CHARACTERISATION AND CONTROL OF THE ZINC ROASTING PROCESS

JENSNYBERG

Faculty of Technology,Department of Process andEnvironmental Engineering,

Systems Engineering Laboratory,

University of Oulu

OULU 2004

JENS NYBERG

CHARACTERISATION AND CONTROL OF THE ZINC ROASTING PROCESS

Academic Dissertation to be presented with the assent ofthe Faculty of Technology, University of Oulu, for publicdiscussion in Kuusamonsal i (Auditorium YB210),Linnanmaa, on December 17th, 2004, at 12 noon.

OULUN YLIOPISTO, OULU 2004

Copyright © 2004University of Oulu, 2004

Supervised byProfessor Urpo Kortela

Reviewed byProfessor Sirkka-Liisa Jämsä-JounelaProfessor Raimo Ylinen

ISBN 951-42-7609-4 (nid.)ISBN 951-42-7610-8 (PDF) http://herkules.oulu.fi/isbn9514276108/

ISSN 0355-3213 http://herkules.oulu.fi/issn03553213/

OULU UNIVERSITY PRESSOULU 2004

Nyberg, Jens, Characterisation and control of the zinc roasting process Faculty of Technology, University of Oulu, P.O.Box 4000, FIN-90014 University of Oulu, Finland,Department of Process and Environmental Engineering, Systems Engineering Laboratory,University of Oulu, P.O.Box 4300, FIN-90014 University of Oulu, Finland 2004Oulu, Finland

AbstractIncreasing efficiency is a necessary target for an industrial roaster nowadays. This thesis presentssome studies on efficiency improvement in the zinc roasting process - process characterisation,control design, implementation and testing.

The thesis focuses on the roaster, i.e. on research regarding the phenomena in the roaster furnace.By learning more about the roasting mechanism, particle size growth and dynamics of the furnace,new control implementations have been developed. More measurements, analyses and calculatedvariables have been added to give more information on the state of the furnace. New control variableshave been introduced to give the operators more opportunities to set the conditions so that they aremore suitable for the actual concentrate feed mixture. Equipment modifications have also been done.In this research, both laboratory and plant experiments have been performed together withthermodynamic evaluations and calculations. It has been necessary to make plant trials in order toobtain information about the impacts of different variables on the process. Only full-scaleexperiments give reliable results of the behaviour of an industrial furnace. The experiments with theroaster furnace have emphasised the study of both the metallurgy and the dynamics of the roastingprocess. The on-line calculated oxygen coefficient and its active control have proved important. Theparticle size distribution analysis of the furnace calcine has been shown to be a significant source ofinformation for evaluating the state of the roasting furnace.

The main target is to improve the economic performance. The key is to be able to be flexible inusing different kinds of raw materials, because the main income is the treatment charge. The trend isthat concentrates are becoming finer, which increases the challenges for roaster furnace control. Thecapability to use low-grade concentrates is also a major challenge and improves the economic result.

Research and development on the boiler and mercury removal has also been part of this work formany reasons. Improved boiler performance and mercury removal gives more freedom in choosingconcentrates and operating the roaster furnace. The approach has been the same as in the roasterfurnace research and development work.

Control improvements based on existing knowledge, such as fuzzy control systems for controllingthe furnace temperature and mercury removal, did stabilize the process, but they did not solve all theproblems regarding process stability. The research and development concept of this thesis hasprovided the extra knowledge needed for further improvement of process control. The results of theprocess characterisation have led to the implementation of a new and effective control strategy.

The research and development carried out has improved performance in a number of ways:increased running time of the furnace and boiler, in-depth knowledge of roasting phenomena whichled to new control methods and instructions for the operators, improved quality of sulphuric acid anda method to control its quality, measurements and analyses that give valuable information of the stateof the process – all of which are now in use.

In the future, the emphasis will be placed on the research and development of roaster furnaceperformance, which will be a great challenge. Control of the roaster furnace is the key to the economicsuccess of the roasting process and more information about these phenomena is needed for improvingand optimising control.

Keywords: control applications, control methods, process dynamics, roasting process

Acknowledgements

The theme of this thesis is to improve zinc roaster efficiency. Characterisation of the pro-cess has led to better control of the process. The process that has been developed is in useat the zinc roaster of Boliden Kokkola Oy in Kokkola, Finland, now a part of the newBoliden Group.

The work was supervised by Professor Urpo Kortela. I would like to thank him for hisadvice, comments and support. He was the initiator of my postgraduate studies. I was alsomade welcome by other postgraduate students and the personnel at the systems engineer-ing laboratory at the University of Oulu. This period in my life has been challenging andinteresting in many ways. It has entailed a lot of work. To combine daily work with stud-ies requires energy, however sometimes it can give energy too.

I wish to thank my boss, Mr. Kullervo Myllykoski, for the trust he showed during theresearch and development period. The freedom to do full-scale plant experiments wasboth challenging and necessary. Kullervo’s support was invaluable, because this kind ofdevelopment requires both time and money.

I would like to thank my colleagues for their dedication and encouragement during thisresearch and development process. I am especially grateful to Dr. Pekka Taskinen andMrs. Maija-Leena Metsärinta for this long-term research co-operation, their comments onmy work and their positive attitude. I would also like to thank other colleagues, such asDr. Antti Roine, Dr. Björn Saxén, Mr. Juha Järvi, Mrs. Heljä Peltola, Mr. Heikki Siirilä,Ms. Aija Rytioja and Mr. Hannu Vainonen. Their contribution has been of great impor-tance. Many other people have also contributed and special thanks goes to the whole per-sonnel of the roaster department. The performance and attitude of the operators, both dur-ing the testing periods and in implementing the results, was excellent. I am also gratefulfor the support and encouragement of my friends and nearest and dearest.

For correcting the text I would like to thank Ms. Margaretha Niemi. The proofreadingof the text into correct English was done by Mr. Mike Jones and Ms. Sue Pearson at PelcSouthbank Languages. I wish to thank Mike and Sue for their excellent work.

My son Daniel and my daughter Johanna have been in my thoughts during this period.

Kokkola, 21 November, 2004 Jens Nyberg

List of abbreviations

ESP Electrostatic PrecipitatorHR Human ResourcesISF Imperial Smelting FurnaceLME London Metal ExchangeORC Outokumpu Research CentreR&D Research and DevelopmentSEM Scanning Electronic MicroscopeSOM Self Organizing MapsTC Treatment ChargeWHB Waste Heat Boiler

List of figures

Fig. 1. Key elements in the development of the new control strategy. . . . . . . . . . . . 19Fig. 2. High roasting performance can only be reached through broad

development. Many factors must be taken into consideration to reach the targets, as shown in the diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Fig. 3. Roasting research and development procedure.. . . . . . . . . . . . . . . . . . . . . . . 21Fig. 4. Boiler research and development procedure has also been diverse. . . . . . . . 24Fig. 5. Experimental apparatus of furnace and boiler (scale 1:20) built for

flow model testing.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Fig. 6. Thermodynamic equilibrium calculations were essential in the mercury

removal research and development procedure. . . . . . . . . . . . . . . . . . . . . . . . 25Fig. 7. A block diagram of the Kokkola zinc process (Takala 1999).

Concentrates are fed both to the roasting process and concentrate leaching, which increases flexibility. Concentrates more suitable for roasting are fed to the roaster, while those more suitable for leaching can be fed to concentrate leaching. . . . . . . . . . . . . . . . 29

Fig. 8. Schematic diagram of a Lurgi type fluid bed roaster furnace (Themelis & Freeman 1983/1984). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

Fig. 9. The particle size distribution varies for different concentrates.. . . . . . . . . . . 34Fig. 10. Microscope picture of Aznacollar concentrate. The pellets contain

mostly small particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Fig. 11. Microscope picture of Trzebionka concentrate. The pellets also

contain large particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Fig. 12. Schematic drawing of the reaction principle. The sulphide mineral

particle reacts with oxygen. The product is calcine (mostly ZnO). During the reaction SO2 gas is formed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Fig. 13. Schematic drawing of a calcine agglomerate formed of many small calcine particles. Bonds are formed between particles due to molten phases. 36

Fig. 14. Agglomerate from the furnace (November 2001) – origin Aznacollar concentrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Fig. 15. Agglomerate from the furnace (November 2001) – origin Polaris concentrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Fig. 16. Grain size distribution of concentrate versus calcines. The formed calcine is coarser than the concentrate, especially the overflow calcine from the furnace. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Fig. 17. The Geldart classification of particles into their fluidisation behaviour in air at ambient conditions (Geldart 1986). The density of calcine in the fluidised bed gives the working area in the upper part of B. . . . . . . . . 40

Fig. 18. Fluidisation characteristic diagram, velocity versus particle diameter (Geldart 1986). The diagram shows the working area for a fluidised bed. . . 41

Fig. 19. Basic and advanced measurements of the furnace. . . . . . . . . . . . . . . . . . . . . 47Fig. 20. The control of the furnace is a multivariable system. Many aspects

must be considered. Knowledge of the dynamics and the thermodynamics of the process is essential. . . . . . . . . . . . . . . . . . . . . . . . . . 50

Fig. 21. Control inputs of the different control levels for the furnace control. . . . . . 51Fig. 22. The calculation scheme for the oxygen coefficient and its use for control.

This has been adopted and is now in use in the control of the furnaces at the Kokkola roaster.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Fig. 23. A furnace control diagram is shown with advanced control and measurements. It includes among others a fuzzy control loop, the calculated oxygen coefficient, oxygen measuring from the gasline, calculated calcine overflow amount and grain size distribution analysis of the furnace calcine. The possibility of feeding water to different spots and oxygen is also available. . . . . . . . . . . . . . . . . . . . . . . . . . 54

Fig. 24. The drop in windbox pressure and the increased proportion of fine calcine indicate an unstable bed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Fig. 25. Temperature drop indicates an unstable bed. Temperatures measured from the sidewall fell while the temperature measured from the bottom remained stable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Fig. 26. Sulphide level of the calcine (WHB, Cyclone and ESP) decreases when the oxygen coefficient is high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Fig. 27. Sulphide level of the WHB calcine is lower when the oxygen coefficient is high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Fig. 28. High oxygen coefficient increases the amount of sulphates. Difference SO4 = the amount of sulphates minus PbSO4 and CaSO4, which are always sulphates in roasting conditions. . . . . . . . . . . . . . 60

Fig. 29. High oxygen level in the WHB gas increases the sulphate amount. . . . . . . . 61Fig. 30. High oxygen level in the WHB gas decreases the sulphide amount.. . . . . . . 61Fig. 31. Oxygen feed versus fine and coarse particles in the bed. During

oxygen feeding the amount of fine particles decreased, while the amount of coarse particles increased. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Fig. 32. The oxygen coefficient versus fine and coarse calcine in the bed. Higher oxygen coefficient decreased the amount of fine particles and increased the amount of coarse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Fig. 33. On-line gas measurement from the roaster bed on the feed side. The O2 level in the bed is very low.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

Fig. 34. On-line gas measurement from the roaster bed on the overflow side. The O2 level is higher on the outlet side of the furnace. . . . . . . . . . . . . . . . . 63

Fig. 35. High moisture of the concentrate mixture decreases the feed. The water makes agglomerates and transfers the burning more to the bed. The feed is reduced because otherwise the temperature would rise in the furnace, due to the furnace temperature control application.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Fig. 36. Water fed to the concentrate before the furnace decreases the concentrate feed. In this case too, water makes agglomerates and transfers the burning more to the bed. The feed is reduced because otherwise the temperature would rise in the furnace, due to the furnace temperature control application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Fig. 37. Water injection into the bed increases the feed. Water sprayed (500 l/h) into the furnace bed cools the bed and more concentrate has to be fed to keep the temperature in the furnace at the set point value due to the furnace temperature control application. . . . . . . . . . . . . . . . 65

Fig. 38. Water fed into the concentrate before the furnace decreases the sulphide level of the boiler and cyclone calcine. The feed decreases due to the water feed and this increases the oxygen coefficient. . . . . . . . . . . 66

Fig. 39. Increasing the water feed (*1) before the day bins increases the sulphate amount in the bed. (*1 = The concentrate feed rate to the day bins is about 350 t/h and water is fed to this stream). . . . . . . . . . . . . . . 66

Fig. 40. SEM picture (left) of an overflow calcine 10.4.2002 (40 x). The coating around the inner part (=ZnO) is PbSO4. The other picture (right) shows that copper behaves differently to lead. A Cu2S core is formed in a copper rich concentrate pellet in a laboratory roasting experiment at 950°C. The coating around the Cu2S core is ZnO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Fig. 41. The Cu level of the feed versus the amount of calcine overflow of the concentrate feed. The proportion of the overflow calcine has decreased when the Cu level has increased. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Fig. 42. Oxygen coefficient versus the amount of calcine overflow of the concentrate feed. The proportion of the overflow calcine has increased when the oxygen coefficient has been increased. . . . . . . . . . . . . . . . . . . . . . 68

Fig. 43. Bed temperatures and oxygen coefficient of roaster furnace No. 2 in August, showing the effect of oxygen coefficient alterations on stabilisation. By increasing the oxygen coefficient, the stabilisation of the bed temperatures can be seen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

Fig. 44. Oxygen coefficient and wind-box pressure in furnace No. 2 (September–October 2002). When the wind-box started to drop, the oxygen coefficient was increased in order to raise the wind-box pressure back to the normal level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Fig. 45. Wind-box pressure and the total moistening water feed in roaster furnace No. 2 (September–October 2002). Water feed correlates with bed pressure, thus the wind-box pressure is low, i.e. unstable bed, when the water feed is low. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Fig. 46. Oxygen coefficient and bed temperatures (# 4 & 5) of roaster No. 1 (July–August 2002). The correlation is not so good, i.e. the temperatures stay stable even if the oxygen coefficient drops. . . . . . . . . 71

Fig. 47. Water feed before the day bins and bed temperatures of roaster No. 1 (July–August 2002). The temperature drops (Thermocouple element No. 5) after the water feed is stopped. After the water is again fed to the concentrate mixture, the temperature value returns back to the normal level.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Fig. 48. Increasing the oxygen coefficient stabilized the bed temperatures (the negative trend stopped). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Fig. 49. Changes of the oxygen coefficient were made by changing the feed rate. This was done by changing the set point value of the furnace temperature. The fuzzy controller then decreased or increased the feed rate. . . . . . . . . . . 72

Fig. 50. Increasing the oxygen coefficient stabilized the bed pressure. The drop in bed pressure was stopped and a slight increase back tothe normal level can be seen. The period is the same as in Fig. 49. . . . . . . . 73

Fig. 51. Increasing the oxygen coefficient brought the bed pressure back to normal. 73Fig. 52. The structure of the installed fuzzy control system (Rauma et al. 2000). . . . 74Fig. 53. The scope of the simulator made for training start-up and shutdown situations.

A dynamic model for the furnace pressure was made (Toskala et al. 2001). 75Fig. 54. Responses of gas temperature and bed temperatures, when water is

fed to the concentrate mixture before the table feeder. The temperature rises in the bed when water is fed to the concentrate mixture. At the beginning there is a slightly inverse reaction in the bed temperatures. The top temperature decreases when water is fed to the concentrate mixture.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Fig. 55. Water feed before the table feeder stabilises the bed, which can be seen from the fact that the temperature measurements in the furnace are closer to each other when more water is fed. . . . . . . . . . . . . . . . . . . . . . . 77

Fig. 56. Response of water injection into the bed. The furnace temperature decreases due to the cooling effect of the water injection. . . . . . . . . . . . . . . 77

Fig. 57. Divergent effect of oxygen on bed temperature measurements. The oxygen seems to have a divergent effect on the temperatures, which could be explained by an increase of the reaction rate in some parts of the furnace. Where there is a temperature drop, the reaction rate decreases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Fig. 58. Tracer content in the furnace overflow calcine for two experiments. The Ti-oxide tracer was fed as an impulse at time 0 h. A first order model with a time constant of 20 h is shown for comparison (Saxén et al. 2004). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Fig. 59. The tracer tests shows that the time constant in the gas line is very short, i.e. the Ti-oxide tracer fed as an impulse to the furnace is almost immediately detected in the calcine from the cyclone after the boiler. . . . . . 79

Fig. 60. The effect of the baffle wall on the boiler behaviour (radiation section of the boiler) is shown here. The velocity vectors are shown.

The Fluent model calculation was made by Foster Wheeler (= the boiler manufacturer). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Fig. 61. The effect of the baffle wall on the boiler behaviour (radiation section of the boiler) is shown in this figure. The temperature gradients are shown. The Fluent model calculation was made by Foster Wheeler (= the boiler manufacturer).. . . . . . . . . . . . . . . . . . . . . . . 81

Fig. 62. The feed increases when the temperature is higher in the furnace. The increased feed decreases the oxygen coefficient and the result is that the sulphide level of the formed calcine is higher, as shown here (Nyberg et al. 2000). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Fig. 63. By increasing the amount of air, the feed increases due to the cooling effect of the air (graph on right). Due to the lower oxygen coefficient the sulphide level of the calcine increases (graph on left). The furnace temperature in the experiments was 950°C (Nyberg et al. 2000). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Fig. 64. By using oxygen the feed increases, due to the cooling effect of the oxygen. In the experiments with additional cooling (C and D), more cooling coils had been added to the furnace. The furnace temperature in the experiments was 950°C (Nyberg et al. 2000). . . . . . . . . . 88

Fig. 65. Higher top temperature increases the feed and due to the lower oxygen coefficient the sulphate level of the formed calcine is lower, thus the sulphate load on the WHB decreases. . . . . . . . . . . . . . . . . . . . . . . . 88

Fig. 66. In both cases the feeds and the oxygen coefficient are the same. By injecting water, the temperature can be lowered in the furnace, without decreasing the concentrate feed. In cases with high impurity levels, a temperature level of about 900°C might be necessary. . . . . . . . . . . 89

Fig. 67. The operating range in the furnace is narrower if the concentrate mixture contains high impurity levels. The temperature and the oxygen coefficient are the main control variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

Fig. 68. The figure shows the developed control strategy. The target is to build a control application that maximizes the feed based on this control strategy. The aim is to optimise the feed and the process conditions (temperature and oxygen coefficient), based on metallurgical knowledge and knowledge of the dynamics. Measurements, analyses, the dynamic model and control variables are also key elements in the control application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

Fig. 69. In order to achieve sustainable results, the issues shown in the diagram must be considered in the development process. . . . . . . . . . . . . . . . . . . . . . . 94

Contents

AbstractAcknowledgementsList of abbreviationsList of figuresContents 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.1 Background of the research and development work . . . . . . . . . . . . . . . . . . . . 171.2 Research problem and asserted hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.3 Content of the research and development work . . . . . . . . . . . . . . . . . . . . . . . 20

1.3.1 Roasting research and development procedure . . . . . . . . . . . . . . . . . . . . 211.3.2 Boiler research and development procedure . . . . . . . . . . . . . . . . . . . . . . 231.3.3 Mercury removal research and development procedure . . . . . . . . . . . . . 25

1.4 Contribution of the author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262 Process description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.1 Zinc process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.2 Roasting process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.3 Gas cleaning and sulphuric acid production . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.4 Roasting and direct leaching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Main phenomena of the roasting process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.1 Mineralogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.2 Main reactions and thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2.1 Sulphates and molten phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.2.2 The sulphide analysis of the formed calcine . . . . . . . . . . . . . . . . . . . . . . 39

3.3 Heat transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.4 Fluidisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.5 Gas cooling and dust handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.6 Gas cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.7 The dynamics of the process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4 Targets for control system development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435 Measurements and control of the fluidised bed furnace . . . . . . . . . . . . . . . . . . . . . . 46

5.1 Measurements and analyses of the fluidised bed furnace . . . . . . . . . . . . . . . . 47

5.2 Control specification of the fluidised bed furnace . . . . . . . . . . . . . . . . . . . . . 495.3 Aspects relating to the boiler and mercury removal . . . . . . . . . . . . . . . . . . . . . 55

6 Results of the control strategy development for the Kokkola zinc roasting process 566.1 Furnace development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.1.1 Factors correlating to bed instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576.1.2 Oxygen coefficient control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586.1.3 Oxygen use to control bed quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616.1.4 Bigger nozzles and oxygen measurements from the bed . . . . . . . . . . . . 626.1.5 Effects of water on the furnace behaviour . . . . . . . . . . . . . . . . . . . . . . . . 646.1.6 Control of the furnace when Pb and/or Cu is high . . . . . . . . . . . . . . . . . 666.1.7 Controlling unstable situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.1.8 The dynamics of the roaster furnace . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6.2 Boiler development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 796.3 Mercury removal development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827.1 Summary of the results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

7.1.1 Summary of roasting development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 847.1.2 Summary of boiler development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857.1.3 Summary of mercury removal development . . . . . . . . . . . . . . . . . . . . . . 86

7.2 Control strategies and further research and development . . . . . . . . . . . . . . . . 867.2.1 Increasing the capacity of the furnace . . . . . . . . . . . . . . . . . . . . . . . . . . . 877.2.2 Example of flexible control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887.2.3 Operation conditions for different feed mixtures . . . . . . . . . . . . . . . . . . 897.2.4 The developed furnace control strategy . . . . . . . . . . . . . . . . . . . . . . . . . 90

8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

1 Introduction

Roasting is an essential part of the electrowinning zinc process. A typical Zn process linecomprises the following sub-processes: roasting, leaching and purification,electrowinning and the last step foundry (Kajiwara et al. 1995, Morris 2000). During thelast few decades, the direct leaching of concentrates, pressure or atmospheric, has beenintroduced as a parallel and/or alternative process stage to roasting. Besides typicalelectrowinning processes there are also other methods to produce zinc, with the ISF(Imperial Smelting Furnace) being a method used by some plants (Bendieck et al. 2002,Brook Hunt 2003, Canham & Charles 1981, James et al. 2000, Lee 2000).

Zinc concentrates from the mines are fed to and roasted in fluidised bed furnaces. Themost common roaster furnace in zinc roasting is the Lurgi furnace (Brook Hunt 2003).The main reaction is ZnS + 1.5 O2 ZnO + SO2. The main components of the roastingprocess are: the furnace, waste heat boiler (WHB), cyclone, electrostatic precipitator(ESP), Hg removal and sulphuric acid plant. Essential process equipment includes:conveyors, bins, calcine cooler, ball mill, fans and steam handling equipment (Magoon etal. 1990). The fluidised bed roaster furnace has been used on commercial scale since themiddle of the 1950s (Heino et al. 1970, Themelis & Freeman 1983/1984). The zincprocess and roasting process are more thoroughly described in chapter 2.

Zinc processes are continuously being developed and the trend is that they arebecoming bigger and more efficient. New possibilities like direct leaching of concentratesis one trend in this development work. Roasters are also being developed and one reasonfor this is that the concentrates are tending to become finer, which has an impact onroasting behaviour. The capability of using impure concentrates is another objective fordevelopment work.

1.1 Background of the research and development work

In this study the roasting process has been developed using intensive and extensiveresearch. The goal has been to improve the process control of the roasting process. Theaim is to understand the behaviour and to evaluate both details (micro level) and the

18

entire roasting process (macro level) to reach the target of increasing the efficiency of theprocess. To be able to do this many aspects must be considered. The focus in the thesis ison the actual fluidised bed roaster furnace behaviour, but developments in the boiler andmercury removal are also discussed to some extent. The roasting mechanism is a verycomplex process and therefore requires knowledge of many fields. Matters like themineralogy of concentrates, thermodynamics, fluidisation, heat transfer and dynamicsmust be studied to gain more knowledge. These are discussed in chapter 3. Normally thefeed to the furnace is a mixture of concentrates and every mixture behaves differently,which increases the difficulty (Brown et al. 1995). Flexibility is required to be able tohandle different feeds. This flexibility means the possibility to run the process in optimalconditions by having the control tools to change the operation conditions. The key toreaching optimal conditions is to have a thorough knowledge of the process.

The behaviour of the roasting furnace is influenced by several parameters and thecomplexity is vast. If, for instance, the impurity levels for certain elements exceed certainlimits or the process conditions are not suitable for the feed, this will lead to instability.

The boiler is closely connected to the roaster furnace and by making the boilerperformance effective, more freedom in roasting conditions is achieved and thus theconstraints due to the boiler are smaller. The same applies to mercury removal; moreefficiency gives more freedom when choosing concentrates.

The reasons for development work are the need to avoid disturbances in the furnace(build-ups, defluidisation and instability), to increase the running time of the furnace andboiler without cleaning and to avoid quality problems with the sulphuric acid (too high alevel of mercury).

The need to increase process knowledge was clear, as the problems could not behandled merely by adopting stabilizing control applications like fuzzy systemapplications. The fuzzy systems developed for temperature control in the furnace and forcontrolling the acid concentration and temperature in mercury removal work well but theydo not solve problems caused by changes in concentrate mixtures. It is important to knowthe optimum process conditions and to have measurements showing the state of theprocess. It is also essential to have the necessary control tools.

The primary goal was to improve the economic result of the operation. Other majorgoals were to decrease the environmental impact of the operation and to reduce safetyproblems. An increase in feed capacity, running time and improved product quality werethe intermediate targets of the work. Achieving this would enable the primary goals to bereached. The treatment charge (TC) is the major input and optimisation of this is the mostimportant factor. The greater quantity of concentrates treated, the higher the revenuesfrom the TC. This means that if concentrates with low Zn levels are used to produce acertain amount of zinc metal, then the income is larger. Decreasing production costs wasanother goal and maintenance costs also had to be optimised. The targets are discussedmore thoroughly in chapter 4.

Flexibility in using different types of concentrates along with the possibility andcapability of using impure concentrates is the key to success. Another major factor is theability to handle fine concentrates. The trend is that concentrates are becoming finer. Thebasic platforms for reaching the goals include process knowledge, working skill, controlpossibilities, and equipment performance.

19

1.2 Research problem and asserted hypothesis

The roasting process of zinc sulphides is very complex. In order to improve the efficiencyof an industrial roaster many aspects must be considered and characterisation of theprocess performed. More metallurgical knowledge is needed, e.g. about the roastingmechanism, calcine particle size growth and the effects of impurities. More knowledgemust also be gained about the state of the process and the dynamics of the furnace.Knowledge about the optimum process conditions is required. New measurements,equipment modifications and control variables are needed. The impacts of changes mustbe studied and experiments have to be performed to provide the necessary information.

Both laboratory and full-scale plant experiments are necessary. Besides knowledgeabout the dynamics knowledge in process metallurgy is also essential. Knowledge aboutthe state of the process requires measurements (both new and old), but also understandingof process phenomena. The effects of control variables must also be studied. Adding morecontrol variables and equipment modifications has also been a part of this development.These additions make it possible to create optimal conditions and increase the flexibilityof the control. The chosen procedure has been to proceed step by step.

In order to reach the goal, i.e. increased efficiency in the control of the roastingprocess, a new control strategy has been developed. Essential parts of the new controlstrategy are process characterisation, dynamic models, new analyses and newmeasurements, new control variables and equipment modifications. The key elements ofthe basis for the new control strategy are shown in Fig. 1.

Fig. 1. Key elements in the development of the new control strategy.

Plant experiments Process Charac- Control Strategy terisation

Emphasis onProcess Metallurgy

Emphasis onProcess Dynamics

ProcessKnowledge- Limits- Phenomena- Effects - Rules

Dynamic Behaviour- Models- Effects

Optimal Control Performance andDeveloping Control Applications

Knowledge about the State of the Process- Measurements- Analyses

ControlVariables

EquipmentModifications

20

1.3 Content of the research and development work

Increasing the efficiency of an industrial roasting process requires that many phenomenaand many variables must be studied, as shown in Fig. 2. The main focus has been onimproving roasting by studying furnace behaviour and roasting mechanism. Muchattention has also been paid simultaneously to boiler and mercury removal performance.These three elements have a great bearing on success. Development of the boiler isessential, because if the boiler performance is good, then there are fewer constraints infurnace control. Mercury removal is necessary to produce sulphuric acid of high quality.Good mercury removal gives more freedom in choosing concentrates, which is essentialwhen evaluating the whole process.

The development that has been made is based on the results of laboratoryexperiments, full-scale plant experiments, calculations, thermodynamic evaluations,experiments with models and extensive chemical analyses.

Fig. 2. High roasting performance can only be reached through broad development. Manyfactors must be taken into consideration to reach the targets, as shown in the diagram.

Control loops and algorithms have been improved by studying the dynamics of theprocess and using advanced methods like fuzzy systems. By doing this, the process hasbeen stabilized. Good examples of such are the fuzzy control of the roaster furnacetemperature and the fuzzy control of the mercury removal towers to stabilize the

Roasting performance

Information Control - Chemical analyses - Control applications (Fuzzy) - On-line analyses - Advanced systems: - Measurements - Water - Oxygen

- Pelletizing - Others

Personnel Equipment - Knowledge - Design - Skill - Maintenance - Activity - Automation

Fluidisation Heat Transfer - Grain size distribution of - Cooling devices the formed calcine - Fluidisation - Viscosity

Concentrates Thermodynamics - Mineralogy - Reactions - Impurities - Phase diagram etc. - Grain size distribution - Temperature

- Oxygen coefficient

21

temperature and acid content (Rauma et al. 2000, Rauma 1999). Both applications workwell and they stabilize the process, i.e. the temperatures and acid content fluctuations aresmall. However they do not solve all the problems: for the furnace i.e. how to handle fineand impure concentrates; for mercury removal how to make good sulphuric acidconsistently, i.e. to keep the mercury level low in the sulphuric acid produced.

New control variables had to be found, as did ways of knowing the right processconditions for different concentrate mixtures and how to maintain them. Besides moreprocess knowledge new measurements, new control tools and methods are needed toimprove the control of the process. Control and control aspects are presented in chapters 5and 6. Measurements and analyses are also presented in chapters 5 and 6. Possiblemethods are discussed in chapter 7.

1.3.1 Roasting research and development procedure

It is essential to have a good and stable bed in the roaster furnace. Defluidisation due to afine bed is one problem. Another problem is the formation of build-ups due to moltenphases or sulphates, which can lead to sintering of the bed or part of the grate area. Heattransfer, oxygen need and fluidisation vary due to differences in the concentrate mixture.The aim of the research and development concerning roaster furnace performance hasbeen to gain more knowledge about the different mechanisms and to design a controlstrategy. The most important research stages are described below and shown in Fig. 3.

Fig. 3. Roasting research and development procedure.

Mineralogical studies of all concentrates used in Kokkola have been made. Chemicalanalysis and microscope analysis were made to get information about the concentratesand to determine what kind of minerals they contain. The basic results from the studieswere the oxygen requirement and the heat energy of different concentrates.

Evaluations of thermodynamic phase diagrams and Kellogg diagrams have been madeon a large scale. In roaster furnace conditions, impurities (Pb, Cu, Si) form molten phases.Sulphates are also formed (Pb, Ca). These have a negative impact on roasting andtherefore it is important to study the phase diagrams of the different combinations ofimpurities. The aim is to find the best roasting conditions and the limits for impurities.Different types of molten phases can be formed: sulphide melts (Pb, Cu, Fe, Zn), sulphatemelts (Na, K, Pb) and oxide melts (Pb, Cu, Si, Zn). The formation depends on the

Mineralogical Analysing Laboratory Plant evaluations calcine experiments experi-

samples ments

Applications and operating rules

Termodynamic Heat balance Active surveillance evaluations calculations of special situations

in the Roaster

22

conditions in the furnace. The main variables are temperature, oxygen pressure andamount of impurities (Appendix 1).

Analysis of calcine samples has also been made. During plant experiments and alsoduring normal operation, calcine samples have been taken mainly from the furnace butalso from the boiler, cyclones and ESPs. Chemical analyses, particle size distribution andmicroscopic analyses have been made (see chapter 6 and Appendix 4). Only by doingthis, can more information about growth mechanisms and what is actually going on in thefurnace be acquired. In particular, analysing samples under the microscope gives valuableinformation.

Thermodynamic heat balance calculations have also been made. The effects ofdifferent variables can be estimated by these calculations. Therefore a model using HSCSoftware has been made for the furnace boiler system (Roine & Nyberg 2000, Björklund2001). The model has gradually been improved to be more user-friendly and accurate.The impacts of different variables can be tested such as temperature, feed composition, airflow, use of oxygen and water injection to the bed. By doing this kind of testing,evaluations of the impacts can be made before real changes in the equipment or operatingparameters are made. This has been useful in the case of both the furnace and the boiler.

At Outokumpu Research Centre experiments have been carried out in small laboratoryfurnaces (Taskinen et al. 2002, Metsärinta et al. 2003, Metsärinta et al. 2005b) wheretests to study the effects and the limits for impurities have been made. Also, differentroasting conditions were tested, for example roasting with a high oxygen coefficient.High impurity levels were also tested. These experiments were very often pre-tests madebefore performing full-scale plant experiments. During these laboratory experiments,samples were taken for chemical and X-ray diffraction analysis.

During the research and development period (especially during 1999–2004) many full-scale roasting experiments were performed (Chapter 6, Lepistö 1999, Nyberg et al. 2000,Metsärinta et al. 2005a). Full-scale experiments are necessary in order to get informationand knowledge before making decisions on how to increase the performance of theroaster furnace; otherwise actual knowledge of how the furnace will behave cannot begained. Different items were tested such as impurity levels, temperature levels, airamount, water feed to different spots and oxygen feed. Risks are involved in plantexperiments. If for instance too high impurity levels are used, it can lead to too manyagglomerates and furnace instability. In the worst case the furnace must then be cleaned,i.e. causing production losses of several days and the ensuing extra costs. Due to thisthere are constraints in full-scale plant trials on how high the impurity levels are that canbe tested and therefore it is safer to make experiments first on laboratory scale.

The plant trials lasted from one to several weeks. The experiment plans includedsampling, an analysis plan and variable changing intervals. During the trials a vastamount of data was collected and active surveillance was made. Measurements andanalyses were collected on the database of the process computer and the plant informationsystem. On-line analysis of the gas composition in the roaster bed was also made duringseveral tests (Chapter 6, Parkkinen et al. 2002). Many samples were taken for chemicaland X-ray diffraction analysis. The results from the tests were thoroughly evaluated andreported.

The emphasis in the full-scale plant experiments has been on the process metallurgy.By learning more about the particle size growth mechanism and the effects of the

23

impurities, new means to control the furnace behaviour have been developed and alsonew process operation conditions have been chosen.

The dynamics of the process have also been tested to get information about the effectsof different variables on the process. By studying the dynamics, information fordeveloping and tuning control loops has been gained. The information has also beeninvaluable in understanding the behaviour of the furnace. By making step trials, i.e.changing the air flow, concentrate feed, water addition, oxygen feed etc. stepwise and byletting the furnace respond to the changes, the dynamics of the furnace have beenidentified. These results have been used to make simulators, a state variable model andtuning control loops (Rauma et al. 2000, Toskala 2000, Toskala et al. 2001, Härkönen2003, Saxén et al. 2004, Valo 2004).

Every now and then the roaster furnace becomes unstable. Active surveillance of thesespecial situations has been done. During those periods active measures have been taken tocorrect the situation back to normal. During such abnormal situations samples havenormally been taken and analyses of the samples made. Often these situations have beensuccessfully corrected by control methods, i.e. using oxygen, decreasing the furnacetemperature, using water etc. Sometimes a change of concentrate mixture has been theonly way to get the furnace back in equilibrium. Such periods have been thoroughlyanalysed and reported. These special situations have provided excellent opportunities fortesting the new control strategies and process knowledge.

During shutdowns, yearly maintenance shutdowns or boiler cleanings, samples havebeen taken from the build-ups and analysed. After analysis, the results have beenevaluated and explanations for the build-ups have been made (Appendix 4).

1.3.2 Boiler research and development procedure

The pattern for boiler development has been similar to that of roasting and is shown inFig. 4. The conditions of the furnace influence the boiler. The main goals are to haveseparation of the calcine coming from the roaster furnace gas through the upper part ofthe furnace, effective gas cooling and a high degree of uptime for the boiler. These are theprerequisites for good results in roasting performance.

Before major modifications were made in the boiler a flow scale model was built. Apicture of it is shown in Fig. 5 and the model has been presented in a paper (Järvi et al.2002). The biggest issue that was tested with this flow model was the impact of the bafflewall. The shape, size and the location of the baffle wall were tested.

24

Fig. 4. Boiler research and development procedure has also been diverse.

Fig. 5. Experimental apparatus of furnace and boiler (scale 1:20) built for flow model testing.

As a complement to the model, numerical calculations were also made to confirm theresults gained from the laboratory experiments. These calculations showed both thetemperature and the flow gradients of the boiler. The calculations gave similar results asthose with the laboratory model. Two commercial programs were used for calculations.Outokumpu Research (ORC) used the CFX programme (Järvi et al. 2002) and FosterWheeler (the boiler manufacturer) used the Fluent programme (Chapter 6). The results ofthe flow model calculations are similar to those in published papers (Peippo et al. 1999,Yang 1996).

During the last decade there has been intensive and close co-operation between theroaster department and the boiler manufacturer. Several meetings and planning sessionshave been held where the results have been evaluated and new modifications have beenplanned. Changes and additions to the boiler have been considered together beforeimplementation. Often changes have an impact on the stress factors in the boiler system.Therefore the risk is that without thorough evaluation new problems can arise. Inaddition, follow-up has been done together and based on the evaluation of the progress,new plans to develop the boiler have been made (Kilju 2004).

Utilisation of plant operation experience has been a part of the boiler developmentwork. The impact of the changes made can only be evaluated after a long period, several

Laboratory Cooperation with Developing of flow tests the manufacturer maintenance

Applications

Mathematical Utilizing of flowmodel plant operation calculations experience

25

months or years. Analysis of malfunctions is also a part of development work. By activemonitoring of the boiler over long periods, valid information for further development hasbeen gained.

Development of maintenance practice is crucial and has a great impact on theperformance of the boiler. Measurements during shutdowns (e.g. X-ray) and monitoringthe boiler during operation by measuring the water flows of the different circulations, arean area that should not be neglected. During shutdowns, maintenance and above allproactive maintenance are important. This means replacement in advance of parts that arenot in good condition. The way the boiler is cleaned also influences the result. Oneexample of an established practise that has increased running time is cleaning the boilerwith water during the yearly shutdowns. This improves the running time of the boiler.

1.3.3 Mercury removal research and development procedure

Likewise, mercury removal research and development has been implemented using thesame pattern. This is shown in Fig. 6.

Fig. 6. Thermodynamic equilibrium calculations were essential in the mercury removalresearch and development procedure.

Material balance calculations had to be made in order to get a clear picture of the entireprocess including the roaster, mercury removal and the sulphuric acid plant. The massbalance contained important elements (Hg, Se, Cl, F, Si, etc.) and all phases wereincluded, i.e. solid, liquid and gas. Some new analyses had to be made and every inputand output was calculated. The chemical analysis of every process stage (and everyphase) and the material balance calculations clearly showed the amounts and, above all,this work resulted in a better picture of the process.

Chemical analyses of process samples were performed in order to get a better pictureof the process. Samples were taken from all process stages. The challenge here wasproper sampling. This is due to the fact that the temperatures are high in some steps andthe content of some elements is very low, both in the liquid and gas phase. If sampling isnot done properly, the analysis is of no value. Due to the low level in some steps thechemical analysis was also demanding. One example of this is the determination of Hg0

and Hg2+ concentrations in the gas phase. This plays a significant role in the process

Material balance Thermodynamic Plant calculations equilibrium experi-

calculations ments

Applications

Chemical Laboratory analyses of experiments process samples

26

chemistry, but due to the low concentration levels in some process stages, analysis isdifficult.

Based on the material balance calculations and chemical analysis, massivethermodynamic equilibrium calculations for every process step were performed bypersonnel from ORC. These calculations were performed using commercial programssuch as HSC, ChemCalc, ThermCalc, etc. The results from the thermodynamiccalculations and the results from the chemical analyses were similar and showed that thegas phase is in a steady state in the different process stages. This information was vital forthe development of the process.

Based on the above, i.e. material balance calculations, chemical analyses andthermodynamic calculations, experiments on laboratory scale were performed. Differentchemicals such as Se solution and thiosulphate were successfully tested (Peltola et al.2000, Berg et al. 2003).

After the laboratory experiments, feeding of Se solution was carried out to differentparts of the process chain, both Hg removal and the acid plant. The most effectivelocation was found to be feeding Se solution to the acid plant process. The quality of thesulphuric acid improved and the process became more stable and manageable.

1.4 Contribution of the author

The author of this thesis worked as manager of the Kokkola roaster department during theresearch and development process presented in the thesis. As manager of the department,the author was responsible not only for production but also for the research anddevelopment of the process. The main partner in this research and development processwas Outokumpu Research (ORC). Universities and others have also participated, e.g.through Master’s theses. The necessity of continuous development is paramount and thetarget has been to increase efficiency. In order to achieve good results, a broaddevelopment in many fields has been necessary. The goals are discussed in chapter 4. Theauthor was the initiator of the research and development projects described in the thesis.

The author has created the control concept presented in this thesis for improving thecontrol of the process and has also been responsible for carrying out the implementation.The main contribution of the thesis has been the process characterisation and thecombination of the process knowledge gained with the control system development.

Active participation and contribution in all the projects has been an essential part ofthe author’s work. The research and development work consists of both basic and appliedresearch. Applied research, i.e. full-scale plant trials, has also involved risk-taking. Asmanager, the author has been responsible for this risk-taking. Therefore, the authorparticipated actively in the planning of the plant trials.

The results of the research have led to applications, e.g. equipment modifications, newcontrol methods, changes in operation practices, new measurements and analyses taken inuse, which have increased efficiency and also improved quality.

Some of the results of the research and development work have been presented inscientific publications. These publications are referred to in the thesis. The aim has beento present some of the key elements and achievements of the research and the

27

developments made. In the papers the procedures of the research and development workhave also been presented. The author has acted as the initiator in the publishing process.The author of this thesis has been active in creating the structures of the papers and in thewriting process of the presented papers.

As previously mentioned, the research and development work has led to practicalapplications. This has also resulted in patents and patent applications have been made.These patents and patent applications are referred to in the thesis. The author is aparticipant in seven patents or patent applications. In the others patents, that are in use atthe Kokkola roaster, the author has played a supportive role and been active in the testingof these systems.

Master’s theses have also been part of the research and development work. TheMaster’s theses are referred to in this thesis. They have been both theoretical andpractical. The theses have been an important part of the research and developmentprocess. They have helped to achieve practical applications and new research resultsvaluable for continuous development. The contribution of the author of this thesis hasbeen as the initiator, helping in the process characterisation, giving feedback andsupervisor of the Master’s theses.

2 Process description

The zinc business is changing all the time. It is a complex network consisting of mines,zinc plants, zinc market, end users and authorities. The price of zinc is of major importan-ce. The London Metal Exchange (LME) zinc price and the size of the LME zinc storageare relevant factors influencing the business. The treatment charge (TC) changes yearlyand is influenced by the zinc price. It is the factor that determines the income and survivalof both mines and zinc plants.

Zinc plants are becoming bigger and more efficient. The major costs, especially inelectrowinning Zn plants, are as follows:

– Zn concentrates– Electricity costs– Labour costs– Maintenance costs– ChemicalsThe major income for zinc plants is received from treating concentrates, i.e. TC pay-

ment. Other incomes are gained from selling zinc plus premium, sulphuric acid and insome cases steam. The demand for more efficiency has led to an increase in the automa-tion level and production capacity but also to the development of the process. Develop-ment work includes all zinc plant departments. Major improvement targets are to increasethe running time and leaching yield, which means better Zn recovery. An important tar-get is to make a purer solution in purification, because then the electric efficiencyincreases in the cellhouse. Challenges in roaster development include new concentrates,which tend to be finer, impurities in the concentrates, the need to increase running timeand reduce maintenance costs.

The continuous development work being done by the zinc industry has also led to newprocess stages such as the direct leaching of concentrates. Environmental aspects havealso been an issue for development with one major issue – how to treat iron residue. Forthis reason, great demands have been placed on the entire zinc industry in many countries.

Cooperation between zinc plants, mines and end users is also necessary in order tomeet the demands from society and the authorities.

29

2.1 Zinc process

There are different types of zinc processes. Among plants treating sulphide zinc concent-rates, the most common type is the electrowinning zinc plant. There are also ISF plantsoperating, though these are diminishing in number. The electrowinning plants differ as tothe kind of leaching process in use and the iron residue precipitated (jarosite, hematite orgoethite) (Brook Hunt 2003). There are also differences in the purification processes. Inthe cellhouse the major difference is the size of the cathode.

There are normally four sections in an electrowinning zinc plant. They are: Roaster,Leaching and Purification, Electrolysis (i.e. the cellhouse) and Foundry. In the roaster theconcentrates are roasted in a fluidised bed furnace. The SO2 gas from the roaster iscleaned (mercury removal) and then sulphuric acid is produced. In leaching the calcinefrom the roaster is leached using sulphuric acid. Iron is precipitated and the solution fromleaching is then purified, i.e. Cu, Co, Ni, Cd, Ge, Sb etc. are precipitated out. The purifiedsolution is then pumped to electrolysis, where zinc is precipitated on aluminium sheetsusing electric power. The Zn is stripped from the aluminium sheets by automatic strippingmachines. The zinc produced is then melted in the foundry.

Fig. 7. A block diagram of the Kokkola zinc process (Takala 1999). Concentrates are fed bothto the roasting process and concentrate leaching, which increases flexibility. Concentrates moresuitable for roasting are fed to the roaster, while those more suitable for leaching can be fed toconcentrate leaching.

As mentioned in chapter 1, direct leaching of concentrates has become an alternativeor parallel process to roasting. In the Kokkola zinc plant both are in use and this givesmore flexibility to the operations (Takala 1999). The advantage lies in the fact that someconcentrates are more suitable for roasting and others are more suitable for leaching. Ablock diagram of the Kokkola plant is shown in Fig. 7. A flow sheet with more details ispresented in Appendix 3.

ROASTING Hg-Se-PROCESS

NEUTRAL LEACHING

SOLUTION PURIFICATION

CONVERSION ELECTROLYSIS

CONCENTRATE LEACHING FOUNDRY

Zn concentrate

ZnSO4

H2SO4

Zn concentrate

Calcine

Ferrites

Jarosite

Jarosite S° + FeS2

Spent acid

SO2

Hg, Se

Cu, Co, Cd

ZnSO4

Zn cathodes

Zn slabs, alloys

NEUTRAL LEACHING

30

2.2 Roasting process

In the thesis the furnace type studied is the Lurgi furnace. Fig. 8 shows a schematic diag-ram of a Lurgi furnace. There are also other types of furnaces used for zinc sulphide roas-ting and some plants use the Dorr Oliver furnace (Yoneoka et al. 1998). The feed to thezinc roaster furnaces is normally a mixture of several concentrates. Concentrate composi-tion is discussed in the mineralogy section of chapter 3. The main phenomena regardingroasting are also described in chapter 3.

In the roaster the nozzle grate area of the furnace has gradually expanded and todaythere are some Lurgi furnaces in operation with a grid area of 123 m2 (Berg & Pape 1978,San Martin et al. 2000). A typical nozzle grate area for Lurgi furnaces is nowadays about72–77 m2. The two furnaces in Kokkola that have been the objects tested in this work areboth 72 m2.

Fig. 8. Schematic diagram of a Lurgi type fluid bed roaster furnace (Themelis & Freeman 1983/1984).

The concentrates are stored in warehouses and the feed mixture is prepared daily andtransported by conveyors to the day bins. Screening and crushing equipment have beeninstalled ahead of the day bins to take care of lumps. From the day bins the concentrate isfed to the table feeder and after that high-speed transport belts, i.e. slinger belts, throw thematerial into the furnace. Air is fed from the bottom of the furnace to the furnace gratethrough nozzles. This air makes the bed in the furnace fluidise. The oxygen in the airreacts with the concentrate mixture. The reaction is exothermic and the heat is generatedto the calcine and SO2 gas that are formed. The roasting temperature is about 920–960oC.

31

The cooling coils in the furnace remove some of the generated heat, but the boilerremoves most of the heat. The heat is recovered, both from furnace and boiler, by thewater-steam system of the boiler. The SO2 gas formed in the furnace leaves the furnacefrom the upper part and enters the boiler. The inlet temperature is about 850–950oC. Theboiler cools the gas to approximately 320–380oC, which is the inlet temperature to thecyclones. The majority (60–70 %) of the calcine is also included in the gas from the fur-nace. The boiler removes part of the calcine from the gas and the rest is removed incyclones and in ESPs. Some (about 30–40%) of the calcine leaves the furnace via theoverflow of the furnace. Some plants also have a furnace underflow to remove the calcinepartly from the furnace.

The calcine from the furnace is cooled in a cooler, normally a section cooler but inKokkola a fluidised bed cooler is used. The calcine from the furnace and boiler is crushedin a ball mill. All calcine from the roaster is collected together in an intermediate silo andtransported pneumatically to the calcine silo.

The SO2 blower that controls the pressure in the roaster furnace is situated after theESPs. A slight underpressure is kept in the furnace.

A good stable bed and good fluidisation are essential for the performance of the roasterfluid bed furnace. Many factors determine the optimal temperature value, but the impuritylevels are the crucial parameters. Therefore the chemical analysis of the calcine formed isimportant. Major impurities in roasting are Cu, Pb, Si, Ca, Na and K. The sulphide levelof the generated calcine, normally in the range of 0.1–0.3 %, is a major indicator of theroasting process. Good control and measurements are required to avoid forming of build-ups, sinters. Issues regarding control and measurements are discussed in chapter 5. In thefurnace there are several temperature measuring devices, pressure and windbox pressuremeasurements. There are several control loops such as draft control, air flow control andconcentrate feed control. The key for the behaviour of the furnace is the calculated feedmixture. The main phenomena regarding the roasting process are discussed in chapter 3.The Kokkola Roaster process is presented in Appendix 2.

2.3 Gas cleaning and sulphuric acid production

After the electrostatic precipitators, the SO2 gas with an SO2 concentration of 7–9 % iscleaned. Mercury is the major impurity to take care of. There are different types of gascleaning processes and many zinc roasters use the Boliden-Norzink process for mercuryremoval (Dyvik 1994). In Kokkola the Outokumpu mercury removal process is in use(Mukherjee 1999). Besides mercury other important elements to take care of are F, Cl andSe. The processes used for mercury removal are discussed in chapter 3.

The incoming temperature of the gas to the cleaning section is over 300oC. The gas iscooled and in Kokkola the outlet temperature of the gas after mercury removal is about60–65oC. The sulphuric acid plant also has washing towers and the final removal of mer-cury occurs in this stage. In the mercury removal process the mercury is removed fromthe gas and ends up in the slurry. The control of mercury is complex and the balancebetween the gas phase, liquid phase and solid phase is influenced by many parameters.The levels of mercury are very low in some phases and it is quite demanding to measure

32

them accurately. In order to produce sulphuric acid with low mercury, the whole process,i.e. mercury removal and the sulphuric acid plant, must be taken into account. Theamount of mercury varies in the concentrates a lot (0.001–0.19 %). The trend is, however,that the product, sulphuric acid, must be low in impurities (Ludtke 2001). Nowadays thetrend is that the mercury level should be lower than 0.1 mg Hg / kg H2SO4.

2.4 Roasting and direct leaching

Direct leaching of zinc concentrates has gradually become more common during the lastdecades as a parallel or alternative process to roasting (De Nys & Terwinghe 1990). The-re are two types of direct leaching, pressure leaching and atmospheric leaching (Buban etal. 2000). Pressure leaching (autoclave leaching) has been in use since the 1980s (Ash-man & Jankola 1990, Collins et al. 1990, Dreisinger et al. 1990, Masters et al. 2002,Mollison & Moore 1990, Ozberk et al. 1995). There are some plants that have addedpressure leaching as a parallel process to increase the capacity, but there is also a zincplant that treats all the concentrates using pressure leaching without any roaster capacity(Collins et al. 1994, Collins et al. 1995, Krysa 1995). In the 1990s, direct atmosphericleaching of concentrates has been adopted and increasing interest towards this processhas arisen (Adams et al. 1990, Takala 1999).

The use of direct leaching (pressure or atmospheric) of concentrates, as a parallel pro-cess to roasting, gives more freedom when choosing concentrates. Some impurities, suchas Cu and Pb, can be higher in leaching while concentrates with high mercury and chlo-rine are more suitable for feeding to the roaster. The sulphur in the concentrate ends up inthe sulphuric acid in the roasting process and in direct leaching the sulphur ends up in thesulphur residue. Here the sulphuric acid market and the possibility to store the additionalresidue influence the process choice.

3 Main phenomena of the roasting process

In the roasting process the following issues are of great importance: mineralogy, thermo-dynamics, heat transfer, fluidisation, gas cooling, dust handling and gas cleaning. Indevelopment work, i.e. when aiming to increase the efficiency of the process, knowledgeabout these issues and their effects is essential. In addition, the dynamics of the processare diverse. The dynamics of some changes are very rapid (pressure, temperature) whilesome changes are slow (formation of build-ups). In this chapter an evaluation of theseissues is made.

3.1 Mineralogy

Normally, the feed to the roaster furnace is mainly a mixture of zinc concentrates. Somesecondary materials like oxides are sometimes blended with the feed. Zinc ash from thecasting department is also fed to the furnace. The main variable for the behaviour of theroasting furnace is the concentrate mixture. Since the feed composition is very signifi-cant for the behaviour of the furnace it is important to know the composition of the con-centrates, of which the basic factor is the mineralogy of the concentrates. Most Zn con-centrates are sulphides but there are also oxides.

The most common Zn minerals are the following:Sphalerite, zinc sulfide, ZnS, (Zn,Fe)S, cubic, trimorphous with martraite and wurtzite.Wurtzite, zinc sulfide, ZnS, (Zn,Fe)S, hexagonal and trigonal polytypes, trimorphouswith martraite sphalerite.Marmatite, (Zn,Fe)S ferroan sphalerite.Martraite, ZnS, trigonal, trimorphous with sphalerite and wurtzite.There are also many other zinc minerals (Fleisher & Mandarino 1991, Mandarino

1999).The chemical analysis and grain size distribution of the concentrate are important.

They influence the behaviour of the roasting and above all the behaviour of the roastingbed. The Zn content for concentrates is in the range of about 45–60%. The S content isabout 30%. The impurity level has a great impact on roasting and important elements

34

include Cu, Pb, Si, Ca and Fe (Taskinen et al. 2002, Metsärinta et al. 2003, Metsärinta etal. 2005b). Other components that have an impact on roasting are the amounts of FeS2and FeS (Taskinen et al. 2002). The most important elements with regard to gas cleaningare Hg, Se, Cl, F and C (Yazawa et al. 1980). For the rest of the plant there are manyimportant elements that must be handled in leaching and purification to avoid problems inelectrolysis and even in the end product. These elements are for instance Co, Ni, Cd, Ge,Sb and Tl.

Typical analysis ranges of zinc concentrates are as follows:Zn 45–60% Cd 0.05–0.35%Fe 1–12% Hg 0.001–0.19%S 28–33% Ge 0.001–0.09%Pb 0.2–4% Ca 0.05–1.1%Cu 0.2–4% Si 0.3–1.5%Because of the analysis and mineralogy, the heat energy and the oxygen requirement

differ for different concentrates, and the range varies (Taskinen et al. 2002). The valuesfor sulphidic zinc concentrates lie in the following range:

The particle size of the concentrates affects roasting significantly. The particle size dis-tribution varies for different concentrates as shown in Fig. 9. There is a tendency towardsfiner concentrates as the mines crush them more nowadays to increase recovery. The d50-values (= mass median diameter) for concentrates are in the range of 5–70 µm.

Fig. 9. The particle size distribution varies for different concentrates.

Figs 10 and 11 show microscope pictures of two concentrates. From these pictures itcan be seen that the Aznacollar concentrate (Fig. 10) contains more fine particles than theTrzebionka concentrate, which contains more coarse particles (Fig. 11).

Heat energy: 980–1430 kWh/tOxygen demand: 280–380 Nm3/t

Particle Size Distribution

0.01 0.1 1 10 100 Particle Size (µm)

0 0.5

1 1.5

2 2.5

3 3.5

4 4.5

5 5.5

6 6.5

Vol

ume

(%)

Antamina 04095, 09/23/04 14:45:43 Pyhäsalmi 04065, 08/04/04 10:14:14Tara 04080, 08/16/04 14:17:51 Red Dog 04092, 09/23/04 13:44:31

Red Dog

Tara

Antamina

Pyhäsalmi

35

Fig. 10. Microscope picture of Aznacollar concentrate. The pellets contain mostly smallparticles.

Fig. 11. Microscope picture of Trzebionka concentrate. The pellets also contain large particles.

3.2 Main reactions and thermodynamics

The product of roasting is called calcine and the reaction is exothermic. The reactionoccurs in the particles through the diffusion of O2 on the surface of the concentrate par-ticles (Ajersch 2000, Beveridge 1962, Dimitrov et al. 1981, Fukunaka et al. 1976, Hat-tori et al. 1980). During roasting ferrites are also formed, such as ZnFe2O4, CuFe2O4 etc.(Graydon & Kirk (1988a, 1988b, 1988c), Richards 1995). Some major reactions are pre-sented below:

36

The main reaction in roasting: ZnS + 1.5 O2 –> ZnO + SO2

Other reactions of importance: 2 FeS + 3.5 O2 –> Fe2O3 + 2 SO22 FeS2 + 5.5 O2 –> Fe2O3 + 4 SO2PbS + 2 O2 –> PbSO4Cu2S + 2 O2 –> 2 CuO + SO2CaCO3 + SO2 + 0.5 O2 –> CaSO4 + CO2

The ferrite formation reaction: ZnO + Fe2O3 –> ZnO.Fe2O3CuO + Fe2O3 –> CuO.Fe2O3

Also, agglomeration between particles takes place in the bed, which leads to largerparticles (Benlyamani & Ajersch 1986, Condina et al. 1980, Richards 1995). The growthmechanism is essential for the behaviour of the bed, but due to its complexity, under-standing of the mechanism is limited (Constantineau et al. 2002). Control of the bed parti-cle size distribution is of major importance for stable roasting conditions. Thermal analy-sis has also been carried out in roasting research (Sarveswara Rao & Ray 1999).

In Fig. 12, a schematic drawing shows how the reaction in a particle occurs. Manysmall particles agglomerate together to form a bigger particle and this is shown schemati-cally in Fig. 13.

Fig. 12. Schematic drawing of the reaction principle. The sulphide mineral particle reacts withoxygen. The product is calcine (mostly ZnO). During the reaction SO2 gas is formed.

Fig. 13. Schematic drawing of a calcine agglomerate formed of many small calcine particles.Bonds are formed between particles due to molten phases.

SO2

ZnO

ZnS ZnS

O2 O2

Formed calcine agglomerate

ZnO

ZnO

ZnO

ZnO

ZnO

ZnO

ZnOZnO

ZnO

ZnO

ZnOZnO

ZnO

ZnOZnO

ZnO

ZnO

ZnO

ZnO

37

Figs 14 and 15 show microscope pictures of agglomerates. The outer layer in themicroscope pictures is ZnO and in the middle there is non-reacted ZnS. Based on micro-scopic studies of concentrates, the origin of the pellets can be recognized. During thesample interval, the feed to the furnace was a mixture containing both Polaris and Azna-collar concentrate. The same structures can be seen in the calcine pellets.

Fig. 14. Agglomerate from the furnace (November 2001) – origin Aznacollar concentrate.

Fig. 15. Agglomerate from the furnace (November 2001) – origin Polaris concentrate.

When the main reaction takes place in the bed and above the furnace bed, calcine isformed. The calcine formed in the bed is coarser than the concentrate as is shown in Fig.16. For good fluidised bed furnace performance, it is significant to have a bed with theright particle size distribution. The bed should not be too fine nor too coarse. If the bed istoo fine, then the furnace becomes unstable due to defluidisation for example. The keyquestion is: How can the particle size distribution of the bed be controlled?

38

Fig. 16. Grain size distribution of concentrate versus calcines. The formed calcine is coarserthan the concentrate, especially the overflow calcine from the furnace.

3.2.1 Sulphates and molten phases

If the feed mixture contains many impurities such as Cu, Pb, Si, Ca and Mg, then the for-mation of sulphates and molten phases occurs more easily (Appendix 1, Freshcorn 1976).This can lead to the formation of build-ups on the roaster furnace grid and walls. In theworst case it leads to sintering of the bed and then the feed to the furnace has to be stop-ped. The furnace must then be emptied, which takes several days. The formation of build-ups and molten phases is a very complex combination of different elements. Thereforethere are limits in roasting for these elements. There are maximum limits for Cu, for Pb,for a combination of Cu and Pb, etc. It is not easy to determine precise limits and mostlythey are based on operating experience in zinc roasters. A commonly accepted “rule ofthumb” in roasting is that “Pb + Cu + SiO2 < 5 %” to avoid bed agglomeration and/or asticky bed (Grant 1994). When the impurity level in the feed is high, then the right roas-ting conditions are important. Variables like temperature and oxygen coefficient must becontrolled carefully. Phase diagrams must be studied to know the dangers and to find theright conditions. There are many important phase diagrams and Kellogg diagrams thatgive valuable information for understanding roasting, which should be evaluated (Elvers& Hawkins 1996, Mäkinen 1972, Appendix 1).

An essential reaction that occurs in the gas phase is SO2 + 0.5 O2 –> SO3. This reac-tion needs some catalysts, such as Cu. Due to this reaction, sulphates are formed in theboiler. The formation reaction of ZnSO4 could be ZnO + SO3 –> ZnSO4 (Gales &Winand 1973). Sulphates are also formed in the furnace due to leakage air on the walls.The formation of sulphates starts when the temperature is reduced to below about 850 oC.

0

20

40

60

80

100

120

0.1 1 10 100 1000 10000Grain size µm

Cum

ulat

ive

pass

vol

-

0

20

40

60

80

100

120

Cum

ulat

ive

pass

w-%

Concentrate 15.4.2002,pass vol-%Boiler calcine 15.4.2002,pass vol-%Cyclone calcine15.4.2002,pass til-%ESP calcine 15.4.2002,pass vol-%Overflow calcine15.4.2002, pass w-%

Overflow calcine

Concentrate

39

3.2.2 The sulphide analysis of the formed calcine

The sulphide level of the calcine is normally in the range of 0.2–0.3 %. Increasing thecapacity increases the sulphide level. Some plants keep the calcine sulphide level in therange of 0.3–0.5 %. Then the danger is that the oxygen coefficient is too low. The oxy-gen coefficient is defined in chapter 5. The sulphide level of the calcine produced is aquality indicator of the roasting process. The sulphide level is one of the main variablesin control aspects and is discussed more in chapter 5. Typical analyses of calcines formedfrom different process stages are shown in Appendix 4. The level of sulphates in diffe-rent process stages can be seen from the analyses. Analyses of samples taken during shut-downs are also shown in Appendix 4.

3.3 Heat transfer

The concentrate mixture is fed to the furnace from the side via slinger belts, and aircomes from the bottom of the furnace through nozzles. The reaction between the concent-rate and the air generates heat. The normal roasting temperature is between 920–960oC.There are cooling coils in the furnace that partly take away the heat that is generated.These elements take about one third of the heat away to the water/steam system. The restof the generated heat is in the outgoing gas with some in the calcine. The gas is cooled inthe boiler and in gas scrubbing. The calcine from the bed is also cooled. To increase thecapacity of the furnace, the area of the cooling coils can be extended.

As mentioned above, the roasting reaction occurs both in the bed and also above thebed. If the concentrate mixture is very fine then more burning takes place above the bedthan when the mixture is coarse. If the bed fluidisation is good then the heat transfer tothe cooling coils is effective.

The temperature gradient of the bed is a good indicator of roasting and bed behaviour.The temperatures are measured from the bottom and the side of the furnace with severalthermocouples. The more homogenous and stable the bed, the more homogenous the val-ues from the different temperature measurements, i.e. only small differences between thevalues of the readings from the different temperature sensors. The gas temperature in theupper part of the furnace, that is, the part between the furnace and boiler, also reveals a lotabout the situation in the furnace. If the top temperature rises, then water can be added tothe concentrate feed mixture before it is fed to the furnace to lower the temperature. Thewater can be fed both before the concentrate enters the day bins and after the day bins butbefore the furnace. The effects of the different locations are somewhat different due totime delays but also due to how the water is absorbed into the concentrate. To increase thecapacity, water injection to the bed can be used. Compared with cooling coils, this givesmore flexibility to control the feed rate. If oxide materials are in the concentrate mixture,then the feed increases because oxide materials have a cooling effect on the bed.

After the furnace some of the calcine, about 30–40 %, comes out via the overflow ofthe furnace. This calcine is cooled in Kokkola in a fluid bed cooler, but more often in zincroasters in a sectional drum cooler (Svens et al. 2003). After this the calcine is transported

40

by water-cooled redlers to the ball mill. Some zinc roasters remove part of the furnacecalcine by some kind of underflow system.

The gas enters the boiler at a temperature of about 850–950oC and is cooled to about320–380oC. The heat transfer that occurs in the boiler, cooling the gas and the calcine inthe gas flow, is both radiation and convection.

3.4 Fluidisation

The amount of air and air velocity are very significant for fluidisation behaviour (Kunii &Levenspiel 1969). The particle size distribution of the bed also affects fluidisation. If thebed has too many fine particles they tend to go with the gas flow to the boiler. A verycoarse bed also affects fluidisation and the heat transfer to the cooling coils. Impurities inthe feed have an effect on the bed and can make it sticky, in which case the fluidisationbehaviour is poor and the fluidity of the calcine is bad. If the fluidising of the bed is weakit can lead to channelling, i.e. a pressure drop in the bed because gas flows through thechannels. In that case the bed will also be badly mixed. The particle size, particle sizedistribution, the shape and density of the particles all influence fluidisation behaviour.The Geldart classification is presented in Fig. 17 (Geldart 1986). The density differenceof particles and gas is shown as a function of particle diameter. Fig. 18 presents a fluidi-sation diagram from the same source.

Fig. 17. The Geldart classification of particles into their fluidisation behaviour in air at ambientconditions (Geldart 1986). The density of calcine in the fluidised bed gives the working area inthe upper part of B.

41

Fig. 18. Fluidisation characteristic diagram, velocity versus particle diameter (Geldart 1986).The diagram shows the working area for a fluidised bed.

3.5 Gas cooling and dust handling

The gas is cooled in the heat recovery boiler from about 850–950oC to about 320–380oC.The calcine dust is removed from the gas in the boiler, cyclones and hot electrostatic fil-ters. The walls in the boiler are membrane cooled and there are also bundles, screens anda baffle wall inside the boiler. Effective spring hammers are installed to remove the calci-ne dust from the walls and bundles. These spring hammers shake the equipment resultingin the calcine falling down to the redler transporter (Peippo et al. 1998). The energy goesto the water and steam system of the membrane boiler. The steam generated goes to thepower plant or is used for heating purposes. Conveyors transport the produced calcinefurther. As mentioned previously, a fluid bed cooler cools the calcine from the furnaceoverflow. In most plants a sectional drum cooler cools this calcine overflow. The furnaceand boiler calcine is coarse and is therefore ground in a ball mill. All the calcine goes intoa bin and then the calcine is normally transported pneumatically to the calcine storagebin.

3.6 Gas cleaning

The SO2 gas contains impurities that must be taken care of in order to make pure sulphu-ric acid at the sulphuric acid plant. In gas cleaning the most important impurity to remo-

42

ve is mercury. For this purpose there are different processes of which the Boliden-Nor-zink method is one of the most common (Mukherjee 1999). Another process is the Outo-kumpu process, which is used in Kokkola. Final mercury cleaning methods are the Dowatower and the Se filter (Abe et al. 1980, Mukherjee 1999). Other important elements ingas cleaning are Cl, F and Se. For Hg removal the important elements are Hg, Se and Cl.These elements have an impact on Hg removal efficiency. The halogens, Cl and F, mustalso be taken care of because they are corrosive elements.

In Kokkola the chain is very long and the gas pipe from the Hg removal towers to theacid plant is 800 m. The process in Kokkola can use concentrates with a very high Hglevel e.g up to 800 ppm in the concentrate mixture has been used. The normal mercurylevel in the concentrate mixture is about 100–400 ppm. The quality of sulphuric acid isalso important and the Hg level should be below 0.1 mg Hg/kg H2SO4. The trend is tomake sulphuric acid with a mercury level of 0.01 to 0.05 mg Hg/kg H2SO4.

3.7 The dynamics of the process

The dynamics of the roasting process are very complex. A slight underpressure is kept inthe furnace. The dynamics of the underpressure are very fast (seconds), i.e. changes in theair feed and valves connected to the fans change this pressure very quickly. Controllingthe underpressure is therefore a challenge for the operators. Start-up and shutdown situa-tions are especially demanding. This was the main reason for building a simulator (Toska-la 2000). The pressure dynamics were studied thoroughly before the simulator for theoperators was built. This simulator enables the operators to practise start-up and shut-down situations (Toskala 2000, Toskala et al. 2001).

Changes in feed rate, air amount and feed water amount influence the gas temperaturevery quickly (seconds) whereas changes in the bed temperature are slower (minutes). Thedynamics of the furnace were studied with the purpose of identifying the process. Basedon this identification a model was built (Härkönen 2003, Saxén et al. 2004). Furtherdevelopment based on the model has been made with the target of optimising the furnacecontrol (Valo 2004).

The dynamics that are much slower are more difficult changes to observe and control.Build-ups in the furnace or instability in the bed due to the fact that the particle size distri-bution in the bed becomes fine are phenomena where the time delay is very long, i.e. sev-eral days. The major challenge is to obtain information about phenomena that occur in thefurnace slowly. Formation of build-ups in the boiler is also a slow process.

In mercury removal the time delay is also very long, several days, before anything canbe seen in the end product. Correcting the process also takes several days.

The dynamics of the process are discussed further in chapter 6. Some results are pre-sented from the step tests, from which it can be seen how quickly the different variableshave an effect e.g. bed and gas temperatures.

4 Targets for control system development

The overall target of the research and development done has been to improve the econo-mic result by running the roaster efficiently, including quality, environment and safetyissues (Nyberg 2001, Fugleberg 1999). The main target is to increase flexibility in the useof concentrates, i.e. increase the treatment charge income, by developing the control stra-tegy. The sub-targets include running time and feed rate. The sub-targets promote themain targets, such as the economic result and quality.

One of the main goals for every zinc plant is to achieve a good economic result. Themajor income of zinc plants is the treatment charge for treating zinc concentrates. Whenselling the zinc ingots they have produced, the plants are paid for the zinc plus they get acertain premium, which is additional income. The price of zinc is a relevant factor, bothwhen concentrates are bought and when zinc is sold. The concentrates are a major costfactor and the purer the concentrates are, i.e. the higher the Zn content, the more expen-sive they are. Besides concentrates the other main costs of the zinc plant are electricitycosts, personnel costs, maintenance costs and costs of the chemicals used. The treatmentcharge (TC) income should be optimised. To achieve this the performance of the roaster iscritical. During the last decade, the TC-value has been in the range of $140–220 /t con-centrate. The capability of using different kinds of concentrates along with the smoothrunning of the process is the key to success. Production of the same amount of zinc usinglow-grade zinc concentrates, i.e. compensating for the low grade with a high feed rate,increases the income. Knowledge and control options are required to increase the feedrate and to be able to use low-grade concentrates.

The general price functions for zinc concentrates are: 1. When the Zn level of the concentrate = Zn % > 53.3% then

Price of the concentrate = Zn price * 0.85 * Zn level (%) – TC – fines2. When the Zn level of the concentrate = Zn % < 53.3% then

Price of the concentrate = (Zn level (%) – 8%) * Zn price/100 – TC – finesIn order to reach the economic target the operation of the roaster should be smooth. A

high concentrate feed rate and high uptime are the main sub-targets. There are severalways of increasing the feed rate to the furnace, such as increasing the temperature, add-ing more cooling coils to the furnace, water injection to the bed and using oxygen. Thechallenge is to know what the optimal conditions are for each concentrate mixture. It is

44

essential to have a good, stable bed in the furnace. The fluidisation should also be good,because it has an impact on the heat transfer. It is therefore important to gain knowledgeabout how the quality of the bed can be controlled. When coarse concentrates and con-centrates with a low amount of impurities are used, this is easier to achieve than when theconcentrates are impure and fine. The main impurities that must be taken into account areCu and Pb. The feed mixture is the most important control variable and planning the feedmixture is the key to success, both in short- and long-term planning. The kinds of concen-trates used influence the roasting performance. The advantage of the zinc plant inKokkola is that concentrates more suitable for roasting are used in the roaster and thosemore suitable for direct leaching are used in leaching. When concentrates are chosen,there are many other viewpoints that must be considered besides the main goal, i.e. agood economic result. In fact, the whole plant must be considered when evaluating theimpacts of the different elements in the concentrates.

The target for the control of the furnace is to optimise both the capacity and the condi-tions in the furnace. To reach a high degree of uptime, it is essential to maintain optimalconditions in the furnace. In zinc roasters the specific capacities of the roaster furnacesare about 6–8.2 t/m2/d (Svens et al. 2003). In the Kokkola zinc roaster the yearly averageis about 7.4 t/m2/d (dry ton). The optimum feed rate depends on the concentrate mixture,the equipment capacity (furnace, boiler and ESPs) and the possibilities of setting the opti-mal conditions, i.e. having the necessary control tools. Too high a feed rate can lead tosinters, bad fluidisation and problems in the boiler, which will influence the uptime. Thetarget is to optimise the capacity, i.e. the yearly concentrate feed amount, by optimisingthe feed rate and running time. Yearly concentrate feed versus feed rate and running timeis shown in Table 1.

Table 1. Yearly concentrate feed as a factor of feed rate and running time.

In order to achieve a high running time there should be few shutdowns and the yearlyshutdown should be as short as possible. The interval between yearly shutdowns shouldalso be as long as possible, however without neglecting the required maintenance of

Running time%

Yearly concentrate feed as a factor of feed rate and running timeMt / a

95.094.594.093.593.092.592.091.591.090.590.0

0.1660.1660.1650.1640.1630.1620.1610.1600.1590.1590.158

0.1710.1700.1690.1680.1670.1660.1650.1640.1630.1630.162

0.1750.1740.1730.1720.1710.1700.1690.1680.1670.1660.166

0.1790.1780.1770.1760.1750.1740.1730.1720.1710.1700.170

0.1830.1820.1810.1800.1790.1780.1770.1760.1750.1740.173

0.1870.1860.1850.1840.1830.1820.1810.1800.1790.1780.177

0.1910.1900.1890.1880.1870.1860.1850.1840.1830.1820.181

0.1960.1950.1940.1920.1910.1900.1890.1880.1870.1860.185

0.2000.1990.1980.1970.1960.1940.1930.1920.1910.1900.189

20.0 20.5 21.0 21.5 22.0 22.5 23.0 23.5 24.0

Concentrate feed ratet/h

45

equipment at proper intervals. The stress and the challenge come from keeping the fur-nace and boiler running without cleaning for long periods. Besides the main equipment,i.e. furnace and boiler, there is also other equipment that must be taken care of, such asthe big fans and ESPs. Back-up equipment (fans, ESPs etc.) is one means of achieving ahigh running time.

Efficient production and minimal process disturbances are the result of process knowl-edge, good control of the process, good equipment, good maintenance and skilled person-nel.

The total concentrate feed is the result of the feed rate versus the running time asshown in Table 1 above. The table shows that increasing the running time from 93% to94% increases the yearly feed rate by about 0.02 Mt/a, at a feed rate of 21.5 t/h. Increas-ing the feed rate from 21.5 t/h to 22.5 t/h increases the yearly feed by 0.08 Mt/a, when therunning time is 93%.

The quality of the steam, calcine and sulphuric acid produced should be acceptable. Inparticular, the quality of sulphuric acid has lately become more important and the trend isthat the mercury level should be low. To be able to produce sulphuric acid with low mer-cury and few disturbances has placed extra stress on process development. The need for agood and manageable process is also important when choosing concentrates. If the pro-cess can handle concentrates with high mercury without problems then flexibility isincreased.

Nowadays, environmental issues are becoming increasingly important. Therefore run-ning the roasting lines with as few disturbances as possible is crucial. Again, processknowledge, maintenance and skilled personnel are essential in reaching this goal. Safetyissues have also become a topic of great importance. Therefore safety issues must be con-sidered all the time.

In roasting, the target has been to obtain more information about the roasting mecha-nism, the impacts of impurities, the dynamics of the process and the influences of differ-ent control variables. More knowledge is needed in order to be able to treat fine andimpure concentrates in the roaster fluid bed furnace. One important target has been toadopt new measurements and analyses that give more information about the state of theroasting process. Adding the means to control the process has also been a significant tar-get. Another goal has been to increase the skill and performance of the operating person-nel by training and creating new instructions using the knowledge gained.

In boiler development, the target has been to modify the boiler based on research anddevelopment work and operational experiences in order to increase efficiency and uptime.Reaching this target improves the overall performance and gives more freedom for roast-ing furnace control.

In mercury removal the target has been to gain more knowledge about the process bydoing research and development work. This has been the platform when creating themeans to control the process.

5 Measurements and control of the fluidised bed furnace

In this chapter measurements and control of the fluidised bed furnace are discussed. Itwas part of this work to establish what kinds of measurements and analyses give valuableinformation for the control system development. Based on more knowledge, moremeasurements and more analyses, the control strategy, loops and means of control couldbe developed. A computer system for monitoring and controlling the process is anecessity. The need to gather measurements and analyses in a system that contains bothshort-term and long-term trends is necessary.

The major goal of the control strategy development is to improve the economic resultand the treatment charge (TC) is the key factor that should be optimised. The target is tobe flexible in the use of concentrates, which is necessary to reach the economic goal.

As discussed in chapter 3, the quality differs between various concentrates which aregenerally known to behave differently in the furnace. The major factors are the impuritylevels and the particle size distribution of the concentrates. The concentrate mixture is themost significant control factor. When no other control tool works, then this control toolmust be used, i.e. the concentrate mixture must be changed.

For example, a broad range of different concentrates is used in the Kokkola roaster.Calculation of the concentrate mixture is made both in short- and long-term planning.Long-term planning means that the concentrate mixture for a period of about one year iscalculated. The key in long-term planning is to optimise the economic result, but also tomake a plan that is realistic, i.e. to keep the impurity levels and particle size distributionwithin a safe range. A short-time plan made on a weekly basis enables changes to bemade when needed, according to the response of the process and the optimisation of theuse of the concentrates that are actually in the concentrate storage.

Control of the roaster furnace is the essential factor allowing flexibility in the use ofconcentrates. There are several limits to the operation of the furnace, as the process isquite varied and many aspects have to be considered. Having the tools to change roasterfurnace behaviour, i.e. control tools, is essential in order to achieve flexibility.

Developing the boiler and mercury removal process has increased the flexibility inrunning the roaster furnace and choosing concentrates. The flexibility in choosingconcentrates is higher when constraints due to these two stages are low. For the boiler thismeans long running periods with few shutdowns due to build-ups or leakages in the

47

boiler. An effective boiler also allows more flexible running of the furnace, i.e. theoperation parameters in the furnace can be chosen so that they are optimal for the furnace,but not necessarily for the boiler. By making the boiler effective it can cope with a highdust load and high sulphate load. The target is to make sulphuric acid with low mercurycontent, but having a process that allows the use of a concentrate mixture with a highmercury level is also essential for optimising the concentrate mix, i.e. optimising the TCincome.

5.1 Measurements and analyses of the fluidised bed furnace

Many measurements and analyses are required to know the state of the roasting processas shown in Fig. 19. Calculated variables give additional information. In this evaluationmainly those measurements and analyses are discussed that are of importance forroasting. Basic measurements give a good overview of the state of the process. The basicmeasurements are:

– Temperature (many measurements in the furnace bed, in the gas line and in calcinecooling).

– Pressure (furnace wind-box, furnace, boiler and gas line).– Flows (air, water, steam, oxygen).

Fig. 19. Basic and advanced measurements of the furnace.

Besides measurements many chemical analyses are required. Chemical analyses of theconcentrates form the basis for making the feed mixture calculation. There are maximumlimits for many impurities. The limits set are based on zinc plant operation experienceand also pilot studies (Grant 1994, Constantineau et al. 2002).

Pressure measurements indicate the state of the process. The wind-box pressure isessential for evaluating furnace behaviour. Many things affect the actual value of the

Basic measurements Advanced measurementsand analyses and analyses

Temperatures More temperatures from the bed (bottom and side) (bottom and side at different

heights) Furnace

Top temperature On-line oxygen after the boiler

Windbox pressure On-line oxygen from the bed

Furnace pressure Furnace calcine analysis

Air flow Furnace calcine particle size distribution

Waterflow Calculated oxygen coefficient

Total calcine analysis Calculated overflow calcine amount

48

wind-box pressure. The actual amount of the bed, the particle size distribution(coarseness), fluidisation and sinters formed in the furnace influence the wind-boxpressure. Thus a change in wind-box pressure can be a result of many factors, and manytimes experience and the ability to analyse trends are required in order to draw the rightconclusion. Draft pressure is another pressure measurement that must be monitored.Pressure measurements in the gas line (boiler, cyclones, ESPs and mercury removal) areindicators of how clean the equipment is and of when manual cleaning of the equipmentis required. The boiler pressure, along with the level control of the boiler drum, isnecessary for the operation of the water steam system.

Different flow measurements are also among the basic measurements that are needed.The amount of concentrate fed to the furnace is normally based on X-ray measurement.Measurements of the airflow to the furnace and water flows to various locations are alsorequired.

Furnace temperatures are measured both from the bottom and the side of the furnace.These measurements are very important when evaluating the state of the furnace. Themore temperature measurements installed in the furnace, the more reliable the picture ofthe state of the furnace gained. At Kokkola more temperature measurements have beeninstalled at the bottom of the furnace and also on different levels of the sidewall to getmore information. Temperature measurements in many places are essential becauseduring long running periods malfunctions due to severe conditions and/or formation ofsinters always cause some of the temperature measurements to stop working.

The calculated overflow calcine amount is in use at Kokkola to give more informationabout the calcine flow. The calculated calcine overflow amount is based on a mass andheat balance calculation of the fluid bed calcine cooler. The calculation is based ontemperature and flow measurements of the calcine, water and air system. Changes infurnace behaviour can often be seen in a changed overflow amount of the calcine flow.

The chemical analysis of the total calcine, for instance the sulphide sulphur (S2-)analysis, gives a picture of the roasting conditions. The sulphide level of the calcine is amajor indicator both of the state of roasting and the quality of the product. This was alsodiscussed in chapter 3. It is important to monitor other analyses like Cu and Pb to avoidthe problems which will arise if their levels are too high. More analyses give moreinformation and at the Kokkola roaster the analysis of the furnace overflow calcine hasproved crucial when evaluating the state of the furnace. Besides chemical analysis (Cu,Pb, S2-, Si, Ca, SO4) of the overflow calcine, the analysis of the particle size distributionalso gives a lot of information. The particle size of the calcine is classified into fourdifferent areas and in particular the amount of fine particles corresponds with furnacebehaviour; thus, for example, there are normally problems when the proportion of fineparticles in the bed is high. The importance of monitoring the particle size distribution isdiscussed further in chapter 6.

Besides temperature, the second most important factor for roasting conditions is theoxygen coefficient. The determination and calculation method of the oxygen coefficientare presented later in this chapter (in the control section). Oxygen pressure is the generalvariable, in addition to temperature, when phase diagrams are evaluated and so theoxygen pressure value is essential.

The on-line oxygen measurement in the off gas line also indicates the oxygencoefficient. It is normally placed at the end of the waste heat boiler. The value of the

49

measurement changes according to changes in the oxygen coefficient, and the correlationis quite good (See chapter 6). However, this requires that the draft control of the furnaceis stable; otherwise the result is disturbed by leakage air.

On-line measurement of the oxygen from the roaster furnace bed has beenimplemented at Kokkola during plant trials. Oxygen has been measured both from thefeed side and the overflow side of the furnace (Parkkinen et al. 2002). Duringmeasurement campaigns done during test periods, changes have been detected in theoxygen content in the bed when making changes in the process variables. The oxygencontent has been lower on the feed side of the furnace compared to the overflow side.This has also been mentioned in a report (Richards 1995). Chapter 6 includes examples ofthese measurements.

5.2 Control specification of the fluidised bed furnace

The control of a fluidised bed furnace and the equipment in connection with the furnaceis quite varied and many aspects have to be considered. One is the necessity for a goodand reliable computer system in order to reach a high level in control aspects. Large fansand many valves must be controlled. It requires a lot of attention and skill by theoperators to control them, especially during start-ups and shutdowns. This was one reasonwhy a simulator was built (Toskala 2000, Toskala et al. 2001) – to enable the operators totrain and learn to understand the dynamics of the process. Pressures and temperaturesmust be within the correct operating range. There is also a lot of external equipment tocontrol. There are conveyors in the roaster both for feeding the concentrate mixture to thefurnace and for transporting the formed calcine to the calcine storage bin. The boilersystem also requires a lot of attention. The high pressure and temperature of the boilermake the dynamics very rapid and there are also many safety aspects that the operatorsmust be aware of. A failure in boiler control may have fatal consequences. During normaloperating conditions the automation system keeps the system quite stable.

The control of the roaster furnace is a multivariable system. Changing one controlvariable has an impact on many variables, as shown in Fig. 20. Both dynamic andthermodynamic aspects must be considered in the development of the furnace controlstrategy.

50

Fig. 20. The control of the furnace is a multivariable system. Many aspects must be considered.Knowledge of the dynamics and the thermodynamics of the process is essential.

The control hierarchy of the furnace control can be classified in three categories and isdescribed as follows (Saxén & Nyberg 2003). Fig. 21 shows the main control inputs ofthe different control levels for the furnace. Conventional Control: When the situation in the furnace is stable and the concentrate (Standard) mixture is easy to handle, then standard control is enough. Changes

in set points, like temperature and air flow, keep the temperaturegradients stable in the furnace, the quality of the product good andalso the production amount at the proper level.

Advanced Control: To increase the control level of the furnace control in order to in-crease flexibility a more advanced control level is needed. This re-quires more measurements and also more control tools, such as thepossibility to add water to different locations and to use oxygen en-richment to keep the furnace in balance. The challenge is to knowthe state of the furnace. Knowing what should be done, i.e. the op-timum process conditions, and having the tools to do it is part ofthis control level. By developing advanced control the chances ofbeing flexible with concentrates increase. The target is to optimisethe production and the process conditions.

The control of the furnace is a multivariable system

Concentrate feed amount Bed temperatures

Concentrate feed composition Top temperature

Air feed Oxygen coefficient

Oxygen feed Calcine composition (S2- , SO4 , grain size distribution etc.)

Water feed before the furnace Behaviour of the bed to the concentrate mixture (wind-box pressure, fluidisation,

temperature gradients etc.)

Water injection into the bed Calcine overflow amount

51

Ultimate Control: When neither standard nor advanced control helps then the onlysolution is often to change the mixture of the concentrate feed, oth-erwise the furnace may remain unstable and lead to formation ofsinters in the furnace. This ultimate control is far more expensive,because purer concentrates must be used and they are more expen-sive. Sometimes such concentrates might not be available at shortnotice.

Fig. 21. Control inputs of the different control levels for the furnace control.

The feed mixture is the most important control input to keep the roasting conditionsstable. The impurity levels and particle size distribution of the concentrate mixtureinfluence the behaviour of the furnace. It is vital that the mixture is homogenous;therefore the blending should be done well. Usually the blend is a calculated mixture ofseveral concentrates and the day bins before the furnace are filled in sequences that makethe content of the bin as homogenous as possible. One possible way to make the feed veryhomogenous is to make the concentrate mixture in the concentrate storage (Ek 1999).Pretreatment of the feed before it is fed to the furnace is essential when the concentrate isfine and contains impurities. One way to handle the problem and to stabilise the furnace isto pelletise part of the feed (Brown & Goosen 1996). Some concentrates are oxides andother materials fed to the furnace are oxides too, for example, the dust from the zincfoundry. Oxide materials increase capacity because they have a cooling effect on the bed.

Additions of water have a significant impact on roasting (Grant (1993,1994), Lightfoot1977). Water addition to the concentrate mixture makes the particles larger and stabilisesthe furnace. Water injection to the bed increases the concentrate feed. In addition,spraying water into the furnace above the bed increases the feed (Longton 1998,Takayama et al. 2000).

The temperature of the furnace is one of the main control variables. When thetemperature is high then the feed increases, but the risk of molten phases also increases

Basic control Advanced control Temperature set point Water feed to different spots

Furnace Feed amount control Oxygen feed Air amount Pretreatment of concentrate mixture Fixed cooling capacity Oxygen coefficient control Bed height at overflow Underflow system Draft control of the furnace Grid area with bigger nozzles (fixed) Calculation of feed mixture Special overflow grid

Ultimate control Changing the feed mixture to a safe mixture

52

(Nyberg et al. 2000). A normal temperature range for the bed in zinc roasters is between920oC and 960oC. One possibility to control the temperature in the furnace is to controlthe concentrate feeding system, either manually or automatically, in order to keep thetemperature set point at the desired value (Rauma et al. 2000). Another possibility is toset a fixed value for the feed and to control the temperature in the bed by automaticallycontrolling the water injection to the bed.

The other main control variable besides the bed temperature is the amount of air fed tothe furnace. At Kokkola the range is normally between 41,000 and 43,000 Nm3/h perfurnace for the furnaces with a grate area of 72 m2. If the concentrate feed controls thetemperature, then the feed increases if the air amount is increased due to the coolingeffect of the air (Nyberg et al. 2000). This can be seen especially during winter when theair is cold. Changing the amount of air also has other effects; it affects fluidising andchanges the oxygen coefficient.

Calculation of the oxygen coefficient and its active control has been in use at theKokkola roaster since 2000. Definition of the oxygen coefficient:

“The oxygen coefficient determines the availability of oxygen for complete roasting ofthe concentrate, i.e. the ratio of total oxygen in the process gas to the oxygen requirementof the feed mixture for the formation of stable oxides and sulphates in the roaster off-gas”(Anjala et al. 1987, Asteljoki et al. 1989, Asteljoki & Muller 1987, Kytö et al. 1989, Rod-olff et al. 1986).

At Kokkola the oxygen coefficient is calculated on-line by an automation systembased on data fed to the system and on measurements (Taskinen et al. 2000b). The use ofthe calculated oxygen coefficient has proved invaluable; especially when the situation inthe furnace has been critical and also during experiments when new process conditionsand different parameters (impurities) have been tested. Fig. 22 shows how the oxygencoefficient is calculated. Mineralogical studies provide the basis for the calculations.

Fig. 22. The calculation scheme for the oxygen coefficient and its use for control. This has beenadopted and is now in use in the control of the furnaces at the Kokkola roaster.

When phase diagrams and Kellogg diagrams are studied there are two significantvariables, namely temperature and oxygen pressure or potential (Mäkinen & Jåfs 1982,

Mineralogical Oxygen requirement Calculated oxygen studies of for different requirement of the concentrates concentrates concentrate mixture.

Analyses of concentrates Measured concen- Calculated oxygen

trate feed to the coefficient in furnace roaster furnace

Measured air feed to the furnace

Calculated oxygen Control decision feed to the furnace

Measured oxygen feed to the furnace

53

Nermes & Talonen 1982). Especially when concentrates containing impurities like Cu,Pb, Si, Na and K are used, the importance of keeping both the temperature and the oxygencoefficient within the right range is highly significant. The oxygen coefficient can becontrolled in many ways, with the key being the feed versus the total oxygen feed.Variables that changes the oxygen coefficient are the set point temperature of the bed, theamount of air used, water mixed into the concentrate mixture feed, water injected to thebed and oxygen added to the air feed. The influence of air for instance, i.e. increasing ordecreasing, is dependent on the control strategy. If, for instance, the feed is fixed, thenincreasing the amount of air increases the oxygen coefficient, but if the temperature iskept fixed by controlling the feed, then increasing the amount decreases the oxygencoefficient. This is due to the fact that an increase of air cools the bed and increases thefeed proportionally more than the amount of air. The influence of the oxygen coefficientis presented in chapter 6. It is essential to know what the actual oxygen coefficient is andto control it. This calculation and control of the oxygen coefficient is one of the mostsignificant parts of this work (Chapter 6, Taskinen et al. 2000b).

Feeding oxygen can be used in roasting for many purposes. The use of oxygenincreases the capacity without changing the fluidisation as much as the use of additionalair does (Filho et al. 1998, Longton 1998, MacLagan et al. 2000, Saha et al. 1989). Theuse of oxygen also increases the oxygen coefficient and thus it changes the operationparameters of the process. In some cases the possibility of using oxygen can be anadvantage for the control of the roaster furnace. The disadvantage is in many cases theprice of oxygen. Increased cooling capacity (increasing the capacity of the cooling coilsor water injection) is needed when oxygen enrichment is used to increase the capacity,otherwise the temperature will increase in the furnace.

On the overflow side of the furnace, the height of the bed is controlled by bars.Depending on the wind-box pressure of the furnace, the height of the bed is adjusted byadding or removing bars. If the bed is high, then the cooling coils in the furnace are usedmore effectively and thus the capacity of the furnace increases.

The cooling coils in the furnace above the grate area cool the bed. By increasing thecooling area, the capacity of the furnace increases. This capacity is fixed and cannot becontrolled during operation, thus a suitable cooling capacity must be decided when thefurnace is built and can only be changed during long maintenance shutdowns. Waterinjection to the bed is a flexible way of controlling the cooling capacity of the furnace.

The draft control must be accurate, and a slight underpressure must be sustained in thefurnace. Overpressure in the furnace leads to SO2 emissions, dust emissions andinsufficient working conditions. The underpressure should be as close to zero as possible;otherwise leakage air increases, which can lead to build-ups on the furnace wall above theslinger belt inlets. These build-ups are sulphates formed due to the low temperature in thispart of the furnace. The SO2 blower and the valves connected to it control the draft.

Some plants use the underflow outlet from the grate bottom at the side of the furnace.This can be an outlet for the very coarse material on the grate surface that does notfluidise properly. There are different types of underflow mechanisms and the advancedversion that is in use in some roasters is based on an automatically controlled lancesystem (Svens et al. 2003). A modified version of this is in use at Kokkola. This systemincludes a cooling system with a system that measures the amount of calcine removed andthe ability to control the amount (Siirilä et al. 2003).

54

The grate area is full of nozzles on the surface. The distance between the nozzles isabout 10 cm and the diameter of the nozzle hole is about 6 mm. To increase the amount ofoxygen at the side where the slinger belts are located, bigger nozzles can be used in thisspecific area (Taskinen et al. 2000a). The advantage of this is that because the concentratecomes into the furnace on this side and reacts here, the oxygen demand is high in this partof the furnace. If the nozzles are of normal type, it leads to a low oxygen coefficient andformation of build-ups on the furnace bottom.

In the overflow area there are nozzles on the surface and the pressure is the wind-boxpressure of the furnace. To be able to control the overflow, the nozzles of this specificarea can be connected to a pressurised system (Siirilä 2000). This makes it possible toincrease the pressure and to increase the amount of calcine removed from the furnace viathe overflow.

An advanced furnace control diagram is shown in Fig. 23. It includes water feed tovarious points, oxygen feed, a fuzzy control loop for the temperature control of the bedtemperature and a top temperature control loop for the boiler inlet temperature. Thediagram includes also advanced measurements (oxygen at the end of the boiler), analyses(grain size distribution analysis of the furnace calcine) and calculated variables such asthe oxygen coefficient and calcine overflow amount.

Fig. 23. A furnace control diagram is shown with advanced control and measurements. Itincludes among others a fuzzy control loop, the calculated oxygen coefficient, oxygen measuringfrom the gasline, calculated calcine overflow amount and grain size distribution analysis of thefurnace calcine. The possibility of feeding water to different spots and oxygen is also available.

T

2H O

AIR 2OO2

P

L T

DAYBIN

H2O

DAYBIN

m

FURNACE

P

2-S

LEACHING

BOILER

STEAM (BOILER+

FEED WATER

T O2

FURNACE)

O2

P

USER SPECIFIEDSETPOINT

FUZZY CONTROL

VACUUM 2H O

H O 2

TP

S 2-

AIR

H2OCHEMICAL ANALYSIS

GRAIN SIZE DISTRIBUTION

H2O

P

P

CALCINE TO

CONCENTRATE MIX

Calculatedoxygen coefficient Calculated calcine

overflow amount

55

5.3 Aspects relating to the boiler and mercury removal

In the heat recovery boiler, the water steam system requires reliable measurements,because otherwise the control of this high-pressure system can lead to severemalfunctions. The pressure control and the level control of the drum are the key controlitems. The feed water, circulating water and steam produced are the other main controlvariables. The goal regarding boiler performance is to have a high degree of uptimewithout shutdowns and start-ups due to manual cleaning of the boiler. To achieve this,rapping or hammering devices must be installed. At Kokkola there are more than 100spring hammers per boiler, which effectively hammer the bundles and the walls. Thehammering devices are controlled sequentially by an automation system.

There are also several measurements in the boiler gas line of temperature and pressure,which indicate how clean the boiler is and if the boiler is working properly.

The control of mercury removal requires measurements (temperature, pressure andflows), analyses of both liquids and slurries (Hg, Se, Cl, and H2SO4) and also controloptions (addition of chemicals and cooling). Often the difficulty is that the quantities arevery low, especially in the gas phase, and controlling them is not so easy. The target, i.e.the quality of the sulphuric acid, is nowadays very high, which means a very low mercurylevel.

6 Results of the control strategy development for the Kokkola zinc roasting process

In the following the control strategy development for the Kokkola zinc roaster isdescribed. The main parts of the development are the furnace, the boiler and mercuryremoval.

6.1 Furnace development

The target of the furnace control system has been to increase the capability of usingdifferent kinds of concentrates, because flexibility with concentrates is essential andeconomically very significant. Low-grade Zn concentrates are less expensive, but treatingthem is more difficult due to the fact that they contain more impurities. To reach thistarget, the optimal conditions for a certain concentrate mixture have to be determined.Control means and knowledge of the influence of different variables are other factors ofimportance. Knowledge of the state of the furnace has to increase by adding moremeasurements and also by taking additional samples for analysis (chemical and particlesize distribution analysis).

The target has also been to run the furnace without the formation of too many build-ups and to prevent the bed from becoming fine by controlling the particle size growth ofthe calcine bed. The research results have led to longer running periods. Formerly theroaster furnaces were emptied once a year, but now the running period of the furnace hasincreased. The target, which now seems attainable, is to run the furnace for two yearswithout cleaning.

Results from furnace research and development work are presented in the followingtext as well as the influence of the various major factors.

57

6.1.1 Factors correlating to bed instability

Factors correlating to instability have been derived from plant experiments and activemonitoring of the roasting process. By adding more measurement devices to the furnaceand by taking samples for analysis, information of great importance for evaluating thesituation in the furnace has been acquired. Placing numerous thermocouples in thefurnace at different heights has increased knowledge of the state of the furnace. Chemicalanalysis of the furnace calcine and above all the grain size distribution of the furnacecalcine have proved to be very informative.

When the fraction of fine calcine increases in the bed, it normally correlates with othervariables and with bed instability. The situation during bed instability is often as follows:

– The percentage of fine grain size of the overflow calcine grows and the fraction < 71µm in the overflow is over 10 %.

– Bed temperature drops in the furnace due to bad fluidisation. During such situations atemperature drop can be seen in some parts of the furnace.

– The windbox pressure drops (bad fluidisation).– The sulphide level analysed from the furnace calcine is very low.– The overflow calcine amount (calculated variable) decreases.– The fluidised bed becomes sticky.Figs 24 and 25 show a situation when the bed was unstable for several days. Fig. 24

shows that the windbox pressure dropped when the amount of fine calcine in the bed, i.e.the fraction less than 71 µm, increased. At the same time, as shown in Fig. 25, the valuesof some bed temperatures dropped.

Fig. 24. The drop in windbox pressure and the increased proportion of fine calcine indicate anunstable bed.

Windbox pressure and fine material in Furnace no 1

0102030405060

1-Sep-2002

6-Sep-2002

11-Sep-2002

16-Sep-2002

21-Sep-2002

26-Sep-2002

Days

Proc

enta

ge fi

ne

mat

eria

l

220

230

240

250

260

270

Pres

sure

m

bar

Fine material Windbox pressure

58

Fig. 25. Temperature drop indicates an unstable bed. Temperatures measured from thesidewall fell while the temperature measured from the bottom remained stable.

6.1.2 Oxygen coefficient control

The importance of the oxygen coefficient has become apparent during this research.Therefore it is important to have the means to control it, to know how it affects roastingand to know the value of the oxygen coefficient. As described in chapter 5, the oxygencoefficient is calculated based on the mineralogical analysis of different concentrates. Byanalysing the concentrates, the oxygen requirement is determined for each concentrateand then the computer system calculates the oxygen coefficient using the available data(concentrate feed, amount of air and oxygen). Thus, the operator knows the oxygencoefficient in the furnace at all times.

There are different ways to either increase or decrease the oxygen coefficient.Different ways of changing the oxygen coefficient include: changing the temperature setpoint, changing the amount of air, using oxygen, using water before the furnace, feedingwater to the furnace, using oxides etc. The plant experiments performed have shown theeffects of different variables. Table 2 shows the alternative methods of changing theoxygen coefficient and their effects. The operators use this information when they maketheir control decisions.

(*1 = The control in this case, as is shown in Table 2, is such that the furnacetemperature set point is fixed and controlled and kept stable by the concentrate feed. *2 =The top temperature is the point between the furnace and the boiler and is controlled bythe amount of water fed to the concentrate feed before the furnace. A top temperaturecontrol loop has been added to the control system. The operator gives the set point andthen the water feed to the concentrate stream before the table feeder controls the toptemperature automatically).

Windbox pressure and temperatures in Furnace no 1

700

750

800

850

900

950

1-Sep-2002

6-Sep-2002

11-Sep-2002

16-Sep-2002

21-Sep-2002

26-Sep-2002

Days

Tem

pera

ture

C

230235240245250255260265270

Pres

sure

m

bar

Temp. no 6. Temp. no 7 Temp. no 2 Windbox pressure

59

Table 2. The effects of different variables are shown in the table. The operators have seve-ral ways to control the oxygen coefficient. Fuzzy temperature and top temperature controlloops are in use in the cases shown in the table.

Some results from plant trials are presented in Figs 26–28, i.e. how different variablesare affected by the oxygen coefficient.

Fig. 26. Sulphide level of the calcine (WHB, Cyclone and ESP) decreases when the oxygencoefficient is high.

The effects of different variables on the oxygen coefficient (*1)Operation Results

1 Oxygen feed to the furnace (=ogygen enrichment of the process air to the furnace).

Increases the oxygen coefficient.

2 Lower bed temperature. Increases the oxygen coefficient due to decreased feed.

3 Water injected into the bed. Decreases the oxygen coefficient due to increased feed.

4 Water feed before the furnace to the concentrate mixture. Increases the oxygen coefficient due to decreased feed.

5 Increasing the airflow. Decreases the oxygen coefficient if the feed increase is proportionally larger than the increased oxygen amount.

6 Decreasing top temperature. Increases the oxygen coefficient (*2).7 Feeding oxide material into the feed mixture. Decreases the oxygen coefficient because

the oxide material has a cooling effect and increases the feed.

8 Increasing the cooling area of the cooling coils in the fur-nace.

Decreases the oxygen coefficient because the feed increases.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35Oxygen coefficient

Sulp

hide

sul

phur

%

WHB sulphide S%Cyclone sulphide S%ESP sulphide S%Linear (WHB sulphide S%)Linear (Cyclone sulphide S%)Linear (ESP sulphide S%)

60

Fig. 27. Sulphide level of the WHB calcine is lower when the oxygen coefficient is high.

Fig. 28. High oxygen coefficient increases the amount of sulphates. Difference SO4 = the amountof sulphates minus PbSO4 and CaSO4, which are always sulphates in roasting conditions.

The oxygen on-line measurement from the end of the WHB indicates, as is shown byFig. 29, that the sulphate amount increases when the oxygen level increases. Fig. 29 alsoshows the need to calibrate the measurement device, because there is a difference in theoxygen level at different times. It is important that the boiler, cyclones and ESPs canhandle the dust and the SO4 load without any trouble, so that the freedom to choose theprocess parameters in the furnace and to use concentrates with high impurity levelsincreases. Fig. 30 shows that the sulphide level decreases when the oxygen level in theWHB gas increases.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.9 1 1.1 1.2 1.3 1.4

Oxygen coefficient

Sulp

hide

sul

phur

%

WHB calcine sulphide sulphur 4/2002

WHB calcine sulphide sulphur 11/2001

WHB calcine sulphide sulphur 5/2002

4

5

6

7

8

9

10

11

0.8 0.9 1 1.1 1.2 1.3 1.4

Oxygen coefficient

Diff

eren

ce S

O4

%

Difference SO4 of the cyclone calcine4/2002Difference SO4 of the cyclone calcine11/2001Difference SO4 of the cyclone calcine5/2002

61

Fig. 29. High oxygen level in the WHB gas increases the sulphate amount.

Fig. 30. High oxygen level in the WHB gas decreases the sulphide amount.

6.1.3 Oxygen use to control bed quality

Oxygen enrichment, i.e. feeding oxygen to the process air, is used by some plants.Oxygen can be used to increase capacity (chapter 5), but also in order to increase theoxygen coefficient. The cost of oxygen is a factor that reduces its use, but if there is thisopportunity, i.e. the oxygen line is installed and oxygen is available, then it is an easyway to control the oxygen coefficient. It is essential to have a bed with a good grain sizedistribution. During operation the bed has often become unstable, i.e. it has become fine,and by using oxygen the situation has several times been successfully corrected back tonormal (Nyberg et al. 2000). Figs 31 and 32 show that by using oxygen, the oxygencoefficient has increased and this has led to an increase in the amount of coarse calcineand a decrease in the amount of fine calcine. The use of oxygen and the increase inoxygen coefficient value have not always solved the situation and other actions have alsobeen needed. More examples of controlling instability situations will be presented later inthis chapter.

4

5

6

7

8

9

10

11

2 2.5 3 3.5 4 4.5 5 5.5 6 6.5

O2-content of the gas after waste heat boiler %

Diff

eren

ce S

O4

%

Difference SO4 of the cyclonecalcine 4/2002Difference SO4 of the cyclonecalcine 11/2001Difference SO4 of the cyclonecalcine 5/2002Li (Diff SO4 f h

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5

O2-content of the gas after WHB %

Sulp

hide

sul

phur

%

WHB calcine sulphide sulphur 4/2002WHB calcine sulphide sulphur 11/2001WHB calcine sulphide sulphur 5/2002Linear (WHB calcine sulphide sulphur 11/2001)

62

Fig. 31. Oxygen feed versus fine and coarse particles in the bed. During oxygen feeding theamount of fine particles decreased, while the amount of coarse particles increased.

Fig. 32. The oxygen coefficient versus fine and coarse calcine in the bed. Higher oxygencoefficient decreased the amount of fine particles and increased the amount of coarse.

6.1.4 Bigger nozzles and oxygen measurements from the bed

The concentrate feed is fed to the furnace by slinger belts from one side of the furnaceand the outlet of the furnace is on the other side. This calcine amount, i.e. the overflowcalcine from the furnace, is about 30–40 % of the total calcine amount. The majority ofthe calcine formed in the furnace is carried by the gas stream to the boiler. In slinger belttests, it became evident that the concentrate feed is not spread very far from the feed side.On many occasions it has been seen during annual maintenance shutdowns, when thefurnace has been cleaned, that there are build-ups on the grate area on the feed side of thefurnace. This can be explained by the fact that most of the concentrate reacts in this partof the furnace and therefore the oxygen need is greater in this part. Because an equalamount of oxygen is fed to the whole grate area there is a lack of oxygen in this area, i.e.the oxygen coefficient is low, which leads to sinters. The lack of oxygen on the feed side

0.0

100.0

200.0

300.0

400.0

500.0

600.0

700.0

800.0

900.0

17.9.20020:00

19.9.20020:00

21.9.20020:00

23.9.20020:00

25.9.20020:00

27.9.20020:00

29.9.20020:00

1.10.20020:00

3.10.20020:00

Time

Oxy

gen

Nm

3/

0.0

10.0

20.0

30.0

40.0

50.0

60.0

Frac

tion

%

Oxygen [Nm3/h]Fraction >1.25 mmFraction <71 µm

0.0

10.0

20.0

30.0

40.0

50.0

60.0

17.9.20020:00

19.9.20020:00

21.9.20020:00

23.9.20020:00

25.9.20020:00

27.9.20020:00

29.9.20020:00

1.10.20020:00

3.10.20020:00

Time

Frac

tion

%

1.06

1.08

1.10

1.12

1.14

1.16

1.18

1.20

1.22

1.24

1.26

Oxy

gen

coe

ffic

ieFraction >1.25 mmFraction <71 µmOxygen coefficient

63

has been mentioned in a paper (Richards 1995). To avoid these build-ups bigger nozzleshave been installed on the feed side of the furnace to keep the oxygen coefficient highenough (Taskinen et al. 2000a). Also, a special nozzle system has been tried to spread theconcentrate better on this side of the furnace (Nyberg et al. 2001). The need for moreoxygen in this side of the furnace has been confirmed by on-line oxygen measurementsfrom the bed. Oxygen has been measured during plant experiments from the feed and theoverflow side of the furnace (Parkkinen et al. 2002). Figs 33 and 34 show the very lowoxygen level on the feed side, while there is oxygen present on the overflow side. TheSO2 level in the bed can also be seen from Figs 33 and 34. SO2 measurements have alsobeen presented in a paper (Dimitrov & Schopov 1985).

Fig. 33. On-line gas measurement from the roaster bed on the feed side. The O2 level in the bedis very low.

Fig. 34. On-line gas measurement from the roaster bed on the overflow side. The O2 level ishigher on the outlet side of the furnace.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

8.11.2001 8:23 8.11.2001 10:23 8.11.2001 12:23 8.11.2001 14:23

O2-

cont

ent o

f the

bed

gas

[ %

0

6

12

18

24

30

SO

2-co

nten

t of t

he b

ed g

as [

%

Input side O2 % Input side SO2 %

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

8.11.2001 8:23 8.11.2001 10:23 8.11.2001 12:23 8.11.2001 14:23

O2-

con

tent

of t

he b

ed g

as [

%

0

6

12

18

24

30

SO

2-co

nten

t of t

he b

ed g

as [

%

Overflow side O2 % Overflow side SO2 %

64

6.1.5 Effects of water on the furnace behaviour

As previously mentioned, the addition of water has many effects. The addition of water tothe concentrate stream particularly before the day bins and also before the table feederhas a pelletising effect. Thus it should be used in cases when the concentrate mixture hasa high impurity level and is fine. Feeding water to the concentrate mixture before thefurnace also reduces the feed and increases the oxygen coefficient. In the cases shown inFigs 35–39, the furnace temperature set point is given and the concentrate mixtureamount controls the temperature. Water can also be used to control the top temperature byfeeding it to the concentrate mixture. By injecting water to the bed the feed can beincreased. The water feed cools the bed and for this reason the control applicationincreases the feed in order to keep the temperature at the set point. Some effects of waterare described in the following text and can be seen in the following figures. Similarresults have also been presented in a published paper (Nyberg et al. 2000).Water fed to the concentrate mixture has the following effects:

– Stabilises sulphide combustion in a fluidised bed. This has been seen during experi-ments. The oxygen on-line measurements have shown that the oxygen level is stabi-lised when water has been fed. Other experiments have shown that the temperaturedifferences between the different furnace temperature measurements have been dec-reased.

– Decreases the feed when the concentrate feed controls the temperature. This is thecase when water is fed before the day bins, as shown in Fig. 35, or is fed after the daybins, as shown in Fig. 36.

– Transfers the combustion of the concentrate mixture more to the bed, i.e. less com-bustion over the bed, which decreases the capacity.

– Increases the total oxygen coefficient.– Decreases the sulphide content of the boiler and cyclone calcine as shown in Fig. 38.– Decreases the temperature of the upper part of the roaster furnace.– Increases the fraction of overflow calcine (= the calcine furnace overflow).

Fig. 35. High moisture of the concentrate mixture decreases the feed. The water makesagglomerates and transfers the burning more to the bed. The feed is reduced because otherwisethe temperature would rise in the furnace, due to the furnace temperature control application.

20

21

22

23

24

25

26

27

28

7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12

Moisture %

Wet

feed

t/h

Test 2000Test 11/2001Linear (Test 2000)Linear (Test 11/2001)

65

Water sprayed into the fluidised bed has the following effects:• Cools the calcine bed in the furnace.• Increases the concentrate feed as shown in Fig. 37.• Decreases the total oxygen coefficient.

Fig. 36. Water fed to the concentrate before the furnace decreases the concentrate feed. In thiscase too, water makes agglomerates and transfers the burning more to the bed. The feed isreduced because otherwise the temperature would rise in the furnace, due to the furnacetemperature control application.

Fig. 37. Water injection into the bed increases the feed. Water sprayed (500 l/h) into the furnacebed cools the bed and more concentrate has to be fed to keep the temperature in the furnace atthe set point value due to the furnace temperature control application.

Fig. 37 shows that the concentrate feed is high when water (500 l/h) is injected into thefurnace bed. When the water injection is stopped, then the feed drops.

20

21

22

23

24

25

26

27

28

29

0 100 200 300 400 500 600 700

Water into concentrate l/h

Feed

t/h

Air 44 000 Nm3/h

Air flow 44 000 Nm3/h; Water feed into furnace

22.0

23.0

24.0

25.0

26.0

27.0

28.0

16.11.200121:36

16.11.200122:48

17.11.200100:00

17.11.200101:12

17.11.200102:24

17.11.200103:36

17.11.200104:48

17.11.200106:00

Time

Feed

rate

t/h

Water feed

No water feed

66

Fig. 38. Water fed into the concentrate before the furnace decreases the sulphide level of theboiler and cyclone calcine. The feed decreases due to the water feed and this increases theoxygen coefficient.

The decrease in the sulphide level of the boiler and cyclone calcine, as shown in Fig.38, is due to the fact that a low feed increases the oxygen coefficient. Fig. 39 shows howwater fed before the furnace increases the SO4 content. The reason for this is that theoxygen coefficient increases with the result that the SO4 level of the furnace calcine rises.

Fig. 39. Increasing the water feed (*1) before the day bins increases the sulphate amount in thebed. (*1 = The concentrate feed rate to the day bins is about 350 t/h and water is fed to thisstream).

6.1.6 Control of the furnace when Pb and/or Cu is high

The major impurities that should be emphasised in roasting are Cu and Pb. Manyconcentrates have a high Pb content, some have a high Cu content. The danger of theseelements is that they can easily form molten phases as described in chapter 3. Besides Cu

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 100 200 300 400 500 600 700

Moistening water of the concentrate l/h

Sulp

hide

sul

phur

con

tent

of t

he c

alci

ne %

Sulphide sulphur of the boilercalcine %Sulphide sulphur of the cyclonecalcine %Linear (Sulphide sulphur of theboiler calcine %)Li (S l hid l h f th

3.95

4

4.05

4.1

4.15

4.2

4.25

4.3

4.35

4.4

4.45

0 1 2 3 4 5

Water feed before day bins m3/h

SO

4-co

nten

t of t

he o

verf

low

cal

cin

67

and Pb, other elements also form molten phases. The molten phases lead to build-ups inthe furnace and even to sintering of the bed in the worst case. The challenge in the controlof the furnace is to avoid molten phases when the impurity levels of the concentratemixture are high.

In plant experiments where the Pb level has been high in the concentrate mixture andespecially when the mixture has also been fine, the need to feed water before the day binshas been obvious. Oxygen measurements from the bed have shown that the water feedhas stabilized the bed. In addition, feeding water after the day bins before the table feederstabilizes the bed, as can be seen from the fact that the temperature gradient is smaller.This will be shown later in this chapter when the dynamics is discussed. In plantexperiments where there has been 3.3% Pb in the concentrate mixture and more than 5%in the bed for some time, the bed has been kept stable successfully without majordifficulties. The crucial keys in keeping the bed stable have been the water feed to theconcentrate mixture before the day bins, keeping the oxygen coefficient over 1.1 andrunning the furnace at a low temperature.

During plant experiments, numerous samples have been taken and analysed, bothchemical and microscopic analyses. Microscopic analyses are important in order to getmore information about what is going on in the bed. Fig. 40 shows how lead behaves inagglomerates. The coating around the particle is mostly PbSO4 while the inner part ismostly ZnO.

Fig. 40. SEM picture (left) of an overflow calcine 10.4.2002 (40 x). The coating around the innerpart (=ZnO) is PbSO4. The other picture (right) shows that copper behaves differently to lead.A Cu2S core is formed in a copper rich concentrate pellet in a laboratory roasting experimentat 950°C. The coating around the Cu2S core is ZnO.

Copper is a critical component in roasting. Copper behaves differently compared withlead. Copper tends to migrate in the agglomerated zinc sulphide concentrates and enrichas the least easily oxidising element in the centre of the agglomerate, as shown in Fig. 40.The Cu2S cores formed, with very high copper concentrations, require long oxidationtimes and a high oxygen coefficient (Taskinen et al. 2002, Metsärinta et al. 2003,Metsärinta et al. 2005b).

During laboratory experiments the copper level in the feed has been up to about 0.9 %.In laboratory experiments with a high Cu content, the need for a very high oxygencoefficient proved essential. With a high oxygen coefficient, over 1.3, the bed remainedstable, i.e. no sintering of the bed occurred, even with a high copper content. Based on the

68

laboratory tests, plant trials were performed with a Cu level of about 0.9 %. From plantexperiments, as shown in Figs 41 and 42, it can be seen that the copper level and oxygencoefficient have an effect on the bed, i.e. the percentage of calcine from the overflowdepends on these variables. Later, based on the good experience from the plant trials withhigh Pb in the feed (about 3.3 %) a plant test with a very high Cu in the feed was plannedand carried out. The test run with a very high Cu level in the feed mixture, which lastedabout two weeks, was performed successfully. The Cu level was about 0.9 –1.5 % in thefeed mixture. The oxygen coefficient must be high and the bed temperature low, when thecopper level of the concentrate feed is high (Metsärinta et al. (2003, 2005a, 2005b)). Thedanger of a high temperature in the bed can be seen when phase diagrams are studied(Appendix 1). In the plant experiments (both in the case with high Pb and in the case withhigh Cu) the furnace bed temperature was kept at the level of about 900oC, which provedto be good, i.e. formation of build-ups on the furnace grate was avoided.

Fig. 41. The Cu level of the feed versus the amount of calcine overflow of the concentrate feed.The proportion of the overflow calcine has decreased when the Cu level has increased.

Fig. 42. Oxygen coefficient versus the amount of calcine overflow of the concentrate feed. Theproportion of the overflow calcine has increased when the oxygen coefficient has beenincreased.

30.00

30.50

31.00

31.50

32.00

32.50

0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Cu-content of the feed-mix %

Prop

ortio

n of

the

over

flow

cal

cine

of t

he fe

ed%

28.00

28.50

29.00

29.50

30.00

30.50

31.00

31.50

32.00

32.50

33.00

33.50

1.18 1.20 1.22 1.24 1.26 1.28 1.30 1.32 1.34

Oxygen coefficient

Prop

ortio

n of

the

over

flow

cal

cine

of t

he fe

ed%

Water onto the feed belt

69

Plant experiments are necessary to gain reliable information on how the furnacebehaves when the impurity levels of the concentrate mixture are high. There is always arisk that such experiments can lead to build-ups in the furnace and to other problems.Therefore, such experiments must be well planned and performed. There are also limitson how high impurity levels can be tested in full-scale furnaces.

6.1.7 Controlling unstable situations

When there is instability in the bed, control actions must be taken to rectify the situation,i.e. to stabilise bed behaviour. Examples of different situations are presented in Figs 43–51 and in an article (Nyberg et al. 2000). In these cases active control based onknowledge gained has been used to correct the situations. Knowledge of the roastingmechanism and also knowledge of how the particle size distribution of the bed can becontrolled is essential for successful control. It can be seen from the figures that controlof the oxygen coefficient has brought the bed back to a stable situation. In some cases,control and increasing the oxygen coefficient have not helped. Feeding water to theconcentrate mixture has also stabilised the bed in some cases. The figures show theimportance and effects of the oxygen coefficient and the water feed. Other aspects to beconsidered are bed temperature and impurity levels.

The fluidised bed was made stable, as shown in Fig. 43, by decreasing the bedtemperature and process airflow and by increasing the oxygen coefficient of the roastingfurnace by feeding oxygen. The bed temperatures rose back to normal during the period(August 2002).

Fig. 43. Bed temperatures and oxygen coefficient of roaster furnace No. 2 in August, showingthe effect of oxygen coefficient alterations on stabilisation. By increasing the oxygen coefficient,the stabilisation of the bed temperatures can be seen.

The fluidised bed also became unstable in September–October 2002. The wind-boxpressure decreased, although the oxygen coefficient was relatively high as shown in Fig.44. In this case the moistening water feed to the concentrate feed before the day bins was

Roaster 2 - Bed temperatures and oxygen coefficient

700

750

800

850

900

950

03.0

8. 0

4:00

03.0

8. 1

6:00

04.0

8. 0

4:00

04.0

8. 1

6:00

05.0

8. 0

4:00

05.0

8. 1

6:00

06.0

8. 0

4:00

06.0

8. 1

6:00

07.0

8. 0

4:00

07.0

8. 1

6:00

08.0

8. 0

4:00

08.0

8. 1

6:00

09.0

8. 0

4:00

09.0

8. 1

6:00

10.0

8. 0

4:00

10.0

8. 1

6:00

11.0

8. 0

4:00

11.0

8. 1

6:00

12.0

8. 0

4:00

12.0

8. 1

6:00

13.0

8. 0

4:00

13.0

8. 1

6:00

[°C]

0.951.001.051.101.151.201.251.301.351.401.45

Temperature measurement 1 [°C] Temperature measurement 5 [°C]

Temperature measurement 2 [°C] Oxygen coefficient

70

decreased at the same time as the wind-box pressure dropped. By increasing the waterfeed, as shown in Fig. 45, the wind-box pressure returned back to normal, thus showingthe importance of the water feed. During this period the grain size of the concentratemixture was fine. Increasing the oxygen coefficient failed to help in this case, but thewater feed solved the situation.

Fig. 44. Oxygen coefficient and wind-box pressure in furnace No. 2 (September–October 2002).When the wind-box started to drop, the oxygen coefficient was increased in order to raise thewind-box pressure back to the normal level.

Fig. 45. Wind-box pressure and the total moistening water feed in roaster furnace No. 2(September–October 2002). Water feed correlates with bed pressure, thus the wind-boxpressure is low, i.e. unstable bed, when the water feed is low.

Figs 46 and 47 show that during this period a stable bed, i.e. stable bed temperatures,correlates more with the water feed and so the water feed is more essential than theoxygen coefficient in this case. The bed temperatures remain quite stable although theoxygen coefficient drops.

230

235

240

245

250

255

260

265

270

275

280

285

25.09.200200:00

27.09.200200:00

29.09.200200:00

01.10.200200:00

03.10.200200:00

05.10.200200:00

07.10.200200:00

Time

Win

dbox

pre

ssur

e m

bar

1.10

1.12

1.14

1.16

1.18

1.20

1.22

1.24

Oxy

gen

coef

ficie

nt

Windbox pressure [mbar]Oxygen coefficient

230

235

240

245

250

255

260

265

270

275

280

285

25.09.2002 00:00 27.09.2002 00:00 29.09.2002 00:00 01.10.2002 00:00 03.10.2002 00:00 05.10.2002 00:00 07.10.2002 00:00

Time

Win

dbox

pre

ssur

e m

bar

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Wat

er t/

h

Windbox pressure [mbar]Total moistening t/h

71

Fig. 46. Oxygen coefficient and bed temperatures (# 4 & 5) of roaster No. 1 (July–August 2002).The correlation is not so good, i.e. the temperatures stay stable even if the oxygen coefficientdrops.

Fig. 47. Water feed before the day bins and bed temperatures of roaster No. 1 (July–August2002). The temperature drops (Thermocouple element No. 5) after the water feed is stopped.After the water is again fed to the concentrate mixture, the temperature value returns back tothe normal level.

Fig. 48 shows that increasing the oxygen coefficient had a stabilising effect on the bedtemperatures, i.e. the values of the measurements stopped falling and rose towards thenormal level. The increase in the oxygen coefficient was achieved by decreasing the feed.The feed was decreased by lowering the temperature set point of the fuzzy controller forthe furnace, as shown in Fig. 49. Increasing the oxygen coefficient can also be seen in thebed pressure (wind-box pressure), the low level on 8.8.2002 moves towards the normallevel when the oxygen coefficient increases, as shown in Fig. 50.

850

860

870

880

890

900

910

920

930

940

950

25.07.200200:00

27.07.200200:00

29.07.200200:00

31.07.200200:00

02.08.200200:00

04.08.200200:00

06.08.200200:00Time

Bed

tem

pera

ture

s °C

1.02

1.04

1.06

1.08

1.10

1.12

1.14

1.16

Oxy

gen

coef

ficie

nt

Temperature measurement 4 [°C]Temperature measurement 5 [°C]Oxygen coefficient

850

860

870

880

890

900

910

920

930

940

950

25.07.2002 00:00 27.07.2002 00:00 29.07.2002 00:00 31.07.2002 00:00 02.08.2002 00:00 04.08.2002 00:00 06.08.2002 00:00

Time

Bed

tem

pera

ture

s °C

0

0.05

0.1

0.15

0.2

0.25

Wat

er t/

h

Temperature measurement 4 [°C]

Temperature measurement 5 [°C]

Moistening before day bins [t/h]

72

Fig. 48. Increasing the oxygen coefficient stabilized the bed temperatures (the negative trendstopped).

Fig. 49. Changes of the oxygen coefficient were made by changing the feed rate. This was doneby changing the set point value of the furnace temperature. The fuzzy controller then decreasedor increased the feed rate.

400.0

500.0

600.0

700.0

800.0

900.0

1000.0

02.08.200200:00

04.08.200200:00

06.08.200200:00

08.08.200200:00

10.08.200200:00

12.08.200200:00

14.08.200200:00

16.08.200200:00

18.08.200200:00

Time

Bed

tem

pera

ture

s °C

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1.35

1.40

Oxy

gen

coef

ficie

nt

Temperature point 1 [°C] Temperature point 2 [°C]Temperature point 3 [°C] Temperature point 4 [°C]Temperature point 5 [°C] Oxygen coefficient

18

20

22

24

26

28

30

02.08.200200:00

04.08.200200:00

06.08.200200:00

08.08.200200:00

10.08.200200:00

12.08.200200:00

14.08.200200:00

16.08.200200:00

18.08.200200:00

Time

Wet

feed

t/h

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1.35

1.40

Oxy

gen

coef

ficie

ntFeed [t/h]Oxygen coefficient

73

Fig. 50. Increasing the oxygen coefficient stabilized the bed pressure. The drop in bed pressurewas stopped and a slight increase back to the normal level can be seen. The period is the sameas in Fig. 49.

Fig. 51 shows that the situation in the furnace was such that increasing the oxygencoefficient had a positive impact on the bed pressure. This resulted in a more stable bed.

Fig. 51. Increasing the oxygen coefficient brought the bed pressure back to normal.

During the period described in Figs 43–51, the feed was fine, i.e. the d50-value (= massmedian diameter) was in the range of about 20–22 µm. The furnace bed becamesomewhat unstable occasionally during the period. The furnaces were brought back tonormal by active control of the process without changing the concentrate mixture to a safemixture, i.e. coarse and fewer impurities. The plant experiments performed haveincreased knowledge of the roasting mechanism. Active surveillance and control indifficult situations have also increased knowledge. Based on this, new instructions havebeen adopted and the operators make their control decisions accordingly.

Means of controlling difficult situations have also been increased. Thus the operatorshave more options to adjust the roasting conditions so that they are suitable for theconcentrate feed mixture. Often it is very complicated to know what is actually takingplace in the furnace. The same kind of indications can have different backgrounds and

200.0

210.0

220.0

230.0

240.0

250.0

260.0

270.0

280.0

290.0

02.08.200200:00

04.08.200200:00

06.08.200200:00

08.08.200200:00

10.08.200200:00

12.08.200200:00

14.08.200200:00

16.08.200200:00

18.08.200200:00

Time

Bed

pre

ssur

e m

bar

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1.35

1.40

Oxy

gen

coef

ficie

nt

Bed pressure [mbar]

Oxygen coefficient

200.0

210.0

220.0

230.0

240.0

250.0

260.0

270.0

280.0

25.07.200200:00

26.07.200200:00

27.07.200200:00

28.07.200200:00

29.07.200200:00

30.07.200200:00

31.07.200200:00

01.08.200200:00

02.08.200200:00

03.08.200200:00

Time

Bed

pre

ssur

e m

bar

0.95

1.00

1.05

1.10

1.15

1.20

1.25

Oxy

gen

Coe

ffici

ent

Bed pressure [mbar]Oxygen coefficient

74

causes. Automation systems that show both short- and long-term trends have madeevaluation more effective. Monitoring and analysing the bed thoroughly is of greatimportance in order to get more information about developments in the furnace. Bothchemical analysis (Cu, Pb, Si etc.) and the particle size distribution analysis of the calcinebed give a lot of information about the situation in the bed together with othermeasurements. Additional measurements are required to get a more accurate picture ofthe roasting furnace behaviour.

Developing systems that visualise the situation, for instance a self-organizing map(SOM) application, and then advise the operators, i.e. expert systems, are possible tools toimprove performance. An on-line application of an SOM has been developed for theKokkola roaster for detecting and predicting instability of the fluidised bed furnaces(Saxén & Nyberg 2003).

6.1.8 The dynamics of the roaster furnace

The dynamics of the roaster furnace have been studied thoroughly. The dynamics arerapid, especially the gas phase, but there are also slower processes where the response isnot so fast. The time delay of the bed is much slower than the time delay for the gasphase. Information about the dynamics has been necessary for developing the furnacecontrol. The fuzzy control of the furnace temperature and development of the simulatorare cases where the dynamics have been studied before implementation (Rauma et al.2000, Toskala et al. 2001). The fuzzy control system that was developed reduced thegeneral deviation of the temperature from 10oC to 3oC 95% of the time (Rauma et al.2000). The structure of the fuzzy control system is shown in Fig. 52.

Fig. 52. The structure of the installed fuzzy control system (Rauma et al. 2000).

In the simulator case the very rapid pressure dynamics, i.e. changes in the pressure ofthe furnace and boiler system, were studied by collecting data from the process during

Basic Controller

RULE 1RULE 2

...RULE N

Additionalmodule

RULE 1RULE 2

...RULE N

GovernedRecovery

RULE 1RULE 2

...RULE N

ZincRoasterFurnace

Scaling Factor

+ x

User Interface

Recovery SpeedInformation ofcontrol actions

Metrics

feeding

Measurements

75

tests. The reason for building the simulator was to give the operators a tool for practisingthe demanding start-up and shutdown situations. The simulator can also be used to see theeffects on the furnace pressure when valves and blowers are regulated. The simulator isintegrated in the automation system that is used for the actual control of the process. Fig.53 shows the scope of the simulator.

Fig. 53. The scope of the simulator made for training start-up and shutdown situations. Adynamic model for the furnace pressure was made (Toskala et al. 2001).

The dynamics of the process have been studied further. Dynamic models wereidentified for input-output relations in the fluidised bed furnace. Based on theseexperiments, a state-space model of the process has been built (Härkönen 2003, Saxén etal. 2004). In this study, step trials were performed where the influence of differentvariables was tested. During the tests the control loops that control the temperature of thefurnace, i.e. the fuzzy control of the bed temperature and the top temperature control loop,were disconnected. The concentrate feed and air amount were fixed during the tests. Incases where the effect of the concentrate feed or of the air amount were tested, they wereof course changed. The main variables that were tested are as follows:

– Water feed to the concentrate mixture before the table feeder.– Water injection to the furnace bed.– Air amount.– Concentrate feed rate.– Oxygen feed.The aim of the tests was to collect data on how the different variables react in order to

build a model, but also to develop a control system to improve the control of the furnace.Some results are presented below in Figs 54–57, which show the dynamics of the furnacebehaviour. There are both rapid and slow responses.

By-pass valve

Airvalve

Air blower

Vane

Furnace + boiler+ cyclones +precipitators

SO2-blower

Vane

Stack valve

Exit gas pipevalve

Dynamic modelof the furnace

pressure

Stack

H2SO4-plant

Ufav

Ubpv

Uabv

Usbv

Usv

Uev

DisturbanceFurnace pressure, Yp

Control variables: - air blower’s vane position Uabv

- by-pass valve position Ubpv

- fluidising air valvei i

Ufav- SO2 blower’s vane position2 Usbv- stack valve position Usv- exit gas pipes valve position Uev

Yfa

Yfa- fluidising air flow

76

Feeding water to the concentrate mixture increases the bed temperature (slowly) anddecreases the top temperature of the furnace (rapid response). At the beginning there is aninverse response of the bed temperature before the temperature starts to rise. Feedingwater to the concentrate mixture stabilises the bed and this can be seen from the fact thatthe bed temperatures start to converge as is shown in Fig. 54, i.e. the differences betweenthe different measurements become smaller. The same thing can be seen in Fig. 55, wherethe difference between different bed temperatures is smaller when more water is used.Water injection cools the bed when water is injected into the bed, as shown in Fig. 56.

Fig. 54. Responses of gas temperature and bed temperatures, when water is fed to theconcentrate mixture before the table feeder. The temperature rises in the bed when water is fedto the concentrate mixture. At the beginning there is a slightly inverse reaction in the bedtemperatures. The top temperature decreases when water is fed to the concentrate mixture.

0100200300400500600

18/8/03 12:00 18/8/03 15:00 18/8/03 18:00 18/8/03 21:00 19/8/03 00:00

L/h

Load conveyor water feed

900

920

940

960

980

18/8/03 12:00 18/8/03 15:00 18/8/03 18:00 18/8/03 21:00 19/8/03 00:00

°C

Furnace temp. 6 Furnace temp. 7

Furnace temp. 12 Furnace temp. 14

850860870880890900910920930

18/8/03 12:00 18/8/03 15:00 18/8/03 18:00 18/8/03 21:00 19/8/03 00:00

°C

Boiler inlet temp. (left)

Boiler inlet temp. (right)

77

Fig. 55. Water feed before the table feeder stabilises the bed, which can be seen from the factthat the temperature measurements in the furnace are closer to each other when more water isfed.

Fig. 56. Response of water injection into the bed. The furnace temperature decreases due to thecooling effect of the water injection.

Fig. 57 shows that when the oxygen feed is started, it has a divergent effect on the bedtemperatures.

0

5

10

15

20

25

30

0 100 200 300 400 500

Load conveyor water (L/h)

max

-min

tem

p. (2

,5,6

,7,1

2,14

) ('C

) Furnace 1 7.-28.6.03

0.0100.0200.0300.0400.0500.0600.0700.0

21/8/03 12:00 21/8/03 15:00 21/8/03 18:00 21/8/03 21:00 22/8/03 00:00

L/h

Furnace water feed

900

920

940

960

980

21/8/03 12:00 21/8/03 15:00 21/8/03 18:00 21/8/03 21:00 22/8/03 00:00

°C

Furnace temp. 6 Furnace temp. 7

Furnace temp. 12 Furnace temp. 14

870

880

890

900

910

920

930

21/8/03 12:00 21/8/03 15:00 21/8/03 18:00 21/8/03 21:00 22/8/03 00:00

°C

Boiler inlet temp. (left)

Boiler inlet temp. (right)

78

Fig. 57. Divergent effect of oxygen on bed temperature measurements. The oxygen seems tohave a divergent effect on the temperatures, which could be explained by an increase of thereaction rate in some parts of the furnace. Where there is a temperature drop, the reaction ratedecreases.

By doing these tests a lot of information was gathered on how different variablesrespond when a variable is changed. The bed behaviour, fluctuation of the temperaturemeasurements when oxygen feed was started to the concentrate mixture and theconvergent effect of the bed temperatures, also provided information about the roastingmechanism and information about the advantage of using water. This research resulted inmore knowledge of process dynamics and also more process know-how.

Tracer tests were made by adding Ti-oxide to the concentrate mixture to determine theresidence times (Saxén et al. 2004). The tracer was fed as an impulse to the concentratefeed. The tests showed that the furnace bed behaved something like an ideally stirredreactor, i.e. a first order system. The time constant for the furnace bed was about 20 h. Forthe gas line the residence time was very short. The peak concentration in the cyclone wasmeasured about one minute after the impulse in the furnace feed. Figs 58 and Fig. 59show the results from the tracer tests.

0

500

1000

1500

2000

9/6/03 18:00 9/6/03 21:00 10/6/03 00:00 10/6/03 03:00 10/6/03 06:00

L/h

Oxygen feed

900

920

940

960

980

9/6/03 18:00 9/6/03 21:00 10/6/03 00:00 10/6/03 03:00 10/6/03 06:00

°C

40.0

45.0

50.0

55.0

60.0Furnace temp. 6Furnace temp. 7

830

840

850

860

870

880

890

9/6/03 18:00 9/6/03 21:00 10/6/03 00:00 10/6/03 03:00 10/6/03 06:00

°C

Boiler inlet temp. (left)

Boiler inlet temp. (right)

79

Fig. 58. Tracer content in the furnace overflow calcine for two experiments. The Ti-oxide tracerwas fed as an impulse at time 0 h. A first order model with a time constant of 20 h is shown forcomparison (Saxén et al. 2004).

Fig. 59. The tracer tests shows that the time constant in the gas line is very short, i.e. the Ti-oxidetracer fed as an impulse to the furnace is almost immediately detected in the calcine from thecyclone after the boiler.

6.2 Boiler development

The work done and changes made to the boilers have been successful (Nyberg et al.2000, Järvi et al. 2002). The running time between manual cleanings has increased fromabout 3–5 months to more than 9 months. Running periods of up to over 1 year have been

Furnace overflow

0

0.05

0.1

0.15

0.2

0.25

-10 0 10 20 30 40 50 60 70 80 90 100

Time (h)

Ti (%

)2.9.2003 Furnace calcine Ti %

22.9.2003 Furnace calcine Ti %

First order model, T= 20h

Gas line

0

0.05

0.1

0.15

0.2

0.25

-10 0 10 20 30

Time (min)

Ti (%

)

1.9.2003 Cyclone Ti %

4.9.2003 Cyclone Ti %

80

achieved (Kilju 2004). The baffle in the radiation section has been a success, but the longrunning periods are also a consequence of other development work (Peippo et al. 1999).The new type of bundles with effective hammering, section walls with hammering andinstallation of more hammering devices have also proved to be good solutions. Inaddition, effective cleaning during yearly shutdowns, i.e. cleaning the walls and bundlesnot only mechanically but also using water, prevents build-ups in the boiler from formingso easily. This is due to the fact that build-ups, in this case sulphates, do not stick soeasily to the boiler walls and bundles when they are metallically clean. The conditions inthe roasting furnace have an impact on the boiler performance. When there are moresulphates in the calcine going to the boiler more stress is exerted on the boiler.

The aim of the boiler improvements has been to reduce the constraints in roasting byhaving a boiler that can handle the calcine and gas load coming from the furnace. Theother aim has been to increase the running time of the boiler. The targets have beenreached, so the bottleneck is now more in the roaster furnace performance.

The effects of the baffle wall on the behaviour of the boiler were calculated in advanceand the resulting Fluent model calculations are shown in Figs 60 and 61.

Fig. 60. The effect of the baffle wall on the boiler behaviour (radiation section of the boiler) isshown here. The velocity vectors are shown. The Fluent model calculation was made by FosterWheeler (= the boiler manufacturer).

81

Fig. 61. The effect of the baffle wall on the boiler behaviour (radiation section of the boiler) isshown in this figure. The temperature gradients are shown. The Fluent model calculation wasmade by Foster Wheeler (= the boiler manufacturer).

6.3 Mercury removal development

The quality of H2SO4 has improved significantly due to the results of the research carriedout (Nyberg et al 2000). The mercury level in the sulphuric acid can be kept at about0.02–0.06 mg Hg/kg H2SO4. Also, stability has improved and there are seldom highmercury levels in the sulphuric acid and if there are, the operators now have better meansto control the quality. The main factor is to feed selenite solution to different processstages and above all to the sulphuric acid process.

Based on material balance calculations and chemical analyses from different processstages, extensive thermodynamic equilibrium calculations were performed. The chemicalanalyses made were very demanding because of the difficulty of making a proper analysisand the fact that the low quantities also complicated the analysis. The equilibriumcalculations correlated well with the chemical analyses. Therefore, new factors ofimportance were learned, i.e. the role of the temperature and H2SO4 were not so criticalas the role of selenium.

In addition, by studying the slurry in the sulphuric acid process it could be detectedthat the mercury was removed as HgSe and the need for Se was critical. Precipitation ofmercury with Se solution was carried out with success in the laboratory. After that, thesame was done at the plant (Peltola et al. 2000). Further research has been done and theuse of thiosulphate is another possible solution (Berg et al. 2003).

Better process know-how and new means of controlling the process are the basicresults of the research. The approach of carrying out both basic and applied research alsoproved in this case to be the right way to improve the control of the process.

7 Discussion

The research and development presented in this study includes and is based on bothlaboratory and full-scale plant experiments, various studies, benchmarking, chemicalanalyses and calculations. Doing research and development in this way has resulted inbetter understanding of the process and in applications that have improved theperformance.

The research must be long-term because the process phenomena are so complicatedthat otherwise good results cannot be achieved. Only by thorough and broad research cannew knowledge be gained. The effects of the applications also often need long-termevaluation before their benefits and effects can be determined. Especially in furnaceresearch, full-scale plant experiments are necessary to gain reliable results. Only by doingthese full-scale experiments can the effects of different control variables and impurities beseen. The performance of these plant experiments must be well planned, because theyinvolve risks. Their evaluation must also be made carefully in order to draw the rightconclusions. The research must be broad in the sense that both detailed and overallresearch must be done. Details must be investigated to get knowledge of the roastingmechanism, boiler behaviour and mercury behaviour. Research must be overall in thesense that the whole process chain must be considered in order to reach the final targets,i.e. enhanced performance. A change in one part of the process has effects on other partsof the process and therefore a holistic view is a necessary. By improving the performancein many areas, i.e. more measurements, new control variables, new instructions for theoperators and equipment modifications, performance is bound to improve. The target hasbeen to eliminate bottlenecks and make changes that do not lead to trouble in other partsof the process.

The research and development carried out has resulted in the implementation of a newoverall control strategy for the roasting process. The control strategy has proved effectiveand is presented later in this chapter. The control strategy will be further developed andthe aim is to raise the level of automation.

83

7.1 Summary of the results

The research carried out has resulted in improved roaster performance in many ways. Therunning time of the furnace and boiler has improved, i.e. there are less shutdowns andlonger periods between cleaning. The quality of the sulphuric acid is better and easier tocontrol. The flexibility in using different kinds of concentrates has increased. Thisflexibility is important because the treatment charge is the major source of income.

The work done to understand the roasting mechanism has been successful in manyways. More knowledge about instability situations has been gained and moremeasurements showing the development of instability have been added. Thus theoperators know at an early stage when the bed starts to become unstable. Many timesinstability in the bed has been successfully corrected by active control operations basedon the knowledge gained.

The target of reaching a running time greater than 95 % due to better furnace practicenow seems within reach. About two years’ running time (i.e. the interval betweencleaning of the furnace) has been achieved and the furnace grate has remained in goodshape, i.e. with only minor build-ups in the grate area. Earlier the normal running timewas one year between cleaning.

More measurements, more analyses and calculated variables have increased theknowledge of the state of the furnace. Above all two things have proved to be of greatimportance, i.e. the calculated oxygen coefficient and the grain size distribution of thefurnace overflow calcine. The calculation of the oxygen coefficient (the oxygencoefficient is defined in chapter 5) and the control of the process based on this are two ofthe major improvements in the control of the roaster furnace. Oxygen coefficient controlhas led to more stable conditions in the furnace and increased running time. Thecalculated oxygen coefficient is one of the main control variables for keeping the furnacestable. Other key variables are the temperature and fluidisation rate. The oxygencoefficient has been an important variable both in plant experiments and in the dailycontrol of the furnace. The grain size distribution analysis of the furnace overflow calcineand above all the amount of fine material are good indicators of the bed quality and bedinstability. Along with others, i.e. wind-box pressure, bed temperatures and calculatedcalcine overflow amount, it indicates changes in the fluidised bed. The grain sizedistribution is also important when evaluating factors that influence grain size growth.

Besides knowing the state of the furnace through measurements, analyses andcalculated variables, one should also have the capability of controlling the furnace. Byincreasing the control options, the chances of keeping the furnace stable are increased. Inthis research the importance has been shown of having different means and tools tocontrol the process and also the effects they have. Feeding water to different spotsinfluences the furnace behaviour in many ways as described earlier. In addition, thechance to use oxygen gives more freedom. Other means such as fixed bigger nozzles insome parts of the furnace are advantageous in order to keep the oxygen coefficient at theproper level, thus decreasing the formation of build-ups. One of the key results of theplant experiments and active surveillance of the process has been new instructions for theoperators on how to control the furnace in different situations. As well as calculating theoxygen coefficient, instructions have been made also on how the oxygen coefficient canbe controlled by different control variables.

84

Better knowledge of the state of the furnace and more control options, along withknowledge of their effect, have led to longer running time of the furnace. Too much build-up in the furnace can be avoided thanks to better control. However, the formation ofbuild-ups also depends on the impurity levels of the concentrate mixture, as too highimpurity levels lead inevitably to formation of build-ups.

7.1.1 Summary of roasting development

Step by step more knowledge about zinc roasting has been gained through research. Theefficiency of the roaster has gradually increased. The results of the focused research anddevelopment program of the roaster furnace can be summarized as follows:

– There is now more knowledge about the state of the process. By adding moretemperature measurements to the furnace more information about the bed behaviour hasbeen gained. Calculated variables like the oxygen coefficient and the amount of overflowcalcine from the furnace have increased the on-line information about the state of thefurnace. Analyses have been also added. The grain size distribution analysis of thefurnace calcine is a valuable indicator of changes in the state of the fluidised bed. Thenew measurements (additional temperature measurements) and analyses (grain sizedistribution) show at an early stage when the bed starts to become unstable.

– There are now more control variables in use, which increases flexibility. There arenow more opportunities to arrange the conditions in the furnace proactively: Water feed todifferent areas. Water agglomerates fine concentrates and thus the top temperature of thefurnace can be controlled. Water injected to the bed increases the feed and thus makes itpossible to control the cooling capacity of the furnace. There is the possibility of usingoxygen enrichment. More oxygen is fed to the feed side of the furnace through biggernozzles.

– More knowledge about the dynamics of the furnace has been gained. Informationnecessary for building simulators, models and designing control applications has beengained through the experiments performed during this work (Rauma et al. 2000,Härkönen 2003, Saxén et al. 2004, Valo 2004).

– A simulator has also been developed which operators use to keep up their skills, i.e.to practise start-ups and shutdowns. The simulator can also be used to test the processdynamics, i.e. the impacts of control variables on the furnace pressure (Toskala 2000,Toskala et al. 2001).

– There is now more knowledge about the effects of different variables on the process.This information has been gained from the experiments performed. There is now moreknowledge regarding both metallurgy and dynamics (Lepistö 1999, Nyberg et al. 2000,chapter 6, Metsärinta et al. 2005a).

– There is now more knowledge about the optimum conditions for differentconcentrate mixtures. Full-scale plant tests with the roaster furnaces have resulted in moreknowledge about the effects of different variables, such as the impact and the size of theimpact on process behaviour. During experiments with impurities such as Pb and Cu inthe feed mixture, new operation conditions have been tested successfully. In cases with

85

high impurity levels (Pb, Cu etc.), it has been found that a low bed temperature (about900 oC) might be necessary.

– A heat balance model has been developed which has been in use to estimate theeffects of different control variables such as water feed, oxygen feed and changes in theair flow (Roine & Nyberg 2000, Björklund 2001). It has also been used to evaluate effectsof changes in equipment, before modifications of equipment have been made.

– The importance of the calculated oxygen coefficient has become apparent and itsactive control is now in use. By doing this, formation of build-ups in the furnace can beminimized or even avoided completely. As a result, a higher on-line availability of theroaster has been achieved. Even up to about two years’ running time of the furnace hasbeen achieved and the furnace has remained in good condition, i.e. with only minor build-ups on the furnace grate.

– Opportunities for using fine concentrates have improved. Water is being used tomake agglomerates in the bins and conveying systems and to control the top temperatureof the furnace.

– There is now more knowledge about the effects of impurities and the optimal processconditions that are possible (Taskinen et al. 2002, Metsärinta et al. (2003, 2005a, 2005b)).Lower temperatures are necessary when the impurity levels of the feed mixture are high.The oxygen coefficient should not be too low in these cases. The limits for impurities,however, are difficult to determine exactly and they vary from one combination toanother. When the levels are too high there are no optimum conditions.

– Some new tools providing data about the state of the roasting furnace process havebeen implemented, of which SOM is one example (Saxén & Nyberg 2003).

7.1.2 Summary of boiler development

The modifications made in the boilers have proved to be effective. The running time forthe boiler was previously about 3 to 5 months between manual cleaning. Now the runningtime is 9 months or more. Over one year of running time without manual cleaning of theboiler has been achieved (Kilju 2004). The baffle wall in the radiation section has provedeffective both for cooling the gas stream and also for removing calcine from the gas flow.In addition, other equipment modifications have contributed to the success, such as neweffective bundles with effective hammering, hammering of the section walls in the lowerpart of the boiler, and installation of screens in the radiation section. Cleaning of theboiler from the inside with water during maintenance shutdowns also delays theformation of build-ups.

Longer running times and a more effective boiler reduce the constraints caused by theboiler, thus freedom in the control of furnace conditions increases, which in many cases isan advantage.

86

7.1.3 Summary of mercury removal development

The extensive thermodynamic calculations along with mass balance calculations,sampling and analyses, laboratory experiments and finally full-scale implementation haveincreased our knowledge of mercury removal. Sampling and analysis of mercury andother elements from the gas phase, where the levels of mercury and selenium are verylow, was a demanding task. The thermodynamic calculations and analyses of samplescorresponded well with each other. Both showed that gas phase of the process hadreached a steady state. The result of the research and development was a new controlmethod, i.e. the feeding of selenium to the process in liquid form. This selenium feedimproved the quality of the sulphuric acid, i.e. the mercury level can be kept low and inthe range of 0.02–0.06 mg Hg/kg H2SO4.

7.2 Control strategies and further research and development

Further research and development is necessary to gain more knowledge. The challengeslie in furnace behaviour, i.e. the roasting mechanism and particle size growth mechanismcertainly need more investigation. The necessity for plant experiments has been obvious.Two types of plant experiments have been done. In some experiments the emphasis hasbeen on studying the process metallurgy and in others on studying the dynamics. Both areneeded for developing the control of the furnace. In future the emphasis will be on doingmore experiments to study the process metallurgy. New measurements that show the stateof the furnace and development of new means of control are some targets for furtherresearch and development. The research and development must be wide-ranging and notfocus only on the control of the furnace, but also take other aspects into consideration.Pretreatment of concentrates is one example of this, which is perhaps in some cases anecessity (Brown & Goosen 1996). The challenge in furnace control is knowing thelimits of the impurities, and predicting the behaviour of different concentrates is not easy.

Installing devices that prevent formation of build-ups and removing build-ups is alsoincluded in the development program (Siirilä et al. 2004). The equipment can keep thearea of the slinger belt clean and prevent build-ups on the wall above the slinger belt inlet.Cleaning the furnace wall of build-ups during maintenance shutdowns can sometimes betime-consuming. Manual cleaning of the area at the top of the furnace between thefurnace and the boiler during operation is a safety issue and should therefore beautomated (Saarinen & Hugg 2000). Another challenge is to sustain reliable temperaturemeasurements in the furnace during long running periods and to develop systems toreplace malfunctioning devices during operation.

87

7.2.1 Increasing the capacity of the furnace

There are several ways to increase the capacity of the furnace (Nyberg et al. 2000). If thebed temperature in the furnace is increased, then the feed will increase. Higher bedtemperature also increases the sulphide level of the produced calcine, as is shown in Fig.62. This is partly due to formation of bigger particles at higher temperatures. Anadditional explanation for the higher sulphide level is that the oxygen coefficient is lowdue to the high feed. However, the danger of high temperatures is that molten phases areformed. This can lead to formation of build-ups, especially if the impurity levels in thefeed mixture are high (Appendix 1). Another way to increase the feed is to add more air,but the limits are set by the fluidisation behaviour of the bed. More air increases thecapacity, but the sulphide level of the calcine also increases in this case due to the loweroxygen coefficient, as the diagrams in Fig. 63 illustrate. Increasing the cooling capacityof the furnace is one possibility. This can be done by increasing the cooling capacity ofcooling coils (fixed) and/or by water injection to the bed (flexible control of the heatbalance). Feeding oxide Zn materials also increases the feed due to their cooling effect.Oxygen enrichment is used by some plants to increase the feed (Filho et al. 1998,Longton 1998, MacLagan et al. 2000, Saha et al. 1989). The increase of feed usingoxygen enrichment is shown in Fig. 64. The costs of using oxygen is, however, adisadvantage. Flexible cooling capacity is advantageous when oxygen enrichment isused.

Fig. 62. The feed increases when the temperature is higher in the furnace. The increased feeddecreases the oxygen coefficient and the result is that the sulphide level of the formed calcine ishigher, as shown here (Nyberg et al. 2000).

Fig. 63. By increasing the amount of air, the feed increases due to the cooling effect of the air(graph on right). Due to the lower oxygen coefficient the sulphide level of the calcine increases(graph on left). The furnace temperature in the experiments was 950°C (Nyberg et al. 2000).

0.00.10.20.30.40.50.60.70.80.9

925 930 935 940 945 950 955 960 965

Bed Temperature °C

Sulp

hide

in C

alci

ne w

t-%

Sulphide in Furnace Calcine

0.0

0.1

0.2

0.3

0.4

0.5

0.6

39000 40000 41000 42000 43000 44000 45000

Process Air Flow Rate Nm3/h

Sulp

hide

in C

alci

ne w

t-% Sulphide in Boiler Calcine

Sulphide in Furnace Calcine26

27

27

28

28

29

29

30

30

39000 40000 41000 42000 43000 44000 45000

Process Air Flow Rate Nm3/h

Con

cent

rate

and

Ste

am t/

h

Steam Production t/h

Concentrate Feed t/h

88

Fig. 64. By using oxygen the feed increases, due to the cooling effect of the oxygen. In the expe-riments with additional cooling (C and D), more cooling coils had been added to the furnace.The furnace temperature in the experiments was 950°C (Nyberg et al. 2000).

By keeping the top temperature of the furnace high, the feed increases and thereforethe oxygen coefficient decreases. The result is that the sulphate level of the formedcalcine going to WHB decreases, as shown in Fig. 65. As mentioned in chapter 4 thetarget is to have both optimal feed and optimal process conditions.

Fig. 65. Higher top temperature increases the feed and due to the lower oxygen coefficient thesulphate level of the formed calcine is lower, thus the sulphate load on the WHB decreases.

7.2.2 Example of flexible control

As mentioned earlier, the opportunities to arrange conditions in the furnace have beenincreased. Fig. 66 shows a situation when this flexibility can be used. For a certainconcentrate mixture, it might be better if the temperature is lower in the furnace. Case 1shows that the feed is A t/h, the air feed is B Nm3/h, the temperature is CoC and theoxygen coefficient is D. Case 2 shows that the same feeds and oxygen coefficient can bekept in the furnace if water (E l/h) is injected into the furnace bed. The desiredtemperature (C-X) oC can be achieved without decreasing the feed.

25

26

27

28

29

30

31

20.5 21 21.5 22 22.5 23 23.5 24

Ozygen Enrichment vol-%

Con

cent

rate

Fee

d t/h

Mixture D & Cooling

Mixture C & Cooling

Mixture B

Mixture A

7

8

9

10

11

12

13

14

780 800 820 840 860 880 900 920 940

WHB inlet temperature °C

Cyc

lone

SO

4 %

4

5

6

7

8

9

10

11

WH

B S

O4

%

Cyclone SO4 %WHB SO4 %Li (WHB SO4 %)

89

Fig. 66. In both cases the feeds and the oxygen coefficient are the same. By injecting water, thetemperature can be lowered in the furnace, without decreasing the concentrate feed. In caseswith high impurity levels, a temperature level of about 900°C might be necessary.

7.2.3 Operation conditions for different feed mixtures

For a good economic result it is important to be able to use different kinds ofconcentrates. Concentrates with lower Zn content are more viable economically, but theiruse is more demanding as is shown in Fig. 67. The operating range is narrower whenconcentrate mixtures with a high degree of impurities are used.

When the impurity levels of the concentrate mixture are very high then there are nooperation ranges, i.e. the furnace will inevitably become unstable. This will lead to theformation of build-ups. The difficulty is to know what are the upper limits for differentimpurities. The more impurities there are in the concentrate mixture, the more difficult itis to estimate the best operating conditions.

Flexible arrangement of the process conditions in the furnace Case 1:Concentrate feed ( A t/h)

Temperature of the furnace

C oCAir feed ( B Nm3/h)

Oxygen coefficientD

Case 2:Concentrate feed ( A t/h)

Temperature of the furnace

(C - X) oCAir feed ( B Nm3/h)

Oxygen coefficientWater injection into the bed D

E l/h

90

Fig. 67. The operating range in the furnace is narrower if the concentrate mixture contains highimpurity levels. The temperature and the oxygen coefficient are the main control variables.

7.2.4 The developed furnace control strategy

The study of the dynamics of the furnace and the construction of a state space model havegiven results that can be used for developing furnace control (Härkönen 2003, Saxén etal. 2004). The aim has been to build a control system that optimises the feed withoutneglecting the results of the metallurgical experiments. The structure of the furnacecontrol strategy that is in use is shown in Fig. 68. The target is to further raise theautomation level. The aim is to build a control application that both optimises the feedand creates optimal process conditions for the concentrate mixture based on thedeveloped control strategy (Valo 2004). The basis is to give a range both for the furnacetemperature and the oxygen coefficient. The optimal values for these (temperature andoxygen coefficient) are estimated based on the knowledge gained during thisdevelopment process. The control application then maximizes the feed. Measurements,analyses and knowledge about the state of the process and knowledge about the dynamicsof the furnace along with control variables are parts of the application.

Control of the furnace bed temperature and the oxygen coefficient

Different types of Safe operating conditions concentrate mixture

Temperature range is wide 900 – 970 oC. Clean and coarse concentrate mixture.

Oxygen coefficient range is wide 1 - 1.35.

Temperature range is narrow, too high temperatures can lead to forming of

Impure and fine molten phases. concentrate mixture.

Oxygen coefficient range is narrow, forming of molten phases is a danger. The optimal oxygen coefficient level depends on the impurity types (Pb, Cu) and amount.

91

Fig. 68. The figure shows the developed control strategy. The target is to build a control appli-cation that maximizes the feed based on this control strategy. The aim is to optimise the feedand the process conditions (temperature and oxygen coefficient), based on metallurgical know-ledge and knowledge of the dynamics. Measurements, analyses, the dynamic model and controlvariables are also key elements in the control application.

Many aspects must be considered in order to achieve the best result in roaster control.The entire process must be taken into account, thus in the control of the furnace boilerefficiency is also important. The control strategy could be the following:

– Optimal process conditions for a certain concentrate mixture should be chosen on thebasis of metallurgical knowledge.

– The aim should be to reach long running times for the furnace-boiler system, thusminimising the number and time of the shutdowns. Therefore it is better not to over-feed, if it leads to trouble and an unstable process. By reducing the number of shut-downs the impacts on the environment are less, and the maintenance costs are alsoreduced.

– The state of the process must always be monitored with care (measurements and ana-lyses) in order to make the corrections in advance.

– Control applications must be developed to maximise the feed rate, as shown in Fig.68.

– If control actions fail to keep the furnace stable, then the concentrate mixture must bechanged.

Some possible ways to further increase performance are mentioned below. Developingfuzzy control systems and/or expert systems based on the knowledge acquired is apossible development path (Ikonen 1996, Isomursu 1995, Viljamaa 2000). Optimisation

Concentrate mixture

Set point range for the Set point range for the furnace temperature oxygen coefficient

Max temperature Max oxygen coefficient Min temperature Min oxygen coefficient

Control application

Target: - maximising the feed - optimising the conditions

Metallurgical knowledge

Knowledge about the dynamics ofthe process anddynamic model

Controlvariables and limits

Measurementsand analyses

State of thefurnace

92

of fluid bed control has been carried out in energy-producing fluidised bed reactors(Leppäkoski et al. 1999, Leppäkoski & Kortela 1999). It could be useful to adopt theirideas, for example Combustion Power Control, or to use Learning Automata in thedevelopment (Leppäkoski et al 2000, Ikonen & Najim 1997, Najim & Ikonen 2002). Theaim is to use the rules learnt from metallurgical experiments and to use the control toolsavailable, e.g. water feed to different spots and using oxygen.

8 Conclusions

The research and development presented in this study has had the target of improving theperformance of the roaster. The main aim is to improve the economic result. To achievethis goal, the capability of being flexible in using different kinds of concentrates is themain sub-target. Other sub-targets are to decrease the amount of disturbances, keep thefurnace stable, increase the running time between maintenance shutdowns, and improvethe quality of the products. To achieve these sub-targets, equipment modifications, newcontrol variables, new measurements and new analyses have been adopted. Newinstructions for the operators, development of maintenance and also improvement of theperformance of the personnel have been the means employed to achieve these targets.

The implemented development concept, i.e. doing an extensive characterisation of theprocess has proved to be an effective way to increase the efficiency of the roastingprocess. The results of the process characterisation have led to the implementation of anew and effective control strategy.

The approach of the thesis has been to carry out extensive and intensive research inorder to learn more about the entire process and process phenomena, i.e. processcharacterisation. Process knowledge is the necessary base for making improvements inprocess control. New applications have been developed through gaining more knowledge.

The tools used to reach the targets are the development of control methods andimprovement of the process, both by equipment modifications and by adding new controlvariables. This improvement can be made by enhancing control, equipment and theprocess based on existing knowledge. Examples of such improvements are the fuzzysystems developed for the control of the furnace temperature and the control of mercuryremoval. These applications stabilise the processes, but are not always optimal in all casesand cannot solve all the process problems. Therefore these kinds of improvements are notenough and in this thesis the emphasis and the approach have been on improvingperformance based on increased knowledge. This is certainly the right approach if betterperformance, i.e. the optimal level, is desired.

There are several challenges for further development. More knowledge is still requiredabout the roasting mechanism. The use of concentrate mixtures with high levels ofimpurities needs further research. The difficulty is to know the limits and how these limits

94

can be tested safely, i.e. without jeopardising the stability of the furnace. Plantexperiments are necessary but they involve risks and are also expensive to perform.Therefore, they must be well planned and analysis of the results must be done with care.When high impurity levels are tested, the safest way to do these experiments is to do themjust before a shutdown. Then the risk of economic and production losses are smaller. Thechallenge is also to determine the optimal conditions. The optimal conditions must besuch that they enable long-term running of the roaster furnace, i.e. minimise the formationof build-ups. More measurements showing the state of the furnace should also bedeveloped and more control variables should be considered. In addition, laboratorystudies are needed in order to gain further knowledge about the roasting mechanism.

The impacts on the boiler system should not be neglected. Therefore, it is important tostudy what the impacts of the different control strategies and conditions in the furnace areon the boiler system. The need to have a better performing boiler may be required, if highimpurity levels in concentrate mixture are used.

Fig. 69. In order to achieve sustainable results, the issues shown in the diagram must be consi-dered in the development process.

There are also other issues that should not be neglected. Fig. 69 shows the issues thatmust be considered when the target is to increase the performance. Human resources(HR) should always be borne in mind during development work. Good performance canonly be achieved if the personnel are skilful. Maintenance must also be of high standardand the work done during shutdowns is of great importance (Cunningham & Connock2003). Longer running time also puts more pressure on maintenance work. Safety issuesare nowadays very important and this factor must also be considered. Environment issuesalong with quality systems are also issues that need more and more attention. Overall, toachieve sustainable results there are many issues that must be considered in thedevelopment process.

Benchmarking, visiting other zinc plants and participating in international conferencesshould also be part of development work, because valuable information can be obtainedfrom these kinds of activities (Brook Hunt 2003). In this development work, ideas and

Process knowledge (Metallurgy, Thermodynamics etc.)

Automation

Human Resources

Maintenance

ProcessEquipment

Safety Quality

Environment

95

knowledge gained from these sources have been one source for the R&D work and havein some cases led to implementation.

References

Abe Hideki, Yamada M, Yamaguchi K (1980) Method of removing mercury containingcontaminations in gases, US Patent No. US4206183.

Adams RW, Mangano P, Roche EG & Carpenter SJ (1990) Direct Leaching of Zinc Concentrates atAtmospheric Pressure, Proceedings of a world symposium on metallurgy and control, Lead~Zinc’90, TMS Annual Meeting, Anaheim, ISBN: 0-87339-111-X, p 351–372.

Ajersch F (2000) Roasting of Zinc Feeds, Proceedings of Lead-Zinc 2000 Symposium, Lead ZincShort Course, Pittsburgh, 37 p.

Anjala Y, Asteljoki J & Hanniala P (1987) The Role of Oxygen in the Outokumpu Flash SmeltingProcess, Proceedings of the Metallurgical Society of the Canadian Institute of Mining andMetallurgy, Vol. 2, edited by G Kachaniwsky & C Newman, Winnipeg, p 87–105.

Ashman DW & Jankola WA (1990) Recent Experience With Zinc Pressure Leaching at Cominco,Proceedings of a world symposium on metallurgy and control, Lead~Zinc ’90, TMS AnnualMeeting, Anaheim, p 253–275.

Asteljoki JA, Krogerus HJ & Mäkinen TSM (1989) Recent Pyrometallurgical Process Developmentsat Outokumpu, TMS-AIME Annual Meeting, Las Vegas, 17 p.

Asteljoki JA & Muller HB (1987) Direct Smelting of Blister Copper – Pilot Flash Smelting Tests ofOlympic Dam Concentrate, Pyrometallurgy ´87 Symposium, London, Publisher: The Institutionof Mining and Metallurgy (IMM), p 19–52.

Bendieck B, Schwab B & Schneider W-D (2002) Rechnergestutzes Integriertes Management System(IMS) zur Prozessoptimierung einer Zinkhutte, Metall, 56. Jahrgang, 6/2002, p 357–363.

Benlyamani M & Ajersch F (1986) Agglomeration of Particles during Roasting Of Zinc SulfideConcentrates, Metallurgical Transactions B, Volume 17B, p 647–656.

Berg A & Pape F (1978) Neuere Erfahrungen bei der Wirbelschichtröstung von Zinkkonzentraten,Erzmetall 31, p 224–228.

Berg D, Nyberg J, Rytioja A, Peltola H & Taskinen P (2003) Method for removing mercury fromsulphuric acid by means of thiosulfate precipitation, FIN Patent Application No 20031146.

Beveridge GSG (1962) The Oxidation Rates of Zinc Sulfide Spheres, Agglomeration InternationalSymposium, 1961, Philadelphia, Edited by WA Knepper, Interscience Publishers a division ofJohn Wiley & Sons, p 303–349.

Björklund P (2001) Development of the Heat and Material Balance Module in HSC ChemistrySoftware, Master’s Thesis, Helsinki University of Technology.

Brook Hunt (2003) Zinc Smelters and Projects: Processes, Costs & Profitability, Mining & MetalIndustry Consultants, England, Website: Brookhunt.com.

Brown MJ & Goosen DW (1996) Fluidized bed roasting process, UStates Patent No. 5 803 949.

97

Brown P, Box T, Brownlee K & Goldsworthy T (1995) Pilot Roasting of Zinc Concentrates, Proceed-ings of the International Symposium on the Extraction and Applications of Zinc and Lead,Symposium organized by The Mining & Materials Processing Institute of Japan and TheMetallurgical Society of CIM, Sendai, Zinc & Lead ’95, Edited by T. Azakami, N. Masuko, J.E.Dutrizac & E. Ozberk, p 239–249.

Buban KR, Collins MJ, Masters IM & Trytten LC (2000) Comparison of Direct Pressure Leachingwith Atmospheric Leaching of Zinc Concentrates, Proceedings of Lead-Zinc 2000 Symposium,TMS Fall Extraction & Process Metallurgy Meeting, The Minerals, Metals & Materials Society,edited by JE Dutrizac et al., Pittsburgh, p 727–738.

Canham DL & Charles JA (1981) Roasting of zinc-lead sulphide concentrates with structured pellets,Paper published in Trans. Instn Min. Metall. 90, p C35–C42.

Collins MJ, Doyle BN, Ozberk E & Masters IM (1990) The Zinc Pressure Leaching ProcessApplications, Lead~Zinc ’90, Proceedings of a World Symposium on Metallurgy and Control,TMS Annual Meeting, Anaheim, p 293–311.

Collins MJ, Masters IM, Ozberk E, Krysa BD & Desroches GJ (1995) Operation of the Sherritt ZincPressure Leach Process at the HBMS Refinery: The First Fifteen Months, Zinc & Lead ’95, Pro-ceedings of the International Symposium on the Extraction and Applications of Zinc and Lead,Sendai, edited by T Azakami et al., ISBN4-9900143-2-4, p 680–696.

Collins MJ, McConaghy EJ, Stauffer RF, Desroches GJ & Krysa BD (1994) Starting up the Sherrittzinc pressure leach process at Hudson Bay, Journal of Metals, p 51–58.

Condina VK, Jorgensen FRA & Turner TM (1980) Basic lead sulfates as agglomerating agents duringthe fluid-bed roasting of zinc concentrates, Metallurgical Transactions B, Volume 11B, p 637–638.

Constantineau JP, Grace JR, Richards GG & Lim CJ, (2002) Factors that influence particle sizedistribution in zinc fluidised bed roasters, Sulfide Smelting 2002, Proceedings of a symposiumsponsored by the Extraction and Processing Division of TMS, edited by RL Stephens & HY Sohn,TMS, Seattle, ISBN Number 0-87339-525-5, p 405–419.

Cunningham C & Connock L (2003) Sulphuric acid plant revamping, Sulphur No 284, published by:British Sulphur Publishing – a Division of CRU Publishing Ltd, www.britishsulphur.com, p 36–41.

De Nys T & Terwinghe F (1990) Integration of Direct Leaching of Zinc-Based Sulphidic Concen-trates and the Vielle-Montagne Goethite Process, Lead~Zinc ’90, Proceedings of a world sympo-sium on metallurgy and control, TMS Annual Meeting, Anaheim, p 335–351.

Dimitrov R, Hekimova A & Balabonov J (1981) Wirbelschichtrösten von Zinkkonzentraten[Fluidised Bed Roasting of Zinc Concentrates], Freiberg Forschungshefte B228, ISSN 0071-9420, p 199–210.

Dimitrov R & Schopov N (1985) Optimierung des Wirbelschichtreaktors zum Rösten vonSulfidkonzentraten [Optimization of Fluid Bed Reactor for Roasting of Sulfide Concentrates),Freiberg Forschungshefte B251, ISSN: 0071 9420, p 55–66.

Dreisinger DB, Peters E, Talaba M, DeGraaf KB, Owusu G & Swiniarski R (1990) The Kinetics ofthe Sherritt Gordon Zinc Pressure Leach Process, Lead~Zinc’90, Proceedings of a worldsymposium on metallurgy and control, TMS Annual Meeting, Anaheim, p 313–333.

Dyvik F (1994) Mercury elimination from process gases by the Boliden/Norzink Process, Edited byN J Roberts, International Minerals & Metals Technology, London, Sterling Publications, ISBN0-89520-358-8, p 160–162.

Ek M (1999) Rönnskär Smelter Expansion, Proceedings of the ninth International Flash SmeltingCongress, Australia, edited by Eila Nieminen, Outokumpu Technology Oy, p 85–94.

Elvers B & Hawkins S (1996) Pretreatment of Raw Materials for the Production of Zinc Metal,Ullmann’s Encyclopedia of Industrial Chemistry, Fifth, completely Revised Edition, Volume A28, Chapter 4, New York, VCH publishers. ISBN 3-527-20128-9, p 513–514.

98

Filho JWM, Filho TT & Welsh J (1998) Debottlenecking Experiences at Companhia Paraibuna deMetais (CPM’s) Smelter in Juiz de Fora, Proceedings of an International Symposium on Zinc andLead Prosessing, Edited by JE Dutrizac, JA Gonzalez, GL Bolton & P Hancock, TheMetallurgical Society of CIM, p 485–499.

Fleischer M & Mandarino JA (1991) Glossary of Mineral Species 1991, Fifth edition, Tucson,Arizona: The Mineralogical Record Inc., 256 p.

Freshcorn DE (1976) Effect of Calcium and Magnesium on Sulfate Sulfur Levels in Zinc Calcine,Proceedings of the Joint MMIJ-AIME Meeting, Denver, p 670–678.

Fugleberg S (1999) Finnish Expert Report on Best Available Techniques in Zinc Production,Helsinki, ISBN 952-11-0505-4, ISSN 1238-7312, Oy Edita Ab.

Fukunaka Y, Monta T, Azaki Z & Kondo Y (1976) Oxidation of zinc sulfide in a fluidized bed,Metallurgical Transactions B, Volume 7B, p 307–314.

Gales F & Winand R (1973) Contribution to the study of the kinetics and mechanism of sulfation ofzinc oxide, Metallurgical Transactions, Volume 4, p 883–885.

Geldart D (1986) Gas Fluidization Technology, University of Bradford, UK, A Wiley-IntersciencePublication, John Wiley & Sons, ISBN 0 471 90806 1, p 1–51.

Grant RM (1993) Zinc Concentrate and Processing Trends, World Zinc ’93, Proceedings of theInternational Symposium on Zinc, Hobart, Tasmania, Edited by Dr Ian G Matthew, TheAustralasian Institute of Mining and Metallurgy, ISBN 0 949106 92 5, p 391–397.

Grant RM (1994) Emerging Developments in Zinc Extraction Metallurgy, Proceedings of the SecondInternational Symposium on Metallurgical Processes for the Year 2000 and Beyond and the 1994TMS Extraction and Process Metallurgy Meeting held in San Diego, edited H Y Sohn, a Publica-tion of TMS, ISBN 0-87339-241-8, p 125–149.

Graydon JW & Kirk DW (1988a) A microscopic study of the transformation of sphalerite particlesduring roasting of zinc concentrate, Metallurgical Transactions B, Volume 19B, p 141–146.

Graydon JW & Kirk DW (1988b) The evidence for a miscibility gap in the Fe3O4-ZnFe2O4 system– A review, Metallurgical Transactions B, Volume 19B, p 919–925.

Graydon JW & Kirk DW (1988c) The mechanism of ferrite formation from iron sulfides during zincroasting, metallurgical transactions B, Volume 19B, p 777–785.

Hattori H, Miyatani J, Asaki Z & Kondo Y (1980) Oxidation of Zinc Sulfide in a Fluidised Bed ofVery Low Oxygen Concentration, Australia Japan Extractive Metallurgy Symposium, Sydney,published by The Australasian Institute of Mining and Metallurgy, p 223–232.

Heino KH, McAndrew RT, Ghatas NE & Morrison BH (1970) Fluid Bed Roasting of ZincConcentrate and Production of Sulphuric Acid and Phosphate Fertilizer at Canadian ElectrolyticZinc, Ltd. Valleyfield, Quebec, AIME World Symposium on Mining & Metallurgy of Lead &Zinc, Volume II, editors: CH Cotterill & JM Cigan, p 144–176.

Härkönen J (2003) Identification of a Zinc Roaster Furnace, Master Thesis, Åbo Akademi.Ikonen E (1996) Algorithms for Process Modelling Using Fuzzy Neural Networks, Department of

Process Engineering, University of Oulu, Acta Univ. Oul. C95, ISBN 951-42-4486-9.Ikonen E & Najim K (1997) Use of learning automata in distributed fuzzy logic processor training,

IEE Proceedings – Control Theory and Applications, Vol. 144, no. 3, p 255–262.Isomursu P (1995) A Software Engineering Approach to the Development of Fuzzy Control Systems,

Doctoral thesis, University of Oulu, ISBN 951-38-4768-3, 79 p.James SE, Watson JL & Peter J (2000) Zinc Production – A Survey of Existing Smelters and Refin-

eries, Proceedings of Lead-Zinc 2000 Symposium, TMS Fall Extraction & Process MetallurgyMeeting, The Minerals, Metals & Materials Society, edited by JE Dutrizac et al., Pittsburgh, p205–225.

Järvi J, Vaarno J, Nyberg J & Siirilä H (2002) Improvements of Zinc Roaster Waste Heat Boiler,Sulfide Smelting 2002, Proceedings of a symposium sponsored by the Extraction and ProcessingDivision of TMS, edited by R L Stephens & H Y Sohn, TMS, Seattle, p 421–431.

99

Kajiwara T, Iida T & Hino T (1995) A Survey of Operating Situation of Zinc Smelters and Refineriesin the World, Proceedings of the International Symposium on the Extraction and Applications onthe Extraction and Applications of Zinc and Lead, ”Zinc & Lead ‘95”, Sendai, publisher: TheMining and Materials Processing Institute of Japan, editors T Azakami et al., ISBN 4-9900143-2-4, p 27 –53.

Kilju M (2004) Working together for 35 years, Megawatt, The Magazine for Foster Wheeler PowerGroup Europe, Editor in chief Peter Herring, Spring 2004, www.fwc.com, p 14–15.

Krysa BD (1995) Zinc pressure leaching at HBMS, Hydrometallurgy 39, Elsevier Science, p 71–77.Kunii D & Levenspiel O (1969) Fluidization Engineering, Series in Chemical Engineering, editor H

Brenner, Massachusetts Institute of Technology, Butterworth-Heinemann, chapter 1 and chapter2, p 1–59.

Kytö SMI, Mäkinen JK & Saarhelo K (1989) On-line Analyzer of the Feed Mixture at the OutokumpuHarjavalta Copper Smelter, International Symposium on Process Control and Automation in Ex-tractive Metallurgy, TMS Annual Meeting, Las Vegas, editors E H Partelpoeg & D C Him-mesoete, ISBN 0-87339-083-0, p 19–40.

Lee RW (2000) The Continuing Evolution of the Imperial Smelting Process, Proceedings of Lead-Zinc 2000 Symposium, TMS Fall Extraction & Process Metallurgy Meeting, The Minerals,Metals & Materials Society, edited by J E Dutrizac et al., Pittsburgh, p 455–466.

Lepistö P (1999) Optimization of Capacity for the Zinc Roaster Furnace, Master of Science Thesis,University of Oulu, Systems Engineering Laboratory.

Leppäkoski K, Jouppila M & Kortela U (1999) An Expert System to Reduce Flue Gas Emissions ina Full Scale Fluidised Bed Boiler, 14th World Congress of IFAC, International Federation ofAutomatic Control, Beijing, Vol. O, p 357–360.

Leppäkoski K & Kortela U (1999) The Utilisation of an Expert System in Minimising Flue GasEmissions in a Full-Scale BFB Boiler, Proceedings of Second IASTED International ConferenceControl and Applications, Banff, Editor M H Hamza, ISBN: 0-88986-253.2, p 493–496.

Leppäkoski K, Mononen J, Kortela U & Kovacs J (2000) Integrated Optimization and ControlSystem to Reduce Flue Gas Emissions, 4th IFAC/CIGRE Symposium on Power Plants & PowerSystems Control 2000, Brussels, p 161–166.

Lightfoot R (1977) Fluid Bed Roasting of Zinc Concentrate at Risdon, Tasmania, The Aus.I.M.M.Conference, Tasmania, Published by The Australasian Institute of Mining and Metallurgy, p 359–365.

Longton RJ (1998) Productivity at Savage Zinc, Proceedings of an International Symposium on Zincand Lead Processing, Calgary, edited by JE Dutrizac, JA Gonzalez, GL Bolton & P Hancock, TheMetallurgical Society of CIM, p 471–483.

Ludtke P (2001) Advanced Sulphuric Acid Plants – Guarantors of the Future of Metal Smelters,Proceedings of EMC 2001 (European Metallurgical Conference), Friedrichshafen, Volume 1(Copper and Nickel, Zinc and Lead, Environment), p 333–341.

MacLagan C, Cloete M, Meyer EHO & Newall A (2000) Oxygen Enrichment of Fluo-Solids Roast-ing at Zincor, Proceedings of Lead-Zinc 2000 Symposium, TMS Fall Extraction & ProcessMetallurgy Meeting, The Minerals, Metals & Materials Society, edited by JE Dutrizac et al.,Pittsburgh, p 417–426.

Magoon DH, Metcalfe KA, Babcock AR & van Beek WA (1990) Equipment and Processes forRoasting Zinc Concentrates at Cominco’s Trail Zinc Smelter, Lead~Zinc ’90, Proceedings of aworld symposium on metallurgy and control, TMS Annual Meeting, Anaheim, p 389–412.

Mandarino JA (1999) Fleischer’s Glossary of Mineral Species 1999, 8th edition, Tucson:Mineralogical Record Inc., 225 p.

Masters IM, Piret NL & Short W (2002) Combined Dynatec and Ausmelt Technologies for ZincRefining, Erzmetall 55 No. 7, p 345–359.

100

Metsärinta M-L, Taskinen P, Jyrkönen S, Nyberg J & Rytioja A (2003) Roasting Mechanisms ofImpure Zinc Concentrates in Fluidized Beds, Yazawa International Symposium at TMS 2003Annual Meeting & Exhibition, edited by F Kongoli et al., TMS, Volume I, ISBN: 0-87339-546-8, San Diego, p 633–646.

Metsärinta M-L, Nyberg J, Ohvo E & Taskinen P (2005a) Industrial Scale Roasting of ImpureConcentrates at Boliden Kokkola, Abstract submitted and paper written for presentation at theLead & Zinc ’05 International Symposium on Lead & Zinc Processing, Kyoto.

Metsärinta M-L, Taskinen P, Nyberg J & Ohvo E (2005b) Mechanisms of Impure Zinc ConcentrateRoasting in Fluidized Beds, Paper submitted for Tsvetnye Metally, 2005.

Mollison AC & Moore GW (1990) Zinc Sulphide Pressure Leaching at Kidd Creek, Lead~Zinc’90,Proceedings of a world symposium on metallurgy and control, TMS Annual Meeting, Anaheim,edited by TS Mackey & RD Prengaman, p 277–291.

Morris AE (2000) Introduction to Lead and Zinc Processing Flowsheets, Proceedings of Lead-Zinc2000 Symposium, Lead Zinc Short Course, Pittsburgh, 51 p.

Mäkinen J (1972) Lic. Thesis, Helsinki University of Technology, 139 p.Mäkinen JK & Jåfs GA (1982) Production of matte, white metal, and blister copper by flash furnace,

Journal of Metals, p 54–59.Mukherjee AB (1999) Advanced Technology Available for the Abatement of Mercury Pollution in

the Metallurgical Industry, Environmental Science, Mercury Contaminated Sites, edited by REbinghaus, RR Turner, LD de Lacerda, O Vasiliev & W Salomons, Springer-Verlag BerlinHeidelberg, p 131–142.

Najim K & Ikonen E (2002) Analysis of Stochastic Iterative Algorithms, University of Oulu, SystemsEngineering Laboratory, Report C27, 85 p.

Nermes EO & Talonen TT (1982) Flash smelting of lead concentrates, Journal of Metals, p 55–59.Nyberg J (2001) The Development of the Zinc Roaster Department by Using System Based

Evaluation, Unpublished MBA Thesis, University of Vaasa, 99 p.Nyberg J, Metsärinta M-L & Roine A (2000) Recent Process Improvements in the Kokkola Zn-

Roaster, Proceedings of Lead-Zinc 2000 Symposium, TMS Fall Extraction & Process MetallurgyMeeting, The Minerals, Metals & Materials Society, edited by J E Dutrizac et al., Pittsburgh, p399–415.

Nyberg J, Siirilä H & Järvi J (2001) Arrangement and method for reducing build-up on a roastingfurnace grate, FIN Patent Application No 20010474, EPC Application No 02703646.2.6.

Ozberk E, Jankola WA, Vecchiarialli M & Krysa BD (1995) Commercial operations of the Sherrittzinc pressure leach process, Hydrometallurgy 39, Elsevier Science, p 49–52.

Parkkinen J, Lilja L, Metsärinta M-L & Saarenmaa J (2002) Method and apparatus for measuring gascomposition in fluidised bed, FIN Application No 20022266.

Peippo R, Holopainen H & Nokelainen J (1999) Copper Smelter Waste Heat Boiler Technology forthe Next Millennium, Proceedings of Copper 99-Cobre 99 International Conference, Volume V –Smelting Operations and Advances, edited by DB George, WJ (Pete) Chen, PJ Mackey & AJWeddick, The Minerals, Metals & Materials Society, ISBN: 0-87339-439-9, p 71–82.

Peippo R, Kemppainen S, Nokelainen J & Holopainen H (1998) Latest Developments in Zinc andLead Smelter Gas Cooling, Proceedings of an international symposium on zinc and leadprocessing, Calgary, The Canadian Institute of Mining, Metallurgy and Petroleum, edited by J EDutrizac et al., p 169–183.

Peltola H, Taskinen PA, Takala H, Nyberg J, Natunen H & Panula J (2000) Method for removingmercury from gas, FIN Application No 20002698, EPC Application No 01999417.7.

Rauma T (1999) Fuzzy modelling for industrial systems, Doctoral thesis. University of Oulu, ISBN951-38-5367-5, 97 p.

101

Rauma T, Nyberg J & Herronen J (2000) Fuzzy Control of Furnace Temperature in a Zinc Roaster,IFAC Workshop on Future Trends in Automation in Mineral and Metal Processing, edited by S-L Jämsä-Jounela & E Vapaavuori, IFAC Preprints, Helsinki, p 161–165.

Richards GG (1995) Reaction Kinetics of Zinc Roasting: Literature Review, Cominco Research,Research Report No. 95PR14, 19 p.

Rodolff DW, Anjala YE & Hanniala PT (1986) Review of Flash Smelting and Flash ConvertingTechnology, TMS Technical Papers from the 115th Annual Meeting, New Orleans, Soc. of AIME,ISSN: 0197-1689, 31 p.

Roine A & Nyberg J (2000) Heat and Material Balance Model of the Kokkola Zinc Roaster Based onthe HSC Chemistry 4.0 Software, General Non-Ferrous Pyrometallurgy, TMS 129th AnnualMeeting, edited by PR Taylor, The Minerals, Metals & Materials Society (TMS), Nashville, p 85–95.

Saarinen R & Hugg E (2000) Apparatus for removing dust accretions from a smelting furnace,Finnish Patent No. 109938.

Saha D, Scott Becker J & Gluns L (1989) Oxygen-Enhanced Fluid Bed Roasting, Proceedings of theInternational Symposium on Productivity and Technology in the Metallurgical Industries, co-sponsored by The Minerals, Metals & Materials Society and Gesellschaft Deutscher Metallhütten-und Bergleute, Cologne, edited by M. Koch & J.C. Taylor, a publication of TMS, p 257–280.

San Martin F, Tamargo F & Lefèvre Y (2000) Asturiana de ZincExpansion at the San Juan de NievaPlant for a Zinc Production of 440,000 Tonnes per Year, Proceedings of Lead-Zinc 2000Symposium, TMS Fall Extraction & Process Metallurgy Meeting, The Minerals, Metals &Materials Society, edited by JE Dutrizac et al., Pittsburgh, p 555–561.

Sarveswara Rao K & Ray HS (1999) Thermal analysis studies on multimetal sulphides: Characteri-sation, Roasting and Leaching, Trans. Indian Inst. Met., Vol. 52, No. 4, p 171–196.

Saxén B & Nyberg J (2003) Data based classification of roaster bed stability, European Symposiumon Computer Aided Process Engineering – 13, edited by A Kraslawski & I Turunen, ElsevierScience B.V., ISBN: 0-444-51368-X, Lappeenranta, p 503–508.

Saxén B, Nyberg J, Härkönen J & Toivonen H (2004) Dynamic Behaviour of a Zinc Roaster Furnace,11th IFAC Symposium on Automation in Mining, Mineral and Metal processing – MMM 2004,IFAC Preprints, http://www.cran.uhp-nancy.fr/ifac-mmm2004, Nancy.

Siirilä H (2000) Method for regulating a roasting furnace, FIN Pat No 109606, US Patent ApplicationNo 10/204937.

Siirilä H, Nyberg J, Saarinen R & Hugg E (2004) Apparatus for preventing build-ups formation andremoving formed build-ups at the slinger belt inlet area of a fluid bed roaster furnace, FIN PatentApplication No 20041026.

Siirilä H, Perälahti H, Nyberg J & Kartsalo V (2003) Method and apparatus for cooling the materialto be removed from the grate of a fluidized bed, FIN Patent Application No 20031146.

Svens K, Kerstiens B & Runkel M (2003) Recent experiences with modern zinc processingtechnology, Erzmetall 56 No. 2, p 94–103.

Takala H (1999) Leaching of zinc concentrates at Outokumpu Kokkola plant, Erzmetall 52 No.1, p37–42.

Takayama T, Magalhães W, Welsh J, Newton T & Thiele S J (2000) Expansion Plans at CPM’s Elec-trolytic Zinc Refinery, Proceedings of Lead-Zinc 2000 Symposium, TMS Fall Extraction &Process Metallurgy Meeting, The Minerals, Metals & Materials Society, edited by JE Dutrizac etal., Pittsburgh, p 261–276.

Taskinen P, Jyrkönen S, Metsärinta M-L, Nyberg J & Rytioja A (2002) The Behavior Of Iron AsImpurity In Zinc Roasting, Sulfide Smelting 2002, Proceedings of a symposium sponsored by theExtraction and Processing Division of TMS, edited by RL Stephens & HY Sohn, TMS, Seattle, p393–404.

102

Taskinen PA, Metsärinta M-L, Järvi J, Nyberg J & Siirilä H (2000a) Method for reducing build-upon a roasting furnace grate, FIN Patent No. 108732, US Application 10/416862.

Taskinen PA, Metsärinta M-L, Nyberg J & Rytioja A (2000b) Method for stabilization of a fluidizedbed in a roasting furnace, FIN Patent No. 111555, US Application 10/416863.

Themelis NJ & Freeman GM (1983 and 1984) Fluid Bed Behavior in Zinc Roasters, Paper presentedat 1983 AIME Annual Meeting, Atlanta, Georgia, 1983, Journal of Metals, August 1984, p 52–57.

Toskala M (2000) Design and Implementation of the Roaster Furnace Simulator, Master’s Thesis,University of Oulu, Systems Engineering Laboratory.

Toskala M, Kortela U, Nyberg J & Kovács J (2001) On-line Simulator of a Zinc Roaster Plant,Proceedings of the IASTED International Conference Modelling, Identification and Control (MIC2001), Edited by MH Hamza, a publication of IASTED, Volume II, ISBN: 0-88986-316-4,Innsbruck, p 735–739.

Valo J (2004) Simulation and Production Optimization of a Zinc Roaster Furnace, Master’s Thesis,University of Oulu, Systems Engineering Laboratory.

Viljamaa P (2000) Fuzzy Gain Scheduling and Tuning of Multivariable Fuzzy Control-Methods ofFuzzy Computing in Control Systems, Doctoral Thesis. Tampere University of Technology, ISBN952-15-0409-9, 101 p.

Yang Y (1996) Computer Simulation of Gas Flow and Heat Transfer in Waste-Heat Boilers of theOutokumpu Copper Flash Smelting Process, Acta Polytechnica Scandinavica, ChemicalTechnology Series No. 242, Helsinki University of Technology, ISBN 952-5148-06-8, 135 p.

Yazawa A, Eguchi M, Itagaki K, Kobayashi T, Shimizu T, Kimura T, Kato T, Seino T & Muto T(1980) Behavior of Minor Elements in the Roasting of Zinc Concentrate, Australia/JapanExtractive Metallurgy Symposium, Sydney, Published by The Australasian Institute of Miningand Metallurgy, p 199–210.

Yoneoka Y, Ohmizo Y, Ishiguro Y & Ochiai N (1998) Continuous Operation of the Roaster at theAnnaka Refinery, Proceedings of International Symposium on Zinc and Lead Processing, Cal-gary, editors J E Dutrizac, J A Gonzalez, G L Bolton, P Hancock, ISBN 0-919086-85-3, p 663–677.

Appendices

104

Appendix 1: Phase and Kellogg diagrams

In this work an evaluation of the phase and Kellogg diagrams has been done in order togain a picture of the phases that are stable or molten phases that can be formed duringzinc roasting conditions. It is essential to be aware of the dangers if the impurity levelsincrease and to seek optimal conditions from the study of these diagrams.

From the phase diagrams it can be seen that different kinds of melts (sulfide melts,oxide melts, sulfate and silicate melts) are formed in conditions typical of a zinc roaster.The probability of melts increases as the impurity levels increase. From the Kelloggdiagrams the stable phases for different elements can be seen. In these diagrams thevariables are temperature, SO2 pressure and O2 pressure.

Some of the basic diagrams are shown below. The data sources for the diagrams aregiven in the reference list at the end of this appendix.

Fig. 1. Zn-S-O phase stability diagram made using the HSC program at fixed temperature. Intypical roasting conditions ZnO is stable (Roine 2002).

0-2-4-6-8-10

2

1

0

-1

-2

-3

-4

-5

Zn-S -O Phase Stability Diagram at 940.000 C

File: C:\HSC5\Lpp\ZnSO940.ips log pO2(g)

log pSO2(g)

ZnO

ZnO*2ZnSO4

ZnS ZnSO4

Zinc RoastingSO2 ~ 10 %O2 ~ 5 %

105

Fig. 2. Predominance diagram for Zn-O-S at fixed oxygen pressure made using HSC (Roine2002). The oval area marks the typical operation conditions for the furnace. Decreasing thetemperature (boiler operation conditions) will result in formation of ZnSO4.

Fig. 3. Cu-S-O phase stability diagram made using the HSC program at fixed temperature. Intypical roasting conditions CuO and Cu2O are stable (Roine 2002).

11001000900800700600

2

1

0

-1

-2

-3

-4

-5

log pSO2(g)

T / °CConstant value:pO2(g) = 5.00E-02

Predominance Diagram for Zn-O-S System

ZnOZnO

ZnO*2ZnSO4ZnO*2ZnSO4

ZnSO4ZnSO4

0-2-4-6-8-10

2

1

0

-1

-2

-3

-4

-5

Cu-O -S Phase Stability Diagram at 940.000 C

File: C:\HSC5\Lpp\CuOS940.ips log pO2(g)

log pSO2(g)

Cu CuOCu2O

CuO*CuSO4Cu2S

CuSO4

Zinc RoastingSO2 ~ 10 %O2 ~5 %

106

Fig. 4. Predominance diagram for Cu-O-S at fixed oxygen pressure made using HSC. The ovalarea marks the typical operation conditions for the furnace. Decreasing the temperature(boiler operation conditions) will result in formation of CuSO4 (Roine 2002).

Fig. 5. Pb-S-O phase stability diagram made using HSC program at fixed temperature. Intypical roasting conditions PbSO4 is stable (Roine 2002).

11001000900800700600

2

1

0

-1

-2

-3

-4

-5

log pSO2(g)

T / °CConstant value:pO2(g) = 5.00E-02

Predominance Diagram for Cu-O-S System

CuOCuO Cu2OCu2O

CuO*CuSO4CuO*CuSO4

CuSO4CuSO4

0-2-4-6-8-10

2

1

0

-1

-2

-3

-4

-5

Pb-O -S Phase Stability Diagram at 940.000 C

File: C:\HSC5\Lpp\PbOS940.ips log pO2(g)

log pSO2(g)

Pb

PbO

PbO*PbSO4

*2PbO*PbSO4

*3PbO*PbSO4

PbSPbSO4

Zinc RoastingSO2 ~10 %O2 ~ 5 %

107

Fig. 6. Predominance diagram for Pb-O-S at fixed oxygen pressure made using HSC (Roine2002). The oval area marks the typical operation conditions for the furnace. PbSO4 remainsstable also when the temperature decreases (boiler operation conditions).

Fig. 7. Cu2O-PbO phase diagram (Gebhart & Obrowski 1964). The danger is that moltenphases are formed even if the temperature is low when there are high concentrations of Cuand Pb present.

11001000900800700600

2

1

0

-1

-2

-3

-4

-5

log pSO2(g)

T / °CConstant value:pO2(g) = 5.00E-02

Predominance Diagram for Pb-O-S System

PbOPbO

PbO*PbSO4PbO*PbSO4

*2PbO*PbSO4*2PbO*PbSO4

*3PbO*PbSO4*3PbO*PbSO4

PbSO4PbSO4

108

Fig. 8. PbSO4-PbO phase diagram (Margulis & Kopylov 1969). The danger of molten phasesincreases when the temperature is high and the concentration of Pb is high.

Fig. 9. Zn2SiO4 and PbZnSiO4 liquids (Jak et al. 1999). The danger of molten phases ispresent when there are high concentrations of silica present.

109

Fig. 10. Na2O-SiO2 system (PED Database 1999). High concentrations of Na and Si form mol-ten phases.

Fig. 11. PbO-ZnO-Fe-O system (Fe 1 %, pO2 =1 %) (Davies et al 2002).

110

Reference list for the sources of the diagrams:Gebhardt E & Obrowski W (1964) In:Phase Diagrams for Ceramists. Ohio, The American Ceramic

Society, p.86.Margulis E V & Kopylov N I (1969) In:Phase Diagrams for Ceramists 1969 Supplement. Ohio: The

American Ceramic Society, p.261.Jak E, Detegterov S A, Wu P, Hayes P C & Pelton A (1999) Thermodynamic Optimization of the

Systems PbO-SiO2, PbO-ZnO, ZnO-SiO2 and PbO-ZnO-SiO2”, Met Trans B, vol 28 B, pp.1011-1018.

Davies R H, Dinsdale A T, JA Gisby J A, Robinson J A J & Martin S M (2002) MTDATA - thermo-dynamic and phase equilibrium software from the national physical laboratory, Calphad Vol. 26,Issue 2, pp. 229-271.

PED Database (1998) vers. 2.1. Columbus, Ohio (USA): The American Ceramic Society.Roine A (2002) Outokumpu HSC Chemistry for Windows-Chemical Reaction and Equi-

librium Software with Extensive Thermochemical Database, User’s Guide, Version5.0”, (Pori, Finland: Outokumpu Research Oy).

111

Appendix 2: The zinc roaster of Kokkola zinc plant (The equipment of one line)

112

Appendix 3: Flowsheet of the Kokkola zinc plant

113

Appendix 4: Analyses of samples from the process

To better understand the process it is necessary to take samples for chemical analysis.Besides the daily samples, extra samples have also been taken in order to get a betterpicture of the process. Extra samples have been taken in particular during the plant trialsand when the process has been unstable. During shutdowns samples have been takenfrom the build-ups in the furnaces and the boilers. Besides chemical analysis, it issometimes necessary to do X-ray analysis too. In order to draw the right conclusionsabout what is going on in the process or what has happened, it is necessary to collect dataover a longer period of time. Gradually more and more knowledge is built up.

Analyses from samples taken during operation and shutdowns are shown in thefollowing tables, i.e. in Tables 3-5. It can be seen that are some differences. Table 3, takenduring normal operation, shows that the sulphate level is lower in the furnace than later inthe process chain. The highest sulphate levels are in the ESP. From the samples takenduring shutdowns, as shown in Tables 4-5, the sulphate levels are very high. This showsthat the build-ups that were formed gradually develop to sulphates.

Table 3. Analyses taken from daily operation show that the sulphate level of the formedcalcine is lower in the furnace than in the boiler, cyclone and ESP. The Pb level has alsobeen quite high in the furnace calcine during some periods.

Date Zn S Fe Pb Cu Si K Na Ca As H2O Sulphide SO4sulphur

% % % % % % % % % % % % %

10.4.2002 Concentrate 54.3 30.2 5.5 2.8 0.12 0.60 0.06 0.03 0.72 0.10 11.915.4.2002 Concentrate 53.7 30.5 5.5 2.8 0.11 0.58 0.09 0.03 0.70 0.10 9.49.9.2003 Concentrate L1 54.5 32.7 8.8 0.7 0.50 0.13 0.06 0.32 0.16 8.19.9.2003 Concentrate L2 55.9 32.7 7.8 0.8 0.62 0.11 0.05 0.40 0.07 8.1

10.4.2002 Furnace ovf. 64.4 1.8 6.4 3.6 0.11 0.84 0.07 0.04 0.72 0.08 0.84 3.9calcine

15.4.2002 Furnace ovf. 64.0 1.9 6.1 4.1 0.10 0.71 0.07 0.05 0.90 0.10 0.41 4.4calcine

9.9.2003 Furnace ovf. 61.0 1.2 9.2 1.4 0.63 0.13 0.06 0.68 0.07 0.28 3.0calcine L1

9.9.2003 Furnace ovf. 62.1 1.1 9.1 1.2 0.65 0.13 0.06 0.61 0.05 0.22 2.8calcine L2

10.4.2002 Boiler 64.2 3.4 6.0 3.4 0.16 0.76 0.08 0.04 0.86 0.11 0.54 8.6calcine

15.4.2002 Boiler 62.9 3.5 5.1 3.0 0.13 0.71 0.10 0.05 1.10 0.15 0.15 10.0calcine

9.9.2003 Boiler 60.6 3.0 8.2 1.0 0.69 0.07 0.04 0.50 0.08 0.35 8.1calcine L1

9.9.2003 Boiler 60.3 3.5 8.1 0.9 0.69 0.12 0.06 0.49 0.07 0.09 10.5calcine L2

10.4.2002 Cyclone 60.3 3.6 6.1 3.4 0.17 0.67 0.08 0.05 0.83 0.19 0.54 9.4calcine

15.4.2002 Cyclone 60.7 4.3 5.4 3.0 0.15 0.62 0.10 0.05 0.94 0.18 0.18 12.5calcine

9.9.2003 Cyclone 60.3 3.2 8.5 1.0 0.71 0.13 0.06 0.46 0.08 0.29 9.0calcine L1

9.9.2003 Cyclone 61.0 3.1 8.3 0.9 0.72 0.11 0.05 0.46 0.07 0.09 9.3calcine L2

10.4.2002 ESP calcine 61.2 4.3 6.1 3.1 0.16 0.67 0.10 0.05 0.72 0.14 0.42 11.615.4.2002 ESP calcine 57.6 6.1 6.4 2.5 0.17 0.68 0.10 0.05 0.48 0.18 0.22 17.59.9.2003 ESP calcine L1 57.9 4.5 8.2 1.0 0.67 0.14 0.06 0.47 0.08 0.19 13.29.9.2003 ESP calcine L2 58.2 4.1 8.5 1.0 0.67 0.09 0.04 0.45 0.08 0.11 12.3

L1 = Roasting line no. 1L2 = Roasting line no. 2ovf. = Overflow

114

Table 4. Samples taken during a shutdown from the furnace and boiler build-ups showthat the sulphate levels are very high, i.e. the build-ups develop gradually into sulphates.One build-up sample from the furnace (19.3.2000/second sample) shows, however, a veryhigh Pb level, while the sulphate level is lower.

Table 5. The table shows the analyses of furnace shutdown samples taken during theyearly shutdowns. The values for the different years show how wide the range of analysisresults has been.

The difficulty, in shutdown samples, is that the equipment (furnace and boiler) is large,the build-ups might be large and the build-ups have been generated during a long period.It is often difficult to take a proper sample. It might be difficult to know from what periodit is. It is often difficult to find the right explanation. However it is essential to takesamples and to analyse them in order to get more information. The result of the intensivecollection of samples at the Kokkola roaster has been that gradually more and moreprocess knowledge has been gained. Whenever there is instability in the process, samplesare taken and analysed. Then the results are compared with similar situations in the past.Step by step more knowledge has been gained and the control actions to correct theprocess have been taken based on this knowledge. Over the last few years, as the result ofimproved furnace control, the amount of build-ups in the furnace grate area hasdecreased. In the shutdown of spring 2004 there were almost no build-ups in the grateareas of the two furnaces.

Date Zn Cu Pb Cd Fe SO4 S2- Ca Mg Al Si As Na K% % % % % % % % % % % % % %

19.3.2000 Furnace 1 55 1.1 0.54 0.29 8.7 21.3 0.01 0.12 0.06 0.13 0.5 0.06 0.07 0.2819.3.2000 Furnace 1 41 1.5 19 3 9.1 15.6 0.01 0.22 0.07 0.13 0.09 0.21 0.05 0.1820.3.2000 Furnace 2 43 0.69 1.5 0.15 6.4 40.5 0.01 0.32 0.11 0.14 0.84 0.14 0.04 0.1720.3.2000 Furnace 2 44 0.65 1.6 0.17 6.6 39 0.01 0.34 0.11 0.11 0.59 0.12 0.07 0.26

1998 WHB 1 41 0.43 1.9 0.27 4.4 39 0.02 0.6 0.2 0.15 0.5 0.1 0.071998 WHB 1 32 2.8 2.4 0.27 3.6 50.7 0.02 0.5 0.2 0.15 0.2 0.2 0.11998 WHB 2 38 0.6 1.6 0.21 3.5 49.8 0.02 0.4 0.2 0.1 0.31 0.2 0.11998 WHB 2 46 0.6 1.4 0.21 7 34.2 0.13 0.4 0.2 0.09 0.22 0.15 0.1

1999 2000 2001

Zn % 33 - 34 41 - 69 45 - 70Cu % 0.33 - 0.35 0.62 - 1.5 0.24 - 0.58Pb % 1.2 0.54 - 19 1.5 - 2.8Cd % 0.11 - 0.14 0.17 - 6.8 0.08 - 0.25Fe % 4.4 - 4.9 7.1 - 11.8 4.1 - 8.2

SO4 % 54 1.1 - 21 3.6 - 36Ca % 0.34 - 0.44 0.12- 0.56 0.45 - 0.84Mg % 0.15 - 0.16 0.04 - 0.12 0.09 - 0.15Al % 0.12 - 0.19 0.14 - 0.18Si % 0.43 0.01 - 0.68 0.41 - 0.74Na % 0.01 0.05 - 0.1 0.07 - 0.09K % 0.01- 0.02 0.18 - 0.33 0.06 - 0.14