Colbornepilot plant has been discussed in detail in Chapter 4. For clarity, a summary of the experimental data together with the conditions under it was taken is shown in Tables 8.1 and 8.2.Gas samples were taken at a variety of axial and radial positions, and under several differentexperimental conditions; solids samples were collected from material which accumulated onthe gas sampling probe. Thus these “solids” samples are more accurately described ascondensed-phase samples, since both solid and liquid products in the flame would adhere tothe gas sampler in this manner.A preliminary analysis of this data was performed in Chapter 4 and indicated that the largequantity of N2 present in the Port Colborne smelter was due to air infiltration, primarily fromthe central gas offtake. Calculations based on these gas samples indicate that approximately25 % of the oxygen needed to combust the concentrate was supplied by this infiltrated air.The dust generation rate in the Port Colborne smelter is quite difficult to quantify, as significant quantities of the dust were leached by quenching water. However, discussion with PortColborne personnel indicate that this dusting rate was approximately 10 % of the solids inputrate.8.2 Factors Complicating Comparison of the Model with ExperimentThere are several factors which make comparison between the mathematical model and theexperimental data from the Port Colborne pilot plant difficult2271. A number of parameters which are important to the mathematical model were eitherunknown, or known only to a limited degree of precision. These included:(a) The inclination angle of the burner. As described in Chapter 4, the concentrateburner was inclined at an angle (approximately 8) to the horizonal, but this anglewas not known precisely. Moreover, the axial and radial location of the sampleports relative to the burner were known only to a precision of at best ±50 mm.(b) The temperatures of both the recirculating gas and of the inside walls of thesmelter. However, good temperature measurements of the bath were available, asPort Colborne personnel measured this periodically.2. More importantly, the design of the Port Colborne flash reactor was such that the assumption of a constant recirculating gas composition was not valid. As shown in Chapter 4, thesignificant air leakage into the flash furnace, the presence of the two natural gas burnersat widely separated positions, and the SO2 arising from the bath together cause a considerable change in the reactor gas composition with position. For example, at sample port 1,the gas outside of the oxygen-concentrate jet was composed of approximately 65 % N2,30 % so2, 5% CO2. At sample port 4, this composition had changed to 45 % N2, 45 %SO2, 10 % CO2 due to the entrainment of SO2 arising from the bath and CO2 from thenorth burner. Quantifying this recirculating gas composition as a function of axial andradial position was not possible with the few sampling locations available for the industrial measurements.These two points suggest that a direct comparison between the model predictions and theexperimental data may be subject to considerable error. In particular, point 1(a) above228suggests that the location of the samples cannot be determined with great precision.However, this difficulty can be overcome by comparing the experimental data with the modelpredictions over a range of locations which bracket the possible positions of the sample ports.Point 1(b) can be overcome similarly, by making predictions for a range of assumed furnacewall and gas temperatures. The compositions of the samples shown in Tables 8.1 and 8.2vary little with run conditions, which suggests there is little sensitivity to changes in assumedgas and wall temperatures. As a result, estimates of these temperatures should providereasonable predictions.The air infiltration, and the subsequent variation in recirculating gas composition as illustrated by the gas samples, causes further problems, because the mathematical model haspreviously assumed that the recirculating gas is of constant composition in is 100 % SO2.Despite this, some comparison between the model and the data from the Port Colborne pilotplant is still possible for the following reasons1. The experimental data of Otero et al. (33) have indicated that combustion of MK concentrate in a 50 % 02 - 50 % N2 gas mixture is virtually identical to that occurring in 50 % 02- 50 % SO2. In effect, the SO2 or N2 behave as an inert diluent in determining MKignition, and therefore only the oxygen concentration of the gas surrounding the particlesis of major importance- it matters little what dilutes the oxygen. Therefore, although theactual gas composition in the Port Colborne smelter may vary substantially with position,it is not likely that this greatly affects the combustion of the MK concentrate, since theoxygen concentration in this gas is low.2292. It is not likely that any oxygen in the entrained air significantly affects the behaviour ofsolids at the centre of the oxygen-concentrate jet. As demonstrated in the UBC flashfurnace, solids at the outer regions of the oxygen-concentrate jet will readily react toCu20 if supplied with sufficient oxygen, and hence will tend to remove oxygen from theentrained air before it reaches the centre of the jet (entrained oxygen may still affectsolids at the edge of the jet however).An additional difficulty is due to the construction of the Port Colborne burner, which differssubstantially from the UBC burner, and may produce an asymrnetical oxygen-concentrate jet.As discussed in Chapter 4, the oxygen enters the Port Colborne burner under a baffle,producing very high local velocities (and generating suction in the concentrate feed tube). Inthe absence of particles, this baffle has been shown (50) to produce an asymmetrical velocitydistribution across the exit of the burner. The effect that concentrate particles would have onthis velocity distribution is unkonwn, but it is likely that the particles would flatten thevelocity distribution considerably due to enhanced radial momentum transport resulting fromgas-particle drag. That is, particles projected from a region of high gas velocity to one oflower gas velocity would result in a net momentum input to the low-velocity region. Similarly, particles entering a high-velocity region from a region of lower gas velocity wouldprovide a net momentum deficit to this high-velocity region. Recent experiments in the PortColborne smelter with a more symmetrical burner have shown no significant change in theperformance of the smelter, indicating that the effects of this asymmetrical oxygen entry arelikely to be slight. Therefore in the simulations presented below, the particles and gas wereassumed to exit from the burner with constant velocity and composition across the crosssection (symmetrical burner).230It is reasonable to expect, therefore, that the model should be capable of predicting thegeneral composition of solids recovered from the Port Colborne smelter within experimentalerror. The model is somewhat more restricted in its ability to calculate the composition of thegas samples. However, assuming that the gas entrained by the jet varies in relative quantitiesof N2, SO2 and CO2. but does not contain any 02, then the model should still be capable ofpredicting the°2 concentration of the gas samples withdrawn from the centre of the jet.8.3 Modifications to the Model to Simulate the Port Colborne SmelterThe following modifications were made to the mathematical model to allow predictions ofthe Port Colborne flash reactor:1. Based on the gas analyses presented in Table 9.1, the recirculating gas in the PortColborne smelter was assumed to be 50 % N2, 40 % SO2 and 10 % CO2. The heatcapacity, density and viscosity of this infiltrated gas was calculated as a function oftemperature based on this assumed composition. The presence of any oxygen in the infiltrated gas was neglected. This differed from the UBC simulations in which the recirculating gas was assumed to be 100 % SO2.2. The horizontal orientation of the burner jet was simulated by removing the gravitationalsource term in the axial particle momentum balance. The radial movement of particlescaused by gravity was ignored as were buoyancy effects. Strictly speaking, this description is valid only over the initial region of the Port Colborne oxygen/concentrate jet. Inregions close to the burner tip (within 10 to 15 nozzle diameters), the momentum of thegas exiting the burner is dominant, and the effects of particle drag and buoyancy in thisregion are minor. As the momentum of the gas dissipates at increasing axial distances231under the influence of turbulence, these factors become more significant. Fortunatelyhowever, much of the combustion of the MK concentrate was found to occur near thefirst sample port, at an axial distance of approximately 8-13 nozzle diameters, and therefore this description is valid.3. Due to the complex flow pattern within the Port Colborne flash reactor caused by airinfiltration and the presence of the natural gas burners, no attempt to close the globalmass balance for the gas was made, and hence calculations were performed assuming thatthe two-phase MK/oxygen jet behaved as an unconfined (free) jet. There is little error inthis assumption (particularly over the initial region of the jet), given the small size of theconcentrate burner compared to the cross-sectional area of the vessel.4. Calculations of particle blockage effects (presented for the UBC pilot plant in Chapter 6and described in detail in Appendix 2) have been extended for the Port Colborne pilotplant. Because of the greater diameter of the Port Colborne burner jet relative to the UBCjet (5.3 - 6.3 cm compared to 1.6 cm in the UBC furnace) particle blockage is moresevere, and the wall-particle view factor was calculated to be 0.5 in this case.8.4 Sensitivity AnalysisThe unknown parameters (furnace gas and wall temperatures) were varied together with theparticle blockage effects to determine the sensitivity of the model calculations to thesevalues. This sensitivity analysis was performed for both burners tested in the Port Colbomesmelter (53 mm and 63 mm ) and little difference was observed. The particle diameter wasassumed to be constant at 20 microns.2328.4.1 Assumed Furnace Gas and Wall TemperaturesAs stated above, little information was available on the gas and wall temperatures in thePort Colborne smelter. The temperature of the semi-blister bath has been measured byPort Colborne personnel to be 1570 - 1620 K. Outside of the oxygen-concentrate jet, thegas temperature in the furnace is more difficult to assess, due to the presence of theoxygen/natural gas burner (when lit) directly adjacent to the oxygen/MK concentrateburner, and the large volumes of infiltrated ambient air. Based on the information fromthe UBC flash furnace, it is likely that the gas and wall temperatures are of the same orderof magnitude and as a first approximation, the gas and wall temperatures were bothassumed to be 1600 K. Given the uncertainty of these values, calculations were alsoperformed for assumed gas and wall temperatures of 1500 K and 1700 K to examine thesensitivity of the model predictions to these values.The predicted gas and solids compositions at the first sample port were not significantly(± 5 % maximum) affected by these changes. This was mainly due to the fact that muchof the solids had combusted by the first sample port (for all three assumed temperatures),and hence similar solids and gas compositions were predicted. This is illustrated by a plotof the predicted particle temperatures within the burner jet (Figure 8.1) which showsmuch of this combustion occurring before the first sample port. Changes in the assumedgas and wall temperature did have a slight effect on the predicted dust generation rate.The effect of temperature on the dust generation rate will be discussed further in Chapter9.2338.4.2 Particle Blockage EffectsCalculations were performed both the particle blockage factor varied from 0.3 to 0.8.Again, little difference (± 5 % maximum) was observed in the predicted solids composition, because in both cases concentrate ignition had commenced by the time the concentrate reached the first sample port. The high gas temperature (1600 K ) was found toprovide sufficient heat to ignite the particles, even when wall-particle radiation wasentirely neglected.8.5 Comparison Between Model and the Experimental Data8.5.1 Initial and Boundary ConditionsThe initial conditions which were used for comparison with the experimental data were asfollowsD = 20 micronsT0 Tg,o = 298 KBurner diameter = 0.05 3 m, 0.063 mUpo = Ug,o = 24 mIS, 38 rn/sOxygen/Concentrate Ratio = 0.18The temperature of the gas surrounding the oxygen-concentrate jet and the temperature ofthe reactor walls were both assumed to be constant at 1600 K.2348.5.2 Comparison Between Predicted and Measured Gas CompositionAs described above, the model is not capable of predicting the compositions of SO2,N2and CO2 in the Port Colbome flash reactor, but should be able to calculate the oxygenconcenation. The oxygen concentrations predicted by the model have been comparedwith the experimental data in Figure 8.2. Because of the error associated with locating thegas sampler, the predictions of the model have been plotted as a solid line representingthe range of predicted gas composition over the radial regions in which the sampler waspositioned. The predictions shown are for the 0.05 m diameter burner, but are virtuallyidentical to the predictions for the 0.0635 m burner. As Figure 8.2 indicates, the agreement between the model and the experimental data is quite reasonable, given the uncertainty in the positioning of the sampler.8.5.3 Comparison Between Predicted and Measured Solids CompositionThe solids sample composition predicted by the model have been compared with theexperimental data in Figure 8.3. As with the gas compositions, the calculated solidssample data has been plotted as a series of vertical lines, indicating the range of valuespredicted by the model over the regions covered by the sampler. The agreement is quitereasonable within the experimental error.It is significant that the solids at the edge of the jet are primarily composed of Cu20,while those at the centre (sample port I) retain relatively high (15 % by mass) amountsof sulphur. Thus, the oxygen-concentrate jet in the Port Colborne smelter clearly exhibitssimilar characteristics to those observed in the UBC flash furnace. In particular, the over-235oxidation of solids at the edge of the jet and the under-oxidation of solids at the centre isseen clearly. This indicates that the dissimilar radial spreading rates of the solids and thegas must play a strong role in combustion of solids in the Port Colbome flash reactor.8.5.4 Dust GenerationUnder all run conditions tested (large and small burners and all gas and wall temperatures), between 10 % and 15 % of the concentrate particles in the Port Colborne flashreactor were predicted to generate dust (Figure 8.1). This contrasts with the calculationsfor UBC flash furnace in which 20 micron particles were not observed to dust. This behaviour is not unexpected, given the higher temperature of the gas in the Port Colbomesmelter relative to the UBC operation. As discussed in Chapter 5, and illustrated byFigure 5.21, an increased temperature should increase the tendency of particles to dust.The dust generation mechanisms within the Port Colborne and UBC smelters will bediscussed in detail in Chapter 9.8.6 SummaryThe mathematical model has been found to predict the compositions of the solids and gassamples taken from the Port Colborne flash furnace with reasonable accuracy. As a result, itmay be concluded that the combustion phenomena in the Port Colborne flame are adequatelydescribed by the model. In addition- and in con strast to the UBC flash furnace- the mathematical model predicts that 10-15 % of the solids in the Port Colborne flame will reach theboiling point of copper. Assuming that all of these solids ultimately fomi dust, the dustgeneration rate is therefore predicted to be between 10-15 % of the solids input rate, which isof the same order of magnitude as the dust generation rate observed experimentally.236The validation of the mathematical model has therefore been completed. The model will beapplied as a predictive tool in Chapter 9, both to elucidate the fundamental phenomenaaffecting the combustion of chalcocite concentrate, and to provide potential solutions to theproblem of dust generation.237Table 8.1: Operating Conditions and Analyses of Gas Samples from Port Colbome FlashReactorRun Sample Probe Flash South North 02 % N2 % SO2 % CO2 % SumPort Depth Gun I.D. Burner Burner(cm) (cm)1 1 108 6.3 on on 1.4 54.5 24.8 19.1 99.81 1 108 6.3 on on 1.2 60.1 23.5 15.2 1002 1 108 6.3 on on 2.9 31.2 22.8 43.1 1002 1 108 6.3 on on 3.3 48.6 28.8 80.73 1 108 6.3 off on 7.3 59.7 22.5 11.7 101.23 1 108 6.3 off on 6.2 53.4 26.1 10.9 96.64 4 76 6.3 on on 2.9 32.5 32.1 32.1 99.65 1 91 6.3 off on 6.6 55.1 24.7 13.6 1005 1 91 6.3 off on 7.3 55.9 21.3 12.8 97.36 2 108 6.3 on on 5.8 53.9 31.2 9.2 100.18 1 108 6.3 off on 16.6 62.9 18.2 1.8 99.59 1 108 6.3 off off 15 60.9 21.2 0.2 97.39 2 91 6.3 off off 4.9 57.7 37.5 0.3 100.49 4 91 6.3 off off 3.2 43.2 49.3 0.4 96.112 1 107 5.3 on on 15.3 58 24.2 10.6 108.112 1 91 5.3 on on 7.7 55 30 12.6 105.314 1 91 5.3 off on 8.4 61.9 26.3 3.4 10014 1 99 5.3 off on 9.6 72 19.3 2.7 103.614 1 107 5.3 off on 16.4 56.7 24.2 2.6 99.914 1 107 5.3 off on 10.5 56 23.9 2.6 9314 1 91 5.3 off on 5.1 58.3 32.7 3.8 99.914 2 107 5.3 off on 13.5 59 21.3 2 95.814 2 91 5.3 off on 4.6 58.6 33.5 3.3 10014 4 107 5.3 off on 1.7 47 47.6 5 101.314 4 91 5.3 off on 4.7 43 48.4 5.6 101.715 1 107 5.3 on on 20.3 57.1 16.3 6.3 10016 1 107 5.3 off on 13.4 58.8 24.9 3 100.116 1 91 5.3 off on 6.7 63.5 24.9 3.7 98.816 1 99 5.3 off on 6.9 62.2 28.2 4.8 102.116 1 107 5.3 off on 7.3 58.1 32.9 5.2 103.516 2 107 5.3 off on 14.8 50.9 28.4 3 97.116 2 107 5.3 off on 3.2 59.6 34.7 5 102.516 4 107 5.3 off on 2.1 41.5 51.1 5.6 100.316 4 91 5.3 oft on 2.3 42 50.1 5.6 100238Table 8.2: Normalized Assays of Solids Samples from Port Colborne Flash ReactorRun Sample Probe Flash South North %Cu %SPort Depth Gun 1.D. Burner Burner(cm) (cm)2 1 108 6.3 on on 74.8 12.5 12.72 1 108 6.3 on on 83.6 7.1 9.33 1 108 6.3 off on 72.9 5.4 21.76 2 108 6.3 on on 89.4 5.5 5.07 2 108 6.3 off on 78.9 17.5 3.67 4 108 6.3 off on 86.9 0.6 12.59 2 91 6.3 off off 83.7 11.5 4.811 2 107 5.3 off on 79.7 16.0 4.314 2 107 5.3 off on 78.0 17.2 4.814 4 107 5.3 off on 87.5 1.7 10.815 1 107 5.3 on on 83.7 16.3 0.016 2 107 5.3 off on 80.0 20.0 0.0239Port. 1 Port 2 Port 42900 K2250 K1600 K950 K298 KFigure 8.1 Predicted Particle Temperatures in Port Colborne Flash Reactor (Top View) [Tg=Tw= 1600K,5.3cmburner]VDusting - Tp > 2800 K2403025 Range of Model Predictions• Experimental Data15 [1oin5.0.5 1’.5 2.5 3Axial Position (m)Figure 8.2: comparison between predicted gas composition and experimental data fromPort Colborne Flash Reactor for 53 mm burner. [Tg = Tw = 1600 K I241100908070C”)0x Experimental Data - Mass Percent Copper5Q. Experimental Data - Mass Percent Sulphur(-)Model Predictions30.20I1 &0.5 1 11.5 21.5Axial Position (m)Figure 8.3: Comparison between predicted solids sample composition and experimentaldata from Port Colborne Flash Reactor ftr 53 mm burner.{TgTw= 1600K]2429 Discussion : Analysis of the MK Flash FlameIn Chapters 7 and 8, the mathematical model was validated by comparison with data from theUBC pilot plant and the Inco, Port Colborne flash reactor. In making these calculations, thefollowing factors were observedlo be of fundamental importance in determining the behaviourof the MK flash flame:1. The structure of the oxygen-concentrate jet produced by the concentrate burner.2. Heat transfer to the concentrate particles.3. Particle size and the particle size distribution.4. The oxygen-to-concentrate ratioThe role that each of these plays in chalcocite combustion and dust generation has been investigated, and is discussed in detail below. Based on this analysis of the MK flash flame, the mathematical model was then run as a predictive tool to suggest improvements in burner design whichcould lead to reduced dust generation rates.9.1 Critical Factors Affecting the Combustion of MK Concentrate9.1.1 Burner Design - The Structure of the Two-Phase JetIn this study, the single most important factor affecting the behaviour of the MK flashflame is the structure of the jet produced by the concentrate burner. In both the PortColborne and the UBC flash furnaces, the two-phase jet issuing from the concentrateburner was found to maintain a consistent structure, even after the concentrate ignites. Inboth cases, it has been found that the composition of the concentrate entering the bath,243and the dust generation rate, are both largely determined by this structure. This is reasonable, because the boundaries and the behaviour of this jet effectively define the “reactor”within which the concentrate is combusted. All other factors significantly affecting thechalcocite flash flame (heat transfer, particle size, etc.) must operate within theconstraints imposed by the behaviour of the burner jet.The structure of the jet emerging from the Port Colbome or the UBC concentrate burners(shown schematically in Figure 9.1) is complicated by many factors, including the presence of particles, density changes caused by temperature, and the effects of chemicalreaction. However, (in agreement with many earlier studies (92)), the model calculationssuggest that the structure of this jet is primarily determined by two competing influences:inertia and turbulence. The inertia of the injected gas (given by the term u) is dissipated by the effects of turbulent eddies (described by i, ). With increasing axial position, the inertia of the gas is depleted as the jet entrains fluid and spreads under theinfluence of turbulence. (Particle drag in the UBC flash furnace affects this somewhat,providing an input of momentum to the gas due to viscous drag, and hence a two-phasejet in this case does not widen as quickly as a single-phase jet). However, the gas temperature and composition fields spread radially at higher rates than momentum. This fact isreflected by the turbulent Prandtl number and the turbulent Schmidt number havingvalues less than unity (0.7). However, the particles within the jet do not spread as rapidlyas gas momentum (turbulent Schmidt number of 2.5 ), because the inertia of the particlesreduces their tendency to follow the turbulent fluctuations of the gas. As a result, theparticles spread radially much more slowly than the concentration.Figure 9.2 illustrates this clearly, in which the gas velocity and concentration fieldspredicted by the model have been plotted together with the predicted particle distribution.244Figure 9.2 was obtained from calculations for the UBC furnace at a distance of 0.12 mfrom the burner tip (8 nozzle diameters). In Figure 9.2, the axial velocity, concentrationand particle fields have been normalized with respect to their centreline values. Thisdifference in the spreading rates of concentration and gas momentum is a consistentfeature of both one-phase and two-phase jets (93). Figure 9.2 is also quite similar to thevelocity and concentration profiles in the Port Colborne burner jet at 0.5 m (8 nozzlediameters)- just before the primary ignition region.The widening and decline of the oxygen concentration with axial distance is illustrated inFigure 9.3, in which the oxygen concentration in the UBC flash furnace is plotted as afunction of axial (8 and 80 nozzle diameters) and radial position. Clearly, the ignition ofconcentrate in the UBC furnace occurs in a region which is very different from that inwhich the solids in the Port Colborne furnace ignite. By the time that the solids ignite inthe UBC flash furnace (> 80 nozzle diameters) the oxygen concentration is very lowacross the entire width of the jet. Effect of Jet Structure on Solids CompositionThe greater radial spread of the oxygen with respect to the concentrate is thereforeprimarily responsible for the over-oxidation of the solids at the edge of the jet (toform Cu20) and the under-oxidation of solids at the centre which has been observedexperimentally in both the UBC and the Port Colborne flash furnaces. It is clear thatmuch of the oxygen input to these vessels enters the bath as Cu20 and as a result, littlemetallic copper is actually produced within the flash flame itself. The bath reactionbetween the under-reacted Cu2S and the over-reacted Cu20 is the primary means bywhich metallic copper is made in this process.245The scavenging of the oxygen at the edge of the jet by the reaction of concentrate tocopper oxide is an important factor in maintaining the overall oxygen utilization efficiency (observed to be 100 % in both the UBC and Port Colborne operations). Thereaction of Cu2S to Cu20requires 50 % more oxygen than the simple conversion tometallic copper. As a result of this, high local oxygen-to-concentrate ratios which arepresent at the edge of the jet do not result in unused oxygen leaving the smelterbecause the concentrate in this region can react with the oxygen to form Cu20. Effect of the Jet Structure on Dust GenerationIn order for particles to create dust (attain the boiling point of copper) in the MK flashflame, the following criteria must be met:1. The effective oxygen-to-concentrate ratio must be above that required by stoichiometry. That is, the combustion of the concentrate particles in a particular regionmust not be able to deplete the available oxygen.2. The temperature of the gas must be sufficiently high so that the particles lie in the“DUST’tregion of Figure 5.21 (predicted dusting diagram for MK concentrate).From the analysis of the jet structure presented above, there are two locations withinthe oxygen-concentrate jet in which criterion I can be met:a) A region close to the edge of the jet, in which the local oxygen-to-concentrateratio is high. This ‘near-edge” dusting area is created by the dissimilar spreadingrates of particles and oxygen as illustrated by Figure 9.2. In the UBC flashfurnace, Figure 7.7 shows that this high local oxygen-to-concentrate ratio occursalong the entire length of the jet prior to combustion.246b) A region closer to the centre, in which there is a net inflow of oxygen due tospreading of the jet. Particles in this “near-centre’ dusting area do not necessarilyrequire a local oxygen-to-concentrate ratio above stoichiometric to produce dust,because sufficient oxygen is supplied by inflow from adjacent regions.Figure 9.4 illustrates conditions that produce a near-centre dusting region within thejet. Figure 9.4 has been obtained for the Port Colborne flash furnace at an axial location of 0.5 m (just prior to concentrate ignition). As this figure indicates, the radial gasvelocity increases from zero at the centre of the jet to a local maximum. The edge ofthe jet is associated with negative radial velocities reflecting the entrainment of thesurrounding gas. The region between the local maximum of the radial velocity and theedge of the jet is the “near-centre” dusting region. The decline in radial velocity fromthe relative maximum implies the existence of a positive net radial mass flowrate, asthe velocity into one of these regions exceeds the velocity out. This positive radialflow of mass reflects the widening of the jet - the centreline velocities and concentrations decline as those toward the edge increase. This increase in oxygen concentrationwith axial position caused by the widening of the jet has been illustrated previously byFigure 7.9.It is important to stress that these radial velocities represent time-averaged values, andthat it is still possible to transport heat and mass against this radial velocity gradientby turbulent diffusion. This is illustrated by the time-averaged conservation of speciesequationc s’ (a2C (9.1)dr Sc 3r2247Provided that the effective viscosity is sufficiently large, a positive value of v cantherefore still be associated with a positive value of. However, the rate of thisturbulent diffusion against the velocity gradient is likely to be much lower than theopposing mass transfer by bulk flow. Dust Generation Mechanism- Port Colborne Flash FurnaceThe model calculations have shown that the generation of dust in the Port Colbornesmelter results from the combustion of particles in the near-centre region. Modelcalculations and experimental data presented in Chapter 8 have shown that theconcentrate in the Port Colborne smelter ignites very close to the burner tip : at0.5-1.0 m or 8-16 nozzle diameters. The model suggests that the centreline oxygenconcentration prior to ignition is not greatly diluted by entrained sulphur dioxide, andhence ample oxygen is available to be transported from the centre of the jet to thisregion.Ignition of the concentrate does not rapidly propagate toward the centre of the jetfrom the near-centre dusting region because the hot gas resulting from this combustion is opposed by positive outfiowing velocities, and therefore this gas can onlymove toward the centre of the jet by turbulent diffusion. Thus the centre of the jetremains relatively cold while oxygen is scavenged by concentrate combusting in thisnear-centre region. When the centre of the jet finally reaches the ignition temperatureof the concentrate, so much oxygen has been removed by this process that onlyincomplete combustion can be achieved.248The near-edge region is not likely to provide a significant source of dust in the PortColborne smelter because the onset of ignition occurs very close to the burner nozzle,and hence very little radial spread of the oxygen or the concentrate takes place prior tothe combustion zone. Effect of Near-Centre Region in the UBC Flash FurnaceThe near-centre region is not likely to be a significant source of dust in the UBC flashfurnace, because the primary combustion zone occurs very far (> 80 nozzle diameters)from the burner tip. By the time the majority of the concentrate reaches the ignitiontemperature, the jet has widened considerably and has entrained large quantities ofsulphur dioxide (Figure 9.3). As a result of this, the oxygen concentration at the centreof the jet is low (< 15 % ), and therefore little oxygen is available to be transportedradially into this near-centre region. Some radial transport of oxygen does indeedoccur, and the combustion of the concentrate in the near-centre region of the UBCsmelter does produce the highest predicted particle temperatures in the combustionzone (Figure 7.3). However, these particle temperatures are some 700 K below thoseneeded to produce dust, and therefore it is likely that the near-edge region of the jet isresponsible for the dust produced in the UBC flash smelter. The means by which dustis generated in this region is discussed below.9.1.2 Heat Transfer9.1.2.1 Location of the Primary Combustion ZoneBased on the analysis presented above, it is clear that the ignition of MK concentrate249in the flash flame is primarily determined by heat transfer. This is supported by theevidence provided by the solids samples recovered from the UBC reactor, in whichsolids at edge of the shaft (in regions of higher temperature but lower oxygen concentration) were consistently more fully oxidized prior to the ignition of the solids at thecentre (lower temperature but higher oxygen concentration). The high ignitiontemperature of the MK concentrate is primarily responsible for this effect, since theconcentrate particles must be heated to well over 1000 K before ignition can occur. The UBC Flash FurnaceThe ratio of heat transferred to the particles by convection to that transferred byradiation in the UBC flash furnace is shown in Figure 9.5. As this figure indicates,particles at the edge of the jet are heated almost entirely by gas convection, whilefor those at the centre, wall radiation is more significant (despite particle blockageeffects). The large number of particles near the centre of the jet requires considerable thermal energy to be heated to their ignition temperature. As discussedpreviously, the ability of the entrained (hot) gas to penetrate to the centre of the jetis limited, and as a result the temperature in this region only rises slowly. The highgas-particle heat transfer coefficient ensures that any entrained gas quickly transfers its heat to the particles, but the rate of convective heating is limited by the rateat which the hot gas can be entrained. As a result, radiated energy from the wallsof the reactor forms an important supplement to gas-particle convection in initiating particle ignition near the centreline. Once the primary combustion zonebegins in the near-centre region, the hot gas released by this combustioneventually triggers ignition of the concentrate at the centre.2509. The Port Colborne Flash FurnaceThe higher gas temperatures and the lower wall-particle view factor in the PortColborne smelter both cause the recirculating gas to be a more significant factordetermining the onset of the combustion zone. Figure 9.6 illustrates this clearly, inwhich the heat transferred by convection dominates that transferred by radiationover much of the flash flame. Role of Recirculating Gas in Dust Generation - UBC FurnaceThe analysis presented above suggests that the near-edge region (high local oxygen-to-concentrate ratio caused by low concentrate spreading rates) is likely to be responsible for the dust generation in the UBC flash furnace. As a result, the temperature ofthe recirculating gas entrained into this region is an important parameter which affectsdust generation by this mechanism. As well as providing much of the heat necessaryto ignite the concentrate, the entrainment of the gas also serves to dilute the localoxygen concentration. Therefore, a sufficiently high gas temperature can triggerparticle ignition without excessive dilution - resulting in dust formation.For example, the kinetic model predictions presented in Figure 5.21 show theboundary between ‘dust’ and ‘no-dust’ regions as a function of particle size, oxygenconcentration and ‘furnace temperature. Assuming that for near-edge regions in theUBC jet, the ‘furnace temperature is effectively the gas temperature (due to the highrates of gas-particle heat transfer in this region), it is possible to calculate theminimum temperature of the recirculating gas such that particles of a given size arelocated in the ‘dust’ region of Figure 5.21. This calculation has been performed251assuming that the oxygen enters the reactor at 298 K, and neglecting both the thermalload of the particles and the oxygen depletion by chemical reaction (due to the lowparticle concentration in this region). The results of these calculations have beenplotted in Figure 9.7. It is interesting that the absolute minimum gas temperaturenecessary to produce dust in this manner is approximately 1200 K, with smallerparticles predicted to dust at lower recirculating gas temperatures than larger particles.Figure 9.7 indicates that as the temperature of the recirculating gas increases, the rateof dust generation at the edge of the jet should increase. However, if the gas temperature were increased sufficiently SO that combustion occurred at a significantly reducedaxial distance, then presumably dusting from the near-centre region would begin tooccur.91.3 Particle Size EffectsFigure 9.7 provides a possible explanation for the observed rates of dust formation in theUBC flash furnace. In the “base case” conditions tested by the model and presented inChapter 7, no particles were predicted to attain the boiling point of copper, whereas theexperimental dust generation rate was approximately 3-5 % of the mass input rate. Tomake these calculations however, the mathematical model assumed a mono-sizedconcentrate 20 microns in diameter. However as shown in Chapter 4, the MK concentrateas received from Inco contains a considerable quantity ( approximately 15 % by mass ) ofvery fine particles 5 microns in diameter or less - the presence of these fines has hithertobeen ignored by the model.252To examine this effect, the model calculations of the UBC flash furnace which assumed amono-sized 20 micron concentrate were re-examined from the point of view of a 5micron particle. All axial and radial locations containing particles were examined, andthose regions in which the temperature and composition of the reaction gas were such thata 5 micron MK particle lay in the “dust” area of Figure 5.21 were recorded. This analysishas shown that 25 % of the particles in the flash reactor occupy such regions. Therefore,if these fines account for 15 % of the total mass input, then the total dust generation ratepredicted by this technique (assuming that all the fines in these regions report as “dust”) isapproximately 3.8 % of the total mass input rate, which is very close to the experimentally measured rate.Although this analysis is very approximate, it does indicate that most of the dust in theUBC flash furnace originates from the combustion of fine particulates in near-edgeregion. The relatively good agreement between the solids sample composition predictedby the mathematical model and that measured experimentally may be due both to the lowmass concentration of the fines and the extremely small size of the dust generated by thefines. As described in Chapter 4, the solids sampler was not capable of capturing veryfine particles, and therefore the model predictions assuming combustion of a 20 micronconcentrate would serendipitously match what could be caught by this sampler. This dustgeneration mechanism is therefore still consistent with the model verification presented inChapter 7.It is important to stress that a mono-sized concentrate, however fine, will have a diminished tendency to produce dust by this mechanism. As particles become very small (< 1micron ) they tend to follow the turbulent fluctuations of the gas more completely, andtend to spread more rapidly than coarser particles. As a result, regions of high local253oxygen-to-concentrate ratio will not tend to occur as readily with this type of concentrate.This is in qualitative agreement with the experimental data, in which the run with thefinely-sized concentrate generated slightly less dust than that observed with the unsizedas-received concentrate. Note however that such a concentrate will still be capable ofgenerating dust in the near-centre region of the jet.9.1.4 SummaryIn both the UBC and the Port Colborne furnaces, the single most significant factoraffecting the behaviour of the MK flash flame is the structure of the oxygen-concentratejet produced by the concentrate burner. The oxygen is consistently observed to spreadmore rapidly than the particles, resulting in a significant radial variation of concentratereaction degree.Depending upon the axial location at which the concentrate combusts, dust can be madein two separate regions of the flash flame- near-edge” regions and regions of high radialoxygen transport (“near centre” ). In Port Colborne, most of the dust is produced byconcentrate combustion in the near-centre region, while the preferential combustion offines in the near-edge region is responsible for the observed dust generation rate in theUBC furnace. The onset of combustion of the MK concentrate is determined primarily byheat transfer, with gas Convection dominating near the edge of the jet and radiationbecoming more important toward the centre.254The oxygen-to-concentrate ratio (factor 4 in section 9.1) is effectively fixed for thisprocess by the desired matte grade, and hence changes in this parameter are of littleindustrial significance. The effects caused by excess oxygen have been discussedpreviously in Chapter 7.9.2 Dust Abatement: Improvements in Burner DesignThe analysis presented above suggests that a complete elimination of the dust formed by theflash reaction of MK concentrate will be very difficult to achieve, due to the variety of mechanisms by which this dust can be made. However, this analysis does suggest that the concentrate burner is the most significant factor affecting the dust generation rate, and thereforesome reduction in dust generation could be achieved through improved burner design.The fundamental difficulty associated with the current single-entry concentrate burner is thatthe oxygen spreads faster than does the concentrate - the ability of a burner of this type todisperse the solids is clearly very limited. This results in a region of high local oxygen-to-concentrate ratio ( near-edge dusting ) and a region of high radial oxygen transport (near-centre dusting). If the concentrate could be induced to spread at the same rate as the oxygen,then a significant reduction in dusting could presumably be achieved. To analyze thishypothesis, the model was run considering the case (in the Port Colborne reactor) of a mono-sized concentrate 20 microns in diameter which spread at exactly the same rate as the gas : nodusting was predicted to occur.In practice, achieving this dispersion with the current burner design is likely to be very difficult. If a dispersing cone or radially-directed jets were employed, then care would have to betaken to ensure that the concentrate was not dispersed too rapidly, as this also would tend tolead to dusting.255Swirling flow, as used in pulverized coal combustion, (112) could possibly provide a solutionto this problem. MK concentrate particles would tend to be projected outward by a swirlingflow regime, potentially resulting in a more even dispersion and therefore reduced dusting.Due to the complexity associated with swirl, a burner of this design would need to besubjected to considerable pilot testing to ascertain its efficacy and to determine appropriatescale-up factors. Ultimately, the burner which is most successful in reducing dust formationwill be one which radially disperses all solids at the same rate as the gas.ii cm(CF.CO tr-1 Ctc.____________—CIi iii’iii’-.4—NII Ot’j2570zU0>0NzO.Figure 9.2: Normalized calculated velocity, concentration and particle density fieldswithin the UBC flash flame at 0.12 m from the burner tip [Base Case ConditionsRadial Position (m)258Calculated radial variation of oxygen concentration within the UBC flashflame at two axial locations [Base Case Conditions]0.CC.0Radial Position (m)Figure 9.3:00C)I-.4)0259c.C0Figure 9.4: Calculated radial velocity and oxygen concentration within the Port Colborneflash flame as a function of radial position j at 0.50 rn from the burner tip,prior to combustion I9260ComparisonRadiation andBetweenConvectionSection BSection CConvectionDominantRadiationDominantFigure 9.5 Ratio of heat transferred by convection to that transferred by radiation in theUBC flash furnace [Base Case ConditionsSection ASectionD261QmvectionDominantRadiationDomin antFigure 9.6 : Comparison between radiative and convective heat transferrates in the Port Colborne Flash Reactor262ICalculated recirculating gas temperature necessary to produce dusting in theUBC flash furnace as a function of diluted gas compositionMole Fraction OxygenFigure 9.7:26310 Conclusions and RecommendationsThe major findings of this study are as follows1. The dust formation associated with the flash smelting of MK concentrate is due to copperboiling within combusting concentrate particles. This boiling leads to particle fragmentationwhich may, in turn, lead to further vaporization of copper. As a result, the “chemical dust”recovered from an MK flash reactor is composed of condensed copper vapour and fragmentsof exploded particles.2. The tendency of a combusting particle to fragment by this mechanism is strongly affected byoxygen concentration, particle size and furnace temperature.3. The mathematical model - supported by the experimental studies in the Port Colbome andUBC pilot plants - has shown that very little metallic copper is formed within the chalcociteflash flame. Concentrate at the centre of the flame tends to remain under-reacted while thattoward the edge reacts to form copper oxide. The bath reaction between the under-reactedand over-reacted concentrate is therefore the primary means by which copper is formed inthis process.4. The mathematical model has indicated that the dissimilar spreading rates of the oxygen andthe MK concentrate are responsible for this uneven reaction degree, and act together toproduce two regions of dust production within the chalcocite flash flame. The “near-edge”region of the flame is associated with high local oxygen-to-concentrate ratio and is primarilyresponsible for the dust production in the UBC flash furnace. The “near-centre” dustingregion is associated with locally high rates of radial oxygen transport, and produces much of264the dust observed in the Port Colborne flash reactor. An improved burner design which iscapable of producing identical spreading rates of oxygen and concentrate would eliminateboth of these dusting regions.Based on this information, the following recommendations can be made:1. An improved concentrate burner design should be sought to reduce the dust generation rateassociated with MK flash smelting. From the point of view of dust reduction, the optimalburner design will be one in which the solids and the gas are dispersed at identical rates.2. The mathematical model should be extended to consider a distribution of particle sizes asencountered industrially, rather than simply assuming a mono-sized concentrate. The experimental studies in the UBC flash furnace have suggested that the preferential combustion offines at the near-edge region is likely to be an important source of dust, and the mathematicalmodel is (currently) limited in its ability to describe this behaviour.3. Additional particle temperature measurements of reacting MK particles should be performedboth to verify Figure 5.21 and to provide information on the combustion of MK concentratein low-oxygen atmospheres.26511 References1. Pannell, D.B., “A Survey of World Copper Smelters” in World Survey of NonferrousSmelters, J.C. Taylor and H. Traulson ed., Pyrometallurgical Soc. of T.M.S., TMS-AIMEmeeting, Phoenix, Jan. 25-29, 1988, pp. 3-1182. Lee, J.H. and Tm, S.J., “Operational Improvements at the Onsan Copper Flash Smelter”,inFlash Reaction Processes, D.G.C. Robertson, H.Y. Sohn, and N.J. Themelis ed., Univ. ofUtah, Salt Lake City, Utah, June 15-17, 1988, Pub. Center for Pyrometallurgy, Univ.Missouri-Rolla, Rolla, Mo.,pp. 19-353. Partelpoeg, E.H., ‘Flash Smelting Reactions at the Phelps Dodge Hidalgo Smelter”, inFlash Reaction Processes, D.G.C. Robertson, H.Y. Sohn, and N.J. Themelis ed., Univ. ofUtah, Salt Lake City, Utah, June 15-17, 1988, Pub. Center for Pyrometallurgy, Univ.Missouri-Rolla, Rolla, Mo.,pp. 35-464. Diaz, C.M.., Presentation Made at Centre for Pyrometallurgy Conference: “Flash Reaction Processes, Salt Lake City, Utah, June 15-17, 1988.5. Makinen, J.K. and Jafs, G.A.,”Production of Matte, White Metal and Blister Copper byFlash Furnace”, J. Metals, 1982, vol.33 (6), pp. 54-596. Asteljoki, J.A., Bailey, L.K., George, D.B. and Rodloff, D.W.,”Flash Converting-Continuous Converting of Copper Mattes”, J. Metals, 1985, vol. 35 (5) pp. 20-237. Bell, M.C., Blanco, J.A, Davies, H. and Sridhar, R., “Oxygen Flash Smelting in aConverter”, J. Metals, 1978, vol. 30 (11), pp. 9-148. Diaz, C.M., Personal Communication, 19899. Antonini , T.N. , “Inco Flash Smelting of Copper Concentrate” in World Survey ofNonferrous Smelter, J.C. Taylor and H. Traulson ed., TMS-AIME meeting, Phoenix,Jan. 25-29, 1988, pp. 3-11810. Landolt, C.A., Fritz, A., Marcuson, S.W., Cowx, R.B. and Miszczak, J.,”Copper makingat Inco’s Copper Cliff smelter”, Copper 91-Cobre 91, C. Diaz, C. Landolt, A. Luraschiand C.J. Newman eds., Pergamon, Elmsford, N.Y., 1991, pp. 15-2911. Solar, M.Y., Neal, R.J., Antonioni, T.N. and Bell, M.C., “Smelting Nickel Concentratesin INCO’s Oxygen Flash Furnace’, J. Metals, 1979, vol. 31(1), pp. 26-3112. Victorovich, G.S., Diaz, C., and Raskauskas, J., “Impurity Distributions, Dusting andControl of Matte Grade in Inco Oxygen Flash Smelting”, Paper presented at the 50thMeeting Chilean Inst. Mining Engineers, Santiago, Chile, Nov. 198013. Otero, A., Brimacombe, J.K. and Richards, G.G.,”Kinetics of the flash reaction ofcopper concentrates”, Copper 91-Cobre 91, C. Diaz, C. Landolt, A. Luraschi and C.J.Newman eds., Pergamon, Elrnsford, N.Y., 1991, pp. 459-47326614. Warczok, A., Utigard, T., Mroz, W., Kowalczyk, J., Warmuz, M., St. Musial,”Modellingof suiphide concentrate smelting in a flash smelter”, Copper 91-Cobre 91, C. Diaz, C.Landolt, A. Luraschi and C.J. Newman eds., Pergamon, Elmsford, N.Y., 1991,pp.637-65015. Asteljoki, J.A. and Mueller, H.B., ‘Direct smelting of blister copper - pilot flashsmelting tests of Olympic Dam concentrate”, in Pyrometallurgy 87, Inst. Mining andMetallurgy, 1987, pp. 19-5216. Smith, T.J.A., Poesner, I. and Williams, C.J., “Oxygen Smelting and the Olympic DamProject”, International Symposium on the Impact of Oxygen on the Productivity of NonFerrous Processes, Winnipeg, Man., Aug. 23-26 1987, pp. 49-5917. Smieszek, Z., Czernecki, J. and Sobierajski, S.,”Modern copper smelting and refiningpractice in Poland”, in Mineral Processing and Extractive Metallurgy, M.J. Jones and P.Gill eds., papers presented at the International Conference on Mineral Processing andExtractive Metallurgy, Kunming, China, Oct. 27 -Nov. 3 1984, pp. 335-34318. Jorgensen, F.R.A., “Single Particle Combustion of Chalcopyrite”, Proc. Australas. Inst.Mi Metall., 1983, vol. 288 (12), pp.37-4619. Jorgensen, F.R.A.,”Heat transfer mechanism in ignition of nickel suiphide under simulated flash smelting conditions”, Proc. Australas. Inst. Mi Metall., 1978, vol. 271 (9),pp.21-2520. Jorgensen, F.R.A.,”Vaporization during the combustion of a complex copper concentrate”, Bull. Proc. Australa. Inst. Mm. Metal!., 1985, vol. 290 (1), pp. 91-9821. Jorgensen, F.R.A., “Combustion of chalcopyrite pyrite galena and shpalerite undersumulated suspension smelting conditions”, Australia Japan Extractive MetallurgySymposium, Sydney, Australia, 1980, pp. 41-5122. Dunn, J.G. and Smith, T.N.,”Flash Smelter Shaft Operations”, Western AustralianMining and Petroleum Research Institute (WAMPRI) Report No. 7, July 198423. Lilja, L. and Makitalo, J., “Means for Producing a Suspension of a Powdery Substanceand a Reaction Gas”, U.S. Patent #4,210,3 1524. Lilja, L. and Makitalo, J.,”Procedure for Producing a Suspension of a PowderySubstance and a Reaction Gas”, U.S. Patent #4,147,53525. Kimura, T., Ojima, Y., Mori, Y. and Ishii, Y., “Reaction Mechanism in a Flash SmeltingReaction Shaft”, in Innovative Technology and Reactor Design in Extraction Metallurgy, Proceedings of the R. Schuhrnann Symposium, eds. D.R. Gaskell. J.P. Hager, J.E.Hoffmann, and P.J. Mackey, TMS/AIME, 1986, pp. 403-4 1826. Kemori, N., Ojima, Y. and Kondo, Y.,”Variation of the Composition and Size of CopperConcentrate Particles in the Reaction Shaft”,in Flash Reaction Processes, D.G.C.Robertson, H.Y. Sohn, and N.J. Themelis ed., Univ. of Utah, Salt Lake City, Utah, June15-17, 1988, Pub. Center for Pyrometallurgy, Univ. Missouri-Rolla, Rolla, Mo.,pp.47-6826727. Sohn, H.Y., Seo, K.W., Chaubal, P.C. and Hahn, Y.B., “Laboratory Studies on the FlashSmelting of Copper Concentrate”,in Flash Reaction Processes, D.G.C. Robertson, H.Y.Sohn, and N.J. Themelis ed., Univ. of Utah, Salt Lake City, Utah, June 15-17, 1988,Pub. Center for Pyrometallurgy, Univ. Missouri-Rolla, Rolla, Mo.,pp. 145-16628. Munroe, N.D.H., “Simulation of Flash Smelting Phenomena in a Laboratory Reactor”,D.Sc. Thesis, Columbia University, 198729. Munroe, N.D.H. and Thernelis, N.J.,”Rate phenomena in a laboratory flash smeltingreactor”, Copper 91-Cobre 91, C. Diaz, C. Landolt, A. Luraschi and C.J. Newman eds.,Pergamon, Elrnsford, N.Y., 1991, pp. 475-49430. Kemori, N., Akada, A., and Kondo, Y.,”Development of a concentrate burner for industrial oxygen flash smelting’, Copper 91-Cobre 91, C. Diaz, C. Landolt, A. Luraschi andC.J. Newman eds., Pergarnon, Elmsford, N.Y., 1991, pp. 141-15431. Inami, T., Baba, K., Kurokawa, H., Nagai, K. and Kondo, Y., “Modification of concentrate burner for a copper flash smelting furnace”, Copper 91-Cobre 91, C. Diaz, C.Landolt, A. Luraschi and C.J. Newman eds., Pergamon, Elmsford, N.Y., 1991, PP. 49-6332. Ruottu, S.,”The Description of a Mathematical Model for the Flash Melting [sic] ofCu-Concentrates”, Combustion and Flame, 1979, vol. 34, pp. 1-1133. Hahn, Y.P. and Sohn, H.Y.,”Prediction of a particle-laden gas jet as related to the flash-smelting process”, in Innovative Technology and Reactor Design in Extraction Metallurgy, Proceedings of the R. Schuhmann Symposium, eds. D.R. Gaskell. J.P. Hager, J.E.Hoffmann, and P.J. Mackey, TMS/AIME, 1986, pp. 469-49934. Hahn, Y.P. and Sohn, H.Y., “Computation of the Flow-Field in a Recirculating Turbulent Gas Jet”, Chem. Eng. Commun., 1977, vol. 61, pp. 39-5735. Hahn, Y.P. and Sohn, H.Y., “Mathematical modelling of the combined turbulent transport phenomena, chemical reactions and thermal radiation in a flash-furnace shaft”, inMathematical Modelling of Materials Processing Operations, J. Szekely ed., TMS, 1987,pp. 799-83436. Hahn, Y.P. and Sohn, H.Y.,”The Trajectories and Distribution of Particles in a Turbulent Axisymmetric Gas Jet Injected into a Flash Furnace Shaft’, Met. Trans. B., 1988,vol. 19B (12), pp. 87 1-88337. Hahn, Y.P. and Sohn, H.Y.,”Radiative heat transfer in a flash-smelting furnace”, inFlash Reaction Processes, D.G.C. Robertson, H.Y. Sohn, and N.J. Themelis ed., Univ.of Utah, Salt Lake City, Utah, June 15-17, 1988, Pub. Center for Pyrometallurgy, Univ.Missouri-Rolla, Rolla, Mo.,pp. 287-3 1738. Hahn, Y.P. and Sohn, H.Y.,”Mathematical Modeling of Sulfide Flash Smelting Process:Part I. Model Development and Verification with Laboratory and Pilot Plant Measurements for Chalcopyrite Concentrate Smelting”, Met. Trans. B., 1990, vol. 21B (12),pp.945-95826839. Hahn, Y.P. and Sohn, H.Y.,’Mathematical Modeling of Sulfide Flash Smelting Process:Part II. Quantitative Analysis of Radiative Heat Transfer”, Met. Trans. B., 1990, vol.21B (12), pp.959-6640. Seo, K.W. and Sohn, H.Y.,”Mathematical Modeling of Sulfide Flash Smelting Process:Part UI. Volatilization of Minor Elements”, Met. Trans. B., 1991, vol. 22B (12),pp.791-941. Gosman, A.D., Pun, W.M., Runchal, A.K., Spalding D.B. and Wolfshtein, M., “Heatand Mass Transfer in Turbulent Recirculating Flows’, 1969, Academic Press, London.42. Crowe, C.T., Sharma, M.P. and Stock, D.E.,’The Particle Source in Cell (PSI-CELL)Model for Gas-Droplet Flows’, Journal of Fluids Engineering, 1977, Trans. ASME,pp.325-33243. Chaubal, P.C. and Sohn, H.Y., “Intrinsic Kinetics of the Oxidation of Chalcopyriteunder Isothermal and Nonisothermal Conditions”, Met. Trans. B, 1986, vol. 17B, pp.5 1-6044. Walsh, P.M., Zhang, M., Farmayan, W.F. and Beer, J.M.,”Ignition and Combustion ofCoal-Water Slurry in a Confined Turbulent Diffusion Flame’, in Proceedings of theTwentieth (International) Symposium on Combustion, The Combustion Institute, 1984,pp. 1401-140745. Koib, T., Farmayan, W.F., Walsh, P.M. and Beer, J.M.,”The Contribution of Radiationto the Ignition of a Confined Turbulent Coal-Water Slurry Diffusion Flame”, Combust.Sci. and Tech., 1988, vol. 58, pp. 77-9546. Kim, Y.H. and Themelis, N.J., “Effect of Phase Transformation and Particle Fragmentation on the Flash Reaction of Complex Metal Sulfides”, in Innovative Technology andReactor Design in Extraction Metallurgy, Proceedings of the R. SchuhmannSymposium, eds. D.R. Gaskell. J.P. Hager, J.E. Hoffmann, and P.J. Mackey,TMS/AIME, 1986, pp. 349-36947. Themelis, N.J., Wu, L. and Jiao, Q., “Some Aspects of Mathematical Modeling of FlashSmelting Phenomena”, in Flash Reaction Processes, D.G.C. Robertson, H.Y. Sohn, andN.J. Themelis ed., Univ. of Utah, Salt Lake City, Utah, June 15-17, 1988, Pub. Centerfor Pyrometallurgy, Univ. Missouri-Rolla, Rolla, Mo., pp. 263-28548. Jiao, Q., Wu, L. and Themelis, N.J.,”Mathematical modeling of flash converting ofcopper matte, in Innovative Technology and Reactor Design in Extraction Metallurgy,Proceedings of the R. Schuhmann Symposium, eds. D.R. Gaskell. J.P. Hager, J.E. Hoffmann, and P.J. Mackey, TMS/AIME, 1986, pp. 835-85949. Shook, A.A., Richards, G.G. and Brirnacombe, J.K., “Pilot Plant Study of Lead FlashSmelting”, in Flash Reaction Processes, D.G.C. Robertson, H.Y. Sohn, and N.J.Themelis ed., Univ. of Utah, Salt Lake City, Utah, June 15-17, 1988, Pub. Center forPyrometallurgy, Univ. Missouri-Rolla, Rolla, Mo., pp. 99-12350. Engineering Technical Report - Flow Evaluation of Inco Burner, Inco Ltd., 198926951. Inco Port Colborne Personnel, No. 2 Research Station Internal Report, Campaign 11,199052. Elliott, J.F., ‘Phase Relationships in the Pyrometallurgy of Copper”, Met Trans B, 1976,vol. 7B, pp. 17-3353 Kellogg, H.H., “Thermochemical properties of the system Cu-S at elevated temperature”, Can. Met. Quart., 1969, vol. 8 (1), PP. 3-2354. Binary Alloy Phase Diagrams, Vol. 1., T.B. Massaiski ed., ASM International, MaterialsPark, Ohio, William R. Scott Jr., 198655. FACT - Facility for the Analysis of Chemical Thermodynamics, Computer Softwareand Database, McGill University, Canada, 199156. Henderson, T.A., ‘The Oxidation of Powder Compacts of Copper-Iron Sulphides”, Bull.Inst. Mining Met, vol. 620, 1947, PP. 497-52057. Lewis, J.R, Hamilton, J.H., Nixon, J.C., and Graversen, C.L., “The Oxidation of Chalcocite in Air Compared with Its Oxidation in Pure Oxygen’, Trans. AIME, 1948,pp.177-18558. Wadsworth, M.E., Leiter, K.L., Porter, W.H., and Lewis, J.R., “Sulfating of CuprousSuiphide and Cuprous Oxide”, Trans. Met. Soc. AIME, 1960, vol. 218, pp.519-2559. Ramakrishna Rao, V.V.V.N.S., and Abraham K.P, “Kinetics of Oxidation of CopperSulfide, Met. Trans., 1971, Vol. 2 (9), pp. 2463-247060. Asaki, Z., Ueguichi, A, Tanabe, T., and Kondo, Y, “Oxidation of Cu2S Pellet”, Trans. J.Inst. Metals, 1986, vol. 27 (5), pp. 361-37161. Kim, Y.H. and Themelis, N.J., “Rate Phenomena in the Oxidation of Zinc, Iron andCopper Sulfide Pellets”. Can. Met. Quart., 1987, vol. 26 (4), pp. 34 1-34962. Otero, A., Tuffrey, N.E., Brirnacombe, J.K. and Richards, G.G., Unpublished Research,198963. Tuffrey, N.E., “Pyrometry studies of the combustion of lead concentrate particles undercontrolled conditions”, PhD Thesis, The University of British Columbia, 198964. Kubaschewski, 0. and Alcock, C.B., Metallurgical Thermochemistry, 51 ed., PergamonPress, New York, 197965. Byerley, J.J., Rempel G.L., and Takebe, N., “Interaction of Copper Sulfides with CopperOxides in the Molten State”, Met. Trans., 1974, vol. 3, Pp. 2501-250666. Friedman, R. and Macek, A., “Combustion studies of single aluminum particles”, 9th(International) Symposium on Combustion, The Combustion Institute, 1953, pp.703-7 1227067. Lasheras, L.C., Femandez-Pello, A.C. and Dryer, F.L., “Initial Observations on the FreeDroplet Combustion Characteristics of Water-In-Fuel Emulsions”, Combustion Scienceand Technology, 1979, vol. 21, pp.-1468. Law, C.K., Chung, S.H., and Srinivasan, N., “Gas-Phase Quasi-Steadiness and FuelVapour Accumulation Effects in Droplet Burning”, Combustion and Flame, 1980, vol.38, pp. 173-19869. Schrnid, R.,”A Thermodynamic Analysis of the Cu-O System with an Associated Solution Model”, Met. Trans. B., 1983, vol. 14B, pp. 473-48 170. Geankopolis, C.J.,Mass Transport Phenomena, Holt, Rinehart and Winston, New York,197271. Elliott, J.F.,”The role of interfaces in pyrometallurgical processes”, Extraction Metallurgy 85, Inst. Mi Metall., 1985, pp. c171-c17872. Pehike, R.D., Unit Processes of Extractive Metallurgy, Elsevier, New York, 1982, pp.137-13973. Jalkanen, H.”Phenomenology of the Oxidation Kinetics of Molten Cuprous Suiphideand Copper”, Scand. J. Met., 1981, vol. 10, pp.257-6274. Barton, R.G., M.A.Sc. Thesis, The University of British Columbia, 1976.75. Barton, R.G. and Brimacombe, J.K., “Influence of Surface Tension-Driven Flow on theKinetics of Oxygen Absorption in Molten Copper”, Met. Trans. B, 1977, vol. 8B, pp417-42776. Brimacombe, J.K. and Weinberg, F.W., “Observations of Surface Movements of LiquidCopper and Tin”, Met. Trans., vol. 3 (8), pp. 2298-229977. “CRC Handbook of Chemistry and Physics”, 61s1 ed.,Eds. R.C. Weast and M.J. Astle,CRC Press Inc., Boca Raton, Florida, 198078. Moore, W.J., Physical Chemist, 4th ed., Prentice-Hall, New York, 197279. NASA, Free Turbulent Shear Flows, Vols. 1-2, Proceedings of the Langley WorkingConference on Free Turbulent Shear Flows, NASA SP-321, 197280. Rudy, D.H. and Bushnell, D.M.,”A rational approach to the use of Prandtl’s mixinglength model in free turbulent shear flow calculations” in NASA, Free Turbulent ShearFlows, Vols. 1-2, Proceedings of the Langley Working Conference on Free TurbulentShear Flows, NASA SP-321, 1972, pp. 67-13581. Morrison, G.I., Tatterson, G.B., and Long, M.W., “Three-Dimensional Laser Velocimeter Investigation of Turbulent, Incompressible Flow in an Axisymmetric SuddenExpansion”, J. Propulsion, 1988, vol. 4 (6), pp. 533-54082. Anderson, D.A., Tannehill, J.C. and Pletcher, R.H., “Computational Fluid Mechanicsand Heat Transfer”, Hemisphere, New York, 198427183. Kwori, O.K. and Pletcher, R.H., ‘Prediction of the Incompressible Flow over aRearward-Facing Step”, in Anderson, D.A., Tannehill, J.C. and Pletcher, R.H., “Computational Fluid Mechanics and Heat Transfer”, Hemisphere, New York, 198484. Kreith, F., Principles of Heat Transfer, 2nd ed., International Textbook Co, Scranton Pa.,196585. Prandtl, L., “Ueber die ausgebildete Turbulenz”, Proceedings of the Second InternationalCongress for Applied Mechanics, Zurich, 1926, pp. 62-7486. Emami, S., Morrison, G. and Tatterson, G., “Eddy Viscosity Distributions Through theTransition of an Incompressible Free Jet”, J. Fluids Eng., Trans. AIME, 1988, vol. 110,pp. 98-100)87. Hwang, S.S.,”Prediction of turbulent jets and plumes in flowing ambients”, Ph.D.Thesis, iowa State University, Ames, Iowa, 197888. Launder, B.E. and Spalding, D.B., Mathematical Models of Turbulence, AcademicPress, London, 197289. Launder, B.E. and Spalding, D.B., “The Numerical Computation of Turbulent Flows”,Comput. Methods Appl. Mech. Eng., 1974, vol. 3, pp. 269-28990. Tennankore, K.N. and Steward, F.R., “Comparison of Several Turbulence Models forPredicting Flow Patterns within Confined Jets”, Can. J. Chem. Eng., 1978, vol. 56, pp.67 3-67891. Mandeibrot, B.B., J. Fluid Mech., vol. 62, p.33192. Melville, W.K. and Bray, K.N.C.,”The Two-Phase Turbulent Jet’, Tnt. J. Heat MassTransfer, 1979, vol. 22., pp. 279-28793. Schetz, J.A., Injection and Mixing in Turbulent Flow, Progress in Astronautics andAeronautics Vol 68, A.I.A.A., New York, 198094. Abbas, A.S, Koussa, S.S. and Lockwood, F.C.,”The Prediction of the Particle Laden GasFlows”, Proceedings of the Eighteenth (International) Symposium on Combustion, TheCombustion Institute, 1980, pp. 1427-143795. Harsha, P.T.,”Free Turbulent Mixing: A Critical Evaluation of Theory and Experiment”, AEDC-TR-71-36, Engine Test Facility, Arnold Engineering DevelopmentCenter, Air Force Systems Command, Arnold Air Force Station, Tn., Feb. 197196. Goldschmidt, V.W. and Eskinazi, S.,”Two-Phase Turbulent Flow in a Plane Jet”, J.Appl. Mech., Trans. AIME, Dec. 1966, pp. 735-74797. Carter, J.E. and Wornom, S.F.,’Forward Marching Procedure for Separated BoundaryLayer Flows”, AIAA J., 1985, Vol. 13, pp. 1101-1 10398. Rehner, T.A. and Flugge-Lotz, I., “The interaction of a Shock Wave with a LaminarBoundary Layer, mt. J. Non-Linear Mech., 1968, vol. 3, pp. 173-19927299. Tuve, G.I., Heating, Piping, Air Conditioning, vol. 25 (1), pp. 181-191100. Wilson, R.A.M. and Danckwerts, P.V.,”Studies in turbulent mixing- II. A hot-air jet”,Chem. Eng. Sci., 1964, vol. 19, pp. 885-895101. Modarress, D., Wuerer, J. and Elghobashi, S., “An experimental study of a turbulentround two-phase jet”, Chem. Eng. Commun., 1984, vol. 28, pp. 341-354102. Keagy, W.R. and Weller, A.E., “A study of freely-expanding inhomogeneous jets”,Proceedings of the Heat Transfer and Fluid Mechanics Institute, 1949, vol. 2, pp. 89-98103. Subramanian, V. and Ganesh, R., “Particle-Gas Dispersion Effects in a Round Jet”,Can. J. Chem. Eng., 1984, vol. 62, pp. 161-164104. Laats, M.K., “Experimental Study of the Dynamics of an Air-Dust Jet”, InzhenemoFizicheskii Zhurnal, 1966, vol. 10 (1), pp. 11-15105. Ivanov, Y.V., Laats, M.K. and Frishman, F.A.,” Dispersion of a heavy admixture in atwo-phase jet’, Inzherno-Fizicheskii Zhurnal, 1970, vol. 18(3), pp. 538-541106. Laats, M.K. and Frishman, F.A., “Scattering of an Inert Admixture of Different GrainSize in a Two-Phase Axisymmetric Jet”, Heat Transfer - Soviet Resarch, 1970, vol. 2(6), pp. 7-12107. Goldschmidt, V.W., Householder, M.K., Ahmadi, G., and Chuang, S.C.,”TurbulentDiffusion of Small Particles Suspended in Turbulent Jets”, in Progress in Heat andMass Transfer, 1972, Vol. 6, G. Hetsroni ed., pp. 487-508108. Tice, C.L. and Smoot, L.D., “Cold Flow Mixing Rates with Recirculation for Pulverized Coal Reactors”, AIChE Journal, 1978, vol. 24 (6), pp. 1029-1035109. Memmot, V.J. and Smoot, L.D., “Cold Flow Mixing Rates with Recirculation forPulverized Coal Reactors”, AIChE Journal, 1978, vol. 24.(6), pp. 466-475110. Sunavala, P.D., Hulse, C., and Thring, M.W. “Mixing and Combustion in Free andEnclosed Turbulent Jet Diffusion Flames”, Comb. and Flame, 1957, 1, pp. 179-193111. Becker, H.A., Hottel, H.C. and Williams, G.C., “The nozzle-fluid concenuation field ofthe round, turbulent, free jet”, J. Fluid Mech., 1967, vol.30(2), pp. 285-303112. Chemical Engineer’s Handbook, 5th ed., R.H. Perry ed., McGraw-Hill, New York,1983273Appendix 1 : Calculated efficiency of non-isokinetic solids samplerThe target efficiency (ri, ) of spheres, cylinders and ribbons have been tabulated in Perry (112).This efficiency represents the fraction of solids in a gas stream which will be collected over thearea of the sampler, and is given by the ratio DJDh (Figure A 1.1).NNNNNN NNNN NN NN NN NNN NN NNDb DsFigure A1.1 : Capture Efficiency of an Impingement SamplerAssuming that the solids sampler behaves as a ‘ribbon and that Stokes law applies, then Perry(112) indicates that the capture efficiency is a function of the separation number (N). A fit of thecurve shown in Perry gives:r= i —e (A1.1)where the separation number, N, is given byu (A1.2)N = u —sS tgDAssuming a 50 % SO2 - 50 % 02 gas stream at 1000 K yeilds the plot shown in Figure 4.9.Changes in the assumed gas composition or temperature had little effect on the calculations.This calculation does assume that all of the solids which strike the sampler are captured, andclearly this is not the case in reactor sections A-C. However, measurements suggest that thesampling efficiency at reactor section D is very high (95-100 %).274275Appendix 2 - Derivation of Particle Blockage EffectsTo calculate the rate of radiant heat transfer to the particles, a mean wall-particle view factor wasrequired. The derivation of this view factor for the UBC flash furnace has been presented below,illustrating the effects of particles blocking the incident radiation from the walls.A2.1 Volume Fraction Solids in UBC Flash FurnaceFrom the UBC flash furnace data presented in Chapter 4, the following can be obtained:Mass Input Rate = 2 kg/mm = thOxygen Input Rate = 0.4 kg/mm = thgAssuming that the MK concentrate has a true density (pr) of 5500 kg/rn3,and that the densityof oxygen (p9) is 1.31 kg/rn3 then:pVolume Fraction Particles (X) =_rP,, PgSubstituting the known values yields X = 0.0012 or 0.12%.A2.2 Relationship Between Volume Fraction and Area FractionConsider the case of a spherical particle exactly surrounded by a cubical region of space(Figure A2.1). The volume fraction of space occupied by the sphere is V/(2r)3 = (4/3)iur3/8rort/6.The area fraction occupied by the sphere is mr2/4 or it/4. Thus, for a given volume fraction,the area fraction will be 50 % larger.2762rFigure A2. I : Diagram of Volume Fraction of Sphere, Radius rA2.3 Particle Blockage Effects - Normally Incident RadiationConsider a ray of emitted radiation passing through a region of space defined by a distance x,over which the volume fraction of solids is constant at X (Figure A2.2).x..S S SFigure A2.2: Schematic Diagram Illustrating Particle Blockage EffectOn average, the volume of space surrounding each particle will be given by Vi,! X,,1, and theside of the equivalent cube by s= ( VP! X ). In traversing the region of space defined byx, the total area not blocked by the particles is therefore given by:277= (1.0— 1.5 )(xls) (A2. 1)For a mono-sized concentrate of 20 micron particles, V, = 4.19x1015m3 and s = 0.152 mm.For radiation oriented normal to the axis of the two-phase jet, and assuming that solids do notspread beyond the initial burner radius, then= (1.0- 1.5 ‘ 0.0012 )(O.0079/O.000152) = 0.91Therefore, 91 % of the wall radiation oriented normal to the axis of the jet can reach theparticles at the centre.A2.4 Calculation of Mean Wall-Particle View FactorConsidering the case of a particle at the centre of the reactor, and assuming that the concentrate particles do not spread beyond the initial burner radius (Figure A2.3), then the distancex that a given beam of radiation incident at 0 to the normal is given byx = — (A2.2)cos 0and the corresponding wall-particle view factor for this angle is given by substitutingEquation (A2.2) into Equation (A2.1)F(0) = (1.0 — I .5Xy’°° (A2.3)Assuming that the particle being considered is relatively far ( > 0.1 m ) from the top orbottom walls, then the mean wall-particle view factor can be calculated from:(A2.4)= J2F,,(O)d0J2dOPerforming this calculation for the conditions of the UBC flash furnace gives a meanwall-particle view factor of 0.803, or approximately 0.8.278A2.5 Calculation of Back-Radiation EffectsThe distance over which significant amounts of heat can be transported upward from thecombustion zone can be calculated from Equation (A2. 1). For example the distance at which90 % of the radiated energy from the combustion zone has been blocked can becalculated by:.00152 (in 0.1 un (1- 1.5 (.0012))) = 0.19 mTherefore, the radiated energy from the combustion zone can only penenate upward to a veryslight extent. In addition, the area of an adjacent particle constitutes a very small fraction(0.18 %) of the total area “seen” by a particle in a given direction, and hence particle-particleradiative heat transfer cannot be a significant factor affecting concentrate ignition.Edge of particlesReactor WallII II I A’I /I /I I /I /I I /I I /I. ‘II ‘II I) II IIII II II 1I II II II II II II IReactor Wall279Figure A2.3 Diagram illustrating the effect of radiation oriented at an angle 0 to thenormal of the jet axis.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items